机器学习4 决策树 概念总结和简单实践

第四章  决策树  概念图解

第一遍sklearn直接上手实践：

习题4.3 基于信息增益的决策树对西瓜集进行分类：

西瓜集表4.3如下：

sklearn直接调参代码：

# first round -- sklearn
import pandas as pd
import numpy as np
data = pd.read_csv('./CH4 TABLE4-3.csv', delimiter=',') 
data1 = data[['色泽','根蒂','敲声','纹理','脐部','触感','密度','含糖率','好坏']]
# change label to data format
data2 = data1.replace(['青绿','蜷缩','浊响','清晰','凹陷','硬滑','乌黑','稍蜷','沉闷','稍糊','稍凹','软粘','浅白','硬挺','清脆','模糊','平坦','好瓜','坏瓜'],[1,1,1,1,1,1,2,2,2,2,2,2,3,3,3,3,3,1,0])

X =np.array(data2.iloc[:,0:8]) #X varible
y = np.array(data2.iloc[:,8])  # target class : bad or good

from sklearn import tree
from sklearn.tree import DecisionTreeClassifier
# 基于信息增益的决策树训练
clf = DecisionTreeClassifier(criterion='entropy',splitter='best')
clf.fit(X,y)
# 决策树可视化
import pydotplus
from IPython.display import Image  
dot_data = tree.export_graphviz(clf, out_file=None, 
                         feature_names=['color','root','knock',' texture','umbilical','feel','density','sweety'],  
                         class_names=['Bad','Good'],  
                         filled=True, rounded=True,  
                         special_characters=True)  
graph = pydotplus.graph_from_dot_data(dot_data)  
Image(graph.create_png()) 

# 预测值与训练值对比---百分百匹配训练集，过拟合
from sklearn import metrics
y_pred = clf.predict(X)
print(metrics.classification_report(y,y_pred))