feature importance type in saved model file (#3220)

* feature importance type in saved model file

* fix nullptr

* fixed formatting

* fix python/R

* Update src/c_api.cpp

* Apply suggestions from code review
Co-authored-by: James Lamb <jaylamb20@gmail.com>

* fix c_api test

* fix swig

* minor docs improvements and added defines for importance types
Co-authored-by: StrikerRUS <nekit94-12@hotmail.com>
Co-authored-by: James Lamb <jaylamb20@gmail.com>

R-package/R/lgb.Booster.R

@@ -424,7 +424,7 @@ Booster <- R6::R6Class(
     },
 
     # Save model
-    save_model = function(filename, num_iteration = NULL) {
+    save_model = function(filename, num_iteration = NULL, feature_importance_type = 0L) {
 
       # Check if number of iteration is non existent
       if (is.null(num_iteration)) {
@@ -437,6 +437,7 @@ Booster <- R6::R6Class(
         , ret = NULL
         , private$handle
         , as.integer(num_iteration)
+        , as.integer(feature_importance_type)
         , lgb.c_str(filename)
       )
 
@@ -445,7 +446,7 @@ Booster <- R6::R6Class(
     },
 
     # Save model to string
-    save_model_to_string = function(num_iteration = NULL) {
+    save_model_to_string = function(num_iteration = NULL, feature_importance_type = 0L) {
 
       # Check if number of iteration is non existent
       if (is.null(num_iteration)) {
@@ -457,12 +458,13 @@ Booster <- R6::R6Class(
         "LGBM_BoosterSaveModelToString_R"
         , private$handle
         , as.integer(num_iteration)
+        , as.integer(feature_importance_type)
       ))
 
     },
 
     # Dump model in memory
-    dump_model = function(num_iteration = NULL) {
+    dump_model = function(num_iteration = NULL, feature_importance_type = 0L) {
 
       # Check if number of iteration is non existent
       if (is.null(num_iteration)) {
@@ -474,6 +476,7 @@ Booster <- R6::R6Class(
         "LGBM_BoosterDumpModel_R"
         , private$handle
         , as.integer(num_iteration)
+        , as.integer(feature_importance_type)
       )
 
     },

R-package/src/lightgbm_R.cpp

@@ -632,15 +632,17 @@ LGBM_SE LGBM_BoosterPredictForMat_R(LGBM_SE handle,
 
 LGBM_SE LGBM_BoosterSaveModel_R(LGBM_SE handle,
   LGBM_SE num_iteration,
+  LGBM_SE feature_importance_type,
   LGBM_SE filename,
   LGBM_SE call_state) {
   R_API_BEGIN();
-  CHECK_CALL(LGBM_BoosterSaveModel(R_GET_PTR(handle), 0, R_AS_INT(num_iteration), R_CHAR_PTR(filename)));
+  CHECK_CALL(LGBM_BoosterSaveModel(R_GET_PTR(handle), 0, R_AS_INT(num_iteration), R_AS_INT(feature_importance_type), R_CHAR_PTR(filename)));
   R_API_END();
 }
 
 LGBM_SE LGBM_BoosterSaveModelToString_R(LGBM_SE handle,
   LGBM_SE num_iteration,
+  LGBM_SE feature_importance_type,
   LGBM_SE buffer_len,
   LGBM_SE actual_len,
   LGBM_SE out_str,
@@ -648,13 +650,14 @@ LGBM_SE LGBM_BoosterSaveModelToString_R(LGBM_SE handle,
   R_API_BEGIN();
   int64_t out_len = 0;
   std::vector<char> inner_char_buf(R_AS_INT(buffer_len));
-  CHECK_CALL(LGBM_BoosterSaveModelToString(R_GET_PTR(handle), 0, R_AS_INT(num_iteration), R_AS_INT(buffer_len), &out_len, inner_char_buf.data()));
+  CHECK_CALL(LGBM_BoosterSaveModelToString(R_GET_PTR(handle), 0, R_AS_INT(num_iteration), R_AS_INT(feature_importance_type), R_AS_INT(buffer_len), &out_len, inner_char_buf.data()));
   EncodeChar(out_str, inner_char_buf.data(), buffer_len, actual_len, static_cast<size_t>(out_len));
   R_API_END();
 }
 
 LGBM_SE LGBM_BoosterDumpModel_R(LGBM_SE handle,
   LGBM_SE num_iteration,
+  LGBM_SE feature_importance_type,
   LGBM_SE buffer_len,
   LGBM_SE actual_len,
   LGBM_SE out_str,
@@ -662,7 +665,7 @@ LGBM_SE LGBM_BoosterDumpModel_R(LGBM_SE handle,
   R_API_BEGIN();
   int64_t out_len = 0;
   std::vector<char> inner_char_buf(R_AS_INT(buffer_len));
-  CHECK_CALL(LGBM_BoosterDumpModel(R_GET_PTR(handle), 0, R_AS_INT(num_iteration), R_AS_INT(buffer_len), &out_len, inner_char_buf.data()));
+  CHECK_CALL(LGBM_BoosterDumpModel(R_GET_PTR(handle), 0, R_AS_INT(num_iteration), R_AS_INT(feature_importance_type), R_AS_INT(buffer_len), &out_len, inner_char_buf.data()));
   EncodeChar(out_str, inner_char_buf.data(), buffer_len, actual_len, static_cast<size_t>(out_len));
   R_API_END();
 }
@@ -707,9 +710,9 @@ static const R_CallMethodDef CallEntries[] = {
   {"LGBM_BoosterCalcNumPredict_R"     , (DL_FUNC) &LGBM_BoosterCalcNumPredict_R     , 8},
   {"LGBM_BoosterPredictForCSC_R"      , (DL_FUNC) &LGBM_BoosterPredictForCSC_R      , 14},
   {"LGBM_BoosterPredictForMat_R"      , (DL_FUNC) &LGBM_BoosterPredictForMat_R      , 11},
-  {"LGBM_BoosterSaveModel_R"          , (DL_FUNC) &LGBM_BoosterSaveModel_R          , 4},
-  {"LGBM_BoosterSaveModelToString_R"  , (DL_FUNC) &LGBM_BoosterSaveModelToString_R  , 6},
-  {"LGBM_BoosterDumpModel_R"          , (DL_FUNC) &LGBM_BoosterDumpModel_R          , 6},
+  {"LGBM_BoosterSaveModel_R"          , (DL_FUNC) &LGBM_BoosterSaveModel_R          , 5},
+  {"LGBM_BoosterSaveModelToString_R"  , (DL_FUNC) &LGBM_BoosterSaveModelToString_R  , 7},
+  {"LGBM_BoosterDumpModel_R"          , (DL_FUNC) &LGBM_BoosterDumpModel_R          , 7},
   {NULL, NULL, 0}
 };
 

R-package/src/lightgbm_R.h

@@ -590,6 +590,7 @@ LIGHTGBM_C_EXPORT LGBM_SE LGBM_BoosterPredictForMat_R(
 LIGHTGBM_C_EXPORT LGBM_SE LGBM_BoosterSaveModel_R(
   LGBM_SE handle,
   LGBM_SE num_iteration,
+  LGBM_SE feature_importance_type,
   LGBM_SE filename,
   LGBM_SE call_state
 );
@@ -604,6 +605,7 @@ LIGHTGBM_C_EXPORT LGBM_SE LGBM_BoosterSaveModel_R(
 LIGHTGBM_C_EXPORT LGBM_SE LGBM_BoosterSaveModelToString_R(
   LGBM_SE handle,
   LGBM_SE num_iteration,
+  LGBM_SE feature_importance_type,
   LGBM_SE buffer_len,
   LGBM_SE actual_len,
   LGBM_SE out_str,
@@ -620,6 +622,7 @@ LIGHTGBM_C_EXPORT LGBM_SE LGBM_BoosterSaveModelToString_R(
 LIGHTGBM_C_EXPORT LGBM_SE LGBM_BoosterDumpModel_R(
   LGBM_SE handle,
   LGBM_SE num_iteration,
+  LGBM_SE feature_importance_type,
   LGBM_SE buffer_len,
   LGBM_SE actual_len,
   LGBM_SE out_str,

docs/Parameters.rst

@@ -574,6 +574,14 @@ Learning Control Parameters
 
    -  **Note**: can be used only in CLI version
 
+-  ``saved_feature_importance_type`` :raw-html:`<a id="saved_feature_importance_type" title="Permalink to this parameter" href="#saved_feature_importance_type">&#x1F517;&#xFE0E;</a>`, default = ``0``, type = int
+
+   -  the feature importance type in the saved model file
+
+   -  ``0``: count-based feature importance (numbers of splits are counted); ``1``: gain-based feature importance (values of gain are counted)
+
+   -  **Note**: can be used only in CLI version
+
 -  ``snapshot_freq`` :raw-html:`<a id="snapshot_freq" title="Permalink to this parameter" href="#snapshot_freq">&#x1F517;&#xFE0E;</a>`, default = ``-1``, type = int, aliases: ``save_period``
 
    -  frequency of saving model file snapshot

include/LightGBM/boosting.h

@@ -176,9 +176,10 @@ class LIGHTGBM_EXPORT Boosting {
   * \brief Dump model to json format string
   * \param start_iteration The model will be saved start from
   * \param num_iteration Number of iterations that want to dump, -1 means dump all
+  * \param feature_importance_type Type of feature importance, 0: split, 1: gain
   * \return Json format string of model
   */
-  virtual std::string DumpModel(int start_iteration, int num_iteration) const = 0;
+  virtual std::string DumpModel(int start_iteration, int num_iteration, int feature_importance_type) const = 0;
 
   /*!
   * \brief Translate model to if-else statement
@@ -199,19 +200,20 @@ class LIGHTGBM_EXPORT Boosting {
   * \brief Save model to file
   * \param start_iteration The model will be saved start from
   * \param num_iterations Number of model that want to save, -1 means save all
-  * \param is_finish Is training finished or not
+  * \param feature_importance_type Type of feature importance, 0: split, 1: gain
   * \param filename Filename that want to save to
   * \return true if succeeded
   */
-  virtual bool SaveModelToFile(int start_iteration, int num_iterations, const char* filename) const = 0;
+  virtual bool SaveModelToFile(int start_iteration, int num_iterations, int feature_importance_type, const char* filename) const = 0;
 
   /*!
   * \brief Save model to string
   * \param start_iteration The model will be saved start from
   * \param num_iterations Number of model that want to save, -1 means save all
+  * \param feature_importance_type Type of feature importance, 0: split, 1: gain
   * \return Non-empty string if succeeded
   */
-  virtual std::string SaveModelToString(int start_iteration, int num_iterations) const = 0;
+  virtual std::string SaveModelToString(int start_iteration, int num_iterations, int feature_importance_type) const = 0;
 
   /*!
   * \brief Restore from a serialized string

include/LightGBM/c_api.h

@@ -36,6 +36,9 @@ typedef void* BoosterHandle;  /*!< \brief Handle of booster. */
 #define C_API_MATRIX_TYPE_CSR (0)  /*!< \brief CSR sparse matrix type. */
 #define C_API_MATRIX_TYPE_CSC (1)  /*!< \brief CSC sparse matrix type. */
 
+#define C_API_FEATURE_IMPORTANCE_SPLIT (0)  /*!< \brief Split type of feature importance. */
+#define C_API_FEATURE_IMPORTANCE_GAIN  (1)  /*!< \brief Gain type of feature importance. */
+
 /*!
  * \brief Get string message of the last error.
  * \return Error information
@@ -996,12 +999,14 @@ LIGHTGBM_C_EXPORT int LGBM_BoosterPredictForMats(BoosterHandle handle,
  * \param handle Handle of booster
  * \param start_iteration Start index of the iteration that should be saved
  * \param num_iteration Index of the iteration that should be saved, <= 0 means save all
+ * \param feature_importance_type Type of feature importance, can be ``C_API_FEATURE_IMPORTANCE_SPLIT`` or ``C_API_FEATURE_IMPORTANCE_GAIN``
  * \param filename The name of the file
  * \return 0 when succeed, -1 when failure happens
  */
 LIGHTGBM_C_EXPORT int LGBM_BoosterSaveModel(BoosterHandle handle,
                                             int start_iteration,
                                             int num_iteration,
+                                            int feature_importance_type,
                                             const char* filename);
 
 /*!
@@ -1009,6 +1014,7 @@ LIGHTGBM_C_EXPORT int LGBM_BoosterSaveModel(BoosterHandle handle,
  * \param handle Handle of booster
  * \param start_iteration Start index of the iteration that should be saved
  * \param num_iteration Index of the iteration that should be saved, <= 0 means save all
+ * \param feature_importance_type Type of feature importance, can be ``C_API_FEATURE_IMPORTANCE_SPLIT`` or ``C_API_FEATURE_IMPORTANCE_GAIN``
  * \param buffer_len String buffer length, if ``buffer_len < out_len``, you should re-allocate buffer
  * \param[out] out_len Actual output length
  * \param[out] out_str String of model, should pre-allocate memory
@@ -1017,6 +1023,7 @@ LIGHTGBM_C_EXPORT int LGBM_BoosterSaveModel(BoosterHandle handle,
 LIGHTGBM_C_EXPORT int LGBM_BoosterSaveModelToString(BoosterHandle handle,
                                                     int start_iteration,
                                                     int num_iteration,
+                                                    int feature_importance_type,
                                                     int64_t buffer_len,
                                                     int64_t* out_len,
                                                     char* out_str);
@@ -1026,6 +1033,7 @@ LIGHTGBM_C_EXPORT int LGBM_BoosterSaveModelToString(BoosterHandle handle,
  * \param handle Handle of booster
  * \param start_iteration Start index of the iteration that should be dumped
  * \param num_iteration Index of the iteration that should be dumped, <= 0 means dump all
+ * \param feature_importance_type Type of feature importance, can be ``C_API_FEATURE_IMPORTANCE_SPLIT`` or ``C_API_FEATURE_IMPORTANCE_GAIN``
  * \param buffer_len String buffer length, if ``buffer_len < out_len``, you should re-allocate buffer
  * \param[out] out_len Actual output length
  * \param[out] out_str JSON format string of model, should pre-allocate memory
@@ -1034,6 +1042,7 @@ LIGHTGBM_C_EXPORT int LGBM_BoosterSaveModelToString(BoosterHandle handle,
 LIGHTGBM_C_EXPORT int LGBM_BoosterDumpModel(BoosterHandle handle,
                                             int start_iteration,
                                             int num_iteration,
+                                            int feature_importance_type,
                                             int64_t buffer_len,
                                             int64_t* out_len,
                                             char* out_str);
@@ -1069,8 +1078,8 @@ LIGHTGBM_C_EXPORT int LGBM_BoosterSetLeafValue(BoosterHandle handle,
  * \param handle Handle of booster
  * \param num_iteration Number of iterations for which feature importance is calculated, <= 0 means use all
  * \param importance_type Method of importance calculation:
- *   - 0 for split, result contains numbers of times the feature is used in a model;
- *   - 1 for gain, result contains total gains of splits which use the feature
+ *   - ``C_API_FEATURE_IMPORTANCE_SPLIT``: result contains numbers of times the feature is used in a model;
+ *   - ``C_API_FEATURE_IMPORTANCE_GAIN``: result contains total gains of splits which use the feature
  * \param[out] out_results Result array with feature importance
  * \return 0 when succeed, -1 when failure happens
  */

include/LightGBM/config.h

@@ -532,6 +532,11 @@ struct Config {
   // desc = **Note**: can be used only in CLI version
   std::string output_model = "LightGBM_model.txt";
 
+  // desc = the feature importance type in the saved model file
+  // desc = ``0``: count-based feature importance (numbers of splits are counted); ``1``: gain-based feature importance (values of gain are counted)
+  // desc = **Note**: can be used only in CLI version
+  int saved_feature_importance_type = 0;
+
   // [no-save]
   // alias = save_period
   // desc = frequency of saving model file snapshot

python-package/lightgbm/basic.py

@@ -284,12 +284,20 @@ C_API_PREDICT_CONTRIB = 3
 C_API_MATRIX_TYPE_CSR = 0
 C_API_MATRIX_TYPE_CSC = 1
 
+"""Macro definition of feature importance type"""
+C_API_FEATURE_IMPORTANCE_SPLIT = 0
+C_API_FEATURE_IMPORTANCE_GAIN = 1
+
 """Data type of data field"""
 FIELD_TYPE_MAPPER = {"label": C_API_DTYPE_FLOAT32,
                      "weight": C_API_DTYPE_FLOAT32,
                      "init_score": C_API_DTYPE_FLOAT64,
                      "group": C_API_DTYPE_INT32}
 
+"""String name to int feature importance type mapper"""
+FEATURE_IMPORTANCE_TYPE_MAPPER = {"split": C_API_FEATURE_IMPORTANCE_SPLIT,
+                                  "gain": C_API_FEATURE_IMPORTANCE_GAIN}
+
 
 def convert_from_sliced_object(data):
     """Fix the memory of multi-dimensional sliced object."""
@@ -2600,7 +2608,7 @@ class Booster(object):
         return [item for i in range_(1, self.__num_dataset)
                 for item in self.__inner_eval(self.name_valid_sets[i - 1], i, feval)]
 
-    def save_model(self, filename, num_iteration=None, start_iteration=0):
+    def save_model(self, filename, num_iteration=None, start_iteration=0, importance_type='split'):
         """Save Booster to file.
 
         Parameters
@@ -2613,6 +2621,10 @@ class Booster(object):
             If <= 0, all iterations are saved.
         start_iteration : int, optional (default=0)
             Start index of the iteration that should be saved.
+        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.
 
         Returns
         -------
@@ -2621,10 +2633,12 @@ class Booster(object):
         """
         if num_iteration is None:
             num_iteration = self.best_iteration
+        importance_type_int = FEATURE_IMPORTANCE_TYPE_MAPPER[importance_type]
         _safe_call(_LIB.LGBM_BoosterSaveModel(
             self.handle,
             ctypes.c_int(start_iteration),
             ctypes.c_int(num_iteration),
+            ctypes.c_int(importance_type_int),
             c_str(filename)))
         _dump_pandas_categorical(self.pandas_categorical, filename)
         return self
@@ -2685,7 +2699,7 @@ class Booster(object):
         self.pandas_categorical = _load_pandas_categorical(model_str=model_str)
         return self
 
-    def model_to_string(self, num_iteration=None, start_iteration=0):
+    def model_to_string(self, num_iteration=None, start_iteration=0, importance_type='split'):
         """Save Booster to string.
 
         Parameters
@@ -2696,6 +2710,10 @@ class Booster(object):
             If <= 0, all iterations are saved.
         start_iteration : int, optional (default=0)
             Start index of the iteration that should be saved.
+        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.
 
         Returns
         -------
@@ -2704,6 +2722,7 @@ class Booster(object):
         """
         if num_iteration is None:
             num_iteration = self.best_iteration
+        importance_type_int = FEATURE_IMPORTANCE_TYPE_MAPPER[importance_type]
         buffer_len = 1 << 20
         tmp_out_len = ctypes.c_int64(0)
         string_buffer = ctypes.create_string_buffer(buffer_len)
@@ -2712,6 +2731,7 @@ class Booster(object):
             self.handle,
             ctypes.c_int(start_iteration),
             ctypes.c_int(num_iteration),
+            ctypes.c_int(importance_type_int),
             ctypes.c_int64(buffer_len),
             ctypes.byref(tmp_out_len),
             ptr_string_buffer))
@@ -2724,6 +2744,7 @@ class Booster(object):
                 self.handle,
                 ctypes.c_int(start_iteration),
                 ctypes.c_int(num_iteration),
+                ctypes.c_int(importance_type_int),
                 ctypes.c_int64(actual_len),
                 ctypes.byref(tmp_out_len),
                 ptr_string_buffer))
@@ -2731,7 +2752,7 @@ class Booster(object):
         ret += _dump_pandas_categorical(self.pandas_categorical)
         return ret
 
-    def dump_model(self, num_iteration=None, start_iteration=0):
+    def dump_model(self, num_iteration=None, start_iteration=0, importance_type='split'):
         """Dump Booster to JSON format.
 
         Parameters
@@ -2742,6 +2763,10 @@ class Booster(object):
             If <= 0, all iterations are dumped.
         start_iteration : int, optional (default=0)
             Start index of the iteration that should be dumped.
+        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.
 
         Returns
         -------
@@ -2750,6 +2775,7 @@ class Booster(object):
         """
         if num_iteration is None:
             num_iteration = self.best_iteration
+        importance_type_int = FEATURE_IMPORTANCE_TYPE_MAPPER[importance_type]
         buffer_len = 1 << 20
         tmp_out_len = ctypes.c_int64(0)
         string_buffer = ctypes.create_string_buffer(buffer_len)
@@ -2758,6 +2784,7 @@ class Booster(object):
             self.handle,
             ctypes.c_int(start_iteration),
             ctypes.c_int(num_iteration),
+            ctypes.c_int(importance_type_int),
             ctypes.c_int64(buffer_len),
             ctypes.byref(tmp_out_len),
             ptr_string_buffer))
@@ -2770,6 +2797,7 @@ class Booster(object):
                 self.handle,
                 ctypes.c_int(start_iteration),
                 ctypes.c_int(num_iteration),
+                ctypes.c_int(importance_type_int),
                 ctypes.c_int64(actual_len),
                 ctypes.byref(tmp_out_len),
                 ptr_string_buffer))
@@ -2969,12 +2997,7 @@ class Booster(object):
         """
         if iteration is None:
             iteration = self.best_iteration
-        if importance_type == "split":
-            importance_type_int = 0
-        elif importance_type == "gain":
-            importance_type_int = 1
-        else:
-            importance_type_int = -1
+        importance_type_int = FEATURE_IMPORTANCE_TYPE_MAPPER[importance_type]
         result = np.zeros(self.num_feature(), dtype=np.float64)
         _safe_call(_LIB.LGBM_BoosterFeatureImportance(
             self.handle,

src/application/application.cpp

@@ -201,7 +201,8 @@ void Application::InitTrain() {
 void Application::Train() {
   Log::Info("Started training...");
   boosting_->Train(config_.snapshot_freq, config_.output_model);
-  boosting_->SaveModelToFile(0, -1, config_.output_model.c_str());
+  boosting_->SaveModelToFile(0, -1, config_.saved_feature_importance_type,
+                             config_.output_model.c_str());
   // convert model to if-else statement code
   if (config_.convert_model_language == std::string("cpp")) {
     boosting_->SaveModelToIfElse(-1, config_.convert_model.c_str());
@@ -233,7 +234,8 @@ void Application::Predict() {
     boosting_->Init(&config_, train_data_.get(), objective_fun_.get(),
                     Common::ConstPtrInVectorWrapper<Metric>(train_metric_));
     boosting_->RefitTree(pred_leaf);
-    boosting_->SaveModelToFile(0, -1, config_.output_model.c_str());
+    boosting_->SaveModelToFile(0, -1, config_.saved_feature_importance_type,
+                               config_.output_model.c_str());
     Log::Info("Finished RefitTree");
   } else {
     // create predictor

src/boosting/gbdt.cpp

@@ -258,7 +258,7 @@ void GBDT::Train(int snapshot_freq, const std::string& model_output_path) {
     if (snapshot_freq > 0
         && (iter + 1) % snapshot_freq == 0) {
       std::string snapshot_out = model_output_path + ".snapshot_iter_" + std::to_string(iter + 1);
-      SaveModelToFile(0, -1, snapshot_out.c_str());
+      SaveModelToFile(0, -1, config_->saved_feature_importance_type, snapshot_out.c_str());
     }
   }
 }

src/boosting/gbdt.h

@@ -249,9 +249,11 @@ class GBDT : public GBDTBase {
   * \brief Dump model to json format string
   * \param start_iteration The model will be saved start from
   * \param num_iteration Number of iterations that want to dump, -1 means dump all
+  * \param feature_importance_type Type of feature importance, 0: split, 1: gain
   * \return Json format string of model
   */
-  std::string DumpModel(int start_iteration, int num_iteration) const override;
+  std::string DumpModel(int start_iteration, int num_iteration,
+                        int feature_importance_type) const override;
 
   /*!
   * \brief Translate model to if-else statement
@@ -272,18 +274,22 @@ class GBDT : public GBDTBase {
   * \brief Save model to file
   * \param start_iteration The model will be saved start from
   * \param num_iterations Number of model that want to save, -1 means save all
+  * \param feature_importance_type Type of feature importance, 0: split, 1: gain
   * \param filename Filename that want to save to
   * \return is_finish Is training finished or not
   */
-  bool SaveModelToFile(int start_iteration, int num_iterations, const char* filename) const override;
+  bool SaveModelToFile(int start_iteration, int num_iterations,
+                       int feature_importance_type,
+                       const char* filename) const override;
 
   /*!
   * \brief Save model to string
   * \param start_iteration The model will be saved start from
   * \param num_iterations Number of model that want to save, -1 means save all
+  * \param feature_importance_type Type of feature importance, 0: split, 1: gain
   * \return Non-empty string if succeeded
   */
-  std::string SaveModelToString(int start_iteration, int num_iterations) const override;
+  std::string SaveModelToString(int start_iteration, int num_iterations, int feature_importance_type) const override;
 
   /*!
   * \brief Restore from a serialized buffer

src/boosting/gbdt_model_text.cpp

@@ -18,7 +18,7 @@ namespace LightGBM {
 
 const char* kModelVersion = "v3";
 
-std::string GBDT::DumpModel(int start_iteration, int num_iteration) const {
+std::string GBDT::DumpModel(int start_iteration, int num_iteration, int feature_importance_type) const {
   std::stringstream str_buf;
 
   str_buf << "{";
@@ -95,7 +95,8 @@ std::string GBDT::DumpModel(int start_iteration, int num_iteration) const {
   }
   str_buf << "]," << '\n';
 
-  std::vector<double> feature_importances = FeatureImportance(num_iteration, 0);
+  std::vector<double> feature_importances = FeatureImportance(
+      num_iteration, feature_importance_type);
   // store the importance first
   std::vector<std::pair<size_t, std::string>> pairs;
   for (size_t i = 0; i < feature_importances.size(); ++i) {
@@ -302,7 +303,7 @@ bool GBDT::SaveModelToIfElse(int num_iteration, const char* filename) const {
   return static_cast<bool>(output_file);
 }
 
-std::string GBDT::SaveModelToString(int start_iteration, int num_iteration) const {
+std::string GBDT::SaveModelToString(int start_iteration, int num_iteration, int feature_importance_type) const {
   std::stringstream ss;
 
   // output model type
@@ -363,8 +364,8 @@ std::string GBDT::SaveModelToString(int start_iteration, int num_iteration) cons
     tree_strs[i].clear();
   }
   ss << "end of trees" << "\n";
-
-  std::vector<double> feature_importances = FeatureImportance(num_iteration, 0);
+  std::vector<double> feature_importances = FeatureImportance(
+      num_iteration, feature_importance_type);
   // store the importance first
   std::vector<std::pair<size_t, std::string>> pairs;
   for (size_t i = 0; i < feature_importances.size(); ++i) {
@@ -395,11 +396,11 @@ std::string GBDT::SaveModelToString(int start_iteration, int num_iteration) cons
   return ss.str();
 }
 
-bool GBDT::SaveModelToFile(int start_iteration, int num_iteration, const char* filename) const {
+bool GBDT::SaveModelToFile(int start_iteration, int num_iteration, int feature_importance_type, const char* filename) const {
   /*! \brief File to write models */
   std::ofstream output_file;
   output_file.open(filename, std::ios::out | std::ios::binary);
-  std::string str_to_write = SaveModelToString(start_iteration, num_iteration);
+  std::string str_to_write = SaveModelToString(start_iteration, num_iteration, feature_importance_type);
   output_file.write(str_to_write.c_str(), str_to_write.size());
   output_file.close();
 

src/c_api.cpp

@@ -689,8 +689,8 @@ class Booster {
     boosting_->GetPredictAt(data_idx, out_result, out_len);
   }
 
-  void SaveModelToFile(int start_iteration, int num_iteration, const char* filename) {
-    boosting_->SaveModelToFile(start_iteration, num_iteration, filename);
+  void SaveModelToFile(int start_iteration, int num_iteration, int feature_importance_type, const char* filename) {
+    boosting_->SaveModelToFile(start_iteration, num_iteration, feature_importance_type, filename);
   }
 
   void LoadModelFromString(const char* model_str) {
@@ -698,12 +698,16 @@ class Booster {
     boosting_->LoadModelFromString(model_str, len);
   }
 
-  std::string SaveModelToString(int start_iteration, int num_iteration) {
-    return boosting_->SaveModelToString(start_iteration, num_iteration);
+  std::string SaveModelToString(int start_iteration, int num_iteration,
+                                int feature_importance_type) {
+    return boosting_->SaveModelToString(start_iteration,
+                                        num_iteration, feature_importance_type);
   }
 
-  std::string DumpModel(int start_iteration, int num_iteration) {
-    return boosting_->DumpModel(start_iteration, num_iteration);
+  std::string DumpModel(int start_iteration, int num_iteration,
+                        int feature_importance_type) {
+    return boosting_->DumpModel(start_iteration, num_iteration,
+                                feature_importance_type);
   }
 
   std::vector<double> FeatureImportance(int num_iteration, int importance_type) {
@@ -2010,22 +2014,26 @@ int LGBM_BoosterPredictForMats(BoosterHandle handle,
 int LGBM_BoosterSaveModel(BoosterHandle handle,
                           int start_iteration,
                           int num_iteration,
+                          int feature_importance_type,
                           const char* filename) {
   API_BEGIN();
   Booster* ref_booster = reinterpret_cast<Booster*>(handle);
-  ref_booster->SaveModelToFile(start_iteration, num_iteration, filename);
+  ref_booster->SaveModelToFile(start_iteration, num_iteration,
+                               feature_importance_type, filename);
   API_END();
 }
 
 int LGBM_BoosterSaveModelToString(BoosterHandle handle,
                                   int start_iteration,
                                   int num_iteration,
+                                  int feature_importance_type,
                                   int64_t buffer_len,
                                   int64_t* out_len,
                                   char* out_str) {
   API_BEGIN();
   Booster* ref_booster = reinterpret_cast<Booster*>(handle);
-  std::string model = ref_booster->SaveModelToString(start_iteration, num_iteration);
+  std::string model = ref_booster->SaveModelToString(
+      start_iteration, num_iteration, feature_importance_type);
   *out_len = static_cast<int64_t>(model.size()) + 1;
   if (*out_len <= buffer_len) {
     std::memcpy(out_str, model.c_str(), *out_len);
@@ -2036,12 +2044,14 @@ int LGBM_BoosterSaveModelToString(BoosterHandle handle,
 int LGBM_BoosterDumpModel(BoosterHandle handle,
                           int start_iteration,
                           int num_iteration,
+                          int feature_importance_type,
                           int64_t buffer_len,
                           int64_t* out_len,
                           char* out_str) {
   API_BEGIN();
   Booster* ref_booster = reinterpret_cast<Booster*>(handle);
-  std::string model = ref_booster->DumpModel(start_iteration, num_iteration);
+  std::string model = ref_booster->DumpModel(start_iteration, num_iteration,
+                                             feature_importance_type);
   *out_len = static_cast<int64_t>(model.size()) + 1;
   if (*out_len <= buffer_len) {
     std::memcpy(out_str, model.c_str(), *out_len);

src/io/config_auto.cpp

@@ -234,6 +234,7 @@ const std::unordered_set<std::string>& Config::parameter_set() {
   "verbosity",
   "input_model",
   "output_model",
+  "saved_feature_importance_type",
   "snapshot_freq",
   "max_bin",
   "max_bin_by_feature",
@@ -463,6 +464,8 @@ void Config::GetMembersFromString(const std::unordered_map<std::string, std::str
 
   GetString(params, "output_model", &output_model);
 
+  GetInt(params, "saved_feature_importance_type", &saved_feature_importance_type);
+
   GetInt(params, "snapshot_freq", &snapshot_freq);
 
   GetInt(params, "max_bin", &max_bin);
@@ -664,6 +667,7 @@ std::string Config::SaveMembersToString() const {
   str_buf << "[path_smooth: " << path_smooth << "]\n";
   str_buf << "[interaction_constraints: " << interaction_constraints << "]\n";
   str_buf << "[verbosity: " << verbosity << "]\n";
+  str_buf << "[saved_feature_importance_type: " << saved_feature_importance_type << "]\n";
   str_buf << "[max_bin: " << max_bin << "]\n";
   str_buf << "[max_bin_by_feature: " << Common::Join(max_bin_by_feature, ",") << "]\n";
   str_buf << "[min_data_in_bin: " << min_data_in_bin << "]\n";

swig/lightgbmlib.i

@@ -37,16 +37,17 @@
   char * LGBM_BoosterSaveModelToStringSWIG(BoosterHandle handle,
                                            int start_iteration,
                                            int num_iteration,
+                                           int feature_importance_type,
                                            int64_t buffer_len,
                                            int64_t* out_len) {
     char* dst = new char[buffer_len];
-    int result = LGBM_BoosterSaveModelToString(handle, start_iteration, num_iteration, buffer_len, out_len, dst);
+    int result = LGBM_BoosterSaveModelToString(handle, start_iteration, num_iteration, feature_importance_type, buffer_len, out_len, dst);
     // Reallocate to use larger length
     if (*out_len > buffer_len) {
       delete [] dst;
       int64_t realloc_len = *out_len;
       dst = new char[realloc_len];
-      result = LGBM_BoosterSaveModelToString(handle, start_iteration, num_iteration, realloc_len, out_len, dst);
+      result = LGBM_BoosterSaveModelToString(handle, start_iteration, num_iteration, feature_importance_type, realloc_len, out_len, dst);
     }
     if (result != 0) {
       return nullptr;
@@ -57,16 +58,17 @@
   char * LGBM_BoosterDumpModelSWIG(BoosterHandle handle,
                                    int start_iteration,
                                    int num_iteration,
+                                   int feature_importance_type,
                                    int64_t buffer_len,
                                    int64_t* out_len) {
     char* dst = new char[buffer_len];
-    int result = LGBM_BoosterDumpModel(handle, start_iteration, num_iteration, buffer_len, out_len, dst);
+    int result = LGBM_BoosterDumpModel(handle, start_iteration, num_iteration, feature_importance_type, buffer_len, out_len, dst);
     // Reallocate to use larger length
     if (*out_len > buffer_len) {
       delete [] dst;
       int64_t realloc_len = *out_len;
       dst = new char[realloc_len];
-      result = LGBM_BoosterDumpModel(handle, start_iteration, num_iteration, realloc_len, out_len, dst);
+      result = LGBM_BoosterDumpModel(handle, start_iteration, num_iteration, feature_importance_type, realloc_len, out_len, dst);
     }
     if (result != 0) {
       return nullptr;

tests/c_api_test/test_.py

@@ -236,7 +236,7 @@ def test_booster():
             result.ctypes.data_as(ctypes.POINTER(ctypes.c_double)))
         if i % 10 == 0:
             print('%d iteration test AUC %f' % (i, result[0]))
-    LIB.LGBM_BoosterSaveModel(booster, 0, -1, c_str('model.txt'))
+    LIB.LGBM_BoosterSaveModel(booster, 0, -1, 0, c_str('model.txt'))
     LIB.LGBM_BoosterFree(booster)
     free_dataset(train)
     free_dataset(test)