sklearn中决策树的应用（python）

摘要：本文使用决策树进行分类，回归。

00 安装scikit-learn库

pip install scikit-learn

---

01 获取sklearn中鸢尾花数据

import numpy as np
import matplotlib.pyplot as plt
from sklearn import datasets
from sklearn.tree import DecisionTreeClassifier
iris=datasets.load_iris()
dex1=np.random.choice(150,size=120,replace=False)
dex2=[]
for i in range(150):
    if i not in dex1:
        dex2.append(i)
train_x=iris.data[dex1,:]
train_y=iris.target[dex1]
test_x=iris.data[dex2,:]
test_y=iris.target[dex2]

02 分类树

classi=DecisionTreeClassifier()
classi.fit(train_x,train_y)
classi.score(test_x,test_y)
Out[112]: 0.9333333333333333

classi.predict(test_x)

classi=DecisionTreeClassifier(criterion='gini')
classi.fit(train_x,train_y)
classi.score(test_x,test_y)
Out[135]: 0.9

classi=DecisionTreeClassifier(criterion='entropy')
classi.fit(train_x,train_y)
classi.score(test_x,test_y)
Out[136]: 0.9333333333333333

classi=DecisionTreeClassifier(splitter='best')
classi.fit(train_x,train_y)
classi.score(test_x,test_y)
Out[139]: 0.9666666666666667

classi=DecisionTreeClassifier(splitter='random')
classi.fit(train_x,train_y)
classi.score(test_x,test_y)
Out[142]: 0.8666666666666667

scor=[]
for i in range(1,20):
    classi=DecisionTreeClassifier(max_depth=i)
    classi.fit(train_x,train_y)
    sco=classi.score(test_x,test_y)
    scor.append(sco)
plt.plot(range(1,20),scor)
plt.grid(axis='both')

---

03 构造数据

import numpy as np
import matplotlib.pyplot as plt
from sklearn.tree import DecisionTreeRegressor

x=np.pi*2*np.random.rand(100)
y=np.sin(x)
y[::5]+=(np.random.rand(20)-0.5)*2

dex1=np.random.choice(100,75,replace=False)
dex2=[]
for i in range(100):
    if i not in dex1:
        dex2.append(i)
train_x=x[dex1].reshape(-1,1)
train_y=y[dex1].reshape(-1,1)
test_x=x[dex2].reshape(-1,1)
test_y=y[dex2].reshape(-1,1)

04 回归树

regre=DecisionTreeRegressor()
regre.fit(train_x,train_y)
regre.score(test_x,test_y)
Out[145]: 0.6619563827409216

regre.predict(test_x)
Out[146]:
array([ 0.98288727, -0.42291139, -0.11918768,  1.21105977, -0.1792897 ,
        0.88046883,  0.77566524, -0.89663976, -0.1565603 , -0.57729087,
       -0.89663976, -1.09538944, -0.97786811,  0.51036697, -1.09538944,
        0.20600983,  0.14411034,  0.88046883, -0.60787944,  0.20600983,
       -0.1792897 ,  0.77566524,  0.41635144,  0.36389462,  0.88046883])

x1=np.arange(0,6,0.1).reshape(-1,1)
y1=regre.predict(x1).reshape(-1,1)

x1=np.arange(0,6,0.1).reshape(-1,1)
y1=regre.predict(x1).reshape(-1,1)
plt.plot(x1,y1,linewidth=0.5,marker='*',ms=10,mfc='g',mec='g',
         alpha=0.4,label='predicted data')
plt.scatter(x,y,s=12,color='r',label='original data')
plt.legend(loc='upper right')

regre=DecisionTreeRegressor(splitter='best')
regre.fit(train_x,train_y)
regre.score(test_x,test_y)
Out[150]: 0.6619563827409216

regre=DecisionTreeRegressor(splitter='random')
regre.fit(train_x,train_y)
regre.score(test_x,test_y)
Out[151]: 0.6543112875182975

scor=[]
for i in range(1,20):
    regre=DecisionTreeRegressor(max_depth=i)
    regre.fit(train_x,train_y)
    sco=regre.score(test_x,test_y)
    scor.append(sco)
plt.plot(range(1,20),scor)
plt.grid(axis='both')

05 总结

01 决策树的关键概念：熵，信息增益，信息增益比；

02 决策树的模型复杂程度（剪枝，max_depth，log2(samples））