Showing function in leaf nodes when using linear regression

advanced_ML/model_tree/models/DT_sklearn_clf.py

@@ -21,11 +21,13 @@ def predict(self, X):
 
     def loss(self, X, y, y_pred):
         return gini_impurity(y)
+    def get_params(self):
+        return None
 
 def gini_impurity(y):
     p2 = 0.0
     y_classes = list(set(y))
     for c in y_classes:
         p2 += (np.sum(y == c) / len(y))**2
     loss = 1.0 - p2
-    return loss
+    return loss

advanced_ML/model_tree/models/DT_sklearn_regr.py

@@ -20,4 +20,6 @@ def predict(self, X):
         return self.model.predict(X)
 
     def loss(self, X, y, y_pred):
-        return mean_squared_error(y, y_pred)
+        return mean_squared_error(y, y_pred)
+    def get_params(self):
+        return None

advanced_ML/model_tree/models/NN_regr.py

@@ -33,4 +33,7 @@ def predict(self, X):
         return self.model.predict(X)
 
     def loss(self, X, y, y_pred):
-        return mean_squared_error(y, y_pred)
+        return mean_squared_error(y, y_pred)
+
+    def get_params(self):
+        return None

advanced_ML/model_tree/models/linear_regr.py

@@ -21,4 +21,5 @@ def predict(self, X):
 
     def loss(self, X, y, y_pred):
         return mean_squared_error(y, y_pred)
-
+    def get_params(self):
+        return self.model.coef_

advanced_ML/model_tree/models/logistic_regr.py

@@ -4,6 +4,7 @@
 
 """
 from sklearn.metrics import mean_squared_error
+import numpy as np
 
 class logistic_regr:
 
@@ -26,6 +27,8 @@ def predict(self, X):
             return self.flag_y_pred * np.ones((len(X),), dtype=int)
         else:
             return self.model.predict(X)
+    def get_params(self):
+        return None
 
     def loss(self, X, y, y_pred):
-        return mean_squared_error(y, y_pred)
+        return mean_squared_error(y, y_pred)

advanced_ML/model_tree/models/mean_regr.py

@@ -19,4 +19,7 @@ def predict(self, X):
         return self.y_mean * np.ones(len(X))
 
     def loss(self, X, y, y_pred):
-        return mean_squared_error(y, y_pred)
+        return mean_squared_error(y, y_pred)
+
+    def get_params(self):
+        return None

advanced_ML/model_tree/models/modal_clf.py

@@ -20,11 +20,14 @@ def predict(self, X):
 
     def loss(self, X, y, y_pred):
         return gini_impurity(y)
+
+    def get_params(self):
+        return None
 
 def gini_impurity(y):
     p2 = 0.0
     y_classes = list(set(y))
     for c in y_classes:
         p2 += (np.sum(y == c) / len(y))**2
     loss = 1.0 - p2
-    return loss
+    return loss

advanced_ML/model_tree/models/svm_regr.py

@@ -18,4 +18,7 @@ def predict(self, X):
         return self.model.predict(X)
 
     def loss(self, X, y, y_pred):
-        return mean_squared_error(y, y_pred)
+        return mean_squared_error(y, y_pred)
+
+    def get_params(self):
+        return None

advanced_ML/model_tree/src/ModelTree.py

@@ -149,6 +149,8 @@ def _explain(node, x, explanation):
             no_children = node["children"]["left"] is None and \
                           node["children"]["right"] is None
             if no_children:
+
+
                 return explanation
             else:
                 if x[node["j_feature"]] <= node["threshold"]:  # x[j] < threshold
@@ -196,7 +198,17 @@ def build_graphviz_recurse(node, parent_node_index=0, parent_depth=0, edge_label
             # Create node
             node_index = node["index"]
             if node["children"]["left"] is None and node["children"]["right"] is None:
-                threshold_str = ""
+
+                params = node['model'].get_params()
+                if params is not None :
+                    threshold_str="y = "
+                    for i in range(len(params)):
+                        threshold_str += str(round(params[i],2))
+                        threshold_str += "*X"+str(i)
+                    threshold_str+="\n"
+                else:
+                    threshold_str=""
+
             else:
                 threshold_str = "{} <= {:.1f}\\n".format(feature_names[node['j_feature']], node["threshold"])