sklearn决策树

调用方式

from sklearn.tree import DecisionTreeClassifier
# 创建样本
X, y = [[9], [8]], [1, 0]
# 建模
model = DecisionTreeClassifier()
# 训练
model.fit(X, y)
# 概率预测
print(model.predict_proba(X))

参数

参数	注释	默认值
criterion	基尼系数或信息增益（gini or entropy）	“gini”
splitter	全部或随机选择特征（best or random）	“best”
max_depth	树最大深度	None
min_samples_split	节点样本数小于该值则不再切分	2
min_samples_leaf	叶子节点的最小样本数	1
min_weight_fraction_leaf	The minimum weighted fraction of the sum total of weights (of all the input samples) required to be at a leaf node. 样本权重默认相等	0.
max_features	最大特征数	None
random_state	设置随机状态，默认用numpy.random	None
max_leaf_nodes	叶子节点的最大个数	None
min_impurity_decrease	A node will be split if this split induces a decrease of the impurity greater than or equal to this value.	0.
min_impurity_split	Threshold for early stopping in tree growth. A node will split if its impurity is above the threshold, otherwise it is a leaf.	None
class_weight	指定样本中各类别的权重	None
presort	是否预分类 以加快 在拟合过程中寻找最优切分 的速度（默认否）	False

概率预测

import pandas as pd
from sklearn.tree import DecisionTreeClassifier
# 创建样本
df = pd.DataFrame({
    '高': [1, 0, 1, 1, 1, 1, 0, 0, 0, 0],
    '富': [1, 1, 0, 0, 0, 0, 0, 0, 0, 0],
    '帅': [1, 0, 1, 0, 0, 0, 1, 1, 1, 0],
    '约': [1, 1, 1, 1, 1, 0, 0, 0, 1, 0]}
)
X = df[['高', '富', '帅']].values
y = df[['约']].values
# 建模训练
model = DecisionTreeClassifier()
model.fit(X, y)
# 约会预测
men = {'高': [1, 0, 0],
       '富': [0, 1, 0],
       '帅': [0, 0, 1]}
for man in men:
    probability = model.predict_proba([men[man]])[0][1]
    print('【{}man】约到女神的概率是：{:.1f}%'.format(man, probability * 100))

【高man】约到女神的概率是：66.7%
【富man】约到女神的概率是：100.0%
【帅man】约到女神的概率是：33.3%

可视化

随机森林

import numpy as np, matplotlib.pyplot as mp
from sklearn.datasets import make_moons
from sklearn.preprocessing import StandardScaler
from sklearn.ensemble import RandomForestClassifier

# 创建随机数据集、数据标准化、训练集测试集切分
X, y = make_moons(n_samples=200, noise=.2)
X = StandardScaler().fit_transform(X)

# 建模、训练、评分
clf = RandomForestClassifier(max_depth=4, min_samples_split=9)
clf.fit(X, y)

# 等高线图
x_min, x_max = X[:, 0].min() - .5, X[:, 0].max() + .5
y_min, y_max = X[:, 1].min() - .5, X[:, 1].max() + .5
xx, yy = np.meshgrid(np.arange(x_min, x_max, .01),
                     np.arange(y_min, y_max, .01))
Z = clf.predict_proba(np.c_[xx.ravel(), yy.ravel()])[:, 1].reshape(xx.shape)
mp.contourf(xx, yy, Z, alpha=.3)
# 散点图
mp.scatter(X[:, 0], X[:, 1], s=50, c=y)
mp.show()

不同树深度的比较

import numpy as np, matplotlib.pyplot as mp
from sklearn.datasets import make_moons
from sklearn.preprocessing import StandardScaler
from sklearn.model_selection import train_test_split
from sklearn.tree import DecisionTreeClassifier

X, y = make_moons(noise=.2)
X = StandardScaler().fit_transform(X)
X_train, X_test, y_train, y_test = train_test_split(X, y)

s, e = 3, 7
for i in range(s, e + 1):
    # 建模、训练、得分
    clf = DecisionTreeClassifier(max_depth=i)
    clf.fit(X_train, y_train)
    score = clf.score(X_test, y_test)
    print(score)
    # 可视化
    mp.subplot(1, e - s + 1, i - s + 1)
    x_min, x_max = X[:, 0].min() - .5, X[:, 0].max() + .5
    y_min, y_max = X[:, 1].min() - .5, X[:, 1].max() + .5
    xx, yy = np.meshgrid(np.arange(x_min, x_max, .01),
                         np.arange(y_min, y_max, .01))
    Z = clf.predict_proba(np.c_[xx.ravel(), yy.ravel()])[:, 1].reshape(xx.shape)
    mp.contourf(xx, yy, Z, alpha=.2)  # 等高线图
    mp.scatter(X_train[:, 0], X_train[:, 1], s=40, c=y_train)
    mp.scatter(X_test[:, 0], X_test[:, 1], s=40, c=y_test, alpha=0.6)
mp.tight_layout()  # 防止重叠
mp.show()

附录

sklearn版本：0.19.1

注释

en	cn
criterion	准则
fraction	分数；部分
probability	概率
impurity	杂质
presort	预分类
feature	n. 特征；vt. 特写
threshold	门槛；阈值
contour	n. 轮廓、等高线；vt. 画轮廓
mesh	网眼
grid	n. 网格；栅格；
meshgrid	用于三维曲面的分格线座标；产生“格点”矩阵
tight	密封的
layout	布局；陈列