决策树的简单应用

决策树（Decision Tree）是一种简单但是广泛使用的分类器。通过训练数据构建决策树，可以高效的对未知的数据进行分类。决策数有两大优点：1）决策树模型可以读性好，具有描述性，有助于人工分析；2）效率高，决策树只需要一次构建，反复使用，每一次预测的最大计算次数不超过决策树的深度。但是决策树对训练集里不存在数据进行预测时，效果很差。

from sklearn.model_selection._split import train_test_split
dataset = [
    ["Outlook","Temperature","Humidity","Wind","Play?"],
    ["sunny","hot","high","false","no"],
    ["sunny","hot","high","true","no"],
    ["overcast","hot","high","false","yes"],
    ["rain","mild","high","false","yes"],
    ["rain","cool","normal","false","yes"],
    ["rain","cool","normal","true","no"],
    ["overcast","cool","normal","true","yes"],
    ["sunny","mild","high","false","no"],
    ["sunny","cool","normal","false","yes"],
    ["rain","mild","normal","false","yes"],
    ["sunny","mild","normal","true","yes"],
    ["overcast","mild","high","true","yes"],
    ["overcast","hot","normal","false","yes"],
    ["rain","mild","high","true","no"]
    ]
'''
处理数据:
第一列；sunny对应0，overcast对应1，rain对应2；
第二列：hot对应0，mild对应1，cool对应2；
第三列：high对应0，normal对应1；
第四列：false对应0，true对应1；
第五列：no对应0，yes对应1.(作为预测结果)
'''
def dealdata():
    #data_name存放原始数据，以坐标作为键，以原始数据的值作为值
    data_name = {}
    #data存放处理后的数据
    data = []
    for row in dataset[1:-1]:
        tmp = []
        for i,item in enumerate(row):
            key = []
            key.append(i)
            if(item == "sunny" or item == "hot" or item == "high" or item=="false" or item == "no"):
                tmp.append(0)
                key.append(0)
            elif(item == "overcast" or item == "mild" or item == "normal" or item=="true" or item == "yes"):
                tmp.append(1)
                key.append(1)
            elif(item =="rain" or item == "cool"):
                tmp.append(2)
                key.append(2)
            data_name[tuple(key)] = item    
        data.append(tmp)
    return data,data_name 
#生成规则:
def rules_format(xc_predict,xc_test,xd_test):
    title = dataset[0]
    rules = []
    for i,row in enumerate(xd_test):
        rule = "If "
        item = row.tolist()
        for j,d in enumerate(item):
            tmp = []
            tmp.append(j)
            tmp.append(d)
            rule += title[j]
            rule += " = "
            rule += data_name[tuple(tmp)]
            if (j != len(item)-1):
                rule += " and "
        rule += " then "
        rule += title[-1]
        rule += " = "
        if xc_predict[i]:
            rule += "yes."
        else:
            rule += "no."
        if xc_predict[i] == xc_test[i]:
            rule += " predict correct!"
        else:
            rule += " predict error!"
        rules.append(rule) 
    return rules

import numpy as np
from sklearn.tree import DecisionTreeClassifier
if __name__ == "__main__":
    data,data_name = dealdata() 
    #产生决策树         
    mode = DecisionTreeClassifier(random_state=14)
    x_d = np.zeros((15,4),dtype="int")
    x_c = np.zeros((15,),dtype="bool")
    for i,row in enumerate(data):
        x_d[i] = row[:-1]
        x_c[i] = row[-1] == 1
    #划分训练集和测试集
    xd_train,xd_test,xc_train,xc_test=train_test_split(x_d,x_c)
    #决策树训练
    mode.fit(xd_train,xc_train)
    #决策树预测
    xc_predict = mode.predict(xd_test)
    rules  = rules_format(xc_predict,xc_test,xd_test) 
    for rule in rules:
        print(rule)  

运行结果：

总结：决策树本已经是封装好了的，使用起来也比较方便，麻烦的是在于原始数据的处理，数据的处理才是关键。