def build_decison_tree(attr_list, df1, heuristic):
treeRoot = TreeNode()
countsPN = count(df1)
if countsPN[0] > countsPN[1]:
treeRoot.add_majority_class(1)
else:
treeRoot.add_majority_class(0)
if countsPN[0] == len(df1.index):
treeRoot.classLabel = 1
elif countsPN[1] == len(df1.index):
treeRoot.classLabel = 0
elif len(attr_list) == 0:
if countsPN[0] > countsPN[1]:
treeRoot.classLabel = 1
else:
treeRoot.classLabel = 0
else:
entropy = entropyFn(df1)
variance_impurity = varianceimpurityFn(df1)
chosen_attr_list = info_gain(entropy, variance_impurity, attr_list,
df1, heuristic)
chosen_attr = chosen_attr_list[heuristic]
if chosen_attr is not None:
treeRoot.add_attribute(chosen_attr)
new_attr_list = []
for attr in attr_list:
if attr != chosen_attr:
new_attr_list.append(attr)
attr_list = new_attr_list
for i in range(2):
df2 = df1[df1[chosen_attr] == i]
if len(df2.index) != 0:
if i == 0:
treeRoot.add_left(
build_decison_tree(attr_list, df2, heuristic))
else:
treeRoot.add_right(
build_decison_tree(attr_list, df2, heuristic))
else:
treeChild = TreeNode()
if countsPN[0] > countsPN[1]:
treeChild.classLabel = 1
else:
treeChild.classLabel = 0
if i == 0:
treeRoot.add_left(treeChild)
else:
treeRoot.add_right(treeChild)
else:
if countsPN[0] > countsPN[1]:
treeRoot.classLabel = 1
else:
treeRoot.classLabel = 0
return treeRoot
def copy_tree(root):
newRoot = TreeNode()
newRoot.add_node_attributes(root)
if root is None:
return None
if root.left is not None:
newRoot.add_left(copy_tree(root.left))
if root.right is not None:
newRoot.add_right(copy_tree(root.right))
node_list.append(newRoot)
return newRoot
