saddle-ml 2.0.0__py3-none-any.whl

saddle/__init__.py

@@ -0,0 +1 @@
+from . import ctr,dl,feature_process,nlp,risk_management,statistical_model,timeseries,comm

saddle/automl/autogluon.py

@@ -0,0 +1,52 @@
+#!/usr/bin/env python
+# -*- coding: UTF-8 -*-
+'''
+@Project ：saddle
+@File    ：automl.py
+@IDE     ：PyCharm
+@Author  ：patrick
+@Date    ：2022/6/1 23:00
+'''
+
+from autogluon.tabular import TabularPredictor
+import logging
+import pandas as pd
+class AutogluonWrapper:
+    def __init__(self,presets="medium_quality_faster_train",model_type="regression"):#  good_quality
+        self.presets    = presets
+        self.model_type = model_type
+    def fit(self,X,y):
+        logging.info('X len [%s],y len[%s]  ',len(X),len(y) )
+        X_df = pd.DataFrame( X,columns=[ 'X'+str(i) for i in range( 0, len(X[0]) )  ] )
+        y_df = pd.DataFrame(y, columns =['y'])
+        self.X_names = list(X_df.columns)
+        self.y_names = list(y_df.columns)
+        train_df     = pd.concat([X_df,y_df],axis=1)
+        self.predictor = TabularPredictor(self.y_names[0],problem_type=self.model_type).fit( train_df,presets=self.presets )
+    def predict(self,X):
+        X_df = pd.DataFrame(X)
+        X_df.columns = self.X_names
+        if self.model_type=="regression":
+            return list(self.predictor.predict(X_df))
+        return "not implemented"
+
+import pandas as pd
+from autogluon.tabular import TabularDataset, TabularPredictor
+def build_train_autogluon_classifier(train_x,train_y,feature_names,presets='best_quality'):
+    df_train = pd.concat([pd.DataFrame(train_x, columns=feature_names), pd.DataFrame(train_y, columns=['label'])], axis=1)
+    # df_test = pd.concat([pd.DataFrame(test_x, columns=feature_names), pd.DataFrame(test_y, columns=['label'])], axis=1)
+    # df_eval = pd.concat([pd.DataFrame(oot_x, columns=feature_names), pd.DataFrame(oot_y, columns=['label'])], axis=1)
+    # X_train = pd.concat([df_train, df_test])
+    # X_train.reset_index(drop=True, inplace=True)
+
+    from autogluon.tabular import TabularDataset, TabularPredictor
+    predictor = TabularPredictor( label='label', problem_type='binary', eval_metric='roc_auc').fit( df_train, presets=presets )
+
+    # predictor = TabularPredictor.load("AutogluonModels/ag-20221025_040314/")
+    # train_y_pred = predictor.predict_proba( df_train.drop(columns=['label']) );
+    # train_y_pred = train_y_pred[1]
+    # test_y_pred = predictor.predict_proba(df_test.drop(columns=['label']));
+    # test_y_pred = test_y_pred[1]
+    # oot_test_y_pre1 = predictor.predict_proba(df_eval.drop(columns=['label']));
+    # oot_test_y_pre1 = oot_test_y_pre1[1]
+    return predictor