[人工智能]python决策树代码

看懂自取

from sklearn import datasets
from sklearn.model_selection import train_test_split
from sklearn.metrics import mean_squared_error
from sklearn import tree
from sklearn.datasets import load_wine
import pandas as pd
import math
import numpy as np

#数据集使用的python自带红酒数据集

wine = pd.read_csv('C:/Users/Dell/Desktop/郑佳重要/不确定理论/决策树/wine.csv')


#分为训练集和测试集
wine_train,wine_test = train_test_split(wine,test_size=0.2)

#重新设置索引
wine_train = wine_train.reset_index(drop=True)
wine_test = wine_test.reset_index(drop=True)

# 统计其中的分类个数，红酒数据集中分三类，计算熵
def tongji(data):
    class0 = 0
    class1 = 0
    class2 = 0
    inn = []
    for i in range(len(data)):
        if data['target'][i] == 0:
            class0 = class0 + 1
        elif data['target'][i] == 1:
            class1 = class1 + 1
        elif data['target'][i] == 2:
            class2 = class2 + 1
    if class0 == 0:
        k0 = 0
    else:
#本代码为不确定理论中定义的熵的公式
        k0 = -(class0 / len(data)) * math.log(class0 / len(data))


    if class1 == 0:
        k1 = 0
    else:
        k1 = -(class1 / len(data)) * math.log(class1 / len(data))

    if class2 == 0:
        k2 = 0
    else:
        k2 = -(class2 / len(data)) * math.log(class2 / len(data))
    entropy = k0 + k1 + k2

    return entropy

#计算某特征的熵
def jisuanshang(feature, dataname):
    for i in range(len(feature)):
        j = feature[i]
        # 将数据集按某特征分为两类

        data3 = dataname[(feature < j)]
        # 重新设置连续的索引
        data3 = data3.reset_index(drop=True)
        shang3 = tongji(data3)
        entr = shang3
        print(feature[i],entr)

#计算某特征的信息增益  

def tiaojianshang(fea,dataset):
    df = pd.DataFrame()
    for i in range(len(fea)):
        j = fea[i]
        # 将数据集按某特征分为两类
        data4 = dataset[(fea >= j)]
        data5 = dataset[(fea < j)]
        # 重新设置连续的索引
        data4 = data4.reset_index(drop=True)
        data5 = data5.reset_index(drop=True)

        shang4 = tongji(data4)
        shang5 = tongji(data5)
        tiaojianentr = (len(data4)/len(fea))*shang4+(len(data5)/len(fea)*shang5)
        # 不考虑特征此时数据集的熵值为
        x = tongji(dataset)
        # 计算信息增益
        gain = x - tiaojianentr

        # 将for循环中得到的数值保存
        df = df.append(pd.DataFrame({'index':[j],'gain': [gain]}), ignore_index=True)
        # 取某列中的最大值
        max1 = np.max(df['gain'])
        df2 = df[df['gain'] == max1]
    return df2


#计算不同特征的信息增益，选择决策树的根结点
def choosebest(dataset):
    fea = ['alcohol', 'malic_acid', 'ash', 'alcalinity_of_ash', 'magnesium',
           'total_phenols', 'flavanoids', 'nonflavanoid_phenols',
           'proanthocyanins', 'color', 'hue', 'od280', 'proline']
    df3 = pd.DataFrame()
    df4 = pd.DataFrame()
    for a in fea:
        gain = tiaojianshang(dataset[a], dataset)

        print(a,gain)



print(choosebest(wine_train))