实现c4.5算法

tree/my_tree.py

@@ -131,6 +131,47 @@ def cal_ent_attr(Xtrain, Ytrain):
             min_mean = mean
     return min_i, min_mean, min_ent
 
+
+def cal_max_ent_attr_c45(Xtrain, Ytrain):
+    max_ent = 0
+    max_mean = 0
+    h = calc_ent(Ytrain)
+    for k in range(len(Xtrain) - 1):
+        left = Xtrain[:k + 1]
+        right = Xtrain[k + 1:]
+
+        left_ent = calc_ent(Ytrain[:k+1])*len(left)/len(Ytrain)
+        right_ent = calc_ent(Ytrain[k + 1:])*len(right)/len(Ytrain)
+
+        iv = -len(left) / len(Ytrain) * np.log2(len(left) / len(Ytrain))
+        iv -= len(right) / len(Ytrain) * np.log2(len(right) / len(Ytrain))
+
+        gain_ent = (h - left_ent - right_ent)/iv
+
+        if gain_ent > max_ent:
+            max_ent = gain_ent
+            max_mean = left[-1]
+    return  max_ent, max_mean
+
+
+# 计算某个属性的信息增益率
+def cal_ent_attr_c45(Xtrain, Ytrain):
+    # 对每个属性
+    max_ent = 0
+    max_i = 0
+    max_mean = 0
+    for i in range(Xtrain.shape[1]): #每个属性
+        argsort = Xtrain[:,i].argsort()
+        x,y = Xtrain[:,i][argsort], Ytrain[argsort]
+
+        gain_ent, mean = cal_max_ent_attr_c45(x, y)
+
+        if gain_ent > max_ent:
+            max_ent = gain_ent
+            max_i = i
+            max_mean = mean
+    return max_i, max_mean, max_ent
+
 # 计算某个属性的基尼指数
 def cal_gini_attr(Xtrain, Ytrain):
     print('sharp', Xtrain.shape)
@@ -203,7 +244,7 @@ def fit(Xtrain, Ytrain, parent_node, depth):
     if depth >= MAX_T:
         return leaf_node(Ytrain)
 
-    i, mean, min_ent = cal_gini_attr(Xtrain, Ytrain)
+    i, mean, min_ent = cal_ent_attr_c45(Xtrain, Ytrain)
     total_ent = calc_ent(Ytrain)
     print("第", i, "个属性,mean:", mean)
     # 生成节点
@@ -265,7 +306,7 @@ if __name__ == '__main__':
 
     print("基尼指数", cal_gini(Ytrain))
 
-    print(cal_ent_attr(Xtrain, Ytrain))
+    print("信息增益率", cal_ent_attr_c45(Xtrain, Ytrain))
 
     node = fit(Xtrain, Ytrain, None, 0)
     print_width([node], 1)