"python-package/vscode:/vscode.git/clone" did not exist on "6f459cee92bd8566be43089b45a0f438976c9961"
basic.py 140 KB
Newer Older
wxchan's avatar
wxchan committed
1
# coding: utf-8
2
"""Wrapper for C API of LightGBM."""
3
import copy
wxchan's avatar
wxchan committed
4
import ctypes
5
import json
6
import os
wxchan's avatar
wxchan committed
7
import warnings
wxchan's avatar
wxchan committed
8
from tempfile import NamedTemporaryFile
9
from collections import OrderedDict
wxchan's avatar
wxchan committed
10
11
12
13

import numpy as np
import scipy.sparse

14
from .compat import PANDAS_INSTALLED, DataFrame, Series, is_dtype_sparse, DataTable
wxchan's avatar
wxchan committed
15
16
from .libpath import find_lib_path

wxchan's avatar
wxchan committed
17

18
19
def _log_callback(msg):
    """Redirect logs from native library into Python console."""
20
    print("{0:s}".format(msg.decode('utf-8')), end='')
21
22


wxchan's avatar
wxchan committed
23
def _load_lib():
24
    """Load LightGBM library."""
wxchan's avatar
wxchan committed
25
26
    lib_path = find_lib_path()
    if len(lib_path) == 0:
27
        return None
wxchan's avatar
wxchan committed
28
29
    lib = ctypes.cdll.LoadLibrary(lib_path[0])
    lib.LGBM_GetLastError.restype = ctypes.c_char_p
30
31
32
    callback = ctypes.CFUNCTYPE(None, ctypes.c_char_p)
    lib.callback = callback(_log_callback)
    if lib.LGBM_RegisterLogCallback(lib.callback) != 0:
33
        raise LightGBMError(lib.LGBM_GetLastError().decode('utf-8'))
wxchan's avatar
wxchan committed
34
35
    return lib

wxchan's avatar
wxchan committed
36

wxchan's avatar
wxchan committed
37
38
_LIB = _load_lib()

wxchan's avatar
wxchan committed
39

40
41
42
NUMERIC_TYPES = (int, float, bool)


wxchan's avatar
wxchan committed
43
def _safe_call(ret):
44
45
    """Check the return value from C API call.

wxchan's avatar
wxchan committed
46
47
48
    Parameters
    ----------
    ret : int
49
        The return value from C API calls.
wxchan's avatar
wxchan committed
50
51
    """
    if ret != 0:
52
        raise LightGBMError(_LIB.LGBM_GetLastError().decode('utf-8'))
wxchan's avatar
wxchan committed
53

wxchan's avatar
wxchan committed
54

wxchan's avatar
wxchan committed
55
def is_numeric(obj):
56
    """Check whether object is a number or not, include numpy number, etc."""
wxchan's avatar
wxchan committed
57
58
59
    try:
        float(obj)
        return True
wxchan's avatar
wxchan committed
60
61
62
    except (TypeError, ValueError):
        # TypeError: obj is not a string or a number
        # ValueError: invalid literal
wxchan's avatar
wxchan committed
63
64
        return False

wxchan's avatar
wxchan committed
65

wxchan's avatar
wxchan committed
66
def is_numpy_1d_array(data):
67
    """Check whether data is a numpy 1-D array."""
68
    return isinstance(data, np.ndarray) and len(data.shape) == 1
wxchan's avatar
wxchan committed
69

wxchan's avatar
wxchan committed
70

wxchan's avatar
wxchan committed
71
def is_1d_list(data):
72
73
    """Check whether data is a 1-D list."""
    return isinstance(data, list) and (not data or is_numeric(data[0]))
wxchan's avatar
wxchan committed
74

wxchan's avatar
wxchan committed
75

76
def list_to_1d_numpy(data, dtype=np.float32, name='list'):
77
    """Convert data to numpy 1-D array."""
wxchan's avatar
wxchan committed
78
79
80
81
82
83
84
    if is_numpy_1d_array(data):
        if data.dtype == dtype:
            return data
        else:
            return data.astype(dtype=dtype, copy=False)
    elif is_1d_list(data):
        return np.array(data, dtype=dtype, copy=False)
85
    elif isinstance(data, Series):
86
87
        if _get_bad_pandas_dtypes([data.dtypes]):
            raise ValueError('Series.dtypes must be int, float or bool')
88
        return np.array(data, dtype=dtype, copy=False)  # SparseArray should be supported as well
wxchan's avatar
wxchan committed
89
    else:
90
91
        raise TypeError("Wrong type({0}) for {1}.\n"
                        "It should be list, numpy 1-D array or pandas Series".format(type(data).__name__, name))
wxchan's avatar
wxchan committed
92

wxchan's avatar
wxchan committed
93

wxchan's avatar
wxchan committed
94
def cfloat32_array_to_numpy(cptr, length):
95
    """Convert a ctypes float pointer array to a numpy array."""
wxchan's avatar
wxchan committed
96
    if isinstance(cptr, ctypes.POINTER(ctypes.c_float)):
97
        return np.fromiter(cptr, dtype=np.float32, count=length)
wxchan's avatar
wxchan committed
98
    else:
99
        raise RuntimeError('Expected float pointer')
wxchan's avatar
wxchan committed
100

Guolin Ke's avatar
Guolin Ke committed
101

Guolin Ke's avatar
Guolin Ke committed
102
def cfloat64_array_to_numpy(cptr, length):
103
    """Convert a ctypes double pointer array to a numpy array."""
Guolin Ke's avatar
Guolin Ke committed
104
105
106
107
108
    if isinstance(cptr, ctypes.POINTER(ctypes.c_double)):
        return np.fromiter(cptr, dtype=np.float64, count=length)
    else:
        raise RuntimeError('Expected double pointer')

wxchan's avatar
wxchan committed
109

wxchan's avatar
wxchan committed
110
def cint32_array_to_numpy(cptr, length):
111
    """Convert a ctypes int pointer array to a numpy array."""
wxchan's avatar
wxchan committed
112
    if isinstance(cptr, ctypes.POINTER(ctypes.c_int32)):
113
        return np.fromiter(cptr, dtype=np.int32, count=length)
wxchan's avatar
wxchan committed
114
    else:
115
116
117
118
119
120
121
122
123
        raise RuntimeError('Expected int32 pointer')


def cint64_array_to_numpy(cptr, length):
    """Convert a ctypes int pointer array to a numpy array."""
    if isinstance(cptr, ctypes.POINTER(ctypes.c_int64)):
        return np.fromiter(cptr, dtype=np.int64, count=length)
    else:
        raise RuntimeError('Expected int64 pointer')
wxchan's avatar
wxchan committed
124

wxchan's avatar
wxchan committed
125

wxchan's avatar
wxchan committed
126
def c_str(string):
127
    """Convert a Python string to C string."""
wxchan's avatar
wxchan committed
128
129
    return ctypes.c_char_p(string.encode('utf-8'))

wxchan's avatar
wxchan committed
130

wxchan's avatar
wxchan committed
131
def c_array(ctype, values):
132
    """Convert a Python array to C array."""
wxchan's avatar
wxchan committed
133
134
    return (ctype * len(values))(*values)

wxchan's avatar
wxchan committed
135

136
137
138
139
140
141
142
143
144
145
def json_default_with_numpy(obj):
    """Convert numpy classes to JSON serializable objects."""
    if isinstance(obj, (np.integer, np.floating, np.bool_)):
        return obj.item()
    elif isinstance(obj, np.ndarray):
        return obj.tolist()
    else:
        return obj


wxchan's avatar
wxchan committed
146
def param_dict_to_str(data):
147
    """Convert Python dictionary to string, which is passed to C API."""
148
    if data is None or not data:
wxchan's avatar
wxchan committed
149
150
151
        return ""
    pairs = []
    for key, val in data.items():
152
        if isinstance(val, (list, tuple, set)) or is_numpy_1d_array(val):
153
154
155
156
157
158
            def to_string(x):
                if isinstance(x, list):
                    return "[{}]".format(','.join(map(str, x)))
                else:
                    return str(x)
            pairs.append(str(key) + '=' + ','.join(map(to_string, val)))
159
        elif isinstance(val, (str, NUMERIC_TYPES)) or is_numeric(val):
wxchan's avatar
wxchan committed
160
            pairs.append(str(key) + '=' + str(val))
161
        elif val is not None:
162
            raise TypeError('Unknown type of parameter:%s, got:%s'
wxchan's avatar
wxchan committed
163
164
                            % (key, type(val).__name__))
    return ' '.join(pairs)
165

wxchan's avatar
wxchan committed
166

167
class _TempFile:
168
169
170
171
    def __enter__(self):
        with NamedTemporaryFile(prefix="lightgbm_tmp_", delete=True) as f:
            self.name = f.name
        return self
wxchan's avatar
wxchan committed
172

173
174
175
    def __exit__(self, exc_type, exc_val, exc_tb):
        if os.path.isfile(self.name):
            os.remove(self.name)
wxchan's avatar
wxchan committed
176

177
178
179
180
    def readlines(self):
        with open(self.name, "r+") as f:
            ret = f.readlines()
        return ret
wxchan's avatar
wxchan committed
181

182
183
    def writelines(self, lines):
        with open(self.name, "w+") as f:
184
            f.writelines(lines)
185

wxchan's avatar
wxchan committed
186

187
class LightGBMError(Exception):
188
189
    """Error thrown by LightGBM."""

190
191
192
    pass


193
194
195
196
197
198
199
200
# DeprecationWarning is not shown by default, so let's create our own with higher level
class LGBMDeprecationWarning(UserWarning):
    """Custom deprecation warning."""

    pass


class _ConfigAliases:
201
202
203
    aliases = {"bin_construct_sample_cnt": {"bin_construct_sample_cnt",
                                            "subsample_for_bin"},
               "boosting": {"boosting",
204
205
206
207
208
209
                            "boosting_type",
                            "boost"},
               "categorical_feature": {"categorical_feature",
                                       "cat_feature",
                                       "categorical_column",
                                       "cat_column"},
210
211
               "data_random_seed": {"data_random_seed",
                                    "data_seed"},
212
213
214
215
               "early_stopping_round": {"early_stopping_round",
                                        "early_stopping_rounds",
                                        "early_stopping",
                                        "n_iter_no_change"},
216
217
218
               "enable_bundle": {"enable_bundle",
                                 "is_enable_bundle",
                                 "bundle"},
219
220
221
222
223
               "eval_at": {"eval_at",
                           "ndcg_eval_at",
                           "ndcg_at",
                           "map_eval_at",
                           "map_at"},
224
225
226
227
228
229
               "group_column": {"group_column",
                                "group",
                                "group_id",
                                "query_column",
                                "query",
                                "query_id"},
230
231
               "header": {"header",
                          "has_header"},
232
233
234
235
236
237
238
239
240
               "ignore_column": {"ignore_column",
                                 "ignore_feature",
                                 "blacklist"},
               "is_enable_sparse": {"is_enable_sparse",
                                    "is_sparse",
                                    "enable_sparse",
                                    "sparse"},
               "label_column": {"label_column",
                                "label"},
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
               "machines": {"machines",
                            "workers",
                            "nodes"},
               "metric": {"metric",
                          "metrics",
                          "metric_types"},
               "num_class": {"num_class",
                             "num_classes"},
               "num_iterations": {"num_iterations",
                                  "num_iteration",
                                  "n_iter",
                                  "num_tree",
                                  "num_trees",
                                  "num_round",
                                  "num_rounds",
                                  "num_boost_round",
                                  "n_estimators"},
               "objective": {"objective",
                             "objective_type",
                             "app",
                             "application"},
262
263
264
265
266
               "pre_partition": {"pre_partition",
                                 "is_pre_partition"},
               "two_round": {"two_round",
                             "two_round_loading",
                             "use_two_round_loading"},
267
               "verbosity": {"verbosity",
268
269
270
                             "verbose"},
               "weight_column": {"weight_column",
                                 "weight"}}
271
272
273
274
275

    @classmethod
    def get(cls, *args):
        ret = set()
        for i in args:
276
            ret |= cls.aliases.get(i, {i})
277
278
279
        return ret


280
281
MAX_INT32 = (1 << 31) - 1

282
"""Macro definition of data type in C API of LightGBM"""
wxchan's avatar
wxchan committed
283
284
285
286
C_API_DTYPE_FLOAT32 = 0
C_API_DTYPE_FLOAT64 = 1
C_API_DTYPE_INT32 = 2
C_API_DTYPE_INT64 = 3
Guolin Ke's avatar
Guolin Ke committed
287

288
"""Matrix is row major in Python"""
wxchan's avatar
wxchan committed
289
290
C_API_IS_ROW_MAJOR = 1

291
"""Macro definition of prediction type in C API of LightGBM"""
wxchan's avatar
wxchan committed
292
293
294
C_API_PREDICT_NORMAL = 0
C_API_PREDICT_RAW_SCORE = 1
C_API_PREDICT_LEAF_INDEX = 2
295
C_API_PREDICT_CONTRIB = 3
wxchan's avatar
wxchan committed
296

297
298
299
300
"""Macro definition of sparse matrix type"""
C_API_MATRIX_TYPE_CSR = 0
C_API_MATRIX_TYPE_CSC = 1

301
302
303
304
"""Macro definition of feature importance type"""
C_API_FEATURE_IMPORTANCE_SPLIT = 0
C_API_FEATURE_IMPORTANCE_GAIN = 1

305
"""Data type of data field"""
wxchan's avatar
wxchan committed
306
307
FIELD_TYPE_MAPPER = {"label": C_API_DTYPE_FLOAT32,
                     "weight": C_API_DTYPE_FLOAT32,
Guolin Ke's avatar
Guolin Ke committed
308
                     "init_score": C_API_DTYPE_FLOAT64,
309
                     "group": C_API_DTYPE_INT32}
wxchan's avatar
wxchan committed
310

311
312
313
314
"""String name to int feature importance type mapper"""
FEATURE_IMPORTANCE_TYPE_MAPPER = {"split": C_API_FEATURE_IMPORTANCE_SPLIT,
                                  "gain": C_API_FEATURE_IMPORTANCE_GAIN}

wxchan's avatar
wxchan committed
315

316
def convert_from_sliced_object(data):
317
    """Fix the memory of multi-dimensional sliced object."""
318
    if isinstance(data, np.ndarray) and isinstance(data.base, np.ndarray):
319
        if not data.flags.c_contiguous:
320
321
            warnings.warn("Usage of np.ndarray subset (sliced data) is not recommended "
                          "due to it will double the peak memory cost in LightGBM.")
322
323
324
325
            return np.copy(data)
    return data


wxchan's avatar
wxchan committed
326
def c_float_array(data):
327
    """Get pointer of float numpy array / list."""
wxchan's avatar
wxchan committed
328
329
330
    if is_1d_list(data):
        data = np.array(data, copy=False)
    if is_numpy_1d_array(data):
331
332
        data = convert_from_sliced_object(data)
        assert data.flags.c_contiguous
wxchan's avatar
wxchan committed
333
334
335
336
337
338
339
        if data.dtype == np.float32:
            ptr_data = data.ctypes.data_as(ctypes.POINTER(ctypes.c_float))
            type_data = C_API_DTYPE_FLOAT32
        elif data.dtype == np.float64:
            ptr_data = data.ctypes.data_as(ctypes.POINTER(ctypes.c_double))
            type_data = C_API_DTYPE_FLOAT64
        else:
340
            raise TypeError("Expected np.float32 or np.float64, met type({})"
wxchan's avatar
wxchan committed
341
342
                            .format(data.dtype))
    else:
343
        raise TypeError("Unknown type({})".format(type(data).__name__))
344
    return (ptr_data, type_data, data)  # return `data` to avoid the temporary copy is freed
wxchan's avatar
wxchan committed
345

wxchan's avatar
wxchan committed
346

wxchan's avatar
wxchan committed
347
def c_int_array(data):
348
    """Get pointer of int numpy array / list."""
wxchan's avatar
wxchan committed
349
350
351
    if is_1d_list(data):
        data = np.array(data, copy=False)
    if is_numpy_1d_array(data):
352
353
        data = convert_from_sliced_object(data)
        assert data.flags.c_contiguous
wxchan's avatar
wxchan committed
354
355
356
357
358
359
360
        if data.dtype == np.int32:
            ptr_data = data.ctypes.data_as(ctypes.POINTER(ctypes.c_int32))
            type_data = C_API_DTYPE_INT32
        elif data.dtype == np.int64:
            ptr_data = data.ctypes.data_as(ctypes.POINTER(ctypes.c_int64))
            type_data = C_API_DTYPE_INT64
        else:
361
            raise TypeError("Expected np.int32 or np.int64, met type({})"
wxchan's avatar
wxchan committed
362
363
                            .format(data.dtype))
    else:
364
        raise TypeError("Unknown type({})".format(type(data).__name__))
365
    return (ptr_data, type_data, data)  # return `data` to avoid the temporary copy is freed
wxchan's avatar
wxchan committed
366

wxchan's avatar
wxchan committed
367

368
369
370
371
372
373
374
375
376
377
378
def _get_bad_pandas_dtypes(dtypes):
    pandas_dtype_mapper = {'int8': 'int', 'int16': 'int', 'int32': 'int',
                           'int64': 'int', 'uint8': 'int', 'uint16': 'int',
                           'uint32': 'int', 'uint64': 'int', 'bool': 'int',
                           'float16': 'float', 'float32': 'float', 'float64': 'float'}
    bad_indices = [i for i, dtype in enumerate(dtypes) if (dtype.name not in pandas_dtype_mapper
                                                           and (not is_dtype_sparse(dtype)
                                                                or dtype.subtype.name not in pandas_dtype_mapper))]
    return bad_indices


379
def _data_from_pandas(data, feature_name, categorical_feature, pandas_categorical):
380
    if isinstance(data, DataFrame):
381
382
        if len(data.shape) != 2 or data.shape[0] < 1:
            raise ValueError('Input data must be 2 dimensional and non empty.')
383
384
        if feature_name == 'auto' or feature_name is None:
            data = data.rename(columns=str)
385
386
        cat_cols = list(data.select_dtypes(include=['category']).columns)
        cat_cols_not_ordered = [col for col in cat_cols if not data[col].cat.ordered]
387
388
389
390
391
        if pandas_categorical is None:  # train dataset
            pandas_categorical = [list(data[col].cat.categories) for col in cat_cols]
        else:
            if len(cat_cols) != len(pandas_categorical):
                raise ValueError('train and valid dataset categorical_feature do not match.')
392
            for col, category in zip(cat_cols, pandas_categorical):
393
394
                if list(data[col].cat.categories) != list(category):
                    data[col] = data[col].cat.set_categories(category)
395
        if len(cat_cols):  # cat_cols is list
396
            data = data.copy()  # not alter origin DataFrame
397
            data[cat_cols] = data[cat_cols].apply(lambda x: x.cat.codes).replace({-1: np.nan})
398
399
400
        if categorical_feature is not None:
            if feature_name is None:
                feature_name = list(data.columns)
401
            if categorical_feature == 'auto':  # use cat cols from DataFrame
402
                categorical_feature = cat_cols_not_ordered
403
404
            else:  # use cat cols specified by user
                categorical_feature = list(categorical_feature)
405
406
        if feature_name == 'auto':
            feature_name = list(data.columns)
407
408
        bad_indices = _get_bad_pandas_dtypes(data.dtypes)
        if bad_indices:
409
            raise ValueError("DataFrame.dtypes for data must be int, float or bool.\n"
410
                             "Did not expect the data types in the following fields: "
411
                             + ', '.join(data.columns[bad_indices]))
412
413
414
        data = data.values
        if data.dtype != np.float32 and data.dtype != np.float64:
            data = data.astype(np.float32)
415
416
417
418
419
420
    else:
        if feature_name == 'auto':
            feature_name = None
        if categorical_feature == 'auto':
            categorical_feature = None
    return data, feature_name, categorical_feature, pandas_categorical
421
422
423
424
425
426


def _label_from_pandas(label):
    if isinstance(label, DataFrame):
        if len(label.columns) > 1:
            raise ValueError('DataFrame for label cannot have multiple columns')
427
        if _get_bad_pandas_dtypes(label.dtypes):
428
            raise ValueError('DataFrame.dtypes for label must be int, float or bool')
429
        label = np.ravel(label.values.astype(np.float32, copy=False))
430
431
432
    return label


433
434
435
436
437
438
439
440
441
442
443
def _dump_pandas_categorical(pandas_categorical, file_name=None):
    pandas_str = ('\npandas_categorical:'
                  + json.dumps(pandas_categorical, default=json_default_with_numpy)
                  + '\n')
    if file_name is not None:
        with open(file_name, 'a') as f:
            f.write(pandas_str)
    return pandas_str


def _load_pandas_categorical(file_name=None, model_str=None):
444
445
    pandas_key = 'pandas_categorical:'
    offset = -len(pandas_key)
446
    if file_name is not None:
447
448
449
450
451
452
453
454
455
456
        max_offset = -os.path.getsize(file_name)
        with open(file_name, 'rb') as f:
            while True:
                if offset < max_offset:
                    offset = max_offset
                f.seek(offset, os.SEEK_END)
                lines = f.readlines()
                if len(lines) >= 2:
                    break
                offset *= 2
457
        last_line = lines[-1].decode('utf-8').strip()
458
        if not last_line.startswith(pandas_key):
459
            last_line = lines[-2].decode('utf-8').strip()
460
    elif model_str is not None:
461
462
463
464
465
466
        idx = model_str.rfind('\n', 0, offset)
        last_line = model_str[idx:].strip()
    if last_line.startswith(pandas_key):
        return json.loads(last_line[len(pandas_key):])
    else:
        return None
467
468


469
class _InnerPredictor:
470
471
472
473
474
    """_InnerPredictor of LightGBM.

    Not exposed to user.
    Used only for prediction, usually used for continued training.

Nikita Titov's avatar
Nikita Titov committed
475
476
477
    .. note::

        Can be converted from Booster, but cannot be converted to Booster.
Guolin Ke's avatar
Guolin Ke committed
478
    """
479

480
    def __init__(self, model_file=None, booster_handle=None, pred_parameter=None):
481
        """Initialize the _InnerPredictor.
wxchan's avatar
wxchan committed
482
483
484

        Parameters
        ----------
485
        model_file : string or None, optional (default=None)
wxchan's avatar
wxchan committed
486
            Path to the model file.
487
488
489
490
        booster_handle : object or None, optional (default=None)
            Handle of Booster.
        pred_parameter: dict or None, optional (default=None)
            Other parameters for the prediciton.
wxchan's avatar
wxchan committed
491
492
493
494
495
        """
        self.handle = ctypes.c_void_p()
        self.__is_manage_handle = True
        if model_file is not None:
            """Prediction task"""
Guolin Ke's avatar
Guolin Ke committed
496
            out_num_iterations = ctypes.c_int(0)
wxchan's avatar
wxchan committed
497
498
499
500
            _safe_call(_LIB.LGBM_BoosterCreateFromModelfile(
                c_str(model_file),
                ctypes.byref(out_num_iterations),
                ctypes.byref(self.handle)))
Guolin Ke's avatar
Guolin Ke committed
501
            out_num_class = ctypes.c_int(0)
wxchan's avatar
wxchan committed
502
503
504
505
            _safe_call(_LIB.LGBM_BoosterGetNumClasses(
                self.handle,
                ctypes.byref(out_num_class)))
            self.num_class = out_num_class.value
506
            self.num_total_iteration = out_num_iterations.value
507
            self.pandas_categorical = _load_pandas_categorical(file_name=model_file)
wxchan's avatar
wxchan committed
508
        elif booster_handle is not None:
Guolin Ke's avatar
Guolin Ke committed
509
            self.__is_manage_handle = False
wxchan's avatar
wxchan committed
510
            self.handle = booster_handle
Guolin Ke's avatar
Guolin Ke committed
511
            out_num_class = ctypes.c_int(0)
wxchan's avatar
wxchan committed
512
513
514
515
            _safe_call(_LIB.LGBM_BoosterGetNumClasses(
                self.handle,
                ctypes.byref(out_num_class)))
            self.num_class = out_num_class.value
516
            self.num_total_iteration = self.current_iteration()
517
            self.pandas_categorical = None
wxchan's avatar
wxchan committed
518
        else:
519
            raise TypeError('Need model_file or booster_handle to create a predictor')
wxchan's avatar
wxchan committed
520

521
522
        pred_parameter = {} if pred_parameter is None else pred_parameter
        self.pred_parameter = param_dict_to_str(pred_parameter)
cbecker's avatar
cbecker committed
523

wxchan's avatar
wxchan committed
524
    def __del__(self):
525
526
527
528
529
        try:
            if self.__is_manage_handle:
                _safe_call(_LIB.LGBM_BoosterFree(self.handle))
        except AttributeError:
            pass
wxchan's avatar
wxchan committed
530

531
532
533
534
535
    def __getstate__(self):
        this = self.__dict__.copy()
        this.pop('handle', None)
        return this

536
    def predict(self, data, start_iteration=0, num_iteration=-1,
537
                raw_score=False, pred_leaf=False, pred_contrib=False, data_has_header=False,
wxchan's avatar
wxchan committed
538
                is_reshape=True):
539
        """Predict logic.
wxchan's avatar
wxchan committed
540
541
542

        Parameters
        ----------
543
        data : string, numpy array, pandas DataFrame, H2O DataTable's Frame or scipy.sparse
544
545
            Data source for prediction.
            When data type is string, it represents the path of txt file.
546
547
        start_iteration : int, optional (default=0)
            Start index of the iteration to predict.
548
549
550
551
552
553
554
555
556
557
558
559
560
        num_iteration : int, optional (default=-1)
            Iteration used for prediction.
        raw_score : bool, optional (default=False)
            Whether to predict raw scores.
        pred_leaf : bool, optional (default=False)
            Whether to predict leaf index.
        pred_contrib : bool, optional (default=False)
            Whether to predict feature contributions.
        data_has_header : bool, optional (default=False)
            Whether data has header.
            Used only for txt data.
        is_reshape : bool, optional (default=True)
            Whether to reshape to (nrow, ncol).
wxchan's avatar
wxchan committed
561
562
563

        Returns
        -------
564
        result : numpy array, scipy.sparse or list of scipy.sparse
565
            Prediction result.
566
            Can be sparse or a list of sparse objects (each element represents predictions for one class) for feature contributions (when ``pred_contrib=True``).
wxchan's avatar
wxchan committed
567
        """
wxchan's avatar
wxchan committed
568
        if isinstance(data, Dataset):
569
            raise TypeError("Cannot use Dataset instance for prediction, please use raw data instead")
570
        data = _data_from_pandas(data, None, None, self.pandas_categorical)[0]
wxchan's avatar
wxchan committed
571
572
573
574
575
        predict_type = C_API_PREDICT_NORMAL
        if raw_score:
            predict_type = C_API_PREDICT_RAW_SCORE
        if pred_leaf:
            predict_type = C_API_PREDICT_LEAF_INDEX
576
577
        if pred_contrib:
            predict_type = C_API_PREDICT_CONTRIB
wxchan's avatar
wxchan committed
578
        int_data_has_header = 1 if data_has_header else 0
cbecker's avatar
cbecker committed
579

580
        if isinstance(data, str):
581
            with _TempFile() as f:
wxchan's avatar
wxchan committed
582
583
584
                _safe_call(_LIB.LGBM_BoosterPredictForFile(
                    self.handle,
                    c_str(data),
Guolin Ke's avatar
Guolin Ke committed
585
586
                    ctypes.c_int(int_data_has_header),
                    ctypes.c_int(predict_type),
587
                    ctypes.c_int(start_iteration),
Guolin Ke's avatar
Guolin Ke committed
588
                    ctypes.c_int(num_iteration),
589
                    c_str(self.pred_parameter),
wxchan's avatar
wxchan committed
590
591
                    c_str(f.name)))
                lines = f.readlines()
592
593
                nrow = len(lines)
                preds = [float(token) for line in lines for token in line.split('\t')]
Guolin Ke's avatar
Guolin Ke committed
594
                preds = np.array(preds, dtype=np.float64, copy=False)
wxchan's avatar
wxchan committed
595
        elif isinstance(data, scipy.sparse.csr_matrix):
596
            preds, nrow = self.__pred_for_csr(data, start_iteration, num_iteration, predict_type)
Guolin Ke's avatar
Guolin Ke committed
597
        elif isinstance(data, scipy.sparse.csc_matrix):
598
            preds, nrow = self.__pred_for_csc(data, start_iteration, num_iteration, predict_type)
wxchan's avatar
wxchan committed
599
        elif isinstance(data, np.ndarray):
600
            preds, nrow = self.__pred_for_np2d(data, start_iteration, num_iteration, predict_type)
601
602
603
        elif isinstance(data, list):
            try:
                data = np.array(data)
604
            except BaseException:
605
                raise ValueError('Cannot convert data list to numpy array.')
606
            preds, nrow = self.__pred_for_np2d(data, start_iteration, num_iteration, predict_type)
607
        elif isinstance(data, DataTable):
608
            preds, nrow = self.__pred_for_np2d(data.to_numpy(), start_iteration, num_iteration, predict_type)
wxchan's avatar
wxchan committed
609
610
        else:
            try:
611
                warnings.warn('Converting data to scipy sparse matrix.')
wxchan's avatar
wxchan committed
612
                csr = scipy.sparse.csr_matrix(data)
613
            except BaseException:
614
                raise TypeError('Cannot predict data for type {}'.format(type(data).__name__))
615
            preds, nrow = self.__pred_for_csr(csr, start_iteration, num_iteration, predict_type)
wxchan's avatar
wxchan committed
616
617
        if pred_leaf:
            preds = preds.astype(np.int32)
618
619
        is_sparse = scipy.sparse.issparse(preds) or isinstance(preds, list)
        if is_reshape and not is_sparse and preds.size != nrow:
wxchan's avatar
wxchan committed
620
            if preds.size % nrow == 0:
621
                preds = preds.reshape(nrow, -1)
wxchan's avatar
wxchan committed
622
            else:
623
                raise ValueError('Length of predict result (%d) cannot be divide nrow (%d)'
wxchan's avatar
wxchan committed
624
625
626
                                 % (preds.size, nrow))
        return preds

627
    def __get_num_preds(self, start_iteration, num_iteration, nrow, predict_type):
628
        """Get size of prediction result."""
629
630
631
632
633
        if nrow > MAX_INT32:
            raise LightGBMError('LightGBM cannot perform prediction for data'
                                'with number of rows greater than MAX_INT32 (%d).\n'
                                'You can split your data into chunks'
                                'and then concatenate predictions for them' % MAX_INT32)
Guolin Ke's avatar
Guolin Ke committed
634
635
636
        n_preds = ctypes.c_int64(0)
        _safe_call(_LIB.LGBM_BoosterCalcNumPredict(
            self.handle,
Guolin Ke's avatar
Guolin Ke committed
637
638
            ctypes.c_int(nrow),
            ctypes.c_int(predict_type),
639
            ctypes.c_int(start_iteration),
Guolin Ke's avatar
Guolin Ke committed
640
            ctypes.c_int(num_iteration),
Guolin Ke's avatar
Guolin Ke committed
641
642
            ctypes.byref(n_preds)))
        return n_preds.value
wxchan's avatar
wxchan committed
643

644
    def __pred_for_np2d(self, mat, start_iteration, num_iteration, predict_type):
645
        """Predict for a 2-D numpy matrix."""
wxchan's avatar
wxchan committed
646
        if len(mat.shape) != 2:
647
            raise ValueError('Input numpy.ndarray or list must be 2 dimensional')
wxchan's avatar
wxchan committed
648

649
        def inner_predict(mat, start_iteration, num_iteration, predict_type, preds=None):
650
651
            if mat.dtype == np.float32 or mat.dtype == np.float64:
                data = np.array(mat.reshape(mat.size), dtype=mat.dtype, copy=False)
652
            else:  # change non-float data to float data, need to copy
653
654
                data = np.array(mat.reshape(mat.size), dtype=np.float32)
            ptr_data, type_ptr_data, _ = c_float_array(data)
655
            n_preds = self.__get_num_preds(start_iteration, num_iteration, mat.shape[0], predict_type)
656
657
658
659
660
661
662
663
664
665
666
667
668
            if preds is None:
                preds = np.zeros(n_preds, dtype=np.float64)
            elif len(preds.shape) != 1 or len(preds) != n_preds:
                raise ValueError("Wrong length of pre-allocated predict array")
            out_num_preds = ctypes.c_int64(0)
            _safe_call(_LIB.LGBM_BoosterPredictForMat(
                self.handle,
                ptr_data,
                ctypes.c_int(type_ptr_data),
                ctypes.c_int(mat.shape[0]),
                ctypes.c_int(mat.shape[1]),
                ctypes.c_int(C_API_IS_ROW_MAJOR),
                ctypes.c_int(predict_type),
669
                ctypes.c_int(start_iteration),
670
671
672
673
674
675
676
677
678
679
680
681
                ctypes.c_int(num_iteration),
                c_str(self.pred_parameter),
                ctypes.byref(out_num_preds),
                preds.ctypes.data_as(ctypes.POINTER(ctypes.c_double))))
            if n_preds != out_num_preds.value:
                raise ValueError("Wrong length for predict results")
            return preds, mat.shape[0]

        nrow = mat.shape[0]
        if nrow > MAX_INT32:
            sections = np.arange(start=MAX_INT32, stop=nrow, step=MAX_INT32)
            # __get_num_preds() cannot work with nrow > MAX_INT32, so calculate overall number of predictions piecemeal
682
            n_preds = [self.__get_num_preds(start_iteration, num_iteration, i, predict_type) for i in np.diff([0] + list(sections) + [nrow])]
683
684
            n_preds_sections = np.array([0] + n_preds, dtype=np.intp).cumsum()
            preds = np.zeros(sum(n_preds), dtype=np.float64)
685
686
            for chunk, (start_idx_pred, end_idx_pred) in zip(np.array_split(mat, sections),
                                                             zip(n_preds_sections, n_preds_sections[1:])):
687
                # avoid memory consumption by arrays concatenation operations
688
                inner_predict(chunk, start_iteration, num_iteration, predict_type, preds[start_idx_pred:end_idx_pred])
689
            return preds, nrow
wxchan's avatar
wxchan committed
690
        else:
691
            return inner_predict(mat, start_iteration, num_iteration, predict_type)
wxchan's avatar
wxchan committed
692

693
694
695
696
697
698
699
700
701
702
703
704
705
706
707
708
709
710
711
712
713
714
715
716
717
718
719
720
721
722
723
724
725
726
727
728
729
730
731
732
733
734
735
736
737
738
    def __create_sparse_native(self, cs, out_shape, out_ptr_indptr, out_ptr_indices, out_ptr_data,
                               indptr_type, data_type, is_csr=True):
        # create numpy array from output arrays
        data_indices_len = out_shape[0]
        indptr_len = out_shape[1]
        if indptr_type == C_API_DTYPE_INT32:
            out_indptr = cint32_array_to_numpy(out_ptr_indptr, indptr_len)
        elif indptr_type == C_API_DTYPE_INT64:
            out_indptr = cint64_array_to_numpy(out_ptr_indptr, indptr_len)
        else:
            raise TypeError("Expected int32 or int64 type for indptr")
        if data_type == C_API_DTYPE_FLOAT32:
            out_data = cfloat32_array_to_numpy(out_ptr_data, data_indices_len)
        elif data_type == C_API_DTYPE_FLOAT64:
            out_data = cfloat64_array_to_numpy(out_ptr_data, data_indices_len)
        else:
            raise TypeError("Expected float32 or float64 type for data")
        out_indices = cint32_array_to_numpy(out_ptr_indices, data_indices_len)
        # break up indptr based on number of rows (note more than one matrix in multiclass case)
        per_class_indptr_shape = cs.indptr.shape[0]
        # for CSC there is extra column added
        if not is_csr:
            per_class_indptr_shape += 1
        out_indptr_arrays = np.split(out_indptr, out_indptr.shape[0] / per_class_indptr_shape)
        # reformat output into a csr or csc matrix or list of csr or csc matrices
        cs_output_matrices = []
        offset = 0
        for cs_indptr in out_indptr_arrays:
            matrix_indptr_len = cs_indptr[cs_indptr.shape[0] - 1]
            cs_indices = out_indices[offset + cs_indptr[0]:offset + matrix_indptr_len]
            cs_data = out_data[offset + cs_indptr[0]:offset + matrix_indptr_len]
            offset += matrix_indptr_len
            # same shape as input csr or csc matrix except extra column for expected value
            cs_shape = [cs.shape[0], cs.shape[1] + 1]
            # note: make sure we copy data as it will be deallocated next
            if is_csr:
                cs_output_matrices.append(scipy.sparse.csr_matrix((cs_data, cs_indices, cs_indptr), cs_shape))
            else:
                cs_output_matrices.append(scipy.sparse.csc_matrix((cs_data, cs_indices, cs_indptr), cs_shape))
        # free the temporary native indptr, indices, and data
        _safe_call(_LIB.LGBM_BoosterFreePredictSparse(out_ptr_indptr, out_ptr_indices, out_ptr_data,
                                                      ctypes.c_int(indptr_type), ctypes.c_int(data_type)))
        if len(cs_output_matrices) == 1:
            return cs_output_matrices[0]
        return cs_output_matrices

739
    def __pred_for_csr(self, csr, start_iteration, num_iteration, predict_type):
740
        """Predict for a CSR data."""
741
        def inner_predict(csr, start_iteration, num_iteration, predict_type, preds=None):
742
            nrow = len(csr.indptr) - 1
743
            n_preds = self.__get_num_preds(start_iteration, num_iteration, nrow, predict_type)
744
745
746
747
748
749
750
751
752
            if preds is None:
                preds = np.zeros(n_preds, dtype=np.float64)
            elif len(preds.shape) != 1 or len(preds) != n_preds:
                raise ValueError("Wrong length of pre-allocated predict array")
            out_num_preds = ctypes.c_int64(0)

            ptr_indptr, type_ptr_indptr, __ = c_int_array(csr.indptr)
            ptr_data, type_ptr_data, _ = c_float_array(csr.data)

753
            assert csr.shape[1] <= MAX_INT32
754
            csr_indices = csr.indices.astype(np.int32, copy=False)
755

756
757
758
759
            _safe_call(_LIB.LGBM_BoosterPredictForCSR(
                self.handle,
                ptr_indptr,
                ctypes.c_int32(type_ptr_indptr),
760
                csr_indices.ctypes.data_as(ctypes.POINTER(ctypes.c_int32)),
761
762
763
764
765
766
                ptr_data,
                ctypes.c_int(type_ptr_data),
                ctypes.c_int64(len(csr.indptr)),
                ctypes.c_int64(len(csr.data)),
                ctypes.c_int64(csr.shape[1]),
                ctypes.c_int(predict_type),
767
                ctypes.c_int(start_iteration),
768
769
770
771
772
773
774
                ctypes.c_int(num_iteration),
                c_str(self.pred_parameter),
                ctypes.byref(out_num_preds),
                preds.ctypes.data_as(ctypes.POINTER(ctypes.c_double))))
            if n_preds != out_num_preds.value:
                raise ValueError("Wrong length for predict results")
            return preds, nrow
wxchan's avatar
wxchan committed
775

776
        def inner_predict_sparse(csr, start_iteration, num_iteration, predict_type):
777
778
779
780
781
782
783
784
785
786
787
788
789
790
791
792
793
794
795
796
797
798
799
800
801
            ptr_indptr, type_ptr_indptr, __ = c_int_array(csr.indptr)
            ptr_data, type_ptr_data, _ = c_float_array(csr.data)
            csr_indices = csr.indices.astype(np.int32, copy=False)
            matrix_type = C_API_MATRIX_TYPE_CSR
            if type_ptr_indptr == C_API_DTYPE_INT32:
                out_ptr_indptr = ctypes.POINTER(ctypes.c_int32)()
            else:
                out_ptr_indptr = ctypes.POINTER(ctypes.c_int64)()
            out_ptr_indices = ctypes.POINTER(ctypes.c_int32)()
            if type_ptr_data == C_API_DTYPE_FLOAT32:
                out_ptr_data = ctypes.POINTER(ctypes.c_float)()
            else:
                out_ptr_data = ctypes.POINTER(ctypes.c_double)()
            out_shape = np.zeros(2, dtype=np.int64)
            _safe_call(_LIB.LGBM_BoosterPredictSparseOutput(
                self.handle,
                ptr_indptr,
                ctypes.c_int32(type_ptr_indptr),
                csr_indices.ctypes.data_as(ctypes.POINTER(ctypes.c_int32)),
                ptr_data,
                ctypes.c_int(type_ptr_data),
                ctypes.c_int64(len(csr.indptr)),
                ctypes.c_int64(len(csr.data)),
                ctypes.c_int64(csr.shape[1]),
                ctypes.c_int(predict_type),
802
                ctypes.c_int(start_iteration),
803
804
805
806
807
808
809
810
811
812
813
814
815
                ctypes.c_int(num_iteration),
                c_str(self.pred_parameter),
                ctypes.c_int(matrix_type),
                out_shape.ctypes.data_as(ctypes.POINTER(ctypes.c_int64)),
                ctypes.byref(out_ptr_indptr),
                ctypes.byref(out_ptr_indices),
                ctypes.byref(out_ptr_data)))
            matrices = self.__create_sparse_native(csr, out_shape, out_ptr_indptr, out_ptr_indices, out_ptr_data,
                                                   type_ptr_indptr, type_ptr_data, is_csr=True)
            nrow = len(csr.indptr) - 1
            return matrices, nrow

        if predict_type == C_API_PREDICT_CONTRIB:
816
            return inner_predict_sparse(csr, start_iteration, num_iteration, predict_type)
817
818
819
820
        nrow = len(csr.indptr) - 1
        if nrow > MAX_INT32:
            sections = [0] + list(np.arange(start=MAX_INT32, stop=nrow, step=MAX_INT32)) + [nrow]
            # __get_num_preds() cannot work with nrow > MAX_INT32, so calculate overall number of predictions piecemeal
821
            n_preds = [self.__get_num_preds(start_iteration, num_iteration, i, predict_type) for i in np.diff(sections)]
822
823
            n_preds_sections = np.array([0] + n_preds, dtype=np.intp).cumsum()
            preds = np.zeros(sum(n_preds), dtype=np.float64)
824
825
            for (start_idx, end_idx), (start_idx_pred, end_idx_pred) in zip(zip(sections, sections[1:]),
                                                                            zip(n_preds_sections, n_preds_sections[1:])):
826
                # avoid memory consumption by arrays concatenation operations
827
                inner_predict(csr[start_idx:end_idx], start_iteration, num_iteration, predict_type, preds[start_idx_pred:end_idx_pred])
828
829
            return preds, nrow
        else:
830
            return inner_predict(csr, start_iteration, num_iteration, predict_type)
Guolin Ke's avatar
Guolin Ke committed
831

832
    def __pred_for_csc(self, csc, start_iteration, num_iteration, predict_type):
833
        """Predict for a CSC data."""
834
        def inner_predict_sparse(csc, start_iteration, num_iteration, predict_type):
835
836
837
838
839
840
841
842
843
844
845
846
847
848
849
850
851
852
853
854
855
856
857
858
859
            ptr_indptr, type_ptr_indptr, __ = c_int_array(csc.indptr)
            ptr_data, type_ptr_data, _ = c_float_array(csc.data)
            csc_indices = csc.indices.astype(np.int32, copy=False)
            matrix_type = C_API_MATRIX_TYPE_CSC
            if type_ptr_indptr == C_API_DTYPE_INT32:
                out_ptr_indptr = ctypes.POINTER(ctypes.c_int32)()
            else:
                out_ptr_indptr = ctypes.POINTER(ctypes.c_int64)()
            out_ptr_indices = ctypes.POINTER(ctypes.c_int32)()
            if type_ptr_data == C_API_DTYPE_FLOAT32:
                out_ptr_data = ctypes.POINTER(ctypes.c_float)()
            else:
                out_ptr_data = ctypes.POINTER(ctypes.c_double)()
            out_shape = np.zeros(2, dtype=np.int64)
            _safe_call(_LIB.LGBM_BoosterPredictSparseOutput(
                self.handle,
                ptr_indptr,
                ctypes.c_int32(type_ptr_indptr),
                csc_indices.ctypes.data_as(ctypes.POINTER(ctypes.c_int32)),
                ptr_data,
                ctypes.c_int(type_ptr_data),
                ctypes.c_int64(len(csc.indptr)),
                ctypes.c_int64(len(csc.data)),
                ctypes.c_int64(csc.shape[0]),
                ctypes.c_int(predict_type),
860
                ctypes.c_int(start_iteration),
861
862
863
864
865
866
867
868
869
870
871
872
                ctypes.c_int(num_iteration),
                c_str(self.pred_parameter),
                ctypes.c_int(matrix_type),
                out_shape.ctypes.data_as(ctypes.POINTER(ctypes.c_int64)),
                ctypes.byref(out_ptr_indptr),
                ctypes.byref(out_ptr_indices),
                ctypes.byref(out_ptr_data)))
            matrices = self.__create_sparse_native(csc, out_shape, out_ptr_indptr, out_ptr_indices, out_ptr_data,
                                                   type_ptr_indptr, type_ptr_data, is_csr=False)
            nrow = csc.shape[0]
            return matrices, nrow

Guolin Ke's avatar
Guolin Ke committed
873
        nrow = csc.shape[0]
874
        if nrow > MAX_INT32:
875
            return self.__pred_for_csr(csc.tocsr(), start_iteration, num_iteration, predict_type)
876
        if predict_type == C_API_PREDICT_CONTRIB:
877
878
            return inner_predict_sparse(csc, start_iteration, num_iteration, predict_type)
        n_preds = self.__get_num_preds(start_iteration, num_iteration, nrow, predict_type)
Guolin Ke's avatar
Guolin Ke committed
879
880
881
        preds = np.zeros(n_preds, dtype=np.float64)
        out_num_preds = ctypes.c_int64(0)

882
883
        ptr_indptr, type_ptr_indptr, __ = c_int_array(csc.indptr)
        ptr_data, type_ptr_data, _ = c_float_array(csc.data)
Guolin Ke's avatar
Guolin Ke committed
884

885
        assert csc.shape[0] <= MAX_INT32
886
        csc_indices = csc.indices.astype(np.int32, copy=False)
887

Guolin Ke's avatar
Guolin Ke committed
888
889
890
        _safe_call(_LIB.LGBM_BoosterPredictForCSC(
            self.handle,
            ptr_indptr,
Guolin Ke's avatar
Guolin Ke committed
891
            ctypes.c_int32(type_ptr_indptr),
892
            csc_indices.ctypes.data_as(ctypes.POINTER(ctypes.c_int32)),
Guolin Ke's avatar
Guolin Ke committed
893
            ptr_data,
Guolin Ke's avatar
Guolin Ke committed
894
895
896
897
898
            ctypes.c_int(type_ptr_data),
            ctypes.c_int64(len(csc.indptr)),
            ctypes.c_int64(len(csc.data)),
            ctypes.c_int64(csc.shape[0]),
            ctypes.c_int(predict_type),
899
            ctypes.c_int(start_iteration),
Guolin Ke's avatar
Guolin Ke committed
900
            ctypes.c_int(num_iteration),
901
            c_str(self.pred_parameter),
Guolin Ke's avatar
Guolin Ke committed
902
            ctypes.byref(out_num_preds),
wxchan's avatar
wxchan committed
903
            preds.ctypes.data_as(ctypes.POINTER(ctypes.c_double))))
wxchan's avatar
wxchan committed
904
        if n_preds != out_num_preds.value:
905
            raise ValueError("Wrong length for predict results")
wxchan's avatar
wxchan committed
906
907
        return preds, nrow

908
909
910
911
912
913
914
915
916
917
918
919
920
921
    def current_iteration(self):
        """Get the index of the current iteration.

        Returns
        -------
        cur_iter : int
            The index of the current iteration.
        """
        out_cur_iter = ctypes.c_int(0)
        _safe_call(_LIB.LGBM_BoosterGetCurrentIteration(
            self.handle,
            ctypes.byref(out_cur_iter)))
        return out_cur_iter.value

wxchan's avatar
wxchan committed
922

923
class Dataset:
wxchan's avatar
wxchan committed
924
    """Dataset in LightGBM."""
925

926
    def __init__(self, data, label=None, reference=None,
927
                 weight=None, group=None, init_score=None, silent=False,
928
                 feature_name='auto', categorical_feature='auto', params=None,
wxchan's avatar
wxchan committed
929
                 free_raw_data=True):
930
        """Initialize Dataset.
931

wxchan's avatar
wxchan committed
932
933
        Parameters
        ----------
934
        data : string, numpy array, pandas DataFrame, H2O DataTable's Frame, scipy.sparse or list of numpy arrays
wxchan's avatar
wxchan committed
935
            Data source of Dataset.
936
            If string, it represents the path to txt file.
937
        label : list, numpy 1-D array, pandas Series / one-column DataFrame or None, optional (default=None)
938
939
940
            Label of the data.
        reference : Dataset or None, optional (default=None)
            If this is Dataset for validation, training data should be used as reference.
941
        weight : list, numpy 1-D array, pandas Series or None, optional (default=None)
wxchan's avatar
wxchan committed
942
            Weight for each instance.
943
        group : list, numpy 1-D array, pandas Series or None, optional (default=None)
944
945
946
947
            Group/query data.
            Only used in the learning-to-rank task.
            sum(group) = n_samples.
            For example, if you have a 100-document dataset with ``group = [10, 20, 40, 10, 10, 10]``, that means that you have 6 groups, where the first 10 records are in the first group, records 11-30 are in the second group, etc.
948
        init_score : list, numpy 1-D array, pandas Series or None, optional (default=None)
949
            Init score for Dataset.
950
951
952
953
954
955
956
957
958
        silent : bool, optional (default=False)
            Whether to print messages during construction.
        feature_name : list of strings or 'auto', optional (default="auto")
            Feature names.
            If 'auto' and data is pandas DataFrame, data columns names are used.
        categorical_feature : list of strings or int, or 'auto', optional (default="auto")
            Categorical features.
            If list of int, interpreted as indices.
            If list of strings, interpreted as feature names (need to specify ``feature_name`` as well).
959
            If 'auto' and data is pandas DataFrame, pandas unordered categorical columns are used.
960
            All values in categorical features should be less than int32 max value (2147483647).
961
            Large values could be memory consuming. Consider using consecutive integers starting from zero.
962
            All negative values in categorical features will be treated as missing values.
963
            The output cannot be monotonically constrained with respect to a categorical feature.
Nikita Titov's avatar
Nikita Titov committed
964
        params : dict or None, optional (default=None)
965
            Other parameters for Dataset.
Nikita Titov's avatar
Nikita Titov committed
966
        free_raw_data : bool, optional (default=True)
967
            If True, raw data is freed after constructing inner Dataset.
wxchan's avatar
wxchan committed
968
        """
wxchan's avatar
wxchan committed
969
970
971
972
973
974
        self.handle = None
        self.data = data
        self.label = label
        self.reference = reference
        self.weight = weight
        self.group = group
975
        self.init_score = init_score
wxchan's avatar
wxchan committed
976
977
        self.silent = silent
        self.feature_name = feature_name
978
        self.categorical_feature = categorical_feature
979
        self.params = copy.deepcopy(params)
wxchan's avatar
wxchan committed
980
981
        self.free_raw_data = free_raw_data
        self.used_indices = None
982
        self.need_slice = True
wxchan's avatar
wxchan committed
983
        self._predictor = None
984
        self.pandas_categorical = None
985
        self.params_back_up = None
986
987
        self.feature_penalty = None
        self.monotone_constraints = None
988
        self.version = 0
wxchan's avatar
wxchan committed
989
990

    def __del__(self):
991
992
993
994
        try:
            self._free_handle()
        except AttributeError:
            pass
995

996
997
998
999
1000
1001
1002
1003
1004
1005
1006
1007
1008
1009
1010
1011
1012
1013
1014
1015
1016
    def get_params(self):
        """Get the used parameters in the Dataset.

        Returns
        -------
        params : dict or None
            The used parameters in this Dataset object.
        """
        if self.params is not None:
            # no min_data, nthreads and verbose in this function
            dataset_params = _ConfigAliases.get("bin_construct_sample_cnt",
                                                "categorical_feature",
                                                "data_random_seed",
                                                "enable_bundle",
                                                "feature_pre_filter",
                                                "forcedbins_filename",
                                                "group_column",
                                                "header",
                                                "ignore_column",
                                                "is_enable_sparse",
                                                "label_column",
1017
                                                "linear_tree",
1018
1019
1020
1021
1022
1023
1024
1025
1026
1027
                                                "max_bin",
                                                "max_bin_by_feature",
                                                "min_data_in_bin",
                                                "pre_partition",
                                                "two_round",
                                                "use_missing",
                                                "weight_column",
                                                "zero_as_missing")
            return {k: v for k, v in self.params.items() if k in dataset_params}

1028
    def _free_handle(self):
1029
        if self.handle is not None:
1030
            _safe_call(_LIB.LGBM_DatasetFree(self.handle))
1031
            self.handle = None
Guolin Ke's avatar
Guolin Ke committed
1032
1033
1034
        self.need_slice = True
        if self.used_indices is not None:
            self.data = None
Nikita Titov's avatar
Nikita Titov committed
1035
        return self
wxchan's avatar
wxchan committed
1036

Guolin Ke's avatar
Guolin Ke committed
1037
1038
    def _set_init_score_by_predictor(self, predictor, data, used_indices=None):
        data_has_header = False
1039
        if isinstance(data, str):
Guolin Ke's avatar
Guolin Ke committed
1040
            # check data has header or not
1041
            data_has_header = any(self.params.get(alias, False) for alias in _ConfigAliases.get("header"))
Guolin Ke's avatar
Guolin Ke committed
1042
        num_data = self.num_data()
1043
1044
1045
1046
1047
1048
1049
        if predictor is not None:
            init_score = predictor.predict(data,
                                           raw_score=True,
                                           data_has_header=data_has_header,
                                           is_reshape=False)
            if used_indices is not None:
                assert not self.need_slice
1050
                if isinstance(data, str):
1051
1052
                    sub_init_score = np.zeros(num_data * predictor.num_class, dtype=np.float32)
                    assert num_data == len(used_indices)
1053
1054
                    for i in range(len(used_indices)):
                        for j in range(predictor.num_class):
1055
1056
1057
1058
1059
                            sub_init_score[i * predictor.num_class + j] = init_score[used_indices[i] * predictor.num_class + j]
                    init_score = sub_init_score
            if predictor.num_class > 1:
                # need to regroup init_score
                new_init_score = np.zeros(init_score.size, dtype=np.float32)
1060
1061
                for i in range(num_data):
                    for j in range(predictor.num_class):
1062
1063
1064
1065
1066
1067
                        new_init_score[j * num_data + i] = init_score[i * predictor.num_class + j]
                init_score = new_init_score
        elif self.init_score is not None:
            init_score = np.zeros(self.init_score.shape, dtype=np.float32)
        else:
            return self
Guolin Ke's avatar
Guolin Ke committed
1068
1069
        self.set_init_score(init_score)

1070
    def _lazy_init(self, data, label=None, reference=None,
1071
                   weight=None, group=None, init_score=None, predictor=None,
wxchan's avatar
wxchan committed
1072
                   silent=False, feature_name='auto',
1073
                   categorical_feature='auto', params=None):
wxchan's avatar
wxchan committed
1074
1075
        if data is None:
            self.handle = None
Nikita Titov's avatar
Nikita Titov committed
1076
            return self
Guolin Ke's avatar
Guolin Ke committed
1077
1078
1079
        if reference is not None:
            self.pandas_categorical = reference.pandas_categorical
            categorical_feature = reference.categorical_feature
1080
1081
1082
1083
        data, feature_name, categorical_feature, self.pandas_categorical = _data_from_pandas(data,
                                                                                             feature_name,
                                                                                             categorical_feature,
                                                                                             self.pandas_categorical)
wxchan's avatar
wxchan committed
1084
        label = _label_from_pandas(label)
Guolin Ke's avatar
Guolin Ke committed
1085

1086
        # process for args
wxchan's avatar
wxchan committed
1087
        params = {} if params is None else params
1088
1089
1090
        args_names = (getattr(self.__class__, '_lazy_init')
                      .__code__
                      .co_varnames[:getattr(self.__class__, '_lazy_init').__code__.co_argcount])
1091
1092
        for key, _ in params.items():
            if key in args_names:
1093
1094
1095
                warnings.warn('{0} keyword has been found in `params` and will be ignored.\n'
                              'Please use {0} argument of the Dataset constructor to pass this parameter.'
                              .format(key))
1096
        # user can set verbose with params, it has higher priority
1097
        if not any(verbose_alias in params for verbose_alias in _ConfigAliases.get("verbosity")) and silent:
1098
            params["verbose"] = -1
1099
        # get categorical features
1100
1101
1102
1103
1104
1105
        if categorical_feature is not None:
            categorical_indices = set()
            feature_dict = {}
            if feature_name is not None:
                feature_dict = {name: i for i, name in enumerate(feature_name)}
            for name in categorical_feature:
1106
                if isinstance(name, str) and name in feature_dict:
1107
                    categorical_indices.add(feature_dict[name])
1108
                elif isinstance(name, int):
1109
1110
1111
1112
                    categorical_indices.add(name)
                else:
                    raise TypeError("Wrong type({}) or unknown name({}) in categorical_feature"
                                    .format(type(name).__name__, name))
1113
            if categorical_indices:
1114
1115
1116
1117
                for cat_alias in _ConfigAliases.get("categorical_feature"):
                    if cat_alias in params:
                        warnings.warn('{} in param dict is overridden.'.format(cat_alias))
                        params.pop(cat_alias, None)
1118
                params['categorical_column'] = sorted(categorical_indices)
1119

wxchan's avatar
wxchan committed
1120
        params_str = param_dict_to_str(params)
1121
        self.params = params
1122
        # process for reference dataset
wxchan's avatar
wxchan committed
1123
        ref_dataset = None
wxchan's avatar
wxchan committed
1124
        if isinstance(reference, Dataset):
1125
            ref_dataset = reference.construct().handle
wxchan's avatar
wxchan committed
1126
1127
        elif reference is not None:
            raise TypeError('Reference dataset should be None or dataset instance')
1128
        # start construct data
1129
        if isinstance(data, str):
wxchan's avatar
wxchan committed
1130
1131
1132
1133
1134
1135
1136
1137
            self.handle = ctypes.c_void_p()
            _safe_call(_LIB.LGBM_DatasetCreateFromFile(
                c_str(data),
                c_str(params_str),
                ref_dataset,
                ctypes.byref(self.handle)))
        elif isinstance(data, scipy.sparse.csr_matrix):
            self.__init_from_csr(data, params_str, ref_dataset)
Guolin Ke's avatar
Guolin Ke committed
1138
1139
        elif isinstance(data, scipy.sparse.csc_matrix):
            self.__init_from_csc(data, params_str, ref_dataset)
wxchan's avatar
wxchan committed
1140
1141
        elif isinstance(data, np.ndarray):
            self.__init_from_np2d(data, params_str, ref_dataset)
1142
1143
        elif isinstance(data, list) and len(data) > 0 and all(isinstance(x, np.ndarray) for x in data):
            self.__init_from_list_np2d(data, params_str, ref_dataset)
1144
1145
        elif isinstance(data, DataTable):
            self.__init_from_np2d(data.to_numpy(), params_str, ref_dataset)
wxchan's avatar
wxchan committed
1146
1147
1148
1149
        else:
            try:
                csr = scipy.sparse.csr_matrix(data)
                self.__init_from_csr(csr, params_str, ref_dataset)
1150
            except BaseException:
wxchan's avatar
wxchan committed
1151
                raise TypeError('Cannot initialize Dataset from {}'.format(type(data).__name__))
wxchan's avatar
wxchan committed
1152
1153
1154
        if label is not None:
            self.set_label(label)
        if self.get_label() is None:
1155
            raise ValueError("Label should not be None")
wxchan's avatar
wxchan committed
1156
1157
1158
1159
        if weight is not None:
            self.set_weight(weight)
        if group is not None:
            self.set_group(group)
1160
1161
1162
        if isinstance(predictor, _InnerPredictor):
            if self._predictor is None and init_score is not None:
                warnings.warn("The init_score will be overridden by the prediction of init_model.")
Guolin Ke's avatar
Guolin Ke committed
1163
            self._set_init_score_by_predictor(predictor, data)
1164
1165
        elif init_score is not None:
            self.set_init_score(init_score)
Guolin Ke's avatar
Guolin Ke committed
1166
1167
        elif predictor is not None:
            raise TypeError('Wrong predictor type {}'.format(type(predictor).__name__))
Guolin Ke's avatar
Guolin Ke committed
1168
        # set feature names
Nikita Titov's avatar
Nikita Titov committed
1169
        return self.set_feature_name(feature_name)
wxchan's avatar
wxchan committed
1170
1171

    def __init_from_np2d(self, mat, params_str, ref_dataset):
1172
        """Initialize data from a 2-D numpy matrix."""
wxchan's avatar
wxchan committed
1173
1174
1175
1176
1177
1178
        if len(mat.shape) != 2:
            raise ValueError('Input numpy.ndarray must be 2 dimensional')

        self.handle = ctypes.c_void_p()
        if mat.dtype == np.float32 or mat.dtype == np.float64:
            data = np.array(mat.reshape(mat.size), dtype=mat.dtype, copy=False)
1179
        else:  # change non-float data to float data, need to copy
wxchan's avatar
wxchan committed
1180
1181
            data = np.array(mat.reshape(mat.size), dtype=np.float32)

1182
        ptr_data, type_ptr_data, _ = c_float_array(data)
wxchan's avatar
wxchan committed
1183
1184
        _safe_call(_LIB.LGBM_DatasetCreateFromMat(
            ptr_data,
Guolin Ke's avatar
Guolin Ke committed
1185
1186
1187
1188
            ctypes.c_int(type_ptr_data),
            ctypes.c_int(mat.shape[0]),
            ctypes.c_int(mat.shape[1]),
            ctypes.c_int(C_API_IS_ROW_MAJOR),
wxchan's avatar
wxchan committed
1189
1190
1191
            c_str(params_str),
            ref_dataset,
            ctypes.byref(self.handle)))
Nikita Titov's avatar
Nikita Titov committed
1192
        return self
wxchan's avatar
wxchan committed
1193

1194
    def __init_from_list_np2d(self, mats, params_str, ref_dataset):
1195
        """Initialize data from a list of 2-D numpy matrices."""
1196
1197
1198
1199
1200
1201
1202
1203
1204
1205
1206
1207
1208
1209
1210
1211
1212
1213
1214
1215
1216
        ncol = mats[0].shape[1]
        nrow = np.zeros((len(mats),), np.int32)
        if mats[0].dtype == np.float64:
            ptr_data = (ctypes.POINTER(ctypes.c_double) * len(mats))()
        else:
            ptr_data = (ctypes.POINTER(ctypes.c_float) * len(mats))()

        holders = []
        type_ptr_data = None

        for i, mat in enumerate(mats):
            if len(mat.shape) != 2:
                raise ValueError('Input numpy.ndarray must be 2 dimensional')

            if mat.shape[1] != ncol:
                raise ValueError('Input arrays must have same number of columns')

            nrow[i] = mat.shape[0]

            if mat.dtype == np.float32 or mat.dtype == np.float64:
                mats[i] = np.array(mat.reshape(mat.size), dtype=mat.dtype, copy=False)
1217
            else:  # change non-float data to float data, need to copy
1218
1219
1220
1221
1222
1223
1224
1225
1226
1227
1228
1229
1230
1231
1232
1233
1234
1235
1236
1237
                mats[i] = np.array(mat.reshape(mat.size), dtype=np.float32)

            chunk_ptr_data, chunk_type_ptr_data, holder = c_float_array(mats[i])
            if type_ptr_data is not None and chunk_type_ptr_data != type_ptr_data:
                raise ValueError('Input chunks must have same type')
            ptr_data[i] = chunk_ptr_data
            type_ptr_data = chunk_type_ptr_data
            holders.append(holder)

        self.handle = ctypes.c_void_p()
        _safe_call(_LIB.LGBM_DatasetCreateFromMats(
            ctypes.c_int(len(mats)),
            ctypes.cast(ptr_data, ctypes.POINTER(ctypes.POINTER(ctypes.c_double))),
            ctypes.c_int(type_ptr_data),
            nrow.ctypes.data_as(ctypes.POINTER(ctypes.c_int32)),
            ctypes.c_int(ncol),
            ctypes.c_int(C_API_IS_ROW_MAJOR),
            c_str(params_str),
            ref_dataset,
            ctypes.byref(self.handle)))
Nikita Titov's avatar
Nikita Titov committed
1238
        return self
1239

wxchan's avatar
wxchan committed
1240
    def __init_from_csr(self, csr, params_str, ref_dataset):
1241
        """Initialize data from a CSR matrix."""
wxchan's avatar
wxchan committed
1242
        if len(csr.indices) != len(csr.data):
1243
            raise ValueError('Length mismatch: {} vs {}'.format(len(csr.indices), len(csr.data)))
wxchan's avatar
wxchan committed
1244
1245
        self.handle = ctypes.c_void_p()

1246
1247
        ptr_indptr, type_ptr_indptr, __ = c_int_array(csr.indptr)
        ptr_data, type_ptr_data, _ = c_float_array(csr.data)
wxchan's avatar
wxchan committed
1248

1249
        assert csr.shape[1] <= MAX_INT32
1250
        csr_indices = csr.indices.astype(np.int32, copy=False)
1251

wxchan's avatar
wxchan committed
1252
1253
        _safe_call(_LIB.LGBM_DatasetCreateFromCSR(
            ptr_indptr,
Guolin Ke's avatar
Guolin Ke committed
1254
            ctypes.c_int(type_ptr_indptr),
1255
            csr_indices.ctypes.data_as(ctypes.POINTER(ctypes.c_int32)),
wxchan's avatar
wxchan committed
1256
            ptr_data,
Guolin Ke's avatar
Guolin Ke committed
1257
1258
1259
1260
            ctypes.c_int(type_ptr_data),
            ctypes.c_int64(len(csr.indptr)),
            ctypes.c_int64(len(csr.data)),
            ctypes.c_int64(csr.shape[1]),
wxchan's avatar
wxchan committed
1261
1262
1263
            c_str(params_str),
            ref_dataset,
            ctypes.byref(self.handle)))
Nikita Titov's avatar
Nikita Titov committed
1264
        return self
wxchan's avatar
wxchan committed
1265

Guolin Ke's avatar
Guolin Ke committed
1266
    def __init_from_csc(self, csc, params_str, ref_dataset):
1267
        """Initialize data from a CSC matrix."""
Guolin Ke's avatar
Guolin Ke committed
1268
1269
1270
1271
        if len(csc.indices) != len(csc.data):
            raise ValueError('Length mismatch: {} vs {}'.format(len(csc.indices), len(csc.data)))
        self.handle = ctypes.c_void_p()

1272
1273
        ptr_indptr, type_ptr_indptr, __ = c_int_array(csc.indptr)
        ptr_data, type_ptr_data, _ = c_float_array(csc.data)
Guolin Ke's avatar
Guolin Ke committed
1274

1275
        assert csc.shape[0] <= MAX_INT32
1276
        csc_indices = csc.indices.astype(np.int32, copy=False)
1277

Guolin Ke's avatar
Guolin Ke committed
1278
1279
        _safe_call(_LIB.LGBM_DatasetCreateFromCSC(
            ptr_indptr,
Guolin Ke's avatar
Guolin Ke committed
1280
            ctypes.c_int(type_ptr_indptr),
1281
            csc_indices.ctypes.data_as(ctypes.POINTER(ctypes.c_int32)),
Guolin Ke's avatar
Guolin Ke committed
1282
            ptr_data,
Guolin Ke's avatar
Guolin Ke committed
1283
1284
1285
1286
            ctypes.c_int(type_ptr_data),
            ctypes.c_int64(len(csc.indptr)),
            ctypes.c_int64(len(csc.data)),
            ctypes.c_int64(csc.shape[0]),
Guolin Ke's avatar
Guolin Ke committed
1287
1288
1289
            c_str(params_str),
            ref_dataset,
            ctypes.byref(self.handle)))
Nikita Titov's avatar
Nikita Titov committed
1290
        return self
Guolin Ke's avatar
Guolin Ke committed
1291

wxchan's avatar
wxchan committed
1292
    def construct(self):
1293
1294
1295
1296
1297
        """Lazy init.

        Returns
        -------
        self : Dataset
Nikita Titov's avatar
Nikita Titov committed
1298
            Constructed Dataset object.
1299
        """
1300
        if self.handle is None:
wxchan's avatar
wxchan committed
1301
            if self.reference is not None:
1302
1303
1304
1305
                reference_params = self.reference.get_params()
                if self.get_params() != reference_params:
                    warnings.warn('Overriding the parameters from Reference Dataset.')
                    self._update_params(reference_params)
wxchan's avatar
wxchan committed
1306
                if self.used_indices is None:
1307
                    # create valid
1308
                    self._lazy_init(self.data, label=self.label, reference=self.reference,
1309
1310
                                    weight=self.weight, group=self.group,
                                    init_score=self.init_score, predictor=self._predictor,
1311
                                    silent=self.silent, feature_name=self.feature_name, params=self.params)
wxchan's avatar
wxchan committed
1312
                else:
1313
                    # construct subset
wxchan's avatar
wxchan committed
1314
                    used_indices = list_to_1d_numpy(self.used_indices, np.int32, name='used_indices')
1315
                    assert used_indices.flags.c_contiguous
Guolin Ke's avatar
Guolin Ke committed
1316
                    if self.reference.group is not None:
1317
                        group_info = np.array(self.reference.group).astype(np.int32, copy=False)
1318
                        _, self.group = np.unique(np.repeat(range(len(group_info)), repeats=group_info)[self.used_indices],
1319
                                                  return_counts=True)
1320
                    self.handle = ctypes.c_void_p()
wxchan's avatar
wxchan committed
1321
1322
                    params_str = param_dict_to_str(self.params)
                    _safe_call(_LIB.LGBM_DatasetGetSubset(
1323
                        self.reference.construct().handle,
wxchan's avatar
wxchan committed
1324
                        used_indices.ctypes.data_as(ctypes.POINTER(ctypes.c_int32)),
Guolin Ke's avatar
Guolin Ke committed
1325
                        ctypes.c_int(used_indices.shape[0]),
wxchan's avatar
wxchan committed
1326
1327
                        c_str(params_str),
                        ctypes.byref(self.handle)))
Guolin Ke's avatar
Guolin Ke committed
1328
1329
                    if not self.free_raw_data:
                        self.get_data()
Guolin Ke's avatar
Guolin Ke committed
1330
1331
                    if self.group is not None:
                        self.set_group(self.group)
wxchan's avatar
wxchan committed
1332
1333
                    if self.get_label() is None:
                        raise ValueError("Label should not be None.")
Guolin Ke's avatar
Guolin Ke committed
1334
1335
1336
                    if isinstance(self._predictor, _InnerPredictor) and self._predictor is not self.reference._predictor:
                        self.get_data()
                        self._set_init_score_by_predictor(self._predictor, self.data, used_indices)
wxchan's avatar
wxchan committed
1337
            else:
1338
                # create train
1339
                self._lazy_init(self.data, label=self.label,
1340
1341
1342
                                weight=self.weight, group=self.group,
                                init_score=self.init_score, predictor=self._predictor,
                                silent=self.silent, feature_name=self.feature_name,
1343
                                categorical_feature=self.categorical_feature, params=self.params)
wxchan's avatar
wxchan committed
1344
1345
1346
            if self.free_raw_data:
                self.data = None
        return self
wxchan's avatar
wxchan committed
1347

wxchan's avatar
wxchan committed
1348
    def create_valid(self, data, label=None, weight=None, group=None,
1349
                     init_score=None, silent=False, params=None):
1350
        """Create validation data align with current Dataset.
wxchan's avatar
wxchan committed
1351
1352
1353

        Parameters
        ----------
1354
        data : string, numpy array, pandas DataFrame, H2O DataTable's Frame, scipy.sparse or list of numpy arrays
wxchan's avatar
wxchan committed
1355
            Data source of Dataset.
1356
            If string, it represents the path to txt file.
1357
        label : list, numpy 1-D array, pandas Series / one-column DataFrame or None, optional (default=None)
1358
1359
            Label of the data.
        weight : list, numpy 1-D array, pandas Series or None, optional (default=None)
wxchan's avatar
wxchan committed
1360
            Weight for each instance.
1361
        group : list, numpy 1-D array, pandas Series or None, optional (default=None)
1362
1363
1364
1365
            Group/query data.
            Only used in the learning-to-rank task.
            sum(group) = n_samples.
            For example, if you have a 100-document dataset with ``group = [10, 20, 40, 10, 10, 10]``, that means that you have 6 groups, where the first 10 records are in the first group, records 11-30 are in the second group, etc.
1366
        init_score : list, numpy 1-D array, pandas Series or None, optional (default=None)
1367
            Init score for Dataset.
1368
1369
        silent : bool, optional (default=False)
            Whether to print messages during construction.
Nikita Titov's avatar
Nikita Titov committed
1370
        params : dict or None, optional (default=None)
1371
            Other parameters for validation Dataset.
1372
1373
1374

        Returns
        -------
Nikita Titov's avatar
Nikita Titov committed
1375
1376
        valid : Dataset
            Validation Dataset with reference to self.
wxchan's avatar
wxchan committed
1377
        """
1378
        ret = Dataset(data, label=label, reference=self,
1379
1380
                      weight=weight, group=group, init_score=init_score,
                      silent=silent, params=params, free_raw_data=self.free_raw_data)
wxchan's avatar
wxchan committed
1381
        ret._predictor = self._predictor
1382
        ret.pandas_categorical = self.pandas_categorical
wxchan's avatar
wxchan committed
1383
        return ret
wxchan's avatar
wxchan committed
1384

wxchan's avatar
wxchan committed
1385
    def subset(self, used_indices, params=None):
1386
        """Get subset of current Dataset.
wxchan's avatar
wxchan committed
1387
1388
1389
1390

        Parameters
        ----------
        used_indices : list of int
1391
            Indices used to create the subset.
Nikita Titov's avatar
Nikita Titov committed
1392
        params : dict or None, optional (default=None)
1393
            These parameters will be passed to Dataset constructor.
1394
1395
1396
1397
1398

        Returns
        -------
        subset : Dataset
            Subset of the current Dataset.
wxchan's avatar
wxchan committed
1399
        """
wxchan's avatar
wxchan committed
1400
1401
        if params is None:
            params = self.params
wxchan's avatar
wxchan committed
1402
        ret = Dataset(None, reference=self, feature_name=self.feature_name,
1403
1404
                      categorical_feature=self.categorical_feature, params=params,
                      free_raw_data=self.free_raw_data)
wxchan's avatar
wxchan committed
1405
        ret._predictor = self._predictor
1406
        ret.pandas_categorical = self.pandas_categorical
1407
        ret.used_indices = sorted(used_indices)
wxchan's avatar
wxchan committed
1408
1409
1410
        return ret

    def save_binary(self, filename):
1411
        """Save Dataset to a binary file.
wxchan's avatar
wxchan committed
1412

1413
1414
1415
1416
1417
        .. note::

            Please note that `init_score` is not saved in binary file.
            If you need it, please set it again after loading Dataset.

wxchan's avatar
wxchan committed
1418
1419
1420
1421
        Parameters
        ----------
        filename : string
            Name of the output file.
Nikita Titov's avatar
Nikita Titov committed
1422
1423
1424
1425
1426

        Returns
        -------
        self : Dataset
            Returns self.
wxchan's avatar
wxchan committed
1427
1428
1429
1430
        """
        _safe_call(_LIB.LGBM_DatasetSaveBinary(
            self.construct().handle,
            c_str(filename)))
Nikita Titov's avatar
Nikita Titov committed
1431
        return self
wxchan's avatar
wxchan committed
1432
1433

    def _update_params(self, params):
1434
1435
        if not params:
            return self
1436
1437
1438
1439
1440
1441
1442
1443
1444
1445
1446
1447
1448
1449
1450
1451
1452
1453
1454
1455
1456
        params = copy.deepcopy(params)

        def update():
            if not self.params:
                self.params = params
            else:
                self.params_back_up = copy.deepcopy(self.params)
                self.params.update(params)

        if self.handle is None:
            update()
        elif params is not None:
            ret = _LIB.LGBM_DatasetUpdateParamChecking(
                c_str(param_dict_to_str(self.params)),
                c_str(param_dict_to_str(params)))
            if ret != 0:
                # could be updated if data is not freed
                if self.data is not None:
                    update()
                    self._free_handle()
                else:
1457
                    raise LightGBMError(_LIB.LGBM_GetLastError().decode('utf-8'))
Nikita Titov's avatar
Nikita Titov committed
1458
        return self
wxchan's avatar
wxchan committed
1459

1460
    def _reverse_update_params(self):
1461
1462
1463
        if self.handle is None:
            self.params = copy.deepcopy(self.params_back_up)
            self.params_back_up = None
Nikita Titov's avatar
Nikita Titov committed
1464
        return self
1465

wxchan's avatar
wxchan committed
1466
    def set_field(self, field_name, data):
wxchan's avatar
wxchan committed
1467
        """Set property into the Dataset.
wxchan's avatar
wxchan committed
1468
1469
1470

        Parameters
        ----------
Nikita Titov's avatar
Nikita Titov committed
1471
        field_name : string
1472
            The field name of the information.
1473
        data : list, numpy 1-D array, pandas Series or None
1474
            The array of data to be set.
Nikita Titov's avatar
Nikita Titov committed
1475
1476
1477
1478
1479

        Returns
        -------
        self : Dataset
            Dataset with set property.
wxchan's avatar
wxchan committed
1480
        """
1481
1482
        if self.handle is None:
            raise Exception("Cannot set %s before construct dataset" % field_name)
wxchan's avatar
wxchan committed
1483
        if data is None:
1484
            # set to None
wxchan's avatar
wxchan committed
1485
1486
1487
1488
            _safe_call(_LIB.LGBM_DatasetSetField(
                self.handle,
                c_str(field_name),
                None,
Guolin Ke's avatar
Guolin Ke committed
1489
1490
                ctypes.c_int(0),
                ctypes.c_int(FIELD_TYPE_MAPPER[field_name])))
Nikita Titov's avatar
Nikita Titov committed
1491
            return self
Guolin Ke's avatar
Guolin Ke committed
1492
1493
1494
1495
1496
        dtype = np.float32
        if field_name == 'group':
            dtype = np.int32
        elif field_name == 'init_score':
            dtype = np.float64
1497
        data = list_to_1d_numpy(data, dtype, name=field_name)
1498
1499
        if data.dtype == np.float32 or data.dtype == np.float64:
            ptr_data, type_data, _ = c_float_array(data)
wxchan's avatar
wxchan committed
1500
        elif data.dtype == np.int32:
1501
            ptr_data, type_data, _ = c_int_array(data)
wxchan's avatar
wxchan committed
1502
        else:
Nikita Titov's avatar
Nikita Titov committed
1503
            raise TypeError("Expected np.float32/64 or np.int32, met type({})".format(data.dtype))
wxchan's avatar
wxchan committed
1504
        if type_data != FIELD_TYPE_MAPPER[field_name]:
1505
            raise TypeError("Input type error for set_field")
wxchan's avatar
wxchan committed
1506
1507
1508
1509
        _safe_call(_LIB.LGBM_DatasetSetField(
            self.handle,
            c_str(field_name),
            ptr_data,
Guolin Ke's avatar
Guolin Ke committed
1510
1511
            ctypes.c_int(len(data)),
            ctypes.c_int(type_data)))
1512
        self.version += 1
Nikita Titov's avatar
Nikita Titov committed
1513
        return self
wxchan's avatar
wxchan committed
1514

wxchan's avatar
wxchan committed
1515
1516
    def get_field(self, field_name):
        """Get property from the Dataset.
wxchan's avatar
wxchan committed
1517
1518
1519

        Parameters
        ----------
Nikita Titov's avatar
Nikita Titov committed
1520
        field_name : string
1521
            The field name of the information.
wxchan's avatar
wxchan committed
1522
1523
1524

        Returns
        -------
1525
1526
        info : numpy array
            A numpy array with information from the Dataset.
Guolin Ke's avatar
Guolin Ke committed
1527
        """
1528
        if self.handle is None:
1529
            raise Exception("Cannot get %s before construct Dataset" % field_name)
Guolin Ke's avatar
Guolin Ke committed
1530
1531
        tmp_out_len = ctypes.c_int()
        out_type = ctypes.c_int()
wxchan's avatar
wxchan committed
1532
1533
1534
1535
1536
1537
1538
1539
1540
1541
1542
1543
1544
1545
1546
        ret = ctypes.POINTER(ctypes.c_void_p)()
        _safe_call(_LIB.LGBM_DatasetGetField(
            self.handle,
            c_str(field_name),
            ctypes.byref(tmp_out_len),
            ctypes.byref(ret),
            ctypes.byref(out_type)))
        if out_type.value != FIELD_TYPE_MAPPER[field_name]:
            raise TypeError("Return type error for get_field")
        if tmp_out_len.value == 0:
            return None
        if out_type.value == C_API_DTYPE_INT32:
            return cint32_array_to_numpy(ctypes.cast(ret, ctypes.POINTER(ctypes.c_int32)), tmp_out_len.value)
        elif out_type.value == C_API_DTYPE_FLOAT32:
            return cfloat32_array_to_numpy(ctypes.cast(ret, ctypes.POINTER(ctypes.c_float)), tmp_out_len.value)
Guolin Ke's avatar
Guolin Ke committed
1547
1548
        elif out_type.value == C_API_DTYPE_FLOAT64:
            return cfloat64_array_to_numpy(ctypes.cast(ret, ctypes.POINTER(ctypes.c_double)), tmp_out_len.value)
1549
        else:
wxchan's avatar
wxchan committed
1550
            raise TypeError("Unknown type")
Guolin Ke's avatar
Guolin Ke committed
1551

1552
    def set_categorical_feature(self, categorical_feature):
1553
        """Set categorical features.
1554
1555
1556

        Parameters
        ----------
1557
1558
        categorical_feature : list of int or strings
            Names or indices of categorical features.
Nikita Titov's avatar
Nikita Titov committed
1559
1560
1561
1562
1563

        Returns
        -------
        self : Dataset
            Dataset with set categorical features.
1564
1565
        """
        if self.categorical_feature == categorical_feature:
Nikita Titov's avatar
Nikita Titov committed
1566
            return self
1567
        if self.data is not None:
1568
1569
            if self.categorical_feature is None:
                self.categorical_feature = categorical_feature
Nikita Titov's avatar
Nikita Titov committed
1570
                return self._free_handle()
1571
1572
            elif categorical_feature == 'auto':
                warnings.warn('Using categorical_feature in Dataset.')
Nikita Titov's avatar
Nikita Titov committed
1573
                return self
1574
            else:
1575
1576
                warnings.warn('categorical_feature in Dataset is overridden.\n'
                              'New categorical_feature is {}'.format(sorted(list(categorical_feature))))
1577
                self.categorical_feature = categorical_feature
Nikita Titov's avatar
Nikita Titov committed
1578
                return self._free_handle()
1579
        else:
1580
1581
            raise LightGBMError("Cannot set categorical feature after freed raw data, "
                                "set free_raw_data=False when construct Dataset to avoid this.")
1582

Guolin Ke's avatar
Guolin Ke committed
1583
    def _set_predictor(self, predictor):
1584
1585
1586
1587
        """Set predictor for continued training.

        It is not recommended for user to call this function.
        Please use init_model argument in engine.train() or engine.cv() instead.
Guolin Ke's avatar
Guolin Ke committed
1588
        """
1589
        if predictor is self._predictor and (predictor is None or predictor.current_iteration() == self._predictor.current_iteration()):
Nikita Titov's avatar
Nikita Titov committed
1590
            return self
1591
        if self.handle is None:
Guolin Ke's avatar
Guolin Ke committed
1592
            self._predictor = predictor
1593
1594
1595
1596
1597
1598
        elif self.data is not None:
            self._predictor = predictor
            self._set_init_score_by_predictor(self._predictor, self.data)
        elif self.used_indices is not None and self.reference is not None and self.reference.data is not None:
            self._predictor = predictor
            self._set_init_score_by_predictor(self._predictor, self.reference.data, self.used_indices)
Guolin Ke's avatar
Guolin Ke committed
1599
        else:
1600
1601
            raise LightGBMError("Cannot set predictor after freed raw data, "
                                "set free_raw_data=False when construct Dataset to avoid this.")
1602
        return self
Guolin Ke's avatar
Guolin Ke committed
1603
1604

    def set_reference(self, reference):
1605
        """Set reference Dataset.
Guolin Ke's avatar
Guolin Ke committed
1606
1607
1608
1609

        Parameters
        ----------
        reference : Dataset
1610
            Reference that is used as a template to construct the current Dataset.
Nikita Titov's avatar
Nikita Titov committed
1611
1612
1613
1614
1615

        Returns
        -------
        self : Dataset
            Dataset with set reference.
Guolin Ke's avatar
Guolin Ke committed
1616
        """
1617
1618
1619
        self.set_categorical_feature(reference.categorical_feature) \
            .set_feature_name(reference.feature_name) \
            ._set_predictor(reference._predictor)
1620
1621
        # we're done if self and reference share a common upstrem reference
        if self.get_ref_chain().intersection(reference.get_ref_chain()):
Nikita Titov's avatar
Nikita Titov committed
1622
            return self
Guolin Ke's avatar
Guolin Ke committed
1623
1624
        if self.data is not None:
            self.reference = reference
Nikita Titov's avatar
Nikita Titov committed
1625
            return self._free_handle()
Guolin Ke's avatar
Guolin Ke committed
1626
        else:
1627
1628
            raise LightGBMError("Cannot set reference after freed raw data, "
                                "set free_raw_data=False when construct Dataset to avoid this.")
Guolin Ke's avatar
Guolin Ke committed
1629
1630

    def set_feature_name(self, feature_name):
1631
        """Set feature name.
Guolin Ke's avatar
Guolin Ke committed
1632
1633
1634

        Parameters
        ----------
1635
1636
        feature_name : list of strings
            Feature names.
Nikita Titov's avatar
Nikita Titov committed
1637
1638
1639
1640
1641

        Returns
        -------
        self : Dataset
            Dataset with set feature name.
Guolin Ke's avatar
Guolin Ke committed
1642
        """
1643
1644
        if feature_name != 'auto':
            self.feature_name = feature_name
1645
        if self.handle is not None and feature_name is not None and feature_name != 'auto':
wxchan's avatar
wxchan committed
1646
            if len(feature_name) != self.num_feature():
1647
1648
                raise ValueError("Length of feature_name({}) and num_feature({}) don't match"
                                 .format(len(feature_name), self.num_feature()))
1649
            c_feature_name = [c_str(name) for name in feature_name]
wxchan's avatar
wxchan committed
1650
1651
1652
            _safe_call(_LIB.LGBM_DatasetSetFeatureNames(
                self.handle,
                c_array(ctypes.c_char_p, c_feature_name),
Guolin Ke's avatar
Guolin Ke committed
1653
                ctypes.c_int(len(feature_name))))
Nikita Titov's avatar
Nikita Titov committed
1654
        return self
Guolin Ke's avatar
Guolin Ke committed
1655
1656

    def set_label(self, label):
1657
        """Set label of Dataset.
Guolin Ke's avatar
Guolin Ke committed
1658
1659
1660

        Parameters
        ----------
1661
        label : list, numpy 1-D array, pandas Series / one-column DataFrame or None
1662
            The label information to be set into Dataset.
Nikita Titov's avatar
Nikita Titov committed
1663
1664
1665
1666
1667

        Returns
        -------
        self : Dataset
            Dataset with set label.
Guolin Ke's avatar
Guolin Ke committed
1668
1669
        """
        self.label = label
1670
        if self.handle is not None:
1671
            label = list_to_1d_numpy(_label_from_pandas(label), name='label')
wxchan's avatar
wxchan committed
1672
            self.set_field('label', label)
1673
            self.label = self.get_field('label')  # original values can be modified at cpp side
Nikita Titov's avatar
Nikita Titov committed
1674
        return self
Guolin Ke's avatar
Guolin Ke committed
1675
1676

    def set_weight(self, weight):
1677
        """Set weight of each instance.
Guolin Ke's avatar
Guolin Ke committed
1678
1679
1680

        Parameters
        ----------
1681
        weight : list, numpy 1-D array, pandas Series or None
1682
            Weight to be set for each data point.
Nikita Titov's avatar
Nikita Titov committed
1683
1684
1685
1686
1687

        Returns
        -------
        self : Dataset
            Dataset with set weight.
Guolin Ke's avatar
Guolin Ke committed
1688
        """
1689
1690
        if weight is not None and np.all(weight == 1):
            weight = None
Guolin Ke's avatar
Guolin Ke committed
1691
        self.weight = weight
1692
        if self.handle is not None and weight is not None:
wxchan's avatar
wxchan committed
1693
1694
            weight = list_to_1d_numpy(weight, name='weight')
            self.set_field('weight', weight)
1695
            self.weight = self.get_field('weight')  # original values can be modified at cpp side
Nikita Titov's avatar
Nikita Titov committed
1696
        return self
Guolin Ke's avatar
Guolin Ke committed
1697
1698

    def set_init_score(self, init_score):
1699
        """Set init score of Booster to start from.
Guolin Ke's avatar
Guolin Ke committed
1700
1701
1702

        Parameters
        ----------
1703
        init_score : list, numpy 1-D array, pandas Series or None
1704
            Init score for Booster.
Nikita Titov's avatar
Nikita Titov committed
1705
1706
1707
1708
1709

        Returns
        -------
        self : Dataset
            Dataset with set init score.
Guolin Ke's avatar
Guolin Ke committed
1710
1711
        """
        self.init_score = init_score
1712
        if self.handle is not None and init_score is not None:
Guolin Ke's avatar
Guolin Ke committed
1713
            init_score = list_to_1d_numpy(init_score, np.float64, name='init_score')
wxchan's avatar
wxchan committed
1714
            self.set_field('init_score', init_score)
1715
            self.init_score = self.get_field('init_score')  # original values can be modified at cpp side
Nikita Titov's avatar
Nikita Titov committed
1716
        return self
Guolin Ke's avatar
Guolin Ke committed
1717
1718

    def set_group(self, group):
1719
        """Set group size of Dataset (used for ranking).
Guolin Ke's avatar
Guolin Ke committed
1720
1721
1722

        Parameters
        ----------
1723
        group : list, numpy 1-D array, pandas Series or None
1724
1725
1726
1727
            Group/query data.
            Only used in the learning-to-rank task.
            sum(group) = n_samples.
            For example, if you have a 100-document dataset with ``group = [10, 20, 40, 10, 10, 10]``, that means that you have 6 groups, where the first 10 records are in the first group, records 11-30 are in the second group, etc.
Nikita Titov's avatar
Nikita Titov committed
1728
1729
1730
1731
1732

        Returns
        -------
        self : Dataset
            Dataset with set group.
Guolin Ke's avatar
Guolin Ke committed
1733
1734
        """
        self.group = group
1735
        if self.handle is not None and group is not None:
wxchan's avatar
wxchan committed
1736
1737
            group = list_to_1d_numpy(group, np.int32, name='group')
            self.set_field('group', group)
Nikita Titov's avatar
Nikita Titov committed
1738
        return self
Guolin Ke's avatar
Guolin Ke committed
1739

1740
1741
1742
1743
1744
1745
1746
1747
1748
1749
1750
1751
1752
1753
    def get_feature_name(self):
        """Get the names of columns (features) in the Dataset.

        Returns
        -------
        feature_names : list
            The names of columns (features) in the Dataset.
        """
        if self.handle is None:
            raise LightGBMError("Cannot get feature_name before construct dataset")
        num_feature = self.num_feature()
        tmp_out_len = ctypes.c_int(0)
        reserved_string_buffer_size = 255
        required_string_buffer_size = ctypes.c_size_t(0)
1754
        string_buffers = [ctypes.create_string_buffer(reserved_string_buffer_size) for i in range(num_feature)]
1755
1756
1757
        ptr_string_buffers = (ctypes.c_char_p * num_feature)(*map(ctypes.addressof, string_buffers))
        _safe_call(_LIB.LGBM_DatasetGetFeatureNames(
            self.handle,
1758
            ctypes.c_int(num_feature),
1759
            ctypes.byref(tmp_out_len),
1760
            ctypes.c_size_t(reserved_string_buffer_size),
1761
1762
1763
1764
1765
1766
1767
1768
1769
            ctypes.byref(required_string_buffer_size),
            ptr_string_buffers))
        if num_feature != tmp_out_len.value:
            raise ValueError("Length of feature names doesn't equal with num_feature")
        if reserved_string_buffer_size < required_string_buffer_size.value:
            raise BufferError(
                "Allocated feature name buffer size ({}) was inferior to the needed size ({})."
                .format(reserved_string_buffer_size, required_string_buffer_size.value)
            )
1770
        return [string_buffers[i].value.decode('utf-8') for i in range(num_feature)]
1771

Guolin Ke's avatar
Guolin Ke committed
1772
    def get_label(self):
1773
        """Get the label of the Dataset.
Guolin Ke's avatar
Guolin Ke committed
1774
1775
1776

        Returns
        -------
Nikita Titov's avatar
Nikita Titov committed
1777
        label : numpy array or None
1778
            The label information from the Dataset.
Guolin Ke's avatar
Guolin Ke committed
1779
        """
1780
        if self.label is None:
wxchan's avatar
wxchan committed
1781
            self.label = self.get_field('label')
Guolin Ke's avatar
Guolin Ke committed
1782
1783
1784
        return self.label

    def get_weight(self):
1785
        """Get the weight of the Dataset.
Guolin Ke's avatar
Guolin Ke committed
1786
1787
1788

        Returns
        -------
Nikita Titov's avatar
Nikita Titov committed
1789
        weight : numpy array or None
1790
            Weight for each data point from the Dataset.
Guolin Ke's avatar
Guolin Ke committed
1791
        """
1792
        if self.weight is None:
wxchan's avatar
wxchan committed
1793
            self.weight = self.get_field('weight')
Guolin Ke's avatar
Guolin Ke committed
1794
1795
1796
        return self.weight

    def get_init_score(self):
1797
        """Get the initial score of the Dataset.
Guolin Ke's avatar
Guolin Ke committed
1798
1799
1800

        Returns
        -------
Nikita Titov's avatar
Nikita Titov committed
1801
        init_score : numpy array or None
1802
            Init score of Booster.
Guolin Ke's avatar
Guolin Ke committed
1803
        """
1804
        if self.init_score is None:
wxchan's avatar
wxchan committed
1805
            self.init_score = self.get_field('init_score')
Guolin Ke's avatar
Guolin Ke committed
1806
1807
        return self.init_score

1808
1809
1810
1811
1812
    def get_data(self):
        """Get the raw data of the Dataset.

        Returns
        -------
1813
        data : string, numpy array, pandas DataFrame, H2O DataTable's Frame, scipy.sparse, list of numpy arrays or None
1814
1815
1816
1817
            Raw data used in the Dataset construction.
        """
        if self.handle is None:
            raise Exception("Cannot get data before construct Dataset")
Guolin Ke's avatar
Guolin Ke committed
1818
1819
1820
1821
1822
1823
1824
1825
1826
1827
1828
1829
        if self.need_slice and self.used_indices is not None and self.reference is not None:
            self.data = self.reference.data
            if self.data is not None:
                if isinstance(self.data, np.ndarray) or scipy.sparse.issparse(self.data):
                    self.data = self.data[self.used_indices, :]
                elif isinstance(self.data, DataFrame):
                    self.data = self.data.iloc[self.used_indices].copy()
                elif isinstance(self.data, DataTable):
                    self.data = self.data[self.used_indices, :]
                else:
                    warnings.warn("Cannot subset {} type of raw data.\n"
                                  "Returning original raw data".format(type(self.data).__name__))
1830
            self.need_slice = False
Guolin Ke's avatar
Guolin Ke committed
1831
1832
1833
        if self.data is None:
            raise LightGBMError("Cannot call `get_data` after freed raw data, "
                                "set free_raw_data=False when construct Dataset to avoid this.")
1834
1835
        return self.data

Guolin Ke's avatar
Guolin Ke committed
1836
    def get_group(self):
1837
        """Get the group of the Dataset.
Guolin Ke's avatar
Guolin Ke committed
1838
1839
1840

        Returns
        -------
Nikita Titov's avatar
Nikita Titov committed
1841
        group : numpy array or None
1842
1843
1844
1845
            Group/query data.
            Only used in the learning-to-rank task.
            sum(group) = n_samples.
            For example, if you have a 100-document dataset with ``group = [10, 20, 40, 10, 10, 10]``, that means that you have 6 groups, where the first 10 records are in the first group, records 11-30 are in the second group, etc.
Guolin Ke's avatar
Guolin Ke committed
1846
        """
1847
        if self.group is None:
wxchan's avatar
wxchan committed
1848
            self.group = self.get_field('group')
Guolin Ke's avatar
Guolin Ke committed
1849
1850
            if self.group is not None:
                # group data from LightGBM is boundaries data, need to convert to group size
Nikita Titov's avatar
Nikita Titov committed
1851
                self.group = np.diff(self.group)
Guolin Ke's avatar
Guolin Ke committed
1852
1853
1854
        return self.group

    def num_data(self):
1855
        """Get the number of rows in the Dataset.
Guolin Ke's avatar
Guolin Ke committed
1856
1857
1858

        Returns
        -------
1859
1860
        number_of_rows : int
            The number of rows in the Dataset.
Guolin Ke's avatar
Guolin Ke committed
1861
        """
1862
        if self.handle is not None:
Guolin Ke's avatar
Guolin Ke committed
1863
            ret = ctypes.c_int()
wxchan's avatar
wxchan committed
1864
1865
1866
            _safe_call(_LIB.LGBM_DatasetGetNumData(self.handle,
                                                   ctypes.byref(ret)))
            return ret.value
Guolin Ke's avatar
Guolin Ke committed
1867
        else:
1868
            raise LightGBMError("Cannot get num_data before construct dataset")
Guolin Ke's avatar
Guolin Ke committed
1869
1870

    def num_feature(self):
1871
        """Get the number of columns (features) in the Dataset.
Guolin Ke's avatar
Guolin Ke committed
1872
1873
1874

        Returns
        -------
1875
1876
        number_of_columns : int
            The number of columns (features) in the Dataset.
Guolin Ke's avatar
Guolin Ke committed
1877
        """
1878
        if self.handle is not None:
Guolin Ke's avatar
Guolin Ke committed
1879
            ret = ctypes.c_int()
wxchan's avatar
wxchan committed
1880
1881
1882
            _safe_call(_LIB.LGBM_DatasetGetNumFeature(self.handle,
                                                      ctypes.byref(ret)))
            return ret.value
Guolin Ke's avatar
Guolin Ke committed
1883
        else:
1884
            raise LightGBMError("Cannot get num_feature before construct dataset")
Guolin Ke's avatar
Guolin Ke committed
1885

1886
    def get_ref_chain(self, ref_limit=100):
1887
1888
1889
1890
1891
        """Get a chain of Dataset objects.

        Starts with r, then goes to r.reference (if exists),
        then to r.reference.reference, etc.
        until we hit ``ref_limit`` or a reference loop.
1892
1893
1894
1895
1896

        Parameters
        ----------
        ref_limit : int, optional (default=100)
            The limit number of references.
1897
1898
1899

        Returns
        -------
1900
1901
1902
        ref_chain : set of Dataset
            Chain of references of the Datasets.
        """
1903
        head = self
1904
        ref_chain = set()
1905
1906
        while len(ref_chain) < ref_limit:
            if isinstance(head, Dataset):
1907
                ref_chain.add(head)
1908
1909
1910
1911
1912
1913
                if (head.reference is not None) and (head.reference not in ref_chain):
                    head = head.reference
                else:
                    break
            else:
                break
Nikita Titov's avatar
Nikita Titov committed
1914
        return ref_chain
1915

1916
1917
1918
1919
1920
1921
1922
1923
1924
1925
1926
1927
1928
1929
1930
1931
1932
1933
    def add_features_from(self, other):
        """Add features from other Dataset to the current Dataset.

        Both Datasets must be constructed before calling this method.

        Parameters
        ----------
        other : Dataset
            The Dataset to take features from.

        Returns
        -------
        self : Dataset
            Dataset with the new features added.
        """
        if self.handle is None or other.handle is None:
            raise ValueError('Both source and target Datasets must be constructed before adding features')
        _safe_call(_LIB.LGBM_DatasetAddFeaturesFrom(self.handle, other.handle))
Guolin Ke's avatar
Guolin Ke committed
1934
1935
1936
1937
1938
1939
1940
1941
1942
1943
1944
1945
1946
1947
1948
1949
1950
1951
1952
1953
1954
1955
1956
1957
1958
1959
1960
1961
1962
1963
1964
1965
1966
1967
1968
1969
1970
1971
1972
1973
1974
1975
1976
1977
1978
1979
1980
1981
1982
1983
1984
1985
1986
1987
1988
1989
1990
1991
1992
1993
1994
1995
1996
1997
1998
1999
2000
2001
2002
2003
        was_none = self.data is None
        old_self_data_type = type(self.data).__name__
        if other.data is None:
            self.data = None
        elif self.data is not None:
            if isinstance(self.data, np.ndarray):
                if isinstance(other.data, np.ndarray):
                    self.data = np.hstack((self.data, other.data))
                elif scipy.sparse.issparse(other.data):
                    self.data = np.hstack((self.data, other.data.toarray()))
                elif isinstance(other.data, DataFrame):
                    self.data = np.hstack((self.data, other.data.values))
                elif isinstance(other.data, DataTable):
                    self.data = np.hstack((self.data, other.data.to_numpy()))
                else:
                    self.data = None
            elif scipy.sparse.issparse(self.data):
                sparse_format = self.data.getformat()
                if isinstance(other.data, np.ndarray) or scipy.sparse.issparse(other.data):
                    self.data = scipy.sparse.hstack((self.data, other.data), format=sparse_format)
                elif isinstance(other.data, DataFrame):
                    self.data = scipy.sparse.hstack((self.data, other.data.values), format=sparse_format)
                elif isinstance(other.data, DataTable):
                    self.data = scipy.sparse.hstack((self.data, other.data.to_numpy()), format=sparse_format)
                else:
                    self.data = None
            elif isinstance(self.data, DataFrame):
                if not PANDAS_INSTALLED:
                    raise LightGBMError("Cannot add features to DataFrame type of raw data "
                                        "without pandas installed")
                from pandas import concat
                if isinstance(other.data, np.ndarray):
                    self.data = concat((self.data, DataFrame(other.data)),
                                       axis=1, ignore_index=True)
                elif scipy.sparse.issparse(other.data):
                    self.data = concat((self.data, DataFrame(other.data.toarray())),
                                       axis=1, ignore_index=True)
                elif isinstance(other.data, DataFrame):
                    self.data = concat((self.data, other.data),
                                       axis=1, ignore_index=True)
                elif isinstance(other.data, DataTable):
                    self.data = concat((self.data, DataFrame(other.data.to_numpy())),
                                       axis=1, ignore_index=True)
                else:
                    self.data = None
            elif isinstance(self.data, DataTable):
                if isinstance(other.data, np.ndarray):
                    self.data = DataTable(np.hstack((self.data.to_numpy(), other.data)))
                elif scipy.sparse.issparse(other.data):
                    self.data = DataTable(np.hstack((self.data.to_numpy(), other.data.toarray())))
                elif isinstance(other.data, DataFrame):
                    self.data = DataTable(np.hstack((self.data.to_numpy(), other.data.values)))
                elif isinstance(other.data, DataTable):
                    self.data = DataTable(np.hstack((self.data.to_numpy(), other.data.to_numpy())))
                else:
                    self.data = None
            else:
                self.data = None
        if self.data is None:
            err_msg = ("Cannot add features from {} type of raw data to "
                       "{} type of raw data.\n").format(type(other.data).__name__,
                                                        old_self_data_type)
            err_msg += ("Set free_raw_data=False when construct Dataset to avoid this"
                        if was_none else "Freeing raw data")
            warnings.warn(err_msg)
        self.feature_name = self.get_feature_name()
        warnings.warn("Reseting categorical features.\n"
                      "You can set new categorical features via ``set_categorical_feature`` method")
        self.categorical_feature = "auto"
        self.pandas_categorical = None
2004
2005
        return self

2006
    def _dump_text(self, filename):
2007
2008
2009
2010
2011
2012
2013
2014
2015
2016
2017
2018
2019
2020
2021
2022
2023
2024
2025
        """Save Dataset to a text file.

        This format cannot be loaded back in by LightGBM, but is useful for debugging purposes.

        Parameters
        ----------
        filename : string
            Name of the output file.

        Returns
        -------
        self : Dataset
            Returns self.
        """
        _safe_call(_LIB.LGBM_DatasetDumpText(
            self.construct().handle,
            c_str(filename)))
        return self

wxchan's avatar
wxchan committed
2026

2027
class Booster:
2028
    """Booster in LightGBM."""
2029

2030
    def __init__(self, params=None, train_set=None, model_file=None, model_str=None, silent=False):
2031
        """Initialize the Booster.
wxchan's avatar
wxchan committed
2032
2033
2034

        Parameters
        ----------
Nikita Titov's avatar
Nikita Titov committed
2035
        params : dict or None, optional (default=None)
2036
2037
2038
2039
            Parameters for Booster.
        train_set : Dataset or None, optional (default=None)
            Training dataset.
        model_file : string or None, optional (default=None)
wxchan's avatar
wxchan committed
2040
            Path to the model file.
2041
2042
        model_str : string or None, optional (default=None)
            Model will be loaded from this string.
2043
2044
        silent : bool, optional (default=False)
            Whether to print messages during construction.
wxchan's avatar
wxchan committed
2045
        """
2046
        self.handle = None
2047
        self.network = False
wxchan's avatar
wxchan committed
2048
        self.__need_reload_eval_info = True
2049
        self._train_data_name = "training"
wxchan's avatar
wxchan committed
2050
        self.__attr = {}
2051
        self.__set_objective_to_none = False
wxchan's avatar
wxchan committed
2052
        self.best_iteration = -1
wxchan's avatar
wxchan committed
2053
        self.best_score = {}
2054
        params = {} if params is None else copy.deepcopy(params)
2055
        # user can set verbose with params, it has higher priority
2056
        if not any(verbose_alias in params for verbose_alias in _ConfigAliases.get("verbosity")) and silent:
2057
            params["verbose"] = -1
wxchan's avatar
wxchan committed
2058
        if train_set is not None:
2059
            # Training task
wxchan's avatar
wxchan committed
2060
            if not isinstance(train_set, Dataset):
2061
2062
                raise TypeError('Training data should be Dataset instance, met {}'
                                .format(type(train_set).__name__))
2063
            # set network if necessary
2064
            for alias in _ConfigAliases.get("machines"):
2065
2066
                if alias in params:
                    machines = params[alias]
2067
                    if isinstance(machines, str):
2068
2069
2070
2071
2072
2073
2074
2075
2076
                        num_machines = len(machines.split(','))
                    elif isinstance(machines, (list, set)):
                        num_machines = len(machines)
                        machines = ','.join(machines)
                    else:
                        raise ValueError("Invalid machines in params.")
                    self.set_network(machines,
                                     local_listen_port=params.get("local_listen_port", 12400),
                                     listen_time_out=params.get("listen_time_out", 120),
2077
                                     num_machines=params.setdefault("num_machines", num_machines))
2078
                    break
2079
            # construct booster object
2080
2081
2082
2083
            train_set.construct()
            # copy the parameters from train_set
            params.update(train_set.get_params())
            params_str = param_dict_to_str(params)
2084
            self.handle = ctypes.c_void_p()
wxchan's avatar
wxchan committed
2085
            _safe_call(_LIB.LGBM_BoosterCreate(
2086
                train_set.handle,
wxchan's avatar
wxchan committed
2087
2088
                c_str(params_str),
                ctypes.byref(self.handle)))
2089
            # save reference to data
wxchan's avatar
wxchan committed
2090
2091
2092
2093
            self.train_set = train_set
            self.valid_sets = []
            self.name_valid_sets = []
            self.__num_dataset = 1
Guolin Ke's avatar
Guolin Ke committed
2094
2095
            self.__init_predictor = train_set._predictor
            if self.__init_predictor is not None:
wxchan's avatar
wxchan committed
2096
2097
                _safe_call(_LIB.LGBM_BoosterMerge(
                    self.handle,
Guolin Ke's avatar
Guolin Ke committed
2098
                    self.__init_predictor.handle))
Guolin Ke's avatar
Guolin Ke committed
2099
            out_num_class = ctypes.c_int(0)
wxchan's avatar
wxchan committed
2100
2101
2102
2103
            _safe_call(_LIB.LGBM_BoosterGetNumClasses(
                self.handle,
                ctypes.byref(out_num_class)))
            self.__num_class = out_num_class.value
2104
            # buffer for inner predict
wxchan's avatar
wxchan committed
2105
2106
2107
            self.__inner_predict_buffer = [None]
            self.__is_predicted_cur_iter = [False]
            self.__get_eval_info()
2108
            self.pandas_categorical = train_set.pandas_categorical
2109
            self.train_set_version = train_set.version
wxchan's avatar
wxchan committed
2110
        elif model_file is not None:
2111
            # Prediction task
Guolin Ke's avatar
Guolin Ke committed
2112
            out_num_iterations = ctypes.c_int(0)
2113
            self.handle = ctypes.c_void_p()
wxchan's avatar
wxchan committed
2114
2115
2116
2117
            _safe_call(_LIB.LGBM_BoosterCreateFromModelfile(
                c_str(model_file),
                ctypes.byref(out_num_iterations),
                ctypes.byref(self.handle)))
Guolin Ke's avatar
Guolin Ke committed
2118
            out_num_class = ctypes.c_int(0)
wxchan's avatar
wxchan committed
2119
2120
2121
2122
            _safe_call(_LIB.LGBM_BoosterGetNumClasses(
                self.handle,
                ctypes.byref(out_num_class)))
            self.__num_class = out_num_class.value
2123
            self.pandas_categorical = _load_pandas_categorical(file_name=model_file)
2124
2125
        elif model_str is not None:
            self.model_from_string(model_str, not silent)
wxchan's avatar
wxchan committed
2126
        else:
2127
2128
            raise TypeError('Need at least one training dataset or model file or model string '
                            'to create Booster instance')
2129
        self.params = params
wxchan's avatar
wxchan committed
2130
2131

    def __del__(self):
2132
2133
2134
2135
2136
2137
2138
2139
2140
2141
        try:
            if self.network:
                self.free_network()
        except AttributeError:
            pass
        try:
            if self.handle is not None:
                _safe_call(_LIB.LGBM_BoosterFree(self.handle))
        except AttributeError:
            pass
wxchan's avatar
wxchan committed
2142

wxchan's avatar
wxchan committed
2143
2144
2145
2146
    def __copy__(self):
        return self.__deepcopy__(None)

    def __deepcopy__(self, _):
2147
        model_str = self.model_to_string(num_iteration=-1)
2148
        booster = Booster(model_str=model_str)
2149
        return booster
wxchan's avatar
wxchan committed
2150
2151
2152
2153
2154
2155
2156

    def __getstate__(self):
        this = self.__dict__.copy()
        handle = this['handle']
        this.pop('train_set', None)
        this.pop('valid_sets', None)
        if handle is not None:
2157
            this["handle"] = self.model_to_string(num_iteration=-1)
wxchan's avatar
wxchan committed
2158
2159
2160
        return this

    def __setstate__(self, state):
2161
2162
        model_str = state.get('handle', None)
        if model_str is not None:
wxchan's avatar
wxchan committed
2163
            handle = ctypes.c_void_p()
Guolin Ke's avatar
Guolin Ke committed
2164
            out_num_iterations = ctypes.c_int(0)
2165
2166
2167
2168
            _safe_call(_LIB.LGBM_BoosterLoadModelFromString(
                c_str(model_str),
                ctypes.byref(out_num_iterations),
                ctypes.byref(handle)))
wxchan's avatar
wxchan committed
2169
2170
2171
            state['handle'] = handle
        self.__dict__.update(state)

wxchan's avatar
wxchan committed
2172
    def free_dataset(self):
Nikita Titov's avatar
Nikita Titov committed
2173
2174
2175
2176
2177
2178
2179
        """Free Booster's Datasets.

        Returns
        -------
        self : Booster
            Booster without Datasets.
        """
wxchan's avatar
wxchan committed
2180
2181
        self.__dict__.pop('train_set', None)
        self.__dict__.pop('valid_sets', None)
2182
        self.__num_dataset = 0
Nikita Titov's avatar
Nikita Titov committed
2183
        return self
wxchan's avatar
wxchan committed
2184

2185
2186
2187
    def _free_buffer(self):
        self.__inner_predict_buffer = []
        self.__is_predicted_cur_iter = []
Nikita Titov's avatar
Nikita Titov committed
2188
        return self
2189

2190
2191
2192
2193
2194
2195
    def set_network(self, machines, local_listen_port=12400,
                    listen_time_out=120, num_machines=1):
        """Set the network configuration.

        Parameters
        ----------
Nikita Titov's avatar
Nikita Titov committed
2196
        machines : list, set or string
2197
            Names of machines.
Nikita Titov's avatar
Nikita Titov committed
2198
        local_listen_port : int, optional (default=12400)
2199
            TCP listen port for local machines.
Nikita Titov's avatar
Nikita Titov committed
2200
        listen_time_out : int, optional (default=120)
2201
            Socket time-out in minutes.
Nikita Titov's avatar
Nikita Titov committed
2202
        num_machines : int, optional (default=1)
2203
            The number of machines for parallel learning application.
Nikita Titov's avatar
Nikita Titov committed
2204
2205
2206
2207
2208

        Returns
        -------
        self : Booster
            Booster with set network.
2209
2210
2211
2212
2213
2214
        """
        _safe_call(_LIB.LGBM_NetworkInit(c_str(machines),
                                         ctypes.c_int(local_listen_port),
                                         ctypes.c_int(listen_time_out),
                                         ctypes.c_int(num_machines)))
        self.network = True
Nikita Titov's avatar
Nikita Titov committed
2215
        return self
2216
2217

    def free_network(self):
Nikita Titov's avatar
Nikita Titov committed
2218
2219
2220
2221
2222
2223
2224
        """Free Booster's network.

        Returns
        -------
        self : Booster
            Booster with freed network.
        """
2225
2226
        _safe_call(_LIB.LGBM_NetworkFree())
        self.network = False
Nikita Titov's avatar
Nikita Titov committed
2227
        return self
2228

2229
2230
2231
    def trees_to_dataframe(self):
        """Parse the fitted model and return in an easy-to-read pandas DataFrame.

2232
2233
2234
2235
2236
2237
2238
2239
2240
2241
2242
2243
2244
2245
2246
2247
2248
2249
        The returned DataFrame has the following columns.

            - ``tree_index`` : int64, which tree a node belongs to. 0-based, so a value of ``6``, for example, means "this node is in the 7th tree".
            - ``node_depth`` : int64, how far a node is from the root of the tree. The root node has a value of ``1``, its direct children are ``2``, etc.
            - ``node_index`` : string, unique identifier for a node.
            - ``left_child`` : string, ``node_index`` of the child node to the left of a split. ``None`` for leaf nodes.
            - ``right_child`` : string, ``node_index`` of the child node to the right of a split. ``None`` for leaf nodes.
            - ``parent_index`` : string, ``node_index`` of this node's parent. ``None`` for the root node.
            - ``split_feature`` : string, name of the feature used for splitting. ``None`` for leaf nodes.
            - ``split_gain`` : float64, gain from adding this split to the tree. ``NaN`` for leaf nodes.
            - ``threshold`` : float64, value of the feature used to decide which side of the split a record will go down. ``NaN`` for leaf nodes.
            - ``decision_type`` : string, logical operator describing how to compare a value to ``threshold``. For example, ``split_feature = "Column_10", threshold = 15, decision_type = "<="`` means that records where ``Column_10 <= 15`` follow the left side of the split, otherwise follows the right side of the split. ``None`` for leaf nodes.
            - ``missing_direction`` : string, split direction that missing values should go to. ``None`` for leaf nodes.
            - ``missing_type`` : string, describes what types of values are treated as missing.
            - ``value`` : float64, predicted value for this leaf node, multiplied by the learning rate.
            - ``weight`` : float64 or int64, sum of hessian (second-order derivative of objective), summed over observations that fall in this node.
            - ``count`` : int64, number of records in the training data that fall into this node.

2250
2251
2252
2253
2254
2255
2256
2257
2258
2259
2260
2261
2262
2263
2264
2265
2266
2267
2268
2269
2270
2271
2272
2273
2274
2275
2276
2277
2278
2279
2280
2281
2282
2283
2284
2285
        Returns
        -------
        result : pandas DataFrame
            Returns a pandas DataFrame of the parsed model.
        """
        if not PANDAS_INSTALLED:
            raise LightGBMError('This method cannot be run without pandas installed')

        if self.num_trees() == 0:
            raise LightGBMError('There are no trees in this Booster and thus nothing to parse')

        def _is_split_node(tree):
            return 'split_index' in tree.keys()

        def create_node_record(tree, node_depth=1, tree_index=None,
                               feature_names=None, parent_node=None):

            def _get_node_index(tree, tree_index):
                tree_num = str(tree_index) + '-' if tree_index is not None else ''
                is_split = _is_split_node(tree)
                node_type = 'S' if is_split else 'L'
                # if a single node tree it won't have `leaf_index` so return 0
                node_num = str(tree.get('split_index' if is_split else 'leaf_index', 0))
                return tree_num + node_type + node_num

            def _get_split_feature(tree, feature_names):
                if _is_split_node(tree):
                    if feature_names is not None:
                        feature_name = feature_names[tree['split_feature']]
                    else:
                        feature_name = tree['split_feature']
                else:
                    feature_name = None
                return feature_name

            def _is_single_node_tree(tree):
2286
                return set(tree.keys()) == {'leaf_value'}
2287
2288
2289
2290
2291
2292
2293
2294
2295
2296
2297
2298
2299
2300
2301
2302
2303
2304
2305
2306
2307
2308
2309
2310
2311
2312
2313
2314
2315
2316
2317
2318
2319
2320
2321
2322
2323
2324
2325
2326
2327
2328
2329
2330
2331
2332
2333
2334
2335
2336
2337
2338
2339
2340
2341
2342
2343
2344
2345
2346
2347
2348
2349
2350
2351
2352
2353
2354
2355
2356
2357
2358
2359
2360
2361

            # Create the node record, and populate universal data members
            node = OrderedDict()
            node['tree_index'] = tree_index
            node['node_depth'] = node_depth
            node['node_index'] = _get_node_index(tree, tree_index)
            node['left_child'] = None
            node['right_child'] = None
            node['parent_index'] = parent_node
            node['split_feature'] = _get_split_feature(tree, feature_names)
            node['split_gain'] = None
            node['threshold'] = None
            node['decision_type'] = None
            node['missing_direction'] = None
            node['missing_type'] = None
            node['value'] = None
            node['weight'] = None
            node['count'] = None

            # Update values to reflect node type (leaf or split)
            if _is_split_node(tree):
                node['left_child'] = _get_node_index(tree['left_child'], tree_index)
                node['right_child'] = _get_node_index(tree['right_child'], tree_index)
                node['split_gain'] = tree['split_gain']
                node['threshold'] = tree['threshold']
                node['decision_type'] = tree['decision_type']
                node['missing_direction'] = 'left' if tree['default_left'] else 'right'
                node['missing_type'] = tree['missing_type']
                node['value'] = tree['internal_value']
                node['weight'] = tree['internal_weight']
                node['count'] = tree['internal_count']
            else:
                node['value'] = tree['leaf_value']
                if not _is_single_node_tree(tree):
                    node['weight'] = tree['leaf_weight']
                    node['count'] = tree['leaf_count']

            return node

        def tree_dict_to_node_list(tree, node_depth=1, tree_index=None,
                                   feature_names=None, parent_node=None):

            node = create_node_record(tree,
                                      node_depth=node_depth,
                                      tree_index=tree_index,
                                      feature_names=feature_names,
                                      parent_node=parent_node)

            res = [node]

            if _is_split_node(tree):
                # traverse the next level of the tree
                children = ['left_child', 'right_child']
                for child in children:
                    subtree_list = tree_dict_to_node_list(
                        tree[child],
                        node_depth=node_depth + 1,
                        tree_index=tree_index,
                        feature_names=feature_names,
                        parent_node=node['node_index'])
                    # In tree format, "subtree_list" is a list of node records (dicts),
                    # and we add node to the list.
                    res.extend(subtree_list)
            return res

        model_dict = self.dump_model()
        feature_names = model_dict['feature_names']
        model_list = []
        for tree in model_dict['tree_info']:
            model_list.extend(tree_dict_to_node_list(tree['tree_structure'],
                                                     tree_index=tree['tree_index'],
                                                     feature_names=feature_names))

        return DataFrame(model_list, columns=model_list[0].keys())

wxchan's avatar
wxchan committed
2362
    def set_train_data_name(self, name):
2363
2364
2365
2366
        """Set the name to the training Dataset.

        Parameters
        ----------
Nikita Titov's avatar
Nikita Titov committed
2367
2368
2369
2370
2371
2372
2373
        name : string
            Name for the training Dataset.

        Returns
        -------
        self : Booster
            Booster with set training Dataset name.
2374
        """
2375
        self._train_data_name = name
Nikita Titov's avatar
Nikita Titov committed
2376
        return self
wxchan's avatar
wxchan committed
2377
2378

    def add_valid(self, data, name):
2379
        """Add validation data.
wxchan's avatar
wxchan committed
2380
2381
2382
2383

        Parameters
        ----------
        data : Dataset
2384
2385
2386
            Validation data.
        name : string
            Name of validation data.
Nikita Titov's avatar
Nikita Titov committed
2387
2388
2389
2390
2391

        Returns
        -------
        self : Booster
            Booster with set validation data.
wxchan's avatar
wxchan committed
2392
        """
Guolin Ke's avatar
Guolin Ke committed
2393
        if not isinstance(data, Dataset):
2394
2395
            raise TypeError('Validation data should be Dataset instance, met {}'
                            .format(type(data).__name__))
Guolin Ke's avatar
Guolin Ke committed
2396
        if data._predictor is not self.__init_predictor:
2397
2398
            raise LightGBMError("Add validation data failed, "
                                "you should use same predictor for these data")
wxchan's avatar
wxchan committed
2399
2400
        _safe_call(_LIB.LGBM_BoosterAddValidData(
            self.handle,
wxchan's avatar
wxchan committed
2401
            data.construct().handle))
wxchan's avatar
wxchan committed
2402
2403
2404
2405
2406
        self.valid_sets.append(data)
        self.name_valid_sets.append(name)
        self.__num_dataset += 1
        self.__inner_predict_buffer.append(None)
        self.__is_predicted_cur_iter.append(False)
Nikita Titov's avatar
Nikita Titov committed
2407
        return self
wxchan's avatar
wxchan committed
2408
2409

    def reset_parameter(self, params):
2410
        """Reset parameters of Booster.
wxchan's avatar
wxchan committed
2411
2412
2413
2414

        Parameters
        ----------
        params : dict
2415
            New parameters for Booster.
Nikita Titov's avatar
Nikita Titov committed
2416
2417
2418
2419
2420

        Returns
        -------
        self : Booster
            Booster with new parameters.
wxchan's avatar
wxchan committed
2421
2422
2423
2424
2425
2426
        """
        params_str = param_dict_to_str(params)
        if params_str:
            _safe_call(_LIB.LGBM_BoosterResetParameter(
                self.handle,
                c_str(params_str)))
Guolin Ke's avatar
Guolin Ke committed
2427
        self.params.update(params)
Nikita Titov's avatar
Nikita Titov committed
2428
        return self
wxchan's avatar
wxchan committed
2429
2430

    def update(self, train_set=None, fobj=None):
Nikita Titov's avatar
Nikita Titov committed
2431
        """Update Booster for one iteration.
2432

wxchan's avatar
wxchan committed
2433
2434
        Parameters
        ----------
2435
2436
2437
2438
        train_set : Dataset or None, optional (default=None)
            Training data.
            If None, last training data is used.
        fobj : callable or None, optional (default=None)
wxchan's avatar
wxchan committed
2439
            Customized objective function.
2440
2441
2442
2443
2444
2445
2446
2447
2448
2449
2450
            Should accept two parameters: preds, train_data,
            and return (grad, hess).

                preds : list or numpy 1-D array
                    The predicted values.
                train_data : Dataset
                    The training dataset.
                grad : list or numpy 1-D array
                    The value of the first order derivative (gradient) for each sample point.
                hess : list or numpy 1-D array
                    The value of the second order derivative (Hessian) for each sample point.
wxchan's avatar
wxchan committed
2451

2452
            For binary task, the preds is probability of positive class (or margin in case of specified ``fobj``).
2453
2454
            For multi-class task, the preds is group by class_id first, then group by row_id.
            If you want to get i-th row preds in j-th class, the access way is score[j * num_data + i]
2455
2456
            and you should group grad and hess in this way as well.

wxchan's avatar
wxchan committed
2457
2458
        Returns
        -------
2459
2460
        is_finished : bool
            Whether the update was successfully finished.
wxchan's avatar
wxchan committed
2461
        """
2462
        # need reset training data
2463
2464
2465
2466
2467
2468
        if train_set is None and self.train_set_version != self.train_set.version:
            train_set = self.train_set
            is_the_same_train_set = False
        else:
            is_the_same_train_set = train_set is self.train_set and self.train_set_version == train_set.version
        if train_set is not None and not is_the_same_train_set:
Guolin Ke's avatar
Guolin Ke committed
2469
            if not isinstance(train_set, Dataset):
2470
2471
                raise TypeError('Training data should be Dataset instance, met {}'
                                .format(type(train_set).__name__))
Guolin Ke's avatar
Guolin Ke committed
2472
            if train_set._predictor is not self.__init_predictor:
2473
2474
                raise LightGBMError("Replace training data failed, "
                                    "you should use same predictor for these data")
wxchan's avatar
wxchan committed
2475
2476
2477
            self.train_set = train_set
            _safe_call(_LIB.LGBM_BoosterResetTrainingData(
                self.handle,
wxchan's avatar
wxchan committed
2478
                self.train_set.construct().handle))
wxchan's avatar
wxchan committed
2479
            self.__inner_predict_buffer[0] = None
2480
            self.train_set_version = self.train_set.version
wxchan's avatar
wxchan committed
2481
2482
        is_finished = ctypes.c_int(0)
        if fobj is None:
2483
            if self.__set_objective_to_none:
2484
                raise LightGBMError('Cannot update due to null objective function.')
wxchan's avatar
wxchan committed
2485
2486
2487
            _safe_call(_LIB.LGBM_BoosterUpdateOneIter(
                self.handle,
                ctypes.byref(is_finished)))
2488
            self.__is_predicted_cur_iter = [False for _ in range(self.__num_dataset)]
wxchan's avatar
wxchan committed
2489
2490
            return is_finished.value == 1
        else:
2491
            if not self.__set_objective_to_none:
Nikita Titov's avatar
Nikita Titov committed
2492
                self.reset_parameter({"objective": "none"}).__set_objective_to_none = True
wxchan's avatar
wxchan committed
2493
2494
2495
2496
            grad, hess = fobj(self.__inner_predict(0), self.train_set)
            return self.__boost(grad, hess)

    def __boost(self, grad, hess):
2497
        """Boost Booster for one iteration with customized gradient statistics.
Nikita Titov's avatar
Nikita Titov committed
2498

Nikita Titov's avatar
Nikita Titov committed
2499
2500
        .. note::

2501
            For binary task, the score is probability of positive class (or margin in case of custom objective).
Nikita Titov's avatar
Nikita Titov committed
2502
2503
2504
            For multi-class task, the score is group by class_id first, then group by row_id.
            If you want to get i-th row score in j-th class, the access way is score[j * num_data + i]
            and you should group grad and hess in this way as well.
2505

wxchan's avatar
wxchan committed
2506
2507
        Parameters
        ----------
2508
        grad : list or numpy 1-D array
Nikita Titov's avatar
Nikita Titov committed
2509
            The first order derivative (gradient).
2510
        hess : list or numpy 1-D array
Nikita Titov's avatar
Nikita Titov committed
2511
            The second order derivative (Hessian).
wxchan's avatar
wxchan committed
2512
2513
2514

        Returns
        -------
Nikita Titov's avatar
Nikita Titov committed
2515
2516
        is_finished : bool
            Whether the boost was successfully finished.
wxchan's avatar
wxchan committed
2517
        """
2518
2519
        grad = list_to_1d_numpy(grad, name='gradient')
        hess = list_to_1d_numpy(hess, name='hessian')
2520
2521
        assert grad.flags.c_contiguous
        assert hess.flags.c_contiguous
wxchan's avatar
wxchan committed
2522
        if len(grad) != len(hess):
2523
2524
            raise ValueError("Lengths of gradient({}) and hessian({}) don't match"
                             .format(len(grad), len(hess)))
wxchan's avatar
wxchan committed
2525
2526
2527
2528
2529
2530
        is_finished = ctypes.c_int(0)
        _safe_call(_LIB.LGBM_BoosterUpdateOneIterCustom(
            self.handle,
            grad.ctypes.data_as(ctypes.POINTER(ctypes.c_float)),
            hess.ctypes.data_as(ctypes.POINTER(ctypes.c_float)),
            ctypes.byref(is_finished)))
2531
        self.__is_predicted_cur_iter = [False for _ in range(self.__num_dataset)]
wxchan's avatar
wxchan committed
2532
2533
2534
        return is_finished.value == 1

    def rollback_one_iter(self):
Nikita Titov's avatar
Nikita Titov committed
2535
2536
2537
2538
2539
2540
2541
        """Rollback one iteration.

        Returns
        -------
        self : Booster
            Booster with rolled back one iteration.
        """
wxchan's avatar
wxchan committed
2542
2543
        _safe_call(_LIB.LGBM_BoosterRollbackOneIter(
            self.handle))
2544
        self.__is_predicted_cur_iter = [False for _ in range(self.__num_dataset)]
Nikita Titov's avatar
Nikita Titov committed
2545
        return self
wxchan's avatar
wxchan committed
2546
2547

    def current_iteration(self):
2548
2549
2550
2551
2552
2553
2554
        """Get the index of the current iteration.

        Returns
        -------
        cur_iter : int
            The index of the current iteration.
        """
Guolin Ke's avatar
Guolin Ke committed
2555
        out_cur_iter = ctypes.c_int(0)
wxchan's avatar
wxchan committed
2556
2557
2558
2559
2560
        _safe_call(_LIB.LGBM_BoosterGetCurrentIteration(
            self.handle,
            ctypes.byref(out_cur_iter)))
        return out_cur_iter.value

2561
2562
2563
2564
2565
2566
2567
2568
2569
2570
2571
2572
2573
2574
2575
2576
2577
2578
2579
2580
2581
2582
2583
2584
2585
2586
2587
2588
    def num_model_per_iteration(self):
        """Get number of models per iteration.

        Returns
        -------
        model_per_iter : int
            The number of models per iteration.
        """
        model_per_iter = ctypes.c_int(0)
        _safe_call(_LIB.LGBM_BoosterNumModelPerIteration(
            self.handle,
            ctypes.byref(model_per_iter)))
        return model_per_iter.value

    def num_trees(self):
        """Get number of weak sub-models.

        Returns
        -------
        num_trees : int
            The number of weak sub-models.
        """
        num_trees = ctypes.c_int(0)
        _safe_call(_LIB.LGBM_BoosterNumberOfTotalModel(
            self.handle,
            ctypes.byref(num_trees)))
        return num_trees.value

2589
2590
2591
2592
2593
2594
2595
2596
2597
2598
2599
2600
2601
2602
2603
2604
2605
2606
2607
2608
2609
2610
2611
2612
2613
2614
2615
2616
    def upper_bound(self):
        """Get upper bound value of a model.

        Returns
        -------
        upper_bound : double
            Upper bound value of the model.
        """
        ret = ctypes.c_double(0)
        _safe_call(_LIB.LGBM_BoosterGetUpperBoundValue(
            self.handle,
            ctypes.byref(ret)))
        return ret.value

    def lower_bound(self):
        """Get lower bound value of a model.

        Returns
        -------
        lower_bound : double
            Lower bound value of the model.
        """
        ret = ctypes.c_double(0)
        _safe_call(_LIB.LGBM_BoosterGetLowerBoundValue(
            self.handle,
            ctypes.byref(ret)))
        return ret.value

wxchan's avatar
wxchan committed
2617
    def eval(self, data, name, feval=None):
2618
        """Evaluate for data.
wxchan's avatar
wxchan committed
2619
2620
2621

        Parameters
        ----------
2622
2623
2624
2625
2626
        data : Dataset
            Data for the evaluating.
        name : string
            Name of the data.
        feval : callable or None, optional (default=None)
2627
            Customized evaluation function.
2628
            Should accept two parameters: preds, eval_data,
2629
            and return (eval_name, eval_result, is_higher_better) or list of such tuples.
2630
2631
2632
2633
2634
2635

                preds : list or numpy 1-D array
                    The predicted values.
                eval_data : Dataset
                    The evaluation dataset.
                eval_name : string
2636
                    The name of evaluation function (without whitespaces).
2637
2638
2639
2640
2641
                eval_result : float
                    The eval result.
                is_higher_better : bool
                    Is eval result higher better, e.g. AUC is ``is_higher_better``.

2642
            For binary task, the preds is probability of positive class (or margin in case of specified ``fobj``).
2643
2644
            For multi-class task, the preds is group by class_id first, then group by row_id.
            If you want to get i-th row preds in j-th class, the access way is preds[j * num_data + i].
2645

wxchan's avatar
wxchan committed
2646
2647
        Returns
        -------
Nikita Titov's avatar
Nikita Titov committed
2648
        result : list
2649
            List with evaluation results.
wxchan's avatar
wxchan committed
2650
        """
Guolin Ke's avatar
Guolin Ke committed
2651
2652
        if not isinstance(data, Dataset):
            raise TypeError("Can only eval for Dataset instance")
wxchan's avatar
wxchan committed
2653
2654
2655
2656
        data_idx = -1
        if data is self.train_set:
            data_idx = 0
        else:
2657
            for i in range(len(self.valid_sets)):
wxchan's avatar
wxchan committed
2658
2659
2660
                if data is self.valid_sets[i]:
                    data_idx = i + 1
                    break
2661
        # need to push new valid data
wxchan's avatar
wxchan committed
2662
2663
2664
2665
2666
2667
2668
        if data_idx == -1:
            self.add_valid(data, name)
            data_idx = self.__num_dataset - 1

        return self.__inner_eval(name, data_idx, feval)

    def eval_train(self, feval=None):
2669
        """Evaluate for training data.
wxchan's avatar
wxchan committed
2670
2671
2672

        Parameters
        ----------
2673
        feval : callable or None, optional (default=None)
2674
            Customized evaluation function.
2675
2676
            Should accept two parameters: preds, train_data,
            and return (eval_name, eval_result, is_higher_better) or list of such tuples.
2677
2678
2679
2680
2681
2682

                preds : list or numpy 1-D array
                    The predicted values.
                train_data : Dataset
                    The training dataset.
                eval_name : string
2683
                    The name of evaluation function (without whitespaces).
2684
2685
2686
2687
2688
                eval_result : float
                    The eval result.
                is_higher_better : bool
                    Is eval result higher better, e.g. AUC is ``is_higher_better``.

2689
            For binary task, the preds is probability of positive class (or margin in case of specified ``fobj``).
2690
2691
            For multi-class task, the preds is group by class_id first, then group by row_id.
            If you want to get i-th row preds in j-th class, the access way is preds[j * num_data + i].
wxchan's avatar
wxchan committed
2692
2693
2694

        Returns
        -------
Nikita Titov's avatar
Nikita Titov committed
2695
        result : list
2696
            List with evaluation results.
wxchan's avatar
wxchan committed
2697
        """
2698
        return self.__inner_eval(self._train_data_name, 0, feval)
wxchan's avatar
wxchan committed
2699
2700

    def eval_valid(self, feval=None):
2701
        """Evaluate for validation data.
wxchan's avatar
wxchan committed
2702
2703
2704

        Parameters
        ----------
2705
        feval : callable or None, optional (default=None)
2706
            Customized evaluation function.
2707
            Should accept two parameters: preds, valid_data,
2708
            and return (eval_name, eval_result, is_higher_better) or list of such tuples.
2709
2710
2711
2712
2713
2714

                preds : list or numpy 1-D array
                    The predicted values.
                valid_data : Dataset
                    The validation dataset.
                eval_name : string
2715
                    The name of evaluation function (without whitespaces).
2716
2717
2718
2719
2720
                eval_result : float
                    The eval result.
                is_higher_better : bool
                    Is eval result higher better, e.g. AUC is ``is_higher_better``.

2721
            For binary task, the preds is probability of positive class (or margin in case of specified ``fobj``).
2722
2723
            For multi-class task, the preds is group by class_id first, then group by row_id.
            If you want to get i-th row preds in j-th class, the access way is preds[j * num_data + i].
wxchan's avatar
wxchan committed
2724
2725
2726

        Returns
        -------
Nikita Titov's avatar
Nikita Titov committed
2727
        result : list
2728
            List with evaluation results.
wxchan's avatar
wxchan committed
2729
        """
2730
        return [item for i in range(1, self.__num_dataset)
wxchan's avatar
wxchan committed
2731
                for item in self.__inner_eval(self.name_valid_sets[i - 1], i, feval)]
wxchan's avatar
wxchan committed
2732

2733
    def save_model(self, filename, num_iteration=None, start_iteration=0, importance_type='split'):
2734
        """Save Booster to file.
wxchan's avatar
wxchan committed
2735
2736
2737

        Parameters
        ----------
2738
2739
        filename : string
            Filename to save Booster.
2740
2741
2742
2743
        num_iteration : int or None, optional (default=None)
            Index of the iteration that should be saved.
            If None, if the best iteration exists, it is saved; otherwise, all iterations are saved.
            If <= 0, all iterations are saved.
Nikita Titov's avatar
Nikita Titov committed
2744
        start_iteration : int, optional (default=0)
2745
            Start index of the iteration that should be saved.
2746
2747
2748
2749
        importance_type : string, optional (default="split")
            What type of feature importance should be saved.
            If "split", result contains numbers of times the feature is used in a model.
            If "gain", result contains total gains of splits which use the feature.
Nikita Titov's avatar
Nikita Titov committed
2750
2751
2752
2753
2754

        Returns
        -------
        self : Booster
            Returns self.
wxchan's avatar
wxchan committed
2755
        """
2756
        if num_iteration is None:
2757
            num_iteration = self.best_iteration
2758
        importance_type_int = FEATURE_IMPORTANCE_TYPE_MAPPER[importance_type]
wxchan's avatar
wxchan committed
2759
2760
        _safe_call(_LIB.LGBM_BoosterSaveModel(
            self.handle,
2761
            ctypes.c_int(start_iteration),
Guolin Ke's avatar
Guolin Ke committed
2762
            ctypes.c_int(num_iteration),
2763
            ctypes.c_int(importance_type_int),
wxchan's avatar
wxchan committed
2764
            c_str(filename)))
2765
        _dump_pandas_categorical(self.pandas_categorical, filename)
Nikita Titov's avatar
Nikita Titov committed
2766
        return self
wxchan's avatar
wxchan committed
2767

2768
    def shuffle_models(self, start_iteration=0, end_iteration=-1):
2769
        """Shuffle models.
Nikita Titov's avatar
Nikita Titov committed
2770

2771
2772
2773
        Parameters
        ----------
        start_iteration : int, optional (default=0)
2774
            The first iteration that will be shuffled.
2775
2776
        end_iteration : int, optional (default=-1)
            The last iteration that will be shuffled.
2777
            If <= 0, means the last available iteration.
2778

Nikita Titov's avatar
Nikita Titov committed
2779
2780
2781
2782
        Returns
        -------
        self : Booster
            Booster with shuffled models.
2783
        """
2784
2785
        _safe_call(_LIB.LGBM_BoosterShuffleModels(
            self.handle,
Guolin Ke's avatar
Guolin Ke committed
2786
2787
            ctypes.c_int(start_iteration),
            ctypes.c_int(end_iteration)))
Nikita Titov's avatar
Nikita Titov committed
2788
        return self
2789
2790
2791
2792
2793
2794

    def model_from_string(self, model_str, verbose=True):
        """Load Booster from a string.

        Parameters
        ----------
Nikita Titov's avatar
Nikita Titov committed
2795
        model_str : string
2796
            Model will be loaded from this string.
Nikita Titov's avatar
Nikita Titov committed
2797
2798
        verbose : bool, optional (default=True)
            Whether to print messages while loading model.
2799
2800
2801

        Returns
        -------
Nikita Titov's avatar
Nikita Titov committed
2802
        self : Booster
2803
2804
            Loaded Booster object.
        """
2805
2806
2807
2808
        if self.handle is not None:
            _safe_call(_LIB.LGBM_BoosterFree(self.handle))
        self._free_buffer()
        self.handle = ctypes.c_void_p()
2809
2810
2811
2812
2813
2814
2815
2816
2817
        out_num_iterations = ctypes.c_int(0)
        _safe_call(_LIB.LGBM_BoosterLoadModelFromString(
            c_str(model_str),
            ctypes.byref(out_num_iterations),
            ctypes.byref(self.handle)))
        out_num_class = ctypes.c_int(0)
        _safe_call(_LIB.LGBM_BoosterGetNumClasses(
            self.handle,
            ctypes.byref(out_num_class)))
2818
        if verbose:
Nikita Titov's avatar
Nikita Titov committed
2819
            print('Finished loading model, total used %d iterations' % int(out_num_iterations.value))
2820
        self.__num_class = out_num_class.value
2821
        self.pandas_categorical = _load_pandas_categorical(model_str=model_str)
2822
2823
        return self

2824
    def model_to_string(self, num_iteration=None, start_iteration=0, importance_type='split'):
2825
        """Save Booster to string.
2826

2827
2828
2829
2830
2831
2832
        Parameters
        ----------
        num_iteration : int or None, optional (default=None)
            Index of the iteration that should be saved.
            If None, if the best iteration exists, it is saved; otherwise, all iterations are saved.
            If <= 0, all iterations are saved.
Nikita Titov's avatar
Nikita Titov committed
2833
        start_iteration : int, optional (default=0)
2834
            Start index of the iteration that should be saved.
2835
2836
2837
2838
        importance_type : string, optional (default="split")
            What type of feature importance should be saved.
            If "split", result contains numbers of times the feature is used in a model.
            If "gain", result contains total gains of splits which use the feature.
2839
2840
2841

        Returns
        -------
Nikita Titov's avatar
Nikita Titov committed
2842
        str_repr : string
2843
2844
            String representation of Booster.
        """
2845
        if num_iteration is None:
2846
            num_iteration = self.best_iteration
2847
        importance_type_int = FEATURE_IMPORTANCE_TYPE_MAPPER[importance_type]
2848
        buffer_len = 1 << 20
2849
        tmp_out_len = ctypes.c_int64(0)
2850
2851
2852
2853
        string_buffer = ctypes.create_string_buffer(buffer_len)
        ptr_string_buffer = ctypes.c_char_p(*[ctypes.addressof(string_buffer)])
        _safe_call(_LIB.LGBM_BoosterSaveModelToString(
            self.handle,
2854
            ctypes.c_int(start_iteration),
2855
            ctypes.c_int(num_iteration),
2856
            ctypes.c_int(importance_type_int),
2857
            ctypes.c_int64(buffer_len),
2858
2859
2860
            ctypes.byref(tmp_out_len),
            ptr_string_buffer))
        actual_len = tmp_out_len.value
2861
        # if buffer length is not long enough, re-allocate a buffer
2862
2863
2864
2865
2866
        if actual_len > buffer_len:
            string_buffer = ctypes.create_string_buffer(actual_len)
            ptr_string_buffer = ctypes.c_char_p(*[ctypes.addressof(string_buffer)])
            _safe_call(_LIB.LGBM_BoosterSaveModelToString(
                self.handle,
2867
                ctypes.c_int(start_iteration),
2868
                ctypes.c_int(num_iteration),
2869
                ctypes.c_int(importance_type_int),
2870
                ctypes.c_int64(actual_len),
2871
2872
                ctypes.byref(tmp_out_len),
                ptr_string_buffer))
2873
        ret = string_buffer.value.decode('utf-8')
2874
2875
        ret += _dump_pandas_categorical(self.pandas_categorical)
        return ret
2876

2877
    def dump_model(self, num_iteration=None, start_iteration=0, importance_type='split'):
Nikita Titov's avatar
Nikita Titov committed
2878
        """Dump Booster to JSON format.
wxchan's avatar
wxchan committed
2879

2880
2881
        Parameters
        ----------
2882
2883
2884
2885
        num_iteration : int or None, optional (default=None)
            Index of the iteration that should be dumped.
            If None, if the best iteration exists, it is dumped; otherwise, all iterations are dumped.
            If <= 0, all iterations are dumped.
Nikita Titov's avatar
Nikita Titov committed
2886
        start_iteration : int, optional (default=0)
2887
            Start index of the iteration that should be dumped.
2888
2889
2890
2891
        importance_type : string, optional (default="split")
            What type of feature importance should be dumped.
            If "split", result contains numbers of times the feature is used in a model.
            If "gain", result contains total gains of splits which use the feature.
2892

wxchan's avatar
wxchan committed
2893
2894
        Returns
        -------
2895
        json_repr : dict
Nikita Titov's avatar
Nikita Titov committed
2896
            JSON format of Booster.
wxchan's avatar
wxchan committed
2897
        """
2898
        if num_iteration is None:
2899
            num_iteration = self.best_iteration
2900
        importance_type_int = FEATURE_IMPORTANCE_TYPE_MAPPER[importance_type]
wxchan's avatar
wxchan committed
2901
        buffer_len = 1 << 20
2902
        tmp_out_len = ctypes.c_int64(0)
wxchan's avatar
wxchan committed
2903
2904
2905
2906
        string_buffer = ctypes.create_string_buffer(buffer_len)
        ptr_string_buffer = ctypes.c_char_p(*[ctypes.addressof(string_buffer)])
        _safe_call(_LIB.LGBM_BoosterDumpModel(
            self.handle,
2907
            ctypes.c_int(start_iteration),
Guolin Ke's avatar
Guolin Ke committed
2908
            ctypes.c_int(num_iteration),
2909
            ctypes.c_int(importance_type_int),
2910
            ctypes.c_int64(buffer_len),
wxchan's avatar
wxchan committed
2911
            ctypes.byref(tmp_out_len),
Guolin Ke's avatar
Guolin Ke committed
2912
            ptr_string_buffer))
wxchan's avatar
wxchan committed
2913
        actual_len = tmp_out_len.value
2914
        # if buffer length is not long enough, reallocate a buffer
wxchan's avatar
wxchan committed
2915
2916
2917
2918
2919
        if actual_len > buffer_len:
            string_buffer = ctypes.create_string_buffer(actual_len)
            ptr_string_buffer = ctypes.c_char_p(*[ctypes.addressof(string_buffer)])
            _safe_call(_LIB.LGBM_BoosterDumpModel(
                self.handle,
2920
                ctypes.c_int(start_iteration),
Guolin Ke's avatar
Guolin Ke committed
2921
                ctypes.c_int(num_iteration),
2922
                ctypes.c_int(importance_type_int),
2923
                ctypes.c_int64(actual_len),
wxchan's avatar
wxchan committed
2924
                ctypes.byref(tmp_out_len),
Guolin Ke's avatar
Guolin Ke committed
2925
                ptr_string_buffer))
2926
        ret = json.loads(string_buffer.value.decode('utf-8'))
2927
2928
2929
        ret['pandas_categorical'] = json.loads(json.dumps(self.pandas_categorical,
                                                          default=json_default_with_numpy))
        return ret
wxchan's avatar
wxchan committed
2930

2931
    def predict(self, data, start_iteration=0, num_iteration=None,
2932
                raw_score=False, pred_leaf=False, pred_contrib=False,
2933
                data_has_header=False, is_reshape=True, **kwargs):
2934
        """Make a prediction.
wxchan's avatar
wxchan committed
2935
2936
2937

        Parameters
        ----------
2938
        data : string, numpy array, pandas DataFrame, H2O DataTable's Frame or scipy.sparse
2939
2940
            Data source for prediction.
            If string, it represents the path to txt file.
2941
        start_iteration : int, optional (default=0)
2942
            Start index of the iteration to predict.
2943
            If <= 0, starts from the first iteration.
2944
        num_iteration : int or None, optional (default=None)
2945
2946
2947
2948
            Total number of iterations used in the prediction.
            If None, if the best iteration exists and start_iteration <= 0, the best iteration is used;
            otherwise, all iterations from ``start_iteration`` are used (no limits).
            If <= 0, all iterations from ``start_iteration`` are used (no limits).
2949
2950
2951
2952
        raw_score : bool, optional (default=False)
            Whether to predict raw scores.
        pred_leaf : bool, optional (default=False)
            Whether to predict leaf index.
2953
2954
        pred_contrib : bool, optional (default=False)
            Whether to predict feature contributions.
2955

Nikita Titov's avatar
Nikita Titov committed
2956
2957
2958
2959
2960
2961
2962
            .. note::

                If you want to get more explanations for your model's predictions using SHAP values,
                like SHAP interaction values,
                you can install the shap package (https://github.com/slundberg/shap).
                Note that unlike the shap package, with ``pred_contrib`` we return a matrix with an extra
                column, where the last column is the expected value.
2963

2964
2965
2966
2967
2968
        data_has_header : bool, optional (default=False)
            Whether the data has header.
            Used only if data is string.
        is_reshape : bool, optional (default=True)
            If True, result is reshaped to [nrow, ncol].
2969
2970
        **kwargs
            Other parameters for the prediction.
wxchan's avatar
wxchan committed
2971
2972
2973

        Returns
        -------
2974
        result : numpy array, scipy.sparse or list of scipy.sparse
2975
            Prediction result.
2976
            Can be sparse or a list of sparse objects (each element represents predictions for one class) for feature contributions (when ``pred_contrib=True``).
wxchan's avatar
wxchan committed
2977
        """
2978
        predictor = self._to_predictor(copy.deepcopy(kwargs))
2979
        if num_iteration is None:
2980
            if start_iteration <= 0:
2981
2982
2983
2984
                num_iteration = self.best_iteration
            else:
                num_iteration = -1
        return predictor.predict(data, start_iteration, num_iteration,
2985
2986
                                 raw_score, pred_leaf, pred_contrib,
                                 data_has_header, is_reshape)
wxchan's avatar
wxchan committed
2987

2988
    def refit(self, data, label, decay_rate=0.9, **kwargs):
Guolin Ke's avatar
Guolin Ke committed
2989
2990
2991
2992
        """Refit the existing Booster by new data.

        Parameters
        ----------
2993
        data : string, numpy array, pandas DataFrame, H2O DataTable's Frame or scipy.sparse
Guolin Ke's avatar
Guolin Ke committed
2994
2995
            Data source for refit.
            If string, it represents the path to txt file.
2996
        label : list, numpy 1-D array or pandas Series / one-column DataFrame
Guolin Ke's avatar
Guolin Ke committed
2997
2998
            Label for refit.
        decay_rate : float, optional (default=0.9)
2999
3000
            Decay rate of refit,
            will use ``leaf_output = decay_rate * old_leaf_output + (1.0 - decay_rate) * new_leaf_output`` to refit trees.
3001
3002
        **kwargs
            Other parameters for refit.
3003
            These parameters will be passed to ``predict`` method.
Guolin Ke's avatar
Guolin Ke committed
3004
3005
3006
3007
3008
3009

        Returns
        -------
        result : Booster
            Refitted Booster.
        """
3010
3011
        if self.__set_objective_to_none:
            raise LightGBMError('Cannot refit due to null objective function.')
3012
        predictor = self._to_predictor(copy.deepcopy(kwargs))
3013
        leaf_preds = predictor.predict(data, -1, pred_leaf=True)
3014
        nrow, ncol = leaf_preds.shape
3015
3016
3017
3018
        out_is_linear = ctypes.c_bool(False)
        _safe_call(_LIB.LGBM_BoosterGetLinear(
            self.handle,
            ctypes.byref(out_is_linear)))
3019
        new_params = copy.deepcopy(self.params)
3020
3021
        new_params["linear_tree"] = out_is_linear.value
        train_set = Dataset(data, label, silent=True, params=new_params)
3022
        new_params['refit_decay_rate'] = decay_rate
3023
        new_booster = Booster(new_params, train_set)
Guolin Ke's avatar
Guolin Ke committed
3024
3025
3026
3027
3028
        # Copy models
        _safe_call(_LIB.LGBM_BoosterMerge(
            new_booster.handle,
            predictor.handle))
        leaf_preds = leaf_preds.reshape(-1)
3029
        ptr_data, _, _ = c_int_array(leaf_preds)
Guolin Ke's avatar
Guolin Ke committed
3030
3031
3032
3033
3034
        _safe_call(_LIB.LGBM_BoosterRefit(
            new_booster.handle,
            ptr_data,
            ctypes.c_int(nrow),
            ctypes.c_int(ncol)))
3035
3036
        new_booster.network = self.network
        new_booster.__attr = self.__attr.copy()
Guolin Ke's avatar
Guolin Ke committed
3037
3038
        return new_booster

3039
    def get_leaf_output(self, tree_id, leaf_id):
3040
3041
3042
3043
3044
3045
3046
3047
3048
3049
3050
3051
3052
3053
        """Get the output of a leaf.

        Parameters
        ----------
        tree_id : int
            The index of the tree.
        leaf_id : int
            The index of the leaf in the tree.

        Returns
        -------
        result : float
            The output of the leaf.
        """
3054
3055
3056
3057
3058
3059
3060
3061
        ret = ctypes.c_double(0)
        _safe_call(_LIB.LGBM_BoosterGetLeafValue(
            self.handle,
            ctypes.c_int(tree_id),
            ctypes.c_int(leaf_id),
            ctypes.byref(ret)))
        return ret.value

3062
    def _to_predictor(self, pred_parameter=None):
3063
        """Convert to predictor."""
3064
        predictor = _InnerPredictor(booster_handle=self.handle, pred_parameter=pred_parameter)
3065
        predictor.pandas_categorical = self.pandas_categorical
wxchan's avatar
wxchan committed
3066
3067
        return predictor

3068
    def num_feature(self):
3069
3070
3071
3072
3073
3074
3075
        """Get number of features.

        Returns
        -------
        num_feature : int
            The number of features.
        """
3076
3077
3078
3079
3080
3081
        out_num_feature = ctypes.c_int(0)
        _safe_call(_LIB.LGBM_BoosterGetNumFeature(
            self.handle,
            ctypes.byref(out_num_feature)))
        return out_num_feature.value

wxchan's avatar
wxchan committed
3082
    def feature_name(self):
3083
        """Get names of features.
wxchan's avatar
wxchan committed
3084
3085
3086

        Returns
        -------
3087
3088
        result : list
            List with names of features.
wxchan's avatar
wxchan committed
3089
        """
3090
        num_feature = self.num_feature()
3091
        # Get name of features
wxchan's avatar
wxchan committed
3092
        tmp_out_len = ctypes.c_int(0)
3093
3094
        reserved_string_buffer_size = 255
        required_string_buffer_size = ctypes.c_size_t(0)
3095
        string_buffers = [ctypes.create_string_buffer(reserved_string_buffer_size) for i in range(num_feature)]
wxchan's avatar
wxchan committed
3096
3097
3098
        ptr_string_buffers = (ctypes.c_char_p * num_feature)(*map(ctypes.addressof, string_buffers))
        _safe_call(_LIB.LGBM_BoosterGetFeatureNames(
            self.handle,
3099
            ctypes.c_int(num_feature),
wxchan's avatar
wxchan committed
3100
            ctypes.byref(tmp_out_len),
3101
            ctypes.c_size_t(reserved_string_buffer_size),
3102
            ctypes.byref(required_string_buffer_size),
wxchan's avatar
wxchan committed
3103
3104
3105
            ptr_string_buffers))
        if num_feature != tmp_out_len.value:
            raise ValueError("Length of feature names doesn't equal with num_feature")
3106
3107
3108
3109
3110
        if reserved_string_buffer_size < required_string_buffer_size.value:
            raise BufferError(
                "Allocated feature name buffer size ({}) was inferior to the needed size ({})."
                .format(reserved_string_buffer_size, required_string_buffer_size.value)
            )
3111
        return [string_buffers[i].value.decode('utf-8') for i in range(num_feature)]
wxchan's avatar
wxchan committed
3112

3113
    def feature_importance(self, importance_type='split', iteration=None):
3114
        """Get feature importances.
3115

3116
3117
        Parameters
        ----------
3118
3119
3120
3121
        importance_type : string, optional (default="split")
            How the importance is calculated.
            If "split", result contains numbers of times the feature is used in a model.
            If "gain", result contains total gains of splits which use the feature.
3122
3123
3124
3125
        iteration : int or None, optional (default=None)
            Limit number of iterations in the feature importance calculation.
            If None, if the best iteration exists, it is used; otherwise, all trees are used.
            If <= 0, all trees are used (no limits).
3126

3127
3128
        Returns
        -------
3129
3130
        result : numpy array
            Array with feature importances.
3131
        """
3132
3133
        if iteration is None:
            iteration = self.best_iteration
3134
        importance_type_int = FEATURE_IMPORTANCE_TYPE_MAPPER[importance_type]
Nikita Titov's avatar
Nikita Titov committed
3135
        result = np.zeros(self.num_feature(), dtype=np.float64)
3136
3137
3138
3139
3140
3141
        _safe_call(_LIB.LGBM_BoosterFeatureImportance(
            self.handle,
            ctypes.c_int(iteration),
            ctypes.c_int(importance_type_int),
            result.ctypes.data_as(ctypes.POINTER(ctypes.c_double))))
        if importance_type_int == 0:
3142
            return result.astype(np.int32)
3143
3144
        else:
            return result
3145

3146
3147
3148
3149
3150
3151
3152
3153
3154
    def get_split_value_histogram(self, feature, bins=None, xgboost_style=False):
        """Get split value histogram for the specified feature.

        Parameters
        ----------
        feature : int or string
            The feature name or index the histogram is calculated for.
            If int, interpreted as index.
            If string, interpreted as name.
3155

Nikita Titov's avatar
Nikita Titov committed
3156
3157
3158
            .. warning::

                Categorical features are not supported.
3159

3160
3161
3162
3163
3164
3165
3166
3167
3168
3169
3170
3171
3172
3173
3174
3175
3176
3177
3178
3179
3180
3181
        bins : int, string or None, optional (default=None)
            The maximum number of bins.
            If None, or int and > number of unique split values and ``xgboost_style=True``,
            the number of bins equals number of unique split values.
            If string, it should be one from the list of the supported values by ``numpy.histogram()`` function.
        xgboost_style : bool, optional (default=False)
            Whether the returned result should be in the same form as it is in XGBoost.
            If False, the returned value is tuple of 2 numpy arrays as it is in ``numpy.histogram()`` function.
            If True, the returned value is matrix, in which the first column is the right edges of non-empty bins
            and the second one is the histogram values.

        Returns
        -------
        result_tuple : tuple of 2 numpy arrays
            If ``xgboost_style=False``, the values of the histogram of used splitting values for the specified feature
            and the bin edges.
        result_array_like : numpy array or pandas DataFrame (if pandas is installed)
            If ``xgboost_style=True``, the histogram of used splitting values for the specified feature.
        """
        def add(root):
            """Recursively add thresholds."""
            if 'split_index' in root:  # non-leaf
3182
                if feature_names is not None and isinstance(feature, str):
3183
3184
3185
3186
                    split_feature = feature_names[root['split_feature']]
                else:
                    split_feature = root['split_feature']
                if split_feature == feature:
3187
                    if isinstance(root['threshold'], str):
3188
3189
3190
                        raise LightGBMError('Cannot compute split value histogram for the categorical feature')
                    else:
                        values.append(root['threshold'])
3191
3192
3193
3194
3195
3196
3197
3198
3199
3200
                add(root['left_child'])
                add(root['right_child'])

        model = self.dump_model()
        feature_names = model.get('feature_names')
        tree_infos = model['tree_info']
        values = []
        for tree_info in tree_infos:
            add(tree_info['tree_structure'])

3201
        if bins is None or isinstance(bins, int) and xgboost_style:
3202
3203
3204
3205
3206
3207
3208
3209
3210
3211
3212
3213
3214
            n_unique = len(np.unique(values))
            bins = max(min(n_unique, bins) if bins is not None else n_unique, 1)
        hist, bin_edges = np.histogram(values, bins=bins)
        if xgboost_style:
            ret = np.column_stack((bin_edges[1:], hist))
            ret = ret[ret[:, 1] > 0]
            if PANDAS_INSTALLED:
                return DataFrame(ret, columns=['SplitValue', 'Count'])
            else:
                return ret
        else:
            return hist, bin_edges

wxchan's avatar
wxchan committed
3215
    def __inner_eval(self, data_name, data_idx, feval=None):
3216
        """Evaluate training or validation data."""
wxchan's avatar
wxchan committed
3217
        if data_idx >= self.__num_dataset:
3218
            raise ValueError("Data_idx should be smaller than number of dataset")
wxchan's avatar
wxchan committed
3219
3220
3221
        self.__get_eval_info()
        ret = []
        if self.__num_inner_eval > 0:
3222
            result = np.zeros(self.__num_inner_eval, dtype=np.float64)
Guolin Ke's avatar
Guolin Ke committed
3223
            tmp_out_len = ctypes.c_int(0)
wxchan's avatar
wxchan committed
3224
3225
            _safe_call(_LIB.LGBM_BoosterGetEval(
                self.handle,
Guolin Ke's avatar
Guolin Ke committed
3226
                ctypes.c_int(data_idx),
wxchan's avatar
wxchan committed
3227
                ctypes.byref(tmp_out_len),
Guolin Ke's avatar
Guolin Ke committed
3228
                result.ctypes.data_as(ctypes.POINTER(ctypes.c_double))))
wxchan's avatar
wxchan committed
3229
            if tmp_out_len.value != self.__num_inner_eval:
3230
                raise ValueError("Wrong length of eval results")
3231
            for i in range(self.__num_inner_eval):
3232
3233
                ret.append((data_name, self.__name_inner_eval[i],
                            result[i], self.__higher_better_inner_eval[i]))
3234
3235
        if callable(feval):
            feval = [feval]
wxchan's avatar
wxchan committed
3236
3237
3238
3239
3240
        if feval is not None:
            if data_idx == 0:
                cur_data = self.train_set
            else:
                cur_data = self.valid_sets[data_idx - 1]
3241
3242
3243
3244
3245
3246
3247
3248
3249
            for eval_function in feval:
                if eval_function is None:
                    continue
                feval_ret = eval_function(self.__inner_predict(data_idx), cur_data)
                if isinstance(feval_ret, list):
                    for eval_name, val, is_higher_better in feval_ret:
                        ret.append((data_name, eval_name, val, is_higher_better))
                else:
                    eval_name, val, is_higher_better = feval_ret
wxchan's avatar
wxchan committed
3250
3251
3252
3253
                    ret.append((data_name, eval_name, val, is_higher_better))
        return ret

    def __inner_predict(self, data_idx):
3254
        """Predict for training and validation dataset."""
wxchan's avatar
wxchan committed
3255
        if data_idx >= self.__num_dataset:
3256
            raise ValueError("Data_idx should be smaller than number of dataset")
wxchan's avatar
wxchan committed
3257
3258
3259
3260
3261
        if self.__inner_predict_buffer[data_idx] is None:
            if data_idx == 0:
                n_preds = self.train_set.num_data() * self.__num_class
            else:
                n_preds = self.valid_sets[data_idx - 1].num_data() * self.__num_class
3262
            self.__inner_predict_buffer[data_idx] = np.zeros(n_preds, dtype=np.float64)
3263
        # avoid to predict many time in one iteration
wxchan's avatar
wxchan committed
3264
3265
        if not self.__is_predicted_cur_iter[data_idx]:
            tmp_out_len = ctypes.c_int64(0)
Guolin Ke's avatar
Guolin Ke committed
3266
            data_ptr = self.__inner_predict_buffer[data_idx].ctypes.data_as(ctypes.POINTER(ctypes.c_double))
wxchan's avatar
wxchan committed
3267
3268
            _safe_call(_LIB.LGBM_BoosterGetPredict(
                self.handle,
Guolin Ke's avatar
Guolin Ke committed
3269
                ctypes.c_int(data_idx),
wxchan's avatar
wxchan committed
3270
3271
3272
                ctypes.byref(tmp_out_len),
                data_ptr))
            if tmp_out_len.value != len(self.__inner_predict_buffer[data_idx]):
3273
                raise ValueError("Wrong length of predict results for data %d" % (data_idx))
wxchan's avatar
wxchan committed
3274
3275
3276
3277
            self.__is_predicted_cur_iter[data_idx] = True
        return self.__inner_predict_buffer[data_idx]

    def __get_eval_info(self):
3278
        """Get inner evaluation count and names."""
wxchan's avatar
wxchan committed
3279
3280
        if self.__need_reload_eval_info:
            self.__need_reload_eval_info = False
Guolin Ke's avatar
Guolin Ke committed
3281
            out_num_eval = ctypes.c_int(0)
3282
            # Get num of inner evals
wxchan's avatar
wxchan committed
3283
3284
3285
3286
3287
            _safe_call(_LIB.LGBM_BoosterGetEvalCounts(
                self.handle,
                ctypes.byref(out_num_eval)))
            self.__num_inner_eval = out_num_eval.value
            if self.__num_inner_eval > 0:
3288
                # Get name of evals
Guolin Ke's avatar
Guolin Ke committed
3289
                tmp_out_len = ctypes.c_int(0)
3290
3291
3292
                reserved_string_buffer_size = 255
                required_string_buffer_size = ctypes.c_size_t(0)
                string_buffers = [
3293
                    ctypes.create_string_buffer(reserved_string_buffer_size) for i in range(self.__num_inner_eval)
3294
                ]
wxchan's avatar
wxchan committed
3295
                ptr_string_buffers = (ctypes.c_char_p * self.__num_inner_eval)(*map(ctypes.addressof, string_buffers))
wxchan's avatar
wxchan committed
3296
3297
                _safe_call(_LIB.LGBM_BoosterGetEvalNames(
                    self.handle,
3298
                    ctypes.c_int(self.__num_inner_eval),
wxchan's avatar
wxchan committed
3299
                    ctypes.byref(tmp_out_len),
3300
                    ctypes.c_size_t(reserved_string_buffer_size),
3301
                    ctypes.byref(required_string_buffer_size),
wxchan's avatar
wxchan committed
3302
3303
                    ptr_string_buffers))
                if self.__num_inner_eval != tmp_out_len.value:
3304
                    raise ValueError("Length of eval names doesn't equal with num_evals")
3305
3306
3307
3308
3309
                if reserved_string_buffer_size < required_string_buffer_size.value:
                    raise BufferError(
                        "Allocated eval name buffer size ({}) was inferior to the needed size ({})."
                        .format(reserved_string_buffer_size, required_string_buffer_size.value)
                    )
3310
                self.__name_inner_eval = \
3311
                    [string_buffers[i].value.decode('utf-8') for i in range(self.__num_inner_eval)]
3312
                self.__higher_better_inner_eval = \
3313
                    [name.startswith(('auc', 'ndcg@', 'map@', 'average_precision')) for name in self.__name_inner_eval]
3314

wxchan's avatar
wxchan committed
3315
    def attr(self, key):
3316
        """Get attribute string from the Booster.
wxchan's avatar
wxchan committed
3317
3318
3319

        Parameters
        ----------
3320
3321
        key : string
            The name of the attribute.
wxchan's avatar
wxchan committed
3322
3323
3324

        Returns
        -------
3325
3326
        value : string or None
            The attribute value.
Nikita Titov's avatar
Nikita Titov committed
3327
            Returns None if attribute does not exist.
wxchan's avatar
wxchan committed
3328
        """
3329
        return self.__attr.get(key, None)
wxchan's avatar
wxchan committed
3330
3331

    def set_attr(self, **kwargs):
3332
        """Set attributes to the Booster.
wxchan's avatar
wxchan committed
3333
3334
3335
3336

        Parameters
        ----------
        **kwargs
3337
3338
            The attributes to set.
            Setting a value to None deletes an attribute.
Nikita Titov's avatar
Nikita Titov committed
3339
3340
3341
3342

        Returns
        -------
        self : Booster
3343
            Booster with set attributes.
wxchan's avatar
wxchan committed
3344
3345
3346
        """
        for key, value in kwargs.items():
            if value is not None:
3347
                if not isinstance(value, str):
Nikita Titov's avatar
Nikita Titov committed
3348
                    raise ValueError("Only string values are accepted")
wxchan's avatar
wxchan committed
3349
3350
3351
                self.__attr[key] = value
            else:
                self.__attr.pop(key, None)
Nikita Titov's avatar
Nikita Titov committed
3352
        return self