basic.py 131 KB
Newer Older
wxchan's avatar
wxchan committed
1
# coding: utf-8
2
"""Wrapper for C API of LightGBM."""
3
from __future__ import absolute_import, print_function
wxchan's avatar
wxchan committed
4

5
import copy
wxchan's avatar
wxchan committed
6
import ctypes
7
import os
wxchan's avatar
wxchan committed
8
import warnings
wxchan's avatar
wxchan committed
9
from tempfile import NamedTemporaryFile
10
from collections import OrderedDict
wxchan's avatar
wxchan committed
11
12
13
14

import numpy as np
import scipy.sparse

15
from .compat import (PANDAS_INSTALLED, DataFrame, Series, is_dtype_sparse,
16
                     DataTable,
17
18
                     decode_string, string_type,
                     integer_types, numeric_types,
19
                     json, json_default_with_numpy,
20
                     range_, zip_)
wxchan's avatar
wxchan committed
21
22
from .libpath import find_lib_path

wxchan's avatar
wxchan committed
23

24
25
26
27
28
def _log_callback(msg):
    """Redirect logs from native library into Python console."""
    print("{0:s}".format(decode_string(msg)), end='')


wxchan's avatar
wxchan committed
29
def _load_lib():
30
    """Load LightGBM library."""
wxchan's avatar
wxchan committed
31
32
    lib_path = find_lib_path()
    if len(lib_path) == 0:
33
        return None
wxchan's avatar
wxchan committed
34
35
    lib = ctypes.cdll.LoadLibrary(lib_path[0])
    lib.LGBM_GetLastError.restype = ctypes.c_char_p
36
37
38
39
    callback = ctypes.CFUNCTYPE(None, ctypes.c_char_p)
    lib.callback = callback(_log_callback)
    if lib.LGBM_RegisterLogCallback(lib.callback) != 0:
        raise LightGBMError(decode_string(lib.LGBM_GetLastError()))
wxchan's avatar
wxchan committed
40
41
    return lib

wxchan's avatar
wxchan committed
42

wxchan's avatar
wxchan committed
43
44
_LIB = _load_lib()

wxchan's avatar
wxchan committed
45

wxchan's avatar
wxchan committed
46
def _safe_call(ret):
47
48
    """Check the return value from C API call.

wxchan's avatar
wxchan committed
49
50
51
    Parameters
    ----------
    ret : int
52
        The return value from C API calls.
wxchan's avatar
wxchan committed
53
54
    """
    if ret != 0:
55
        raise LightGBMError(decode_string(_LIB.LGBM_GetLastError()))
wxchan's avatar
wxchan committed
56

wxchan's avatar
wxchan committed
57

wxchan's avatar
wxchan committed
58
def is_numeric(obj):
59
    """Check whether object is a number or not, include numpy number, etc."""
wxchan's avatar
wxchan committed
60
61
62
    try:
        float(obj)
        return True
wxchan's avatar
wxchan committed
63
64
65
    except (TypeError, ValueError):
        # TypeError: obj is not a string or a number
        # ValueError: invalid literal
wxchan's avatar
wxchan committed
66
67
        return False

wxchan's avatar
wxchan committed
68

wxchan's avatar
wxchan committed
69
def is_numpy_1d_array(data):
70
    """Check whether data is a numpy 1-D array."""
71
    return isinstance(data, np.ndarray) and len(data.shape) == 1
wxchan's avatar
wxchan committed
72

wxchan's avatar
wxchan committed
73

wxchan's avatar
wxchan committed
74
def is_1d_list(data):
75
76
    """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
77

wxchan's avatar
wxchan committed
78

79
def list_to_1d_numpy(data, dtype=np.float32, name='list'):
80
    """Convert data to numpy 1-D array."""
wxchan's avatar
wxchan committed
81
82
83
84
85
86
87
    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)
88
    elif isinstance(data, Series):
89
90
        if _get_bad_pandas_dtypes([data.dtypes]):
            raise ValueError('Series.dtypes must be int, float or bool')
91
        return np.array(data, dtype=dtype, copy=False)  # SparseArray should be supported as well
wxchan's avatar
wxchan committed
92
    else:
93
94
        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
95

wxchan's avatar
wxchan committed
96

wxchan's avatar
wxchan committed
97
def cfloat32_array_to_numpy(cptr, length):
98
    """Convert a ctypes float pointer array to a numpy array."""
wxchan's avatar
wxchan committed
99
    if isinstance(cptr, ctypes.POINTER(ctypes.c_float)):
100
        return np.fromiter(cptr, dtype=np.float32, count=length)
wxchan's avatar
wxchan committed
101
    else:
102
        raise RuntimeError('Expected float pointer')
wxchan's avatar
wxchan committed
103

Guolin Ke's avatar
Guolin Ke committed
104

Guolin Ke's avatar
Guolin Ke committed
105
def cfloat64_array_to_numpy(cptr, length):
106
    """Convert a ctypes double pointer array to a numpy array."""
Guolin Ke's avatar
Guolin Ke committed
107
108
109
110
111
    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
112

wxchan's avatar
wxchan committed
113
def cint32_array_to_numpy(cptr, length):
114
    """Convert a ctypes int pointer array to a numpy array."""
wxchan's avatar
wxchan committed
115
    if isinstance(cptr, ctypes.POINTER(ctypes.c_int32)):
116
        return np.fromiter(cptr, dtype=np.int32, count=length)
wxchan's avatar
wxchan committed
117
    else:
118
119
120
121
122
123
124
125
126
        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
127

wxchan's avatar
wxchan committed
128

wxchan's avatar
wxchan committed
129
def c_str(string):
130
    """Convert a Python string to C string."""
wxchan's avatar
wxchan committed
131
132
    return ctypes.c_char_p(string.encode('utf-8'))

wxchan's avatar
wxchan committed
133

wxchan's avatar
wxchan committed
134
def c_array(ctype, values):
135
    """Convert a Python array to C array."""
wxchan's avatar
wxchan committed
136
137
    return (ctype * len(values))(*values)

wxchan's avatar
wxchan committed
138

wxchan's avatar
wxchan committed
139
def param_dict_to_str(data):
140
    """Convert Python dictionary to string, which is passed to C API."""
141
    if data is None or not data:
wxchan's avatar
wxchan committed
142
143
144
        return ""
    pairs = []
    for key, val in data.items():
145
        if isinstance(val, (list, tuple, set)) or is_numpy_1d_array(val):
146
147
148
149
150
151
            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)))
wxchan's avatar
wxchan committed
152
        elif isinstance(val, string_type) or isinstance(val, numeric_types) or is_numeric(val):
wxchan's avatar
wxchan committed
153
            pairs.append(str(key) + '=' + str(val))
154
        elif val is not None:
155
            raise TypeError('Unknown type of parameter:%s, got:%s'
wxchan's avatar
wxchan committed
156
157
                            % (key, type(val).__name__))
    return ' '.join(pairs)
158

wxchan's avatar
wxchan committed
159

160
class _TempFile(object):
161
162
163
164
    def __enter__(self):
        with NamedTemporaryFile(prefix="lightgbm_tmp_", delete=True) as f:
            self.name = f.name
        return self
wxchan's avatar
wxchan committed
165

166
167
168
    def __exit__(self, exc_type, exc_val, exc_tb):
        if os.path.isfile(self.name):
            os.remove(self.name)
wxchan's avatar
wxchan committed
169

170
171
172
173
    def readlines(self):
        with open(self.name, "r+") as f:
            ret = f.readlines()
        return ret
wxchan's avatar
wxchan committed
174

175
176
    def writelines(self, lines):
        with open(self.name, "w+") as f:
177
            f.writelines(lines)
178

wxchan's avatar
wxchan committed
179

180
class LightGBMError(Exception):
181
182
    """Error thrown by LightGBM."""

183
184
185
    pass


186
class _ConfigAliases(object):
187
188
189
    aliases = {"bin_construct_sample_cnt": {"bin_construct_sample_cnt",
                                            "subsample_for_bin"},
               "boosting": {"boosting",
190
191
192
193
194
195
                            "boosting_type",
                            "boost"},
               "categorical_feature": {"categorical_feature",
                                       "cat_feature",
                                       "categorical_column",
                                       "cat_column"},
196
197
               "data_random_seed": {"data_random_seed",
                                    "data_seed"},
198
199
200
201
               "early_stopping_round": {"early_stopping_round",
                                        "early_stopping_rounds",
                                        "early_stopping",
                                        "n_iter_no_change"},
202
203
204
               "enable_bundle": {"enable_bundle",
                                 "is_enable_bundle",
                                 "bundle"},
205
206
207
208
209
               "eval_at": {"eval_at",
                           "ndcg_eval_at",
                           "ndcg_at",
                           "map_eval_at",
                           "map_at"},
210
211
212
213
214
215
               "group_column": {"group_column",
                                "group",
                                "group_id",
                                "query_column",
                                "query",
                                "query_id"},
216
217
               "header": {"header",
                          "has_header"},
218
219
220
221
222
223
224
225
226
               "ignore_column": {"ignore_column",
                                 "ignore_feature",
                                 "blacklist"},
               "is_enable_sparse": {"is_enable_sparse",
                                    "is_sparse",
                                    "enable_sparse",
                                    "sparse"},
               "label_column": {"label_column",
                                "label"},
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
               "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"},
248
249
250
251
252
               "pre_partition": {"pre_partition",
                                 "is_pre_partition"},
               "two_round": {"two_round",
                             "two_round_loading",
                             "use_two_round_loading"},
253
               "verbosity": {"verbosity",
254
255
256
                             "verbose"},
               "weight_column": {"weight_column",
                                 "weight"}}
257
258
259
260
261

    @classmethod
    def get(cls, *args):
        ret = set()
        for i in args:
262
            ret |= cls.aliases.get(i, {i})
263
264
265
        return ret


266
267
MAX_INT32 = (1 << 31) - 1

268
"""Macro definition of data type in C API of LightGBM"""
wxchan's avatar
wxchan committed
269
270
271
272
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
273

274
"""Matrix is row major in Python"""
wxchan's avatar
wxchan committed
275
276
C_API_IS_ROW_MAJOR = 1

277
"""Macro definition of prediction type in C API of LightGBM"""
wxchan's avatar
wxchan committed
278
279
280
C_API_PREDICT_NORMAL = 0
C_API_PREDICT_RAW_SCORE = 1
C_API_PREDICT_LEAF_INDEX = 2
281
C_API_PREDICT_CONTRIB = 3
wxchan's avatar
wxchan committed
282

283
284
285
286
"""Macro definition of sparse matrix type"""
C_API_MATRIX_TYPE_CSR = 0
C_API_MATRIX_TYPE_CSC = 1

287
288
289
290
"""Macro definition of feature importance type"""
C_API_FEATURE_IMPORTANCE_SPLIT = 0
C_API_FEATURE_IMPORTANCE_GAIN = 1

291
"""Data type of data field"""
wxchan's avatar
wxchan committed
292
293
FIELD_TYPE_MAPPER = {"label": C_API_DTYPE_FLOAT32,
                     "weight": C_API_DTYPE_FLOAT32,
Guolin Ke's avatar
Guolin Ke committed
294
                     "init_score": C_API_DTYPE_FLOAT64,
295
                     "group": C_API_DTYPE_INT32}
wxchan's avatar
wxchan committed
296

297
298
299
300
"""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
301

302
def convert_from_sliced_object(data):
303
    """Fix the memory of multi-dimensional sliced object."""
304
    if isinstance(data, np.ndarray) and isinstance(data.base, np.ndarray):
305
        if not data.flags.c_contiguous:
306
307
            warnings.warn("Usage of np.ndarray subset (sliced data) is not recommended "
                          "due to it will double the peak memory cost in LightGBM.")
308
309
310
311
            return np.copy(data)
    return data


wxchan's avatar
wxchan committed
312
def c_float_array(data):
313
    """Get pointer of float numpy array / list."""
wxchan's avatar
wxchan committed
314
315
316
    if is_1d_list(data):
        data = np.array(data, copy=False)
    if is_numpy_1d_array(data):
317
318
        data = convert_from_sliced_object(data)
        assert data.flags.c_contiguous
wxchan's avatar
wxchan committed
319
320
321
322
323
324
325
        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:
326
            raise TypeError("Expected np.float32 or np.float64, met type({})"
wxchan's avatar
wxchan committed
327
328
                            .format(data.dtype))
    else:
329
        raise TypeError("Unknown type({})".format(type(data).__name__))
330
    return (ptr_data, type_data, data)  # return `data` to avoid the temporary copy is freed
wxchan's avatar
wxchan committed
331

wxchan's avatar
wxchan committed
332

wxchan's avatar
wxchan committed
333
def c_int_array(data):
334
    """Get pointer of int numpy array / list."""
wxchan's avatar
wxchan committed
335
336
337
    if is_1d_list(data):
        data = np.array(data, copy=False)
    if is_numpy_1d_array(data):
338
339
        data = convert_from_sliced_object(data)
        assert data.flags.c_contiguous
wxchan's avatar
wxchan committed
340
341
342
343
344
345
346
        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:
347
            raise TypeError("Expected np.int32 or np.int64, met type({})"
wxchan's avatar
wxchan committed
348
349
                            .format(data.dtype))
    else:
350
        raise TypeError("Unknown type({})".format(type(data).__name__))
351
    return (ptr_data, type_data, data)  # return `data` to avoid the temporary copy is freed
wxchan's avatar
wxchan committed
352

wxchan's avatar
wxchan committed
353

354
355
356
357
358
359
360
361
362
363
364
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


365
def _data_from_pandas(data, feature_name, categorical_feature, pandas_categorical):
366
    if isinstance(data, DataFrame):
367
368
        if len(data.shape) != 2 or data.shape[0] < 1:
            raise ValueError('Input data must be 2 dimensional and non empty.')
369
370
        if feature_name == 'auto' or feature_name is None:
            data = data.rename(columns=str)
371
372
        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]
373
374
375
376
377
        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.')
378
            for col, category in zip_(cat_cols, pandas_categorical):
379
380
                if list(data[col].cat.categories) != list(category):
                    data[col] = data[col].cat.set_categories(category)
381
        if len(cat_cols):  # cat_cols is list
382
            data = data.copy()  # not alter origin DataFrame
383
            data[cat_cols] = data[cat_cols].apply(lambda x: x.cat.codes).replace({-1: np.nan})
384
385
386
        if categorical_feature is not None:
            if feature_name is None:
                feature_name = list(data.columns)
387
            if categorical_feature == 'auto':  # use cat cols from DataFrame
388
                categorical_feature = cat_cols_not_ordered
389
390
            else:  # use cat cols specified by user
                categorical_feature = list(categorical_feature)
391
392
        if feature_name == 'auto':
            feature_name = list(data.columns)
393
394
        bad_indices = _get_bad_pandas_dtypes(data.dtypes)
        if bad_indices:
395
            raise ValueError("DataFrame.dtypes for data must be int, float or bool.\n"
396
                             "Did not expect the data types in the following fields: "
397
                             + ', '.join(data.columns[bad_indices]))
398
399
400
        data = data.values
        if data.dtype != np.float32 and data.dtype != np.float64:
            data = data.astype(np.float32)
401
402
403
404
405
406
    else:
        if feature_name == 'auto':
            feature_name = None
        if categorical_feature == 'auto':
            categorical_feature = None
    return data, feature_name, categorical_feature, pandas_categorical
407
408
409
410
411
412


def _label_from_pandas(label):
    if isinstance(label, DataFrame):
        if len(label.columns) > 1:
            raise ValueError('DataFrame for label cannot have multiple columns')
413
        if _get_bad_pandas_dtypes(label.dtypes):
414
            raise ValueError('DataFrame.dtypes for label must be int, float or bool')
415
        label = np.ravel(label.values.astype(np.float32, copy=False))
416
417
418
    return label


419
420
421
422
423
424
425
426
427
428
429
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):
430
431
    pandas_key = 'pandas_categorical:'
    offset = -len(pandas_key)
432
    if file_name is not None:
433
434
435
436
437
438
439
440
441
442
443
444
445
        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
        last_line = decode_string(lines[-1]).strip()
        if not last_line.startswith(pandas_key):
            last_line = decode_string(lines[-2]).strip()
446
    elif model_str is not None:
447
448
449
450
451
452
        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
453
454


Guolin Ke's avatar
Guolin Ke committed
455
class _InnerPredictor(object):
456
457
458
459
460
    """_InnerPredictor of LightGBM.

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

Nikita Titov's avatar
Nikita Titov committed
461
462
463
    .. note::

        Can be converted from Booster, but cannot be converted to Booster.
Guolin Ke's avatar
Guolin Ke committed
464
    """
465

466
    def __init__(self, model_file=None, booster_handle=None, pred_parameter=None):
467
        """Initialize the _InnerPredictor.
wxchan's avatar
wxchan committed
468
469
470

        Parameters
        ----------
471
        model_file : string or None, optional (default=None)
wxchan's avatar
wxchan committed
472
            Path to the model file.
473
474
475
476
        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
477
478
479
480
481
        """
        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
482
            out_num_iterations = ctypes.c_int(0)
wxchan's avatar
wxchan committed
483
484
485
486
            _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
487
            out_num_class = ctypes.c_int(0)
wxchan's avatar
wxchan committed
488
489
490
491
            _safe_call(_LIB.LGBM_BoosterGetNumClasses(
                self.handle,
                ctypes.byref(out_num_class)))
            self.num_class = out_num_class.value
492
            self.num_total_iteration = out_num_iterations.value
493
            self.pandas_categorical = _load_pandas_categorical(file_name=model_file)
wxchan's avatar
wxchan committed
494
        elif booster_handle is not None:
Guolin Ke's avatar
Guolin Ke committed
495
            self.__is_manage_handle = False
wxchan's avatar
wxchan committed
496
            self.handle = booster_handle
Guolin Ke's avatar
Guolin Ke committed
497
            out_num_class = ctypes.c_int(0)
wxchan's avatar
wxchan committed
498
499
500
501
            _safe_call(_LIB.LGBM_BoosterGetNumClasses(
                self.handle,
                ctypes.byref(out_num_class)))
            self.num_class = out_num_class.value
502
            self.num_total_iteration = self.current_iteration()
503
            self.pandas_categorical = None
wxchan's avatar
wxchan committed
504
        else:
505
            raise TypeError('Need model_file or booster_handle to create a predictor')
wxchan's avatar
wxchan committed
506

507
508
        pred_parameter = {} if pred_parameter is None else pred_parameter
        self.pred_parameter = param_dict_to_str(pred_parameter)
cbecker's avatar
cbecker committed
509

wxchan's avatar
wxchan committed
510
    def __del__(self):
511
512
513
514
515
        try:
            if self.__is_manage_handle:
                _safe_call(_LIB.LGBM_BoosterFree(self.handle))
        except AttributeError:
            pass
wxchan's avatar
wxchan committed
516

517
518
519
520
521
    def __getstate__(self):
        this = self.__dict__.copy()
        this.pop('handle', None)
        return this

wxchan's avatar
wxchan committed
522
    def predict(self, data, num_iteration=-1,
523
                raw_score=False, pred_leaf=False, pred_contrib=False, data_has_header=False,
wxchan's avatar
wxchan committed
524
                is_reshape=True):
525
        """Predict logic.
wxchan's avatar
wxchan committed
526
527
528

        Parameters
        ----------
529
        data : string, numpy array, pandas DataFrame, H2O DataTable's Frame or scipy.sparse
530
531
532
533
534
535
536
537
538
539
540
541
542
543
544
            Data source for prediction.
            When data type is string, it represents the path of txt file.
        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
545
546
547

        Returns
        -------
548
        result : numpy array, scipy.sparse or list of scipy.sparse
549
            Prediction result.
550
            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
551
        """
wxchan's avatar
wxchan committed
552
        if isinstance(data, Dataset):
553
            raise TypeError("Cannot use Dataset instance for prediction, please use raw data instead")
554
        data = _data_from_pandas(data, None, None, self.pandas_categorical)[0]
wxchan's avatar
wxchan committed
555
556
557
558
559
        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
560
561
        if pred_contrib:
            predict_type = C_API_PREDICT_CONTRIB
wxchan's avatar
wxchan committed
562
        int_data_has_header = 1 if data_has_header else 0
563
564
        if num_iteration > self.num_total_iteration:
            num_iteration = self.num_total_iteration
cbecker's avatar
cbecker committed
565

wxchan's avatar
wxchan committed
566
        if isinstance(data, string_type):
567
            with _TempFile() as f:
wxchan's avatar
wxchan committed
568
569
570
                _safe_call(_LIB.LGBM_BoosterPredictForFile(
                    self.handle,
                    c_str(data),
Guolin Ke's avatar
Guolin Ke committed
571
572
573
                    ctypes.c_int(int_data_has_header),
                    ctypes.c_int(predict_type),
                    ctypes.c_int(num_iteration),
574
                    c_str(self.pred_parameter),
wxchan's avatar
wxchan committed
575
576
                    c_str(f.name)))
                lines = f.readlines()
577
578
                nrow = len(lines)
                preds = [float(token) for line in lines for token in line.split('\t')]
Guolin Ke's avatar
Guolin Ke committed
579
                preds = np.array(preds, dtype=np.float64, copy=False)
wxchan's avatar
wxchan committed
580
        elif isinstance(data, scipy.sparse.csr_matrix):
581
            preds, nrow = self.__pred_for_csr(data, num_iteration, predict_type)
Guolin Ke's avatar
Guolin Ke committed
582
        elif isinstance(data, scipy.sparse.csc_matrix):
583
            preds, nrow = self.__pred_for_csc(data, num_iteration, predict_type)
wxchan's avatar
wxchan committed
584
        elif isinstance(data, np.ndarray):
585
            preds, nrow = self.__pred_for_np2d(data, num_iteration, predict_type)
586
587
588
        elif isinstance(data, list):
            try:
                data = np.array(data)
589
            except BaseException:
590
                raise ValueError('Cannot convert data list to numpy array.')
591
            preds, nrow = self.__pred_for_np2d(data, num_iteration, predict_type)
592
593
        elif isinstance(data, DataTable):
            preds, nrow = self.__pred_for_np2d(data.to_numpy(), num_iteration, predict_type)
wxchan's avatar
wxchan committed
594
595
        else:
            try:
596
                warnings.warn('Converting data to scipy sparse matrix.')
wxchan's avatar
wxchan committed
597
                csr = scipy.sparse.csr_matrix(data)
598
            except BaseException:
599
                raise TypeError('Cannot predict data for type {}'.format(type(data).__name__))
600
            preds, nrow = self.__pred_for_csr(csr, num_iteration, predict_type)
wxchan's avatar
wxchan committed
601
602
        if pred_leaf:
            preds = preds.astype(np.int32)
603
604
        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
605
            if preds.size % nrow == 0:
606
                preds = preds.reshape(nrow, -1)
wxchan's avatar
wxchan committed
607
            else:
608
                raise ValueError('Length of predict result (%d) cannot be divide nrow (%d)'
wxchan's avatar
wxchan committed
609
610
611
612
                                 % (preds.size, nrow))
        return preds

    def __get_num_preds(self, num_iteration, nrow, predict_type):
613
        """Get size of prediction result."""
614
615
616
617
618
        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
619
620
621
        n_preds = ctypes.c_int64(0)
        _safe_call(_LIB.LGBM_BoosterCalcNumPredict(
            self.handle,
Guolin Ke's avatar
Guolin Ke committed
622
623
624
            ctypes.c_int(nrow),
            ctypes.c_int(predict_type),
            ctypes.c_int(num_iteration),
Guolin Ke's avatar
Guolin Ke committed
625
626
            ctypes.byref(n_preds)))
        return n_preds.value
wxchan's avatar
wxchan committed
627
628

    def __pred_for_np2d(self, mat, num_iteration, predict_type):
629
        """Predict for a 2-D numpy matrix."""
wxchan's avatar
wxchan committed
630
        if len(mat.shape) != 2:
631
            raise ValueError('Input numpy.ndarray or list must be 2 dimensional')
wxchan's avatar
wxchan committed
632

633
634
635
        def inner_predict(mat, num_iteration, predict_type, preds=None):
            if mat.dtype == np.float32 or mat.dtype == np.float64:
                data = np.array(mat.reshape(mat.size), dtype=mat.dtype, copy=False)
636
            else:  # change non-float data to float data, need to copy
637
638
639
640
641
642
643
644
645
646
647
648
649
650
651
652
653
654
655
656
657
658
659
660
661
662
663
664
665
666
667
                data = np.array(mat.reshape(mat.size), dtype=np.float32)
            ptr_data, type_ptr_data, _ = c_float_array(data)
            n_preds = self.__get_num_preds(num_iteration, mat.shape[0], predict_type)
            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),
                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
            n_preds = [self.__get_num_preds(num_iteration, i, predict_type) for i in np.diff([0] + list(sections) + [nrow])]
            n_preds_sections = np.array([0] + n_preds, dtype=np.intp).cumsum()
            preds = np.zeros(sum(n_preds), dtype=np.float64)
668
669
            for chunk, (start_idx_pred, end_idx_pred) in zip_(np.array_split(mat, sections),
                                                              zip_(n_preds_sections, n_preds_sections[1:])):
670
671
672
                # avoid memory consumption by arrays concatenation operations
                inner_predict(chunk, num_iteration, predict_type, preds[start_idx_pred:end_idx_pred])
            return preds, nrow
wxchan's avatar
wxchan committed
673
        else:
674
            return inner_predict(mat, num_iteration, predict_type)
wxchan's avatar
wxchan committed
675

676
677
678
679
680
681
682
683
684
685
686
687
688
689
690
691
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
    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

wxchan's avatar
wxchan committed
722
    def __pred_for_csr(self, csr, num_iteration, predict_type):
723
        """Predict for a CSR data."""
724
725
726
727
728
729
730
731
732
733
734
735
        def inner_predict(csr, num_iteration, predict_type, preds=None):
            nrow = len(csr.indptr) - 1
            n_preds = self.__get_num_preds(num_iteration, nrow, predict_type)
            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)

736
            assert csr.shape[1] <= MAX_INT32
737
            csr_indices = csr.indices.astype(np.int32, copy=False)
738

739
740
741
742
            _safe_call(_LIB.LGBM_BoosterPredictForCSR(
                self.handle,
                ptr_indptr,
                ctypes.c_int32(type_ptr_indptr),
743
                csr_indices.ctypes.data_as(ctypes.POINTER(ctypes.c_int32)),
744
745
746
747
748
749
750
751
752
753
754
755
756
                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),
                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
757

758
759
760
761
762
763
764
765
766
767
768
769
770
771
772
773
774
775
776
777
778
779
780
781
782
783
784
785
786
787
788
789
790
791
792
793
794
795
796
797
        def inner_predict_sparse(csr, num_iteration, predict_type):
            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),
                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:
            return inner_predict_sparse(csr, num_iteration, predict_type)
798
799
800
801
802
803
804
805
806
807
808
809
810
811
        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
            n_preds = [self.__get_num_preds(num_iteration, i, predict_type) for i in np.diff(sections)]
            n_preds_sections = np.array([0] + n_preds, dtype=np.intp).cumsum()
            preds = np.zeros(sum(n_preds), dtype=np.float64)
            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:])):
                # avoid memory consumption by arrays concatenation operations
                inner_predict(csr[start_idx:end_idx], num_iteration, predict_type, preds[start_idx_pred:end_idx_pred])
            return preds, nrow
        else:
            return inner_predict(csr, num_iteration, predict_type)
Guolin Ke's avatar
Guolin Ke committed
812
813

    def __pred_for_csc(self, csc, num_iteration, predict_type):
814
        """Predict for a CSC data."""
815
816
817
818
819
820
821
822
823
824
825
826
827
828
829
830
831
832
833
834
835
836
837
838
839
840
841
842
843
844
845
846
847
848
849
850
851
852
        def inner_predict_sparse(csc, num_iteration, predict_type):
            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),
                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
853
        nrow = csc.shape[0]
854
855
        if nrow > MAX_INT32:
            return self.__pred_for_csr(csc.tocsr(), num_iteration, predict_type)
856
857
        if predict_type == C_API_PREDICT_CONTRIB:
            return inner_predict_sparse(csc, num_iteration, predict_type)
Guolin Ke's avatar
Guolin Ke committed
858
859
860
861
        n_preds = self.__get_num_preds(num_iteration, nrow, predict_type)
        preds = np.zeros(n_preds, dtype=np.float64)
        out_num_preds = ctypes.c_int64(0)

862
863
        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
864

865
        assert csc.shape[0] <= MAX_INT32
866
        csc_indices = csc.indices.astype(np.int32, copy=False)
867

Guolin Ke's avatar
Guolin Ke committed
868
869
870
        _safe_call(_LIB.LGBM_BoosterPredictForCSC(
            self.handle,
            ptr_indptr,
Guolin Ke's avatar
Guolin Ke committed
871
            ctypes.c_int32(type_ptr_indptr),
872
            csc_indices.ctypes.data_as(ctypes.POINTER(ctypes.c_int32)),
Guolin Ke's avatar
Guolin Ke committed
873
            ptr_data,
Guolin Ke's avatar
Guolin Ke committed
874
875
876
877
878
879
            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),
            ctypes.c_int(num_iteration),
880
            c_str(self.pred_parameter),
Guolin Ke's avatar
Guolin Ke committed
881
            ctypes.byref(out_num_preds),
wxchan's avatar
wxchan committed
882
            preds.ctypes.data_as(ctypes.POINTER(ctypes.c_double))))
wxchan's avatar
wxchan committed
883
        if n_preds != out_num_preds.value:
884
            raise ValueError("Wrong length for predict results")
wxchan's avatar
wxchan committed
885
886
        return preds, nrow

887
888
889
890
891
892
893
894
895
896
897
898
899
900
    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
901

wxchan's avatar
wxchan committed
902
903
class Dataset(object):
    """Dataset in LightGBM."""
904

905
    def __init__(self, data, label=None, reference=None,
906
                 weight=None, group=None, init_score=None, silent=False,
907
                 feature_name='auto', categorical_feature='auto', params=None,
wxchan's avatar
wxchan committed
908
                 free_raw_data=True):
909
        """Initialize Dataset.
910

wxchan's avatar
wxchan committed
911
912
        Parameters
        ----------
913
        data : string, numpy array, pandas DataFrame, H2O DataTable's Frame, scipy.sparse or list of numpy arrays
wxchan's avatar
wxchan committed
914
            Data source of Dataset.
915
            If string, it represents the path to txt file.
916
        label : list, numpy 1-D array, pandas Series / one-column DataFrame or None, optional (default=None)
917
918
919
            Label of the data.
        reference : Dataset or None, optional (default=None)
            If this is Dataset for validation, training data should be used as reference.
920
        weight : list, numpy 1-D array, pandas Series or None, optional (default=None)
wxchan's avatar
wxchan committed
921
            Weight for each instance.
922
        group : list, numpy 1-D array, pandas Series or None, optional (default=None)
923
            Group/query size for Dataset.
924
        init_score : list, numpy 1-D array, pandas Series or None, optional (default=None)
925
            Init score for Dataset.
926
927
928
929
930
931
932
933
934
        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).
935
            If 'auto' and data is pandas DataFrame, pandas unordered categorical columns are used.
936
            All values in categorical features should be less than int32 max value (2147483647).
937
            Large values could be memory consuming. Consider using consecutive integers starting from zero.
938
            All negative values in categorical features will be treated as missing values.
939
            The output cannot be monotonically constrained with respect to a categorical feature.
Nikita Titov's avatar
Nikita Titov committed
940
        params : dict or None, optional (default=None)
941
            Other parameters for Dataset.
Nikita Titov's avatar
Nikita Titov committed
942
        free_raw_data : bool, optional (default=True)
943
            If True, raw data is freed after constructing inner Dataset.
wxchan's avatar
wxchan committed
944
        """
wxchan's avatar
wxchan committed
945
946
947
948
949
950
        self.handle = None
        self.data = data
        self.label = label
        self.reference = reference
        self.weight = weight
        self.group = group
951
        self.init_score = init_score
wxchan's avatar
wxchan committed
952
953
        self.silent = silent
        self.feature_name = feature_name
954
        self.categorical_feature = categorical_feature
955
        self.params = copy.deepcopy(params)
wxchan's avatar
wxchan committed
956
957
        self.free_raw_data = free_raw_data
        self.used_indices = None
958
        self.need_slice = True
wxchan's avatar
wxchan committed
959
        self._predictor = None
960
        self.pandas_categorical = None
961
        self.params_back_up = None
962
963
        self.feature_penalty = None
        self.monotone_constraints = None
964
        self.version = 0
wxchan's avatar
wxchan committed
965
966

    def __del__(self):
967
968
969
970
        try:
            self._free_handle()
        except AttributeError:
            pass
971

972
973
974
975
976
977
978
979
980
981
982
983
984
985
986
987
988
989
990
991
992
993
994
995
996
997
998
999
1000
1001
1002
    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",
                                                "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}

1003
    def _free_handle(self):
1004
        if self.handle is not None:
1005
            _safe_call(_LIB.LGBM_DatasetFree(self.handle))
1006
            self.handle = None
Guolin Ke's avatar
Guolin Ke committed
1007
1008
1009
        self.need_slice = True
        if self.used_indices is not None:
            self.data = None
Nikita Titov's avatar
Nikita Titov committed
1010
        return self
wxchan's avatar
wxchan committed
1011

Guolin Ke's avatar
Guolin Ke committed
1012
1013
1014
1015
    def _set_init_score_by_predictor(self, predictor, data, used_indices=None):
        data_has_header = False
        if isinstance(data, string_type):
            # check data has header or not
1016
            data_has_header = any(self.params.get(alias, False) for alias in _ConfigAliases.get("header"))
Guolin Ke's avatar
Guolin Ke committed
1017
        num_data = self.num_data()
1018
1019
1020
1021
1022
1023
1024
1025
1026
1027
1028
1029
1030
1031
1032
1033
1034
1035
        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
                if isinstance(data, string_type):
                    sub_init_score = np.zeros(num_data * predictor.num_class, dtype=np.float32)
                    assert num_data == len(used_indices)
                    for i in range_(len(used_indices)):
                        for j in range_(predictor.num_class):
                            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)
                for i in range_(num_data):
Guolin Ke's avatar
Guolin Ke committed
1036
                    for j in range_(predictor.num_class):
1037
1038
1039
1040
1041
1042
                        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
1043
1044
        self.set_init_score(init_score)

1045
    def _lazy_init(self, data, label=None, reference=None,
1046
                   weight=None, group=None, init_score=None, predictor=None,
wxchan's avatar
wxchan committed
1047
                   silent=False, feature_name='auto',
1048
                   categorical_feature='auto', params=None):
wxchan's avatar
wxchan committed
1049
1050
        if data is None:
            self.handle = None
Nikita Titov's avatar
Nikita Titov committed
1051
            return self
Guolin Ke's avatar
Guolin Ke committed
1052
1053
1054
        if reference is not None:
            self.pandas_categorical = reference.pandas_categorical
            categorical_feature = reference.categorical_feature
1055
1056
1057
1058
        data, feature_name, categorical_feature, self.pandas_categorical = _data_from_pandas(data,
                                                                                             feature_name,
                                                                                             categorical_feature,
                                                                                             self.pandas_categorical)
wxchan's avatar
wxchan committed
1059
        label = _label_from_pandas(label)
Guolin Ke's avatar
Guolin Ke committed
1060

1061
        # process for args
wxchan's avatar
wxchan committed
1062
        params = {} if params is None else params
1063
1064
1065
        args_names = (getattr(self.__class__, '_lazy_init')
                      .__code__
                      .co_varnames[:getattr(self.__class__, '_lazy_init').__code__.co_argcount])
1066
1067
        for key, _ in params.items():
            if key in args_names:
1068
1069
1070
                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))
1071
        # user can set verbose with params, it has higher priority
1072
        if not any(verbose_alias in params for verbose_alias in _ConfigAliases.get("verbosity")) and silent:
1073
            params["verbose"] = -1
1074
        # get categorical features
1075
1076
1077
1078
1079
1080
1081
1082
1083
1084
1085
1086
1087
        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:
                if isinstance(name, string_type) and name in feature_dict:
                    categorical_indices.add(feature_dict[name])
                elif isinstance(name, integer_types):
                    categorical_indices.add(name)
                else:
                    raise TypeError("Wrong type({}) or unknown name({}) in categorical_feature"
                                    .format(type(name).__name__, name))
1088
            if categorical_indices:
1089
1090
1091
1092
                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)
1093
                params['categorical_column'] = sorted(categorical_indices)
1094

wxchan's avatar
wxchan committed
1095
        params_str = param_dict_to_str(params)
1096
        self.params = params
1097
        # process for reference dataset
wxchan's avatar
wxchan committed
1098
        ref_dataset = None
wxchan's avatar
wxchan committed
1099
        if isinstance(reference, Dataset):
1100
            ref_dataset = reference.construct().handle
wxchan's avatar
wxchan committed
1101
1102
        elif reference is not None:
            raise TypeError('Reference dataset should be None or dataset instance')
1103
        # start construct data
wxchan's avatar
wxchan committed
1104
        if isinstance(data, string_type):
wxchan's avatar
wxchan committed
1105
1106
1107
1108
1109
1110
1111
1112
            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
1113
1114
        elif isinstance(data, scipy.sparse.csc_matrix):
            self.__init_from_csc(data, params_str, ref_dataset)
wxchan's avatar
wxchan committed
1115
1116
        elif isinstance(data, np.ndarray):
            self.__init_from_np2d(data, params_str, ref_dataset)
1117
1118
        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)
1119
1120
        elif isinstance(data, DataTable):
            self.__init_from_np2d(data.to_numpy(), params_str, ref_dataset)
wxchan's avatar
wxchan committed
1121
1122
1123
1124
        else:
            try:
                csr = scipy.sparse.csr_matrix(data)
                self.__init_from_csr(csr, params_str, ref_dataset)
1125
            except BaseException:
wxchan's avatar
wxchan committed
1126
                raise TypeError('Cannot initialize Dataset from {}'.format(type(data).__name__))
wxchan's avatar
wxchan committed
1127
1128
1129
        if label is not None:
            self.set_label(label)
        if self.get_label() is None:
1130
            raise ValueError("Label should not be None")
wxchan's avatar
wxchan committed
1131
1132
1133
1134
        if weight is not None:
            self.set_weight(weight)
        if group is not None:
            self.set_group(group)
1135
1136
1137
        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
1138
            self._set_init_score_by_predictor(predictor, data)
1139
1140
        elif init_score is not None:
            self.set_init_score(init_score)
Guolin Ke's avatar
Guolin Ke committed
1141
1142
        elif predictor is not None:
            raise TypeError('Wrong predictor type {}'.format(type(predictor).__name__))
Guolin Ke's avatar
Guolin Ke committed
1143
        # set feature names
Nikita Titov's avatar
Nikita Titov committed
1144
        return self.set_feature_name(feature_name)
wxchan's avatar
wxchan committed
1145
1146

    def __init_from_np2d(self, mat, params_str, ref_dataset):
1147
        """Initialize data from a 2-D numpy matrix."""
wxchan's avatar
wxchan committed
1148
1149
1150
1151
1152
1153
        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)
1154
        else:  # change non-float data to float data, need to copy
wxchan's avatar
wxchan committed
1155
1156
            data = np.array(mat.reshape(mat.size), dtype=np.float32)

1157
        ptr_data, type_ptr_data, _ = c_float_array(data)
wxchan's avatar
wxchan committed
1158
1159
        _safe_call(_LIB.LGBM_DatasetCreateFromMat(
            ptr_data,
Guolin Ke's avatar
Guolin Ke committed
1160
1161
1162
1163
            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
1164
1165
1166
            c_str(params_str),
            ref_dataset,
            ctypes.byref(self.handle)))
Nikita Titov's avatar
Nikita Titov committed
1167
        return self
wxchan's avatar
wxchan committed
1168

1169
    def __init_from_list_np2d(self, mats, params_str, ref_dataset):
1170
        """Initialize data from a list of 2-D numpy matrices."""
1171
1172
1173
1174
1175
1176
1177
1178
1179
1180
1181
1182
1183
1184
1185
1186
1187
1188
1189
1190
1191
        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)
1192
            else:  # change non-float data to float data, need to copy
1193
1194
1195
1196
1197
1198
1199
1200
1201
1202
1203
1204
1205
1206
1207
1208
1209
1210
1211
1212
                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
1213
        return self
1214

wxchan's avatar
wxchan committed
1215
    def __init_from_csr(self, csr, params_str, ref_dataset):
1216
        """Initialize data from a CSR matrix."""
wxchan's avatar
wxchan committed
1217
        if len(csr.indices) != len(csr.data):
1218
            raise ValueError('Length mismatch: {} vs {}'.format(len(csr.indices), len(csr.data)))
wxchan's avatar
wxchan committed
1219
1220
        self.handle = ctypes.c_void_p()

1221
1222
        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
1223

1224
        assert csr.shape[1] <= MAX_INT32
1225
        csr_indices = csr.indices.astype(np.int32, copy=False)
1226

wxchan's avatar
wxchan committed
1227
1228
        _safe_call(_LIB.LGBM_DatasetCreateFromCSR(
            ptr_indptr,
Guolin Ke's avatar
Guolin Ke committed
1229
            ctypes.c_int(type_ptr_indptr),
1230
            csr_indices.ctypes.data_as(ctypes.POINTER(ctypes.c_int32)),
wxchan's avatar
wxchan committed
1231
            ptr_data,
Guolin Ke's avatar
Guolin Ke committed
1232
1233
1234
1235
            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
1236
1237
1238
            c_str(params_str),
            ref_dataset,
            ctypes.byref(self.handle)))
Nikita Titov's avatar
Nikita Titov committed
1239
        return self
wxchan's avatar
wxchan committed
1240

Guolin Ke's avatar
Guolin Ke committed
1241
    def __init_from_csc(self, csc, params_str, ref_dataset):
1242
        """Initialize data from a CSC matrix."""
Guolin Ke's avatar
Guolin Ke committed
1243
1244
1245
1246
        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()

1247
1248
        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
1249

1250
        assert csc.shape[0] <= MAX_INT32
1251
        csc_indices = csc.indices.astype(np.int32, copy=False)
1252

Guolin Ke's avatar
Guolin Ke committed
1253
1254
        _safe_call(_LIB.LGBM_DatasetCreateFromCSC(
            ptr_indptr,
Guolin Ke's avatar
Guolin Ke committed
1255
            ctypes.c_int(type_ptr_indptr),
1256
            csc_indices.ctypes.data_as(ctypes.POINTER(ctypes.c_int32)),
Guolin Ke's avatar
Guolin Ke committed
1257
            ptr_data,
Guolin Ke's avatar
Guolin Ke committed
1258
1259
1260
1261
            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
1262
1263
1264
            c_str(params_str),
            ref_dataset,
            ctypes.byref(self.handle)))
Nikita Titov's avatar
Nikita Titov committed
1265
        return self
Guolin Ke's avatar
Guolin Ke committed
1266

wxchan's avatar
wxchan committed
1267
    def construct(self):
1268
1269
1270
1271
1272
        """Lazy init.

        Returns
        -------
        self : Dataset
Nikita Titov's avatar
Nikita Titov committed
1273
            Constructed Dataset object.
1274
        """
1275
        if self.handle is None:
wxchan's avatar
wxchan committed
1276
            if self.reference is not None:
1277
1278
1279
1280
                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
1281
                if self.used_indices is None:
1282
                    # create valid
1283
                    self._lazy_init(self.data, label=self.label, reference=self.reference,
1284
1285
                                    weight=self.weight, group=self.group,
                                    init_score=self.init_score, predictor=self._predictor,
1286
                                    silent=self.silent, feature_name=self.feature_name, params=self.params)
wxchan's avatar
wxchan committed
1287
                else:
1288
                    # construct subset
wxchan's avatar
wxchan committed
1289
                    used_indices = list_to_1d_numpy(self.used_indices, np.int32, name='used_indices')
1290
                    assert used_indices.flags.c_contiguous
Guolin Ke's avatar
Guolin Ke committed
1291
                    if self.reference.group is not None:
1292
                        group_info = np.array(self.reference.group).astype(np.int32, copy=False)
1293
1294
                        _, self.group = np.unique(np.repeat(range_(len(group_info)), repeats=group_info)[self.used_indices],
                                                  return_counts=True)
1295
                    self.handle = ctypes.c_void_p()
wxchan's avatar
wxchan committed
1296
1297
                    params_str = param_dict_to_str(self.params)
                    _safe_call(_LIB.LGBM_DatasetGetSubset(
1298
                        self.reference.construct().handle,
wxchan's avatar
wxchan committed
1299
                        used_indices.ctypes.data_as(ctypes.POINTER(ctypes.c_int32)),
Guolin Ke's avatar
Guolin Ke committed
1300
                        ctypes.c_int(used_indices.shape[0]),
wxchan's avatar
wxchan committed
1301
1302
                        c_str(params_str),
                        ctypes.byref(self.handle)))
Guolin Ke's avatar
Guolin Ke committed
1303
1304
                    if not self.free_raw_data:
                        self.get_data()
Guolin Ke's avatar
Guolin Ke committed
1305
1306
                    if self.group is not None:
                        self.set_group(self.group)
wxchan's avatar
wxchan committed
1307
1308
                    if self.get_label() is None:
                        raise ValueError("Label should not be None.")
Guolin Ke's avatar
Guolin Ke committed
1309
1310
1311
                    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
1312
            else:
1313
                # create train
1314
                self._lazy_init(self.data, label=self.label,
1315
1316
1317
                                weight=self.weight, group=self.group,
                                init_score=self.init_score, predictor=self._predictor,
                                silent=self.silent, feature_name=self.feature_name,
1318
                                categorical_feature=self.categorical_feature, params=self.params)
wxchan's avatar
wxchan committed
1319
1320
1321
            if self.free_raw_data:
                self.data = None
        return self
wxchan's avatar
wxchan committed
1322

wxchan's avatar
wxchan committed
1323
    def create_valid(self, data, label=None, weight=None, group=None,
1324
                     init_score=None, silent=False, params=None):
1325
        """Create validation data align with current Dataset.
wxchan's avatar
wxchan committed
1326
1327
1328

        Parameters
        ----------
1329
        data : string, numpy array, pandas DataFrame, H2O DataTable's Frame, scipy.sparse or list of numpy arrays
wxchan's avatar
wxchan committed
1330
            Data source of Dataset.
1331
            If string, it represents the path to txt file.
1332
        label : list, numpy 1-D array, pandas Series / one-column DataFrame or None, optional (default=None)
1333
1334
            Label of the data.
        weight : list, numpy 1-D array, pandas Series or None, optional (default=None)
wxchan's avatar
wxchan committed
1335
            Weight for each instance.
1336
        group : list, numpy 1-D array, pandas Series or None, optional (default=None)
1337
            Group/query size for Dataset.
1338
        init_score : list, numpy 1-D array, pandas Series or None, optional (default=None)
1339
            Init score for Dataset.
1340
1341
        silent : bool, optional (default=False)
            Whether to print messages during construction.
Nikita Titov's avatar
Nikita Titov committed
1342
        params : dict or None, optional (default=None)
1343
            Other parameters for validation Dataset.
1344
1345
1346

        Returns
        -------
Nikita Titov's avatar
Nikita Titov committed
1347
1348
        valid : Dataset
            Validation Dataset with reference to self.
wxchan's avatar
wxchan committed
1349
        """
1350
        ret = Dataset(data, label=label, reference=self,
1351
1352
                      weight=weight, group=group, init_score=init_score,
                      silent=silent, params=params, free_raw_data=self.free_raw_data)
wxchan's avatar
wxchan committed
1353
        ret._predictor = self._predictor
1354
        ret.pandas_categorical = self.pandas_categorical
wxchan's avatar
wxchan committed
1355
        return ret
wxchan's avatar
wxchan committed
1356

wxchan's avatar
wxchan committed
1357
    def subset(self, used_indices, params=None):
1358
        """Get subset of current Dataset.
wxchan's avatar
wxchan committed
1359
1360
1361
1362

        Parameters
        ----------
        used_indices : list of int
1363
            Indices used to create the subset.
Nikita Titov's avatar
Nikita Titov committed
1364
        params : dict or None, optional (default=None)
1365
            These parameters will be passed to Dataset constructor.
1366
1367
1368
1369
1370

        Returns
        -------
        subset : Dataset
            Subset of the current Dataset.
wxchan's avatar
wxchan committed
1371
        """
wxchan's avatar
wxchan committed
1372
1373
        if params is None:
            params = self.params
wxchan's avatar
wxchan committed
1374
        ret = Dataset(None, reference=self, feature_name=self.feature_name,
1375
1376
                      categorical_feature=self.categorical_feature, params=params,
                      free_raw_data=self.free_raw_data)
wxchan's avatar
wxchan committed
1377
        ret._predictor = self._predictor
1378
        ret.pandas_categorical = self.pandas_categorical
1379
        ret.used_indices = sorted(used_indices)
wxchan's avatar
wxchan committed
1380
1381
1382
        return ret

    def save_binary(self, filename):
1383
        """Save Dataset to a binary file.
wxchan's avatar
wxchan committed
1384

1385
1386
1387
1388
1389
        .. 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
1390
1391
1392
1393
        Parameters
        ----------
        filename : string
            Name of the output file.
Nikita Titov's avatar
Nikita Titov committed
1394
1395
1396
1397
1398

        Returns
        -------
        self : Dataset
            Returns self.
wxchan's avatar
wxchan committed
1399
1400
1401
1402
        """
        _safe_call(_LIB.LGBM_DatasetSaveBinary(
            self.construct().handle,
            c_str(filename)))
Nikita Titov's avatar
Nikita Titov committed
1403
        return self
wxchan's avatar
wxchan committed
1404
1405

    def _update_params(self, params):
1406
1407
        if not params:
            return self
1408
1409
1410
1411
1412
1413
1414
1415
1416
1417
1418
1419
1420
1421
1422
1423
1424
1425
1426
1427
1428
1429
        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:
                    raise LightGBMError(decode_string(_LIB.LGBM_GetLastError()))
Nikita Titov's avatar
Nikita Titov committed
1430
        return self
wxchan's avatar
wxchan committed
1431

1432
    def _reverse_update_params(self):
1433
1434
1435
        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
1436
        return self
1437

wxchan's avatar
wxchan committed
1438
    def set_field(self, field_name, data):
wxchan's avatar
wxchan committed
1439
        """Set property into the Dataset.
wxchan's avatar
wxchan committed
1440
1441
1442

        Parameters
        ----------
Nikita Titov's avatar
Nikita Titov committed
1443
        field_name : string
1444
            The field name of the information.
1445
        data : list, numpy 1-D array, pandas Series or None
1446
            The array of data to be set.
Nikita Titov's avatar
Nikita Titov committed
1447
1448
1449
1450
1451

        Returns
        -------
        self : Dataset
            Dataset with set property.
wxchan's avatar
wxchan committed
1452
        """
1453
1454
        if self.handle is None:
            raise Exception("Cannot set %s before construct dataset" % field_name)
wxchan's avatar
wxchan committed
1455
        if data is None:
1456
            # set to None
wxchan's avatar
wxchan committed
1457
1458
1459
1460
            _safe_call(_LIB.LGBM_DatasetSetField(
                self.handle,
                c_str(field_name),
                None,
Guolin Ke's avatar
Guolin Ke committed
1461
1462
                ctypes.c_int(0),
                ctypes.c_int(FIELD_TYPE_MAPPER[field_name])))
Nikita Titov's avatar
Nikita Titov committed
1463
            return self
Guolin Ke's avatar
Guolin Ke committed
1464
1465
1466
1467
1468
        dtype = np.float32
        if field_name == 'group':
            dtype = np.int32
        elif field_name == 'init_score':
            dtype = np.float64
1469
        data = list_to_1d_numpy(data, dtype, name=field_name)
1470
1471
        if data.dtype == np.float32 or data.dtype == np.float64:
            ptr_data, type_data, _ = c_float_array(data)
wxchan's avatar
wxchan committed
1472
        elif data.dtype == np.int32:
1473
            ptr_data, type_data, _ = c_int_array(data)
wxchan's avatar
wxchan committed
1474
        else:
Nikita Titov's avatar
Nikita Titov committed
1475
            raise TypeError("Expected np.float32/64 or np.int32, met type({})".format(data.dtype))
wxchan's avatar
wxchan committed
1476
        if type_data != FIELD_TYPE_MAPPER[field_name]:
1477
            raise TypeError("Input type error for set_field")
wxchan's avatar
wxchan committed
1478
1479
1480
1481
        _safe_call(_LIB.LGBM_DatasetSetField(
            self.handle,
            c_str(field_name),
            ptr_data,
Guolin Ke's avatar
Guolin Ke committed
1482
1483
            ctypes.c_int(len(data)),
            ctypes.c_int(type_data)))
1484
        self.version += 1
Nikita Titov's avatar
Nikita Titov committed
1485
        return self
wxchan's avatar
wxchan committed
1486

wxchan's avatar
wxchan committed
1487
1488
    def get_field(self, field_name):
        """Get property from the Dataset.
wxchan's avatar
wxchan committed
1489
1490
1491

        Parameters
        ----------
Nikita Titov's avatar
Nikita Titov committed
1492
        field_name : string
1493
            The field name of the information.
wxchan's avatar
wxchan committed
1494
1495
1496

        Returns
        -------
1497
1498
        info : numpy array
            A numpy array with information from the Dataset.
Guolin Ke's avatar
Guolin Ke committed
1499
        """
1500
        if self.handle is None:
1501
            raise Exception("Cannot get %s before construct Dataset" % field_name)
Guolin Ke's avatar
Guolin Ke committed
1502
1503
        tmp_out_len = ctypes.c_int()
        out_type = ctypes.c_int()
wxchan's avatar
wxchan committed
1504
1505
1506
1507
1508
1509
1510
1511
1512
1513
1514
1515
1516
1517
1518
        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
1519
1520
        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)
1521
        else:
wxchan's avatar
wxchan committed
1522
            raise TypeError("Unknown type")
Guolin Ke's avatar
Guolin Ke committed
1523

1524
    def set_categorical_feature(self, categorical_feature):
1525
        """Set categorical features.
1526
1527
1528

        Parameters
        ----------
1529
1530
        categorical_feature : list of int or strings
            Names or indices of categorical features.
Nikita Titov's avatar
Nikita Titov committed
1531
1532
1533
1534
1535

        Returns
        -------
        self : Dataset
            Dataset with set categorical features.
1536
1537
        """
        if self.categorical_feature == categorical_feature:
Nikita Titov's avatar
Nikita Titov committed
1538
            return self
1539
        if self.data is not None:
1540
1541
            if self.categorical_feature is None:
                self.categorical_feature = categorical_feature
Nikita Titov's avatar
Nikita Titov committed
1542
                return self._free_handle()
1543
1544
            elif categorical_feature == 'auto':
                warnings.warn('Using categorical_feature in Dataset.')
Nikita Titov's avatar
Nikita Titov committed
1545
                return self
1546
            else:
1547
1548
                warnings.warn('categorical_feature in Dataset is overridden.\n'
                              'New categorical_feature is {}'.format(sorted(list(categorical_feature))))
1549
                self.categorical_feature = categorical_feature
Nikita Titov's avatar
Nikita Titov committed
1550
                return self._free_handle()
1551
        else:
1552
1553
            raise LightGBMError("Cannot set categorical feature after freed raw data, "
                                "set free_raw_data=False when construct Dataset to avoid this.")
1554

Guolin Ke's avatar
Guolin Ke committed
1555
    def _set_predictor(self, predictor):
1556
1557
1558
1559
        """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
1560
        """
1561
        if predictor is self._predictor and (predictor is None or predictor.current_iteration() == self._predictor.current_iteration()):
Nikita Titov's avatar
Nikita Titov committed
1562
            return self
1563
        if self.handle is None:
Guolin Ke's avatar
Guolin Ke committed
1564
            self._predictor = predictor
1565
1566
1567
1568
1569
1570
        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
1571
        else:
1572
1573
            raise LightGBMError("Cannot set predictor after freed raw data, "
                                "set free_raw_data=False when construct Dataset to avoid this.")
1574
        return self
Guolin Ke's avatar
Guolin Ke committed
1575
1576

    def set_reference(self, reference):
1577
        """Set reference Dataset.
Guolin Ke's avatar
Guolin Ke committed
1578
1579
1580
1581

        Parameters
        ----------
        reference : Dataset
1582
            Reference that is used as a template to construct the current Dataset.
Nikita Titov's avatar
Nikita Titov committed
1583
1584
1585
1586
1587

        Returns
        -------
        self : Dataset
            Dataset with set reference.
Guolin Ke's avatar
Guolin Ke committed
1588
        """
1589
1590
1591
        self.set_categorical_feature(reference.categorical_feature) \
            .set_feature_name(reference.feature_name) \
            ._set_predictor(reference._predictor)
1592
1593
        # 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
1594
            return self
Guolin Ke's avatar
Guolin Ke committed
1595
1596
        if self.data is not None:
            self.reference = reference
Nikita Titov's avatar
Nikita Titov committed
1597
            return self._free_handle()
Guolin Ke's avatar
Guolin Ke committed
1598
        else:
1599
1600
            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
1601
1602

    def set_feature_name(self, feature_name):
1603
        """Set feature name.
Guolin Ke's avatar
Guolin Ke committed
1604
1605
1606

        Parameters
        ----------
1607
1608
        feature_name : list of strings
            Feature names.
Nikita Titov's avatar
Nikita Titov committed
1609
1610
1611
1612
1613

        Returns
        -------
        self : Dataset
            Dataset with set feature name.
Guolin Ke's avatar
Guolin Ke committed
1614
        """
1615
1616
        if feature_name != 'auto':
            self.feature_name = feature_name
1617
        if self.handle is not None and feature_name is not None and feature_name != 'auto':
wxchan's avatar
wxchan committed
1618
            if len(feature_name) != self.num_feature():
1619
1620
                raise ValueError("Length of feature_name({}) and num_feature({}) don't match"
                                 .format(len(feature_name), self.num_feature()))
1621
            c_feature_name = [c_str(name) for name in feature_name]
wxchan's avatar
wxchan committed
1622
1623
1624
            _safe_call(_LIB.LGBM_DatasetSetFeatureNames(
                self.handle,
                c_array(ctypes.c_char_p, c_feature_name),
Guolin Ke's avatar
Guolin Ke committed
1625
                ctypes.c_int(len(feature_name))))
Nikita Titov's avatar
Nikita Titov committed
1626
        return self
Guolin Ke's avatar
Guolin Ke committed
1627
1628

    def set_label(self, label):
1629
        """Set label of Dataset.
Guolin Ke's avatar
Guolin Ke committed
1630
1631
1632

        Parameters
        ----------
1633
        label : list, numpy 1-D array, pandas Series / one-column DataFrame or None
1634
            The label information to be set into Dataset.
Nikita Titov's avatar
Nikita Titov committed
1635
1636
1637
1638
1639

        Returns
        -------
        self : Dataset
            Dataset with set label.
Guolin Ke's avatar
Guolin Ke committed
1640
1641
        """
        self.label = label
1642
        if self.handle is not None:
1643
            label = list_to_1d_numpy(_label_from_pandas(label), name='label')
wxchan's avatar
wxchan committed
1644
            self.set_field('label', label)
1645
            self.label = self.get_field('label')  # original values can be modified at cpp side
Nikita Titov's avatar
Nikita Titov committed
1646
        return self
Guolin Ke's avatar
Guolin Ke committed
1647
1648

    def set_weight(self, weight):
1649
        """Set weight of each instance.
Guolin Ke's avatar
Guolin Ke committed
1650
1651
1652

        Parameters
        ----------
1653
        weight : list, numpy 1-D array, pandas Series or None
1654
            Weight to be set for each data point.
Nikita Titov's avatar
Nikita Titov committed
1655
1656
1657
1658
1659

        Returns
        -------
        self : Dataset
            Dataset with set weight.
Guolin Ke's avatar
Guolin Ke committed
1660
        """
1661
1662
        if weight is not None and np.all(weight == 1):
            weight = None
Guolin Ke's avatar
Guolin Ke committed
1663
        self.weight = weight
1664
        if self.handle is not None and weight is not None:
wxchan's avatar
wxchan committed
1665
1666
            weight = list_to_1d_numpy(weight, name='weight')
            self.set_field('weight', weight)
1667
            self.weight = self.get_field('weight')  # original values can be modified at cpp side
Nikita Titov's avatar
Nikita Titov committed
1668
        return self
Guolin Ke's avatar
Guolin Ke committed
1669
1670

    def set_init_score(self, init_score):
1671
        """Set init score of Booster to start from.
Guolin Ke's avatar
Guolin Ke committed
1672
1673
1674

        Parameters
        ----------
1675
        init_score : list, numpy 1-D array, pandas Series or None
1676
            Init score for Booster.
Nikita Titov's avatar
Nikita Titov committed
1677
1678
1679
1680
1681

        Returns
        -------
        self : Dataset
            Dataset with set init score.
Guolin Ke's avatar
Guolin Ke committed
1682
1683
        """
        self.init_score = init_score
1684
        if self.handle is not None and init_score is not None:
Guolin Ke's avatar
Guolin Ke committed
1685
            init_score = list_to_1d_numpy(init_score, np.float64, name='init_score')
wxchan's avatar
wxchan committed
1686
            self.set_field('init_score', init_score)
1687
            self.init_score = self.get_field('init_score')  # original values can be modified at cpp side
Nikita Titov's avatar
Nikita Titov committed
1688
        return self
Guolin Ke's avatar
Guolin Ke committed
1689
1690

    def set_group(self, group):
1691
        """Set group size of Dataset (used for ranking).
Guolin Ke's avatar
Guolin Ke committed
1692
1693
1694

        Parameters
        ----------
1695
        group : list, numpy 1-D array, pandas Series or None
1696
            Group size of each group.
Nikita Titov's avatar
Nikita Titov committed
1697
1698
1699
1700
1701

        Returns
        -------
        self : Dataset
            Dataset with set group.
Guolin Ke's avatar
Guolin Ke committed
1702
1703
        """
        self.group = group
1704
        if self.handle is not None and group is not None:
wxchan's avatar
wxchan committed
1705
1706
            group = list_to_1d_numpy(group, np.int32, name='group')
            self.set_field('group', group)
Nikita Titov's avatar
Nikita Titov committed
1707
        return self
Guolin Ke's avatar
Guolin Ke committed
1708

1709
1710
1711
1712
1713
1714
1715
1716
1717
1718
1719
1720
1721
1722
1723
1724
1725
1726
1727
1728
1729
1730
1731
1732
1733
1734
1735
1736
1737
1738
1739
1740
    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)
        string_buffers = [ctypes.create_string_buffer(reserved_string_buffer_size) for i in range_(num_feature)]
        ptr_string_buffers = (ctypes.c_char_p * num_feature)(*map(ctypes.addressof, string_buffers))
        _safe_call(_LIB.LGBM_DatasetGetFeatureNames(
            self.handle,
            num_feature,
            ctypes.byref(tmp_out_len),
            reserved_string_buffer_size,
            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)
            )
        return [string_buffers[i].value.decode('utf-8') for i in range_(num_feature)]

Guolin Ke's avatar
Guolin Ke committed
1741
    def get_label(self):
1742
        """Get the label of the Dataset.
Guolin Ke's avatar
Guolin Ke committed
1743
1744
1745

        Returns
        -------
Nikita Titov's avatar
Nikita Titov committed
1746
        label : numpy array or None
1747
            The label information from the Dataset.
Guolin Ke's avatar
Guolin Ke committed
1748
        """
1749
        if self.label is None:
wxchan's avatar
wxchan committed
1750
            self.label = self.get_field('label')
Guolin Ke's avatar
Guolin Ke committed
1751
1752
1753
        return self.label

    def get_weight(self):
1754
        """Get the weight of the Dataset.
Guolin Ke's avatar
Guolin Ke committed
1755
1756
1757

        Returns
        -------
Nikita Titov's avatar
Nikita Titov committed
1758
        weight : numpy array or None
1759
            Weight for each data point from the Dataset.
Guolin Ke's avatar
Guolin Ke committed
1760
        """
1761
        if self.weight is None:
wxchan's avatar
wxchan committed
1762
            self.weight = self.get_field('weight')
Guolin Ke's avatar
Guolin Ke committed
1763
1764
1765
        return self.weight

    def get_init_score(self):
1766
        """Get the initial score of the Dataset.
Guolin Ke's avatar
Guolin Ke committed
1767
1768
1769

        Returns
        -------
Nikita Titov's avatar
Nikita Titov committed
1770
        init_score : numpy array or None
1771
            Init score of Booster.
Guolin Ke's avatar
Guolin Ke committed
1772
        """
1773
        if self.init_score is None:
wxchan's avatar
wxchan committed
1774
            self.init_score = self.get_field('init_score')
Guolin Ke's avatar
Guolin Ke committed
1775
1776
        return self.init_score

1777
1778
1779
1780
1781
    def get_data(self):
        """Get the raw data of the Dataset.

        Returns
        -------
1782
        data : string, numpy array, pandas DataFrame, H2O DataTable's Frame, scipy.sparse, list of numpy arrays or None
1783
1784
1785
1786
            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
1787
1788
1789
1790
1791
1792
1793
1794
1795
1796
1797
1798
        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__))
1799
            self.need_slice = False
Guolin Ke's avatar
Guolin Ke committed
1800
1801
1802
        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.")
1803
1804
        return self.data

Guolin Ke's avatar
Guolin Ke committed
1805
    def get_group(self):
1806
        """Get the group of the Dataset.
Guolin Ke's avatar
Guolin Ke committed
1807
1808
1809

        Returns
        -------
Nikita Titov's avatar
Nikita Titov committed
1810
        group : numpy array or None
1811
            Group size of each group.
Guolin Ke's avatar
Guolin Ke committed
1812
        """
1813
        if self.group is None:
wxchan's avatar
wxchan committed
1814
            self.group = self.get_field('group')
Guolin Ke's avatar
Guolin Ke committed
1815
1816
            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
1817
                self.group = np.diff(self.group)
Guolin Ke's avatar
Guolin Ke committed
1818
1819
1820
        return self.group

    def num_data(self):
1821
        """Get the number of rows in the Dataset.
Guolin Ke's avatar
Guolin Ke committed
1822
1823
1824

        Returns
        -------
1825
1826
        number_of_rows : int
            The number of rows in the Dataset.
Guolin Ke's avatar
Guolin Ke committed
1827
        """
1828
        if self.handle is not None:
Guolin Ke's avatar
Guolin Ke committed
1829
            ret = ctypes.c_int()
wxchan's avatar
wxchan committed
1830
1831
1832
            _safe_call(_LIB.LGBM_DatasetGetNumData(self.handle,
                                                   ctypes.byref(ret)))
            return ret.value
Guolin Ke's avatar
Guolin Ke committed
1833
        else:
1834
            raise LightGBMError("Cannot get num_data before construct dataset")
Guolin Ke's avatar
Guolin Ke committed
1835
1836

    def num_feature(self):
1837
        """Get the number of columns (features) in the Dataset.
Guolin Ke's avatar
Guolin Ke committed
1838
1839
1840

        Returns
        -------
1841
1842
        number_of_columns : int
            The number of columns (features) in the Dataset.
Guolin Ke's avatar
Guolin Ke committed
1843
        """
1844
        if self.handle is not None:
Guolin Ke's avatar
Guolin Ke committed
1845
            ret = ctypes.c_int()
wxchan's avatar
wxchan committed
1846
1847
1848
            _safe_call(_LIB.LGBM_DatasetGetNumFeature(self.handle,
                                                      ctypes.byref(ret)))
            return ret.value
Guolin Ke's avatar
Guolin Ke committed
1849
        else:
1850
            raise LightGBMError("Cannot get num_feature before construct dataset")
Guolin Ke's avatar
Guolin Ke committed
1851

1852
    def get_ref_chain(self, ref_limit=100):
1853
1854
1855
1856
1857
        """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.
1858
1859
1860
1861
1862

        Parameters
        ----------
        ref_limit : int, optional (default=100)
            The limit number of references.
1863
1864
1865

        Returns
        -------
1866
1867
1868
        ref_chain : set of Dataset
            Chain of references of the Datasets.
        """
1869
        head = self
1870
        ref_chain = set()
1871
1872
        while len(ref_chain) < ref_limit:
            if isinstance(head, Dataset):
1873
                ref_chain.add(head)
1874
1875
1876
1877
1878
1879
                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
1880
        return ref_chain
1881

1882
1883
1884
1885
1886
1887
1888
1889
1890
1891
1892
1893
1894
1895
1896
1897
1898
1899
1900
1901
    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))
        return self

1902
    def _dump_text(self, filename):
1903
1904
1905
1906
1907
1908
1909
1910
1911
1912
1913
1914
1915
1916
1917
1918
1919
1920
1921
        """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
1922

wxchan's avatar
wxchan committed
1923
class Booster(object):
1924
    """Booster in LightGBM."""
1925

1926
    def __init__(self, params=None, train_set=None, model_file=None, model_str=None, silent=False):
1927
        """Initialize the Booster.
wxchan's avatar
wxchan committed
1928
1929
1930

        Parameters
        ----------
Nikita Titov's avatar
Nikita Titov committed
1931
        params : dict or None, optional (default=None)
1932
1933
1934
1935
            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
1936
            Path to the model file.
1937
1938
        model_str : string or None, optional (default=None)
            Model will be loaded from this string.
1939
1940
        silent : bool, optional (default=False)
            Whether to print messages during construction.
wxchan's avatar
wxchan committed
1941
        """
1942
        self.handle = None
1943
        self.network = False
wxchan's avatar
wxchan committed
1944
        self.__need_reload_eval_info = True
1945
        self._train_data_name = "training"
wxchan's avatar
wxchan committed
1946
        self.__attr = {}
1947
        self.__set_objective_to_none = False
wxchan's avatar
wxchan committed
1948
        self.best_iteration = -1
wxchan's avatar
wxchan committed
1949
        self.best_score = {}
1950
        params = {} if params is None else copy.deepcopy(params)
1951
        # user can set verbose with params, it has higher priority
1952
        if not any(verbose_alias in params for verbose_alias in _ConfigAliases.get("verbosity")) and silent:
1953
            params["verbose"] = -1
wxchan's avatar
wxchan committed
1954
        if train_set is not None:
1955
            # Training task
wxchan's avatar
wxchan committed
1956
            if not isinstance(train_set, Dataset):
1957
1958
                raise TypeError('Training data should be Dataset instance, met {}'
                                .format(type(train_set).__name__))
1959
            # set network if necessary
1960
            for alias in _ConfigAliases.get("machines"):
1961
1962
1963
1964
1965
1966
1967
1968
1969
1970
1971
1972
                if alias in params:
                    machines = params[alias]
                    if isinstance(machines, string_type):
                        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),
1973
                                     num_machines=params.setdefault("num_machines", num_machines))
1974
                    break
1975
            # construct booster object
1976
1977
1978
1979
            train_set.construct()
            # copy the parameters from train_set
            params.update(train_set.get_params())
            params_str = param_dict_to_str(params)
1980
            self.handle = ctypes.c_void_p()
wxchan's avatar
wxchan committed
1981
            _safe_call(_LIB.LGBM_BoosterCreate(
1982
                train_set.handle,
wxchan's avatar
wxchan committed
1983
1984
                c_str(params_str),
                ctypes.byref(self.handle)))
1985
            # save reference to data
wxchan's avatar
wxchan committed
1986
1987
1988
1989
            self.train_set = train_set
            self.valid_sets = []
            self.name_valid_sets = []
            self.__num_dataset = 1
Guolin Ke's avatar
Guolin Ke committed
1990
1991
            self.__init_predictor = train_set._predictor
            if self.__init_predictor is not None:
wxchan's avatar
wxchan committed
1992
1993
                _safe_call(_LIB.LGBM_BoosterMerge(
                    self.handle,
Guolin Ke's avatar
Guolin Ke committed
1994
                    self.__init_predictor.handle))
Guolin Ke's avatar
Guolin Ke committed
1995
            out_num_class = ctypes.c_int(0)
wxchan's avatar
wxchan committed
1996
1997
1998
1999
            _safe_call(_LIB.LGBM_BoosterGetNumClasses(
                self.handle,
                ctypes.byref(out_num_class)))
            self.__num_class = out_num_class.value
2000
            # buffer for inner predict
wxchan's avatar
wxchan committed
2001
2002
2003
            self.__inner_predict_buffer = [None]
            self.__is_predicted_cur_iter = [False]
            self.__get_eval_info()
2004
            self.pandas_categorical = train_set.pandas_categorical
2005
            self.train_set_version = train_set.version
wxchan's avatar
wxchan committed
2006
        elif model_file is not None:
2007
            # Prediction task
Guolin Ke's avatar
Guolin Ke committed
2008
            out_num_iterations = ctypes.c_int(0)
2009
            self.handle = ctypes.c_void_p()
wxchan's avatar
wxchan committed
2010
2011
2012
2013
            _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
2014
            out_num_class = ctypes.c_int(0)
wxchan's avatar
wxchan committed
2015
2016
2017
2018
            _safe_call(_LIB.LGBM_BoosterGetNumClasses(
                self.handle,
                ctypes.byref(out_num_class)))
            self.__num_class = out_num_class.value
2019
            self.pandas_categorical = _load_pandas_categorical(file_name=model_file)
2020
2021
        elif model_str is not None:
            self.model_from_string(model_str, not silent)
wxchan's avatar
wxchan committed
2022
        else:
2023
2024
            raise TypeError('Need at least one training dataset or model file or model string '
                            'to create Booster instance')
2025
        self.params = params
wxchan's avatar
wxchan committed
2026
2027

    def __del__(self):
2028
2029
2030
2031
2032
2033
2034
2035
2036
2037
        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
2038

wxchan's avatar
wxchan committed
2039
2040
2041
2042
    def __copy__(self):
        return self.__deepcopy__(None)

    def __deepcopy__(self, _):
2043
        model_str = self.model_to_string(num_iteration=-1)
2044
        booster = Booster(model_str=model_str)
2045
        return booster
wxchan's avatar
wxchan committed
2046
2047
2048
2049
2050
2051
2052

    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:
2053
            this["handle"] = self.model_to_string(num_iteration=-1)
wxchan's avatar
wxchan committed
2054
2055
2056
        return this

    def __setstate__(self, state):
2057
2058
        model_str = state.get('handle', None)
        if model_str is not None:
wxchan's avatar
wxchan committed
2059
            handle = ctypes.c_void_p()
Guolin Ke's avatar
Guolin Ke committed
2060
            out_num_iterations = ctypes.c_int(0)
2061
2062
2063
2064
            _safe_call(_LIB.LGBM_BoosterLoadModelFromString(
                c_str(model_str),
                ctypes.byref(out_num_iterations),
                ctypes.byref(handle)))
wxchan's avatar
wxchan committed
2065
2066
2067
            state['handle'] = handle
        self.__dict__.update(state)

wxchan's avatar
wxchan committed
2068
    def free_dataset(self):
Nikita Titov's avatar
Nikita Titov committed
2069
2070
2071
2072
2073
2074
2075
        """Free Booster's Datasets.

        Returns
        -------
        self : Booster
            Booster without Datasets.
        """
wxchan's avatar
wxchan committed
2076
2077
        self.__dict__.pop('train_set', None)
        self.__dict__.pop('valid_sets', None)
2078
        self.__num_dataset = 0
Nikita Titov's avatar
Nikita Titov committed
2079
        return self
wxchan's avatar
wxchan committed
2080

2081
2082
2083
    def _free_buffer(self):
        self.__inner_predict_buffer = []
        self.__is_predicted_cur_iter = []
Nikita Titov's avatar
Nikita Titov committed
2084
        return self
2085

2086
2087
2088
2089
2090
2091
    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
2092
        machines : list, set or string
2093
            Names of machines.
Nikita Titov's avatar
Nikita Titov committed
2094
        local_listen_port : int, optional (default=12400)
2095
            TCP listen port for local machines.
Nikita Titov's avatar
Nikita Titov committed
2096
        listen_time_out : int, optional (default=120)
2097
            Socket time-out in minutes.
Nikita Titov's avatar
Nikita Titov committed
2098
        num_machines : int, optional (default=1)
2099
            The number of machines for parallel learning application.
Nikita Titov's avatar
Nikita Titov committed
2100
2101
2102
2103
2104

        Returns
        -------
        self : Booster
            Booster with set network.
2105
2106
2107
2108
2109
2110
        """
        _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
2111
        return self
2112
2113

    def free_network(self):
Nikita Titov's avatar
Nikita Titov committed
2114
2115
2116
2117
2118
2119
2120
        """Free Booster's network.

        Returns
        -------
        self : Booster
            Booster with freed network.
        """
2121
2122
        _safe_call(_LIB.LGBM_NetworkFree())
        self.network = False
Nikita Titov's avatar
Nikita Titov committed
2123
        return self
2124

2125
2126
2127
2128
2129
2130
2131
2132
2133
2134
2135
2136
2137
2138
2139
2140
2141
2142
2143
2144
2145
2146
2147
2148
2149
2150
2151
2152
2153
2154
2155
2156
2157
2158
2159
2160
2161
2162
2163
    def trees_to_dataframe(self):
        """Parse the fitted model and return in an easy-to-read pandas DataFrame.

        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):
2164
                return set(tree.keys()) == {'leaf_value'}
2165
2166
2167
2168
2169
2170
2171
2172
2173
2174
2175
2176
2177
2178
2179
2180
2181
2182
2183
2184
2185
2186
2187
2188
2189
2190
2191
2192
2193
2194
2195
2196
2197
2198
2199
2200
2201
2202
2203
2204
2205
2206
2207
2208
2209
2210
2211
2212
2213
2214
2215
2216
2217
2218
2219
2220
2221
2222
2223
2224
2225
2226
2227
2228
2229
2230
2231
2232
2233
2234
2235
2236
2237
2238
2239

            # 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
2240
    def set_train_data_name(self, name):
2241
2242
2243
2244
        """Set the name to the training Dataset.

        Parameters
        ----------
Nikita Titov's avatar
Nikita Titov committed
2245
2246
2247
2248
2249
2250
2251
        name : string
            Name for the training Dataset.

        Returns
        -------
        self : Booster
            Booster with set training Dataset name.
2252
        """
2253
        self._train_data_name = name
Nikita Titov's avatar
Nikita Titov committed
2254
        return self
wxchan's avatar
wxchan committed
2255
2256

    def add_valid(self, data, name):
2257
        """Add validation data.
wxchan's avatar
wxchan committed
2258
2259
2260
2261

        Parameters
        ----------
        data : Dataset
2262
2263
2264
            Validation data.
        name : string
            Name of validation data.
Nikita Titov's avatar
Nikita Titov committed
2265
2266
2267
2268
2269

        Returns
        -------
        self : Booster
            Booster with set validation data.
wxchan's avatar
wxchan committed
2270
        """
Guolin Ke's avatar
Guolin Ke committed
2271
        if not isinstance(data, Dataset):
2272
2273
            raise TypeError('Validation data should be Dataset instance, met {}'
                            .format(type(data).__name__))
Guolin Ke's avatar
Guolin Ke committed
2274
        if data._predictor is not self.__init_predictor:
2275
2276
            raise LightGBMError("Add validation data failed, "
                                "you should use same predictor for these data")
wxchan's avatar
wxchan committed
2277
2278
        _safe_call(_LIB.LGBM_BoosterAddValidData(
            self.handle,
wxchan's avatar
wxchan committed
2279
            data.construct().handle))
wxchan's avatar
wxchan committed
2280
2281
2282
2283
2284
        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
2285
        return self
wxchan's avatar
wxchan committed
2286
2287

    def reset_parameter(self, params):
2288
        """Reset parameters of Booster.
wxchan's avatar
wxchan committed
2289
2290
2291
2292

        Parameters
        ----------
        params : dict
2293
            New parameters for Booster.
Nikita Titov's avatar
Nikita Titov committed
2294
2295
2296
2297
2298

        Returns
        -------
        self : Booster
            Booster with new parameters.
wxchan's avatar
wxchan committed
2299
2300
2301
2302
2303
2304
        """
        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
2305
        self.params.update(params)
Nikita Titov's avatar
Nikita Titov committed
2306
        return self
wxchan's avatar
wxchan committed
2307
2308

    def update(self, train_set=None, fobj=None):
Nikita Titov's avatar
Nikita Titov committed
2309
        """Update Booster for one iteration.
2310

wxchan's avatar
wxchan committed
2311
2312
        Parameters
        ----------
2313
2314
2315
2316
        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
2317
            Customized objective function.
2318
2319
2320
2321
2322
2323
2324
2325
2326
2327
2328
            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
2329

2330
            For binary task, the preds is probability of positive class (or margin in case of specified ``fobj``).
2331
2332
            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]
2333
2334
            and you should group grad and hess in this way as well.

wxchan's avatar
wxchan committed
2335
2336
        Returns
        -------
2337
2338
        is_finished : bool
            Whether the update was successfully finished.
wxchan's avatar
wxchan committed
2339
        """
2340
        # need reset training data
2341
2342
2343
2344
2345
2346
        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
2347
            if not isinstance(train_set, Dataset):
2348
2349
                raise TypeError('Training data should be Dataset instance, met {}'
                                .format(type(train_set).__name__))
Guolin Ke's avatar
Guolin Ke committed
2350
            if train_set._predictor is not self.__init_predictor:
2351
2352
                raise LightGBMError("Replace training data failed, "
                                    "you should use same predictor for these data")
wxchan's avatar
wxchan committed
2353
2354
2355
            self.train_set = train_set
            _safe_call(_LIB.LGBM_BoosterResetTrainingData(
                self.handle,
wxchan's avatar
wxchan committed
2356
                self.train_set.construct().handle))
wxchan's avatar
wxchan committed
2357
            self.__inner_predict_buffer[0] = None
2358
            self.train_set_version = self.train_set.version
wxchan's avatar
wxchan committed
2359
2360
        is_finished = ctypes.c_int(0)
        if fobj is None:
2361
            if self.__set_objective_to_none:
2362
                raise LightGBMError('Cannot update due to null objective function.')
wxchan's avatar
wxchan committed
2363
2364
2365
            _safe_call(_LIB.LGBM_BoosterUpdateOneIter(
                self.handle,
                ctypes.byref(is_finished)))
wxchan's avatar
wxchan committed
2366
            self.__is_predicted_cur_iter = [False for _ in range_(self.__num_dataset)]
wxchan's avatar
wxchan committed
2367
2368
            return is_finished.value == 1
        else:
2369
            if not self.__set_objective_to_none:
Nikita Titov's avatar
Nikita Titov committed
2370
                self.reset_parameter({"objective": "none"}).__set_objective_to_none = True
wxchan's avatar
wxchan committed
2371
2372
2373
2374
            grad, hess = fobj(self.__inner_predict(0), self.train_set)
            return self.__boost(grad, hess)

    def __boost(self, grad, hess):
2375
        """Boost Booster for one iteration with customized gradient statistics.
Nikita Titov's avatar
Nikita Titov committed
2376

Nikita Titov's avatar
Nikita Titov committed
2377
2378
        .. note::

2379
            For binary task, the score is probability of positive class (or margin in case of custom objective).
Nikita Titov's avatar
Nikita Titov committed
2380
2381
2382
            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.
2383

wxchan's avatar
wxchan committed
2384
2385
        Parameters
        ----------
2386
        grad : list or numpy 1-D array
Nikita Titov's avatar
Nikita Titov committed
2387
            The first order derivative (gradient).
2388
        hess : list or numpy 1-D array
Nikita Titov's avatar
Nikita Titov committed
2389
            The second order derivative (Hessian).
wxchan's avatar
wxchan committed
2390
2391
2392

        Returns
        -------
Nikita Titov's avatar
Nikita Titov committed
2393
2394
        is_finished : bool
            Whether the boost was successfully finished.
wxchan's avatar
wxchan committed
2395
        """
2396
2397
        grad = list_to_1d_numpy(grad, name='gradient')
        hess = list_to_1d_numpy(hess, name='hessian')
2398
2399
        assert grad.flags.c_contiguous
        assert hess.flags.c_contiguous
wxchan's avatar
wxchan committed
2400
        if len(grad) != len(hess):
2401
2402
            raise ValueError("Lengths of gradient({}) and hessian({}) don't match"
                             .format(len(grad), len(hess)))
wxchan's avatar
wxchan committed
2403
2404
2405
2406
2407
2408
        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)))
wxchan's avatar
wxchan committed
2409
        self.__is_predicted_cur_iter = [False for _ in range_(self.__num_dataset)]
wxchan's avatar
wxchan committed
2410
2411
2412
        return is_finished.value == 1

    def rollback_one_iter(self):
Nikita Titov's avatar
Nikita Titov committed
2413
2414
2415
2416
2417
2418
2419
        """Rollback one iteration.

        Returns
        -------
        self : Booster
            Booster with rolled back one iteration.
        """
wxchan's avatar
wxchan committed
2420
2421
        _safe_call(_LIB.LGBM_BoosterRollbackOneIter(
            self.handle))
wxchan's avatar
wxchan committed
2422
        self.__is_predicted_cur_iter = [False for _ in range_(self.__num_dataset)]
Nikita Titov's avatar
Nikita Titov committed
2423
        return self
wxchan's avatar
wxchan committed
2424
2425

    def current_iteration(self):
2426
2427
2428
2429
2430
2431
2432
        """Get the index of the current iteration.

        Returns
        -------
        cur_iter : int
            The index of the current iteration.
        """
Guolin Ke's avatar
Guolin Ke committed
2433
        out_cur_iter = ctypes.c_int(0)
wxchan's avatar
wxchan committed
2434
2435
2436
2437
2438
        _safe_call(_LIB.LGBM_BoosterGetCurrentIteration(
            self.handle,
            ctypes.byref(out_cur_iter)))
        return out_cur_iter.value

2439
2440
2441
2442
2443
2444
2445
2446
2447
2448
2449
2450
2451
2452
2453
2454
2455
2456
2457
2458
2459
2460
2461
2462
2463
2464
2465
2466
    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

2467
2468
2469
2470
2471
2472
2473
2474
2475
2476
2477
2478
2479
2480
2481
2482
2483
2484
2485
2486
2487
2488
2489
2490
2491
2492
2493
2494
    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
2495
    def eval(self, data, name, feval=None):
2496
        """Evaluate for data.
wxchan's avatar
wxchan committed
2497
2498
2499

        Parameters
        ----------
2500
2501
2502
2503
2504
        data : Dataset
            Data for the evaluating.
        name : string
            Name of the data.
        feval : callable or None, optional (default=None)
2505
            Customized evaluation function.
2506
            Should accept two parameters: preds, eval_data,
2507
            and return (eval_name, eval_result, is_higher_better) or list of such tuples.
2508
2509
2510
2511
2512
2513

                preds : list or numpy 1-D array
                    The predicted values.
                eval_data : Dataset
                    The evaluation dataset.
                eval_name : string
2514
                    The name of evaluation function (without whitespaces).
2515
2516
2517
2518
2519
                eval_result : float
                    The eval result.
                is_higher_better : bool
                    Is eval result higher better, e.g. AUC is ``is_higher_better``.

2520
            For binary task, the preds is probability of positive class (or margin in case of specified ``fobj``).
2521
2522
            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].
2523

wxchan's avatar
wxchan committed
2524
2525
        Returns
        -------
Nikita Titov's avatar
Nikita Titov committed
2526
        result : list
2527
            List with evaluation results.
wxchan's avatar
wxchan committed
2528
        """
Guolin Ke's avatar
Guolin Ke committed
2529
2530
        if not isinstance(data, Dataset):
            raise TypeError("Can only eval for Dataset instance")
wxchan's avatar
wxchan committed
2531
2532
2533
2534
        data_idx = -1
        if data is self.train_set:
            data_idx = 0
        else:
wxchan's avatar
wxchan committed
2535
            for i in range_(len(self.valid_sets)):
wxchan's avatar
wxchan committed
2536
2537
2538
                if data is self.valid_sets[i]:
                    data_idx = i + 1
                    break
2539
        # need to push new valid data
wxchan's avatar
wxchan committed
2540
2541
2542
2543
2544
2545
2546
        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):
2547
        """Evaluate for training data.
wxchan's avatar
wxchan committed
2548
2549
2550

        Parameters
        ----------
2551
        feval : callable or None, optional (default=None)
2552
            Customized evaluation function.
2553
2554
            Should accept two parameters: preds, train_data,
            and return (eval_name, eval_result, is_higher_better) or list of such tuples.
2555
2556
2557
2558
2559
2560

                preds : list or numpy 1-D array
                    The predicted values.
                train_data : Dataset
                    The training dataset.
                eval_name : string
2561
                    The name of evaluation function (without whitespaces).
2562
2563
2564
2565
2566
                eval_result : float
                    The eval result.
                is_higher_better : bool
                    Is eval result higher better, e.g. AUC is ``is_higher_better``.

2567
            For binary task, the preds is probability of positive class (or margin in case of specified ``fobj``).
2568
2569
            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
2570
2571
2572

        Returns
        -------
Nikita Titov's avatar
Nikita Titov committed
2573
        result : list
2574
            List with evaluation results.
wxchan's avatar
wxchan committed
2575
        """
2576
        return self.__inner_eval(self._train_data_name, 0, feval)
wxchan's avatar
wxchan committed
2577
2578

    def eval_valid(self, feval=None):
2579
        """Evaluate for validation data.
wxchan's avatar
wxchan committed
2580
2581
2582

        Parameters
        ----------
2583
        feval : callable or None, optional (default=None)
2584
            Customized evaluation function.
2585
            Should accept two parameters: preds, valid_data,
2586
            and return (eval_name, eval_result, is_higher_better) or list of such tuples.
2587
2588
2589
2590
2591
2592

                preds : list or numpy 1-D array
                    The predicted values.
                valid_data : Dataset
                    The validation dataset.
                eval_name : string
2593
                    The name of evaluation function (without whitespaces).
2594
2595
2596
2597
2598
                eval_result : float
                    The eval result.
                is_higher_better : bool
                    Is eval result higher better, e.g. AUC is ``is_higher_better``.

2599
            For binary task, the preds is probability of positive class (or margin in case of specified ``fobj``).
2600
2601
            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
2602
2603
2604

        Returns
        -------
Nikita Titov's avatar
Nikita Titov committed
2605
        result : list
2606
            List with evaluation results.
wxchan's avatar
wxchan committed
2607
        """
wxchan's avatar
wxchan committed
2608
        return [item for i in range_(1, self.__num_dataset)
wxchan's avatar
wxchan committed
2609
                for item in self.__inner_eval(self.name_valid_sets[i - 1], i, feval)]
wxchan's avatar
wxchan committed
2610

2611
    def save_model(self, filename, num_iteration=None, start_iteration=0, importance_type='split'):
2612
        """Save Booster to file.
wxchan's avatar
wxchan committed
2613
2614
2615

        Parameters
        ----------
2616
2617
        filename : string
            Filename to save Booster.
2618
2619
2620
2621
        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
2622
        start_iteration : int, optional (default=0)
2623
            Start index of the iteration that should be saved.
2624
2625
2626
2627
        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
2628
2629
2630
2631
2632

        Returns
        -------
        self : Booster
            Returns self.
wxchan's avatar
wxchan committed
2633
        """
2634
        if num_iteration is None:
2635
            num_iteration = self.best_iteration
2636
        importance_type_int = FEATURE_IMPORTANCE_TYPE_MAPPER[importance_type]
wxchan's avatar
wxchan committed
2637
2638
        _safe_call(_LIB.LGBM_BoosterSaveModel(
            self.handle,
2639
            ctypes.c_int(start_iteration),
Guolin Ke's avatar
Guolin Ke committed
2640
            ctypes.c_int(num_iteration),
2641
            ctypes.c_int(importance_type_int),
wxchan's avatar
wxchan committed
2642
            c_str(filename)))
2643
        _dump_pandas_categorical(self.pandas_categorical, filename)
Nikita Titov's avatar
Nikita Titov committed
2644
        return self
wxchan's avatar
wxchan committed
2645

2646
    def shuffle_models(self, start_iteration=0, end_iteration=-1):
2647
        """Shuffle models.
Nikita Titov's avatar
Nikita Titov committed
2648

2649
2650
2651
        Parameters
        ----------
        start_iteration : int, optional (default=0)
2652
            The first iteration that will be shuffled.
2653
2654
        end_iteration : int, optional (default=-1)
            The last iteration that will be shuffled.
2655
            If <= 0, means the last available iteration.
2656

Nikita Titov's avatar
Nikita Titov committed
2657
2658
2659
2660
        Returns
        -------
        self : Booster
            Booster with shuffled models.
2661
        """
2662
2663
        _safe_call(_LIB.LGBM_BoosterShuffleModels(
            self.handle,
Guolin Ke's avatar
Guolin Ke committed
2664
2665
            ctypes.c_int(start_iteration),
            ctypes.c_int(end_iteration)))
Nikita Titov's avatar
Nikita Titov committed
2666
        return self
2667
2668
2669
2670
2671
2672

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

        Parameters
        ----------
Nikita Titov's avatar
Nikita Titov committed
2673
        model_str : string
2674
            Model will be loaded from this string.
Nikita Titov's avatar
Nikita Titov committed
2675
2676
        verbose : bool, optional (default=True)
            Whether to print messages while loading model.
2677
2678
2679

        Returns
        -------
Nikita Titov's avatar
Nikita Titov committed
2680
        self : Booster
2681
2682
            Loaded Booster object.
        """
2683
2684
2685
2686
        if self.handle is not None:
            _safe_call(_LIB.LGBM_BoosterFree(self.handle))
        self._free_buffer()
        self.handle = ctypes.c_void_p()
2687
2688
2689
2690
2691
2692
2693
2694
2695
        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)))
2696
        if verbose:
Nikita Titov's avatar
Nikita Titov committed
2697
            print('Finished loading model, total used %d iterations' % int(out_num_iterations.value))
2698
        self.__num_class = out_num_class.value
2699
        self.pandas_categorical = _load_pandas_categorical(model_str=model_str)
2700
2701
        return self

2702
    def model_to_string(self, num_iteration=None, start_iteration=0, importance_type='split'):
2703
        """Save Booster to string.
2704

2705
2706
2707
2708
2709
2710
        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
2711
        start_iteration : int, optional (default=0)
2712
            Start index of the iteration that should be saved.
2713
2714
2715
2716
        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.
2717
2718
2719

        Returns
        -------
Nikita Titov's avatar
Nikita Titov committed
2720
        str_repr : string
2721
2722
            String representation of Booster.
        """
2723
        if num_iteration is None:
2724
            num_iteration = self.best_iteration
2725
        importance_type_int = FEATURE_IMPORTANCE_TYPE_MAPPER[importance_type]
2726
        buffer_len = 1 << 20
2727
        tmp_out_len = ctypes.c_int64(0)
2728
2729
2730
2731
        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,
2732
            ctypes.c_int(start_iteration),
2733
            ctypes.c_int(num_iteration),
2734
            ctypes.c_int(importance_type_int),
2735
            ctypes.c_int64(buffer_len),
2736
2737
2738
            ctypes.byref(tmp_out_len),
            ptr_string_buffer))
        actual_len = tmp_out_len.value
2739
        # if buffer length is not long enough, re-allocate a buffer
2740
2741
2742
2743
2744
        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,
2745
                ctypes.c_int(start_iteration),
2746
                ctypes.c_int(num_iteration),
2747
                ctypes.c_int(importance_type_int),
2748
                ctypes.c_int64(actual_len),
2749
2750
                ctypes.byref(tmp_out_len),
                ptr_string_buffer))
2751
        ret = string_buffer.value.decode('utf-8')
2752
2753
        ret += _dump_pandas_categorical(self.pandas_categorical)
        return ret
2754

2755
    def dump_model(self, num_iteration=None, start_iteration=0, importance_type='split'):
Nikita Titov's avatar
Nikita Titov committed
2756
        """Dump Booster to JSON format.
wxchan's avatar
wxchan committed
2757

2758
2759
        Parameters
        ----------
2760
2761
2762
2763
        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
2764
        start_iteration : int, optional (default=0)
2765
            Start index of the iteration that should be dumped.
2766
2767
2768
2769
        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.
2770

wxchan's avatar
wxchan committed
2771
2772
        Returns
        -------
2773
        json_repr : dict
Nikita Titov's avatar
Nikita Titov committed
2774
            JSON format of Booster.
wxchan's avatar
wxchan committed
2775
        """
2776
        if num_iteration is None:
2777
            num_iteration = self.best_iteration
2778
        importance_type_int = FEATURE_IMPORTANCE_TYPE_MAPPER[importance_type]
wxchan's avatar
wxchan committed
2779
        buffer_len = 1 << 20
2780
        tmp_out_len = ctypes.c_int64(0)
wxchan's avatar
wxchan committed
2781
2782
2783
2784
        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,
2785
            ctypes.c_int(start_iteration),
Guolin Ke's avatar
Guolin Ke committed
2786
            ctypes.c_int(num_iteration),
2787
            ctypes.c_int(importance_type_int),
2788
            ctypes.c_int64(buffer_len),
wxchan's avatar
wxchan committed
2789
            ctypes.byref(tmp_out_len),
Guolin Ke's avatar
Guolin Ke committed
2790
            ptr_string_buffer))
wxchan's avatar
wxchan committed
2791
        actual_len = tmp_out_len.value
2792
        # if buffer length is not long enough, reallocate a buffer
wxchan's avatar
wxchan committed
2793
2794
2795
2796
2797
        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,
2798
                ctypes.c_int(start_iteration),
Guolin Ke's avatar
Guolin Ke committed
2799
                ctypes.c_int(num_iteration),
2800
                ctypes.c_int(importance_type_int),
2801
                ctypes.c_int64(actual_len),
wxchan's avatar
wxchan committed
2802
                ctypes.byref(tmp_out_len),
Guolin Ke's avatar
Guolin Ke committed
2803
                ptr_string_buffer))
2804
        ret = json.loads(string_buffer.value.decode('utf-8'))
2805
2806
2807
        ret['pandas_categorical'] = json.loads(json.dumps(self.pandas_categorical,
                                                          default=json_default_with_numpy))
        return ret
wxchan's avatar
wxchan committed
2808

2809
2810
    def predict(self, data, num_iteration=None,
                raw_score=False, pred_leaf=False, pred_contrib=False,
2811
                data_has_header=False, is_reshape=True, **kwargs):
2812
        """Make a prediction.
wxchan's avatar
wxchan committed
2813
2814
2815

        Parameters
        ----------
2816
        data : string, numpy array, pandas DataFrame, H2O DataTable's Frame or scipy.sparse
2817
2818
            Data source for prediction.
            If string, it represents the path to txt file.
2819
2820
2821
2822
        num_iteration : int or None, optional (default=None)
            Limit number of iterations in the prediction.
            If None, if the best iteration exists, it is used; otherwise, all iterations are used.
            If <= 0, all iterations are used (no limits).
2823
2824
2825
2826
        raw_score : bool, optional (default=False)
            Whether to predict raw scores.
        pred_leaf : bool, optional (default=False)
            Whether to predict leaf index.
2827
2828
        pred_contrib : bool, optional (default=False)
            Whether to predict feature contributions.
2829

Nikita Titov's avatar
Nikita Titov committed
2830
2831
2832
2833
2834
2835
2836
            .. 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.
2837

2838
2839
2840
2841
2842
        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].
2843
2844
        **kwargs
            Other parameters for the prediction.
wxchan's avatar
wxchan committed
2845
2846
2847

        Returns
        -------
2848
        result : numpy array, scipy.sparse or list of scipy.sparse
2849
            Prediction result.
2850
            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
2851
        """
2852
        predictor = self._to_predictor(copy.deepcopy(kwargs))
2853
        if num_iteration is None:
2854
            num_iteration = self.best_iteration
2855
2856
2857
        return predictor.predict(data, num_iteration,
                                 raw_score, pred_leaf, pred_contrib,
                                 data_has_header, is_reshape)
wxchan's avatar
wxchan committed
2858

2859
    def refit(self, data, label, decay_rate=0.9, **kwargs):
Guolin Ke's avatar
Guolin Ke committed
2860
2861
2862
2863
        """Refit the existing Booster by new data.

        Parameters
        ----------
2864
        data : string, numpy array, pandas DataFrame, H2O DataTable's Frame or scipy.sparse
Guolin Ke's avatar
Guolin Ke committed
2865
2866
            Data source for refit.
            If string, it represents the path to txt file.
2867
        label : list, numpy 1-D array or pandas Series / one-column DataFrame
Guolin Ke's avatar
Guolin Ke committed
2868
2869
            Label for refit.
        decay_rate : float, optional (default=0.9)
2870
2871
            Decay rate of refit,
            will use ``leaf_output = decay_rate * old_leaf_output + (1.0 - decay_rate) * new_leaf_output`` to refit trees.
2872
2873
        **kwargs
            Other parameters for refit.
2874
            These parameters will be passed to ``predict`` method.
Guolin Ke's avatar
Guolin Ke committed
2875
2876
2877
2878
2879
2880

        Returns
        -------
        result : Booster
            Refitted Booster.
        """
2881
2882
        if self.__set_objective_to_none:
            raise LightGBMError('Cannot refit due to null objective function.')
2883
        predictor = self._to_predictor(copy.deepcopy(kwargs))
2884
        leaf_preds = predictor.predict(data, -1, pred_leaf=True)
2885
        nrow, ncol = leaf_preds.shape
2886
        train_set = Dataset(data, label, silent=True)
2887
2888
        new_params = copy.deepcopy(self.params)
        new_params['refit_decay_rate'] = decay_rate
2889
        new_booster = Booster(new_params, train_set)
Guolin Ke's avatar
Guolin Ke committed
2890
2891
2892
2893
2894
2895
2896
2897
2898
2899
2900
        # Copy models
        _safe_call(_LIB.LGBM_BoosterMerge(
            new_booster.handle,
            predictor.handle))
        leaf_preds = leaf_preds.reshape(-1)
        ptr_data, type_ptr_data, _ = c_int_array(leaf_preds)
        _safe_call(_LIB.LGBM_BoosterRefit(
            new_booster.handle,
            ptr_data,
            ctypes.c_int(nrow),
            ctypes.c_int(ncol)))
2901
2902
        new_booster.network = self.network
        new_booster.__attr = self.__attr.copy()
Guolin Ke's avatar
Guolin Ke committed
2903
2904
        return new_booster

2905
    def get_leaf_output(self, tree_id, leaf_id):
2906
2907
2908
2909
2910
2911
2912
2913
2914
2915
2916
2917
2918
2919
        """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.
        """
2920
2921
2922
2923
2924
2925
2926
2927
        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

2928
    def _to_predictor(self, pred_parameter=None):
2929
        """Convert to predictor."""
2930
        predictor = _InnerPredictor(booster_handle=self.handle, pred_parameter=pred_parameter)
2931
        predictor.pandas_categorical = self.pandas_categorical
wxchan's avatar
wxchan committed
2932
2933
        return predictor

2934
    def num_feature(self):
2935
2936
2937
2938
2939
2940
2941
        """Get number of features.

        Returns
        -------
        num_feature : int
            The number of features.
        """
2942
2943
2944
2945
2946
2947
        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
2948
    def feature_name(self):
2949
        """Get names of features.
wxchan's avatar
wxchan committed
2950
2951
2952

        Returns
        -------
2953
2954
        result : list
            List with names of features.
wxchan's avatar
wxchan committed
2955
        """
2956
        num_feature = self.num_feature()
2957
        # Get name of features
wxchan's avatar
wxchan committed
2958
        tmp_out_len = ctypes.c_int(0)
2959
2960
2961
        reserved_string_buffer_size = 255
        required_string_buffer_size = ctypes.c_size_t(0)
        string_buffers = [ctypes.create_string_buffer(reserved_string_buffer_size) for i in range_(num_feature)]
wxchan's avatar
wxchan committed
2962
2963
2964
        ptr_string_buffers = (ctypes.c_char_p * num_feature)(*map(ctypes.addressof, string_buffers))
        _safe_call(_LIB.LGBM_BoosterGetFeatureNames(
            self.handle,
2965
            num_feature,
wxchan's avatar
wxchan committed
2966
            ctypes.byref(tmp_out_len),
2967
2968
            reserved_string_buffer_size,
            ctypes.byref(required_string_buffer_size),
wxchan's avatar
wxchan committed
2969
2970
2971
            ptr_string_buffers))
        if num_feature != tmp_out_len.value:
            raise ValueError("Length of feature names doesn't equal with num_feature")
2972
2973
2974
2975
2976
        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)
            )
2977
        return [string_buffers[i].value.decode('utf-8') for i in range_(num_feature)]
wxchan's avatar
wxchan committed
2978

2979
    def feature_importance(self, importance_type='split', iteration=None):
2980
        """Get feature importances.
2981

2982
2983
        Parameters
        ----------
2984
2985
2986
2987
        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.
2988
2989
2990
2991
        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).
2992

2993
2994
        Returns
        -------
2995
2996
        result : numpy array
            Array with feature importances.
2997
        """
2998
2999
        if iteration is None:
            iteration = self.best_iteration
3000
        importance_type_int = FEATURE_IMPORTANCE_TYPE_MAPPER[importance_type]
Nikita Titov's avatar
Nikita Titov committed
3001
        result = np.zeros(self.num_feature(), dtype=np.float64)
3002
3003
3004
3005
3006
3007
        _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:
3008
            return result.astype(np.int32)
3009
3010
        else:
            return result
3011

3012
3013
3014
3015
3016
3017
3018
3019
3020
    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.
3021

Nikita Titov's avatar
Nikita Titov committed
3022
3023
3024
            .. warning::

                Categorical features are not supported.
3025

3026
3027
3028
3029
3030
3031
3032
3033
3034
3035
3036
3037
3038
3039
3040
3041
3042
3043
3044
3045
3046
3047
3048
3049
3050
3051
3052
        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
                if feature_names is not None and isinstance(feature, string_type):
                    split_feature = feature_names[root['split_feature']]
                else:
                    split_feature = root['split_feature']
                if split_feature == feature:
3053
3054
3055
3056
                    if isinstance(root['threshold'], string_type):
                        raise LightGBMError('Cannot compute split value histogram for the categorical feature')
                    else:
                        values.append(root['threshold'])
3057
3058
3059
3060
3061
3062
3063
3064
3065
3066
3067
3068
3069
3070
3071
3072
3073
3074
3075
3076
3077
3078
3079
3080
                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'])

        if bins is None or isinstance(bins, integer_types) and xgboost_style:
            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
3081
    def __inner_eval(self, data_name, data_idx, feval=None):
3082
        """Evaluate training or validation data."""
wxchan's avatar
wxchan committed
3083
        if data_idx >= self.__num_dataset:
3084
            raise ValueError("Data_idx should be smaller than number of dataset")
wxchan's avatar
wxchan committed
3085
3086
3087
        self.__get_eval_info()
        ret = []
        if self.__num_inner_eval > 0:
3088
            result = np.zeros(self.__num_inner_eval, dtype=np.float64)
Guolin Ke's avatar
Guolin Ke committed
3089
            tmp_out_len = ctypes.c_int(0)
wxchan's avatar
wxchan committed
3090
3091
            _safe_call(_LIB.LGBM_BoosterGetEval(
                self.handle,
Guolin Ke's avatar
Guolin Ke committed
3092
                ctypes.c_int(data_idx),
wxchan's avatar
wxchan committed
3093
                ctypes.byref(tmp_out_len),
Guolin Ke's avatar
Guolin Ke committed
3094
                result.ctypes.data_as(ctypes.POINTER(ctypes.c_double))))
wxchan's avatar
wxchan committed
3095
            if tmp_out_len.value != self.__num_inner_eval:
3096
                raise ValueError("Wrong length of eval results")
wxchan's avatar
wxchan committed
3097
            for i in range_(self.__num_inner_eval):
3098
3099
                ret.append((data_name, self.__name_inner_eval[i],
                            result[i], self.__higher_better_inner_eval[i]))
wxchan's avatar
wxchan committed
3100
3101
3102
3103
3104
3105
3106
3107
3108
3109
3110
3111
3112
3113
3114
        if feval is not None:
            if data_idx == 0:
                cur_data = self.train_set
            else:
                cur_data = self.valid_sets[data_idx - 1]
            feval_ret = feval(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
                ret.append((data_name, eval_name, val, is_higher_better))
        return ret

    def __inner_predict(self, data_idx):
3115
        """Predict for training and validation dataset."""
wxchan's avatar
wxchan committed
3116
        if data_idx >= self.__num_dataset:
3117
            raise ValueError("Data_idx should be smaller than number of dataset")
wxchan's avatar
wxchan committed
3118
3119
3120
3121
3122
        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
3123
            self.__inner_predict_buffer[data_idx] = np.zeros(n_preds, dtype=np.float64)
3124
        # avoid to predict many time in one iteration
wxchan's avatar
wxchan committed
3125
3126
        if not self.__is_predicted_cur_iter[data_idx]:
            tmp_out_len = ctypes.c_int64(0)
Guolin Ke's avatar
Guolin Ke committed
3127
            data_ptr = self.__inner_predict_buffer[data_idx].ctypes.data_as(ctypes.POINTER(ctypes.c_double))
wxchan's avatar
wxchan committed
3128
3129
            _safe_call(_LIB.LGBM_BoosterGetPredict(
                self.handle,
Guolin Ke's avatar
Guolin Ke committed
3130
                ctypes.c_int(data_idx),
wxchan's avatar
wxchan committed
3131
3132
3133
                ctypes.byref(tmp_out_len),
                data_ptr))
            if tmp_out_len.value != len(self.__inner_predict_buffer[data_idx]):
3134
                raise ValueError("Wrong length of predict results for data %d" % (data_idx))
wxchan's avatar
wxchan committed
3135
3136
3137
3138
            self.__is_predicted_cur_iter[data_idx] = True
        return self.__inner_predict_buffer[data_idx]

    def __get_eval_info(self):
3139
        """Get inner evaluation count and names."""
wxchan's avatar
wxchan committed
3140
3141
        if self.__need_reload_eval_info:
            self.__need_reload_eval_info = False
Guolin Ke's avatar
Guolin Ke committed
3142
            out_num_eval = ctypes.c_int(0)
3143
            # Get num of inner evals
wxchan's avatar
wxchan committed
3144
3145
3146
3147
3148
            _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:
3149
                # Get name of evals
Guolin Ke's avatar
Guolin Ke committed
3150
                tmp_out_len = ctypes.c_int(0)
3151
3152
3153
3154
3155
                reserved_string_buffer_size = 255
                required_string_buffer_size = ctypes.c_size_t(0)
                string_buffers = [
                    ctypes.create_string_buffer(reserved_string_buffer_size) for i in range_(self.__num_inner_eval)
                ]
wxchan's avatar
wxchan committed
3156
                ptr_string_buffers = (ctypes.c_char_p * self.__num_inner_eval)(*map(ctypes.addressof, string_buffers))
wxchan's avatar
wxchan committed
3157
3158
                _safe_call(_LIB.LGBM_BoosterGetEvalNames(
                    self.handle,
3159
                    self.__num_inner_eval,
wxchan's avatar
wxchan committed
3160
                    ctypes.byref(tmp_out_len),
3161
3162
                    reserved_string_buffer_size,
                    ctypes.byref(required_string_buffer_size),
wxchan's avatar
wxchan committed
3163
3164
                    ptr_string_buffers))
                if self.__num_inner_eval != tmp_out_len.value:
3165
                    raise ValueError("Length of eval names doesn't equal with num_evals")
3166
3167
3168
3169
3170
                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)
                    )
3171
                self.__name_inner_eval = \
3172
                    [string_buffers[i].value.decode('utf-8') for i in range_(self.__num_inner_eval)]
3173
                self.__higher_better_inner_eval = \
3174
                    [name.startswith(('auc', 'ndcg@', 'map@')) for name in self.__name_inner_eval]
3175

wxchan's avatar
wxchan committed
3176
    def attr(self, key):
3177
        """Get attribute string from the Booster.
wxchan's avatar
wxchan committed
3178
3179
3180

        Parameters
        ----------
3181
3182
        key : string
            The name of the attribute.
wxchan's avatar
wxchan committed
3183
3184
3185

        Returns
        -------
3186
3187
        value : string or None
            The attribute value.
Nikita Titov's avatar
Nikita Titov committed
3188
            Returns None if attribute does not exist.
wxchan's avatar
wxchan committed
3189
        """
3190
        return self.__attr.get(key, None)
wxchan's avatar
wxchan committed
3191
3192

    def set_attr(self, **kwargs):
3193
        """Set attributes to the Booster.
wxchan's avatar
wxchan committed
3194
3195
3196
3197

        Parameters
        ----------
        **kwargs
3198
3199
            The attributes to set.
            Setting a value to None deletes an attribute.
Nikita Titov's avatar
Nikita Titov committed
3200
3201
3202
3203

        Returns
        -------
        self : Booster
3204
            Booster with set attributes.
wxchan's avatar
wxchan committed
3205
3206
3207
        """
        for key, value in kwargs.items():
            if value is not None:
wxchan's avatar
wxchan committed
3208
                if not isinstance(value, string_type):
Nikita Titov's avatar
Nikita Titov committed
3209
                    raise ValueError("Only string values are accepted")
wxchan's avatar
wxchan committed
3210
3211
3212
                self.__attr[key] = value
            else:
                self.__attr.pop(key, None)
Nikita Titov's avatar
Nikita Titov committed
3213
        return self