e53

SmirkCao · Oct 5, 2018 · fcc6170 · fcc6170
1 parent f54d844
commit fcc6170
Show file tree

Hide file tree

Showing 3 changed files with 53 additions and 5 deletions.
diff --git a/CH05/README.md b/CH05/README.md
@@ -86,7 +86,7 @@ $$
 > 输入：训练数据集$D$, 特征集$A$，阈值$\epsilon$
 > 输出：决策树$T$
 >
-> 1. $D$属于同一类$C_k$，$ T$为单节点树，类$C_k$作为该节点的类标记，返回$T$
+> 1. $D$属于同一类$C_k$，$T$为单节点树，类$C_k$作为该节点的类标记，返回$T$
 > 1. $A$是空集->T为单节点树，实例数最多的作为该节点类标记，返回T
 > 1. 计算$g$, 选择信息增益最大的特征$A_g$
 > 1. 如果$A_g$的信息增益小于$\epsilon$，$T$为单节点树，$D$中实例数最大的类$C_k$作为类标记，返回$T$
@@ -97,6 +97,16 @@ $$
 
 ### 算法5.3 C4.5生成算法
 
+> 输入：训练数据集$D$, 特征集$A$，阈值$\epsilon$
+> 输出：决策树$T$
+>
+> 1. D属于同一类C_k -> T为单节点树，类Ck作为该节点的类标记，返回T
+> 1. A是空集->T为单节点树，实例数最多的作为该节点类标记，返回T
+> 1. 计算g, 选择信息增益最大的特征Ag
+> 1. 如果Ag的信息增益小于ε，T为单节点树，D中实例数最大的类Ck作为类标记，返回T
+> 1. Ag划分若干非空子集Di，
+> 1. Di训练集，A-{Ag}为特征集，递归调用前面步骤，得到Ti，返回Ti
+
 ### 算法5.4 树的剪枝算法
 
 ### 算法5.5 最小二乘回归树生成算法

diff --git a/CH05/dt.py b/CH05/dt.py
@@ -12,12 +12,14 @@
 
 class dt(object):
     def __init__(self,
-                 tol=10e-3):
-        self.tree = None
+                 tol=10e-3,
+                 criterion = 'ID3'):
+        self.tree = dict()
         self.tol = tol
+        self.criterion = criterion
 
     def fit(self, X, y):
-        pass
+        self._build_tree(X, y)
 
     def predict(self, X):
         pass
@@ -54,6 +56,33 @@ def _gain_ratio(X, y):
     def _cal_gini(X, y):
         pass
 
+    def _build_tree(self, X, y):
+        ck, cnts = np.unique(y, return_counts=True)
+        # same y
+        if ck.shape[0] == 1:
+            return {ck[0]: None}
+        elif X.shape[1] == 0:
+            return {ck[np.argmax(cnts)]: None}
+        else:
+            rst = 0
+            cols = X.columns.tolist()
+            rst_col = cols[0]
+            for col in cols:
+                gain = dt._gain(X[col], y)
+                if gain >= rst:
+                    rst, rst_col = gain, col
+            if gain < self.tol:
+                return self.tree
+
+            cols.remove(rst_col)
+            rst = dict()
+            X_sub = X[cols]
+            for x in np.unique(X[rst_col]):
+                mask = X[rst_col] == x
+                rst.update({x: self._build_tree(X_sub[mask], y[mask])})
+            self.tree = {rst_col: rst}
+        return self.tree
+
 
 if __name__ == '__main__':
     logging.basicConfig(level=logging.INFO, format='%(asctime)s - %(name)s - %(levelname)s - %(message)s')

diff --git a/CH05/unit_test.py b/CH05/unit_test.py
@@ -34,7 +34,16 @@ def test_e52(self):
         self.assertEqual(np.argmax(rst), 2) # index = 2 -> A3
 
     def test_e53(self):
-        pass
+        raw_data = pd.read_csv("./Input/data_5-1.txt")
+        cols = raw_data.columns
+        X = raw_data[cols[1:-1]]
+        y = raw_data[cols[-1]]
+
+        clf = dt()
+        clf.fit(X, y)
+        logger.info(clf.tree)
+        rst = {'有自己的房子': {'否': {'有工作': {'否': {'否': None}, '是': {'是': None}}}, '是': {'是': None}}}
+        self.assertEqual(rst, clf.tree)
 
     def test_e54(self):
         pass