决策树与随机森林的sklearn实现（update on going）

决策树

对iris数据进行决策树分类，采用sklearn库实现。决策树算法不需要对原始数据进行归一化处理，缺点是容易陷入过拟合，此时可以通过剪枝来控制树的深度。

# Author: Daniel Geng
from sklearn import tree
from sklearn.datasets import load_iris
import matplotlib.pyplot as plt
import numpy as np

X, y = load_iris(return_X_y=True)
sample = y.shape[0]
y = y.reshape(sample, 1)  # 原来y为一个向量，变为150行1列

# random the order and get 120 samples for training
C = np.concatenate((X, y), axis=1)  # 拼接为150行5列
np.random.shuffle(C)

# train
train_sample = 120
X_train = C[:train_sample, :4]
y_train = C[:train_sample, -1]
clf = tree.DecisionTreeClassifier()
clf = clf.fit(X_train, y_train)

# test
X_test = C[train_sample:, :4]
y_test = C[train_sample:, -1]
y_predict = clf.predict(X_test)
error = 0
error_position = []
for i in range(sample - train_sample):
    if y_predict[i] != y_test[i]:
        error += 1
        error_position.append(i)
accuracy = 1 - error / (sample - train_sample)
print('accuracy = ' + '%.2f%%' % (accuracy * 100))
print(error_position)

# plot
tree.plot_tree(clf, filled=True)
plt.show()

随机森林

随机生成多棵决策树，并随机抽取部分特征作为其中某一棵决策树的特征，最后汇总分类结果作为集成输出的分类结果。

# Author: Daniel Geng
from sklearn import tree
from sklearn import ensemble
from sklearn.datasets import load_iris
import matplotlib.pyplot as plt
import numpy as np

X, y = load_iris(return_X_y=True)
sample = y.shape[0]
y = y.reshape(sample, 1)  # 原来y为一个向量，变为150行1列

# random the order and get 120 samples for training
C = np.concatenate((X, y), axis=1)  # 拼接为150行5列
np.random.shuffle(C)

# train
train_sample = 120
X_train = C[:train_sample, :4]
y_train = C[:train_sample, -1]
clf = ensemble.RandomForestClassifier(n_estimators=10)
clf = clf.fit(X_train, y_train)

# test
X_test = C[train_sample:, :4]
y_test = C[train_sample:, -1]
y_predict = clf.predict(X_test)
error = 0A
error_position = []
for i in range(sample - train_sample):
    if y_predict[i] != y_test[i]:
        error += 1
        error_position.append(i)
accuracy = 1 - error / (sample - train_sample)
print('accuracy = ' + '%.2f%%' % (accuracy * 100))
print(error_position)