Python机器学习之决策树

if(len(set(D.好瓜)) == 1):
        #标记返回 
        return D.好瓜.iloc[0]
    elif((len(A) == 0) or Check(D, A[:-1])):
        #选择D中结果最多的为标记
        cnt = D.groupby('好瓜').size()
        maxValue = cnt[cnt == cnt.max()].index[0]
        return maxValue
    else:
        A1 = copy.deepcopy(A)
        attr = Choose(D, A1[:-1])
        tree = {attr:{}}
        for value in set(D[attr]):
            tree[attr][value] = TreeGen(D[D[attr] == value], A1)
    return tree

def Choose(D:pd.DataFrame, A:list):
    result = 0.0
    resultAttr = ''
    for attr in A:
        tmpVal = CalcZengYi(D, attr)
        if(tmpVal > result):
            resultAttr = attr
            result = tmpVal
    A.remove(resultAttr)
    return resultAttr

#!/usr/bin/python3
# -*- encoding: utf-8 -*-
'''
@Description:决策树:
@Date     :2021/04/25 15:57:14
@Author      :willpower
@version      :1.0
'''
import pandas as pd
import numpy as np
import treeplot
import copy
import math
"""
@description  :计算熵值
---------
@param  :输入为基本pandas类型dataFrame,其中输入最后一行为实际结果
-------
@Returns  :返回熵值，类型为浮点型
-------
"""
def CalcShang(D:pd.DataFrame):
    setCnt = D.shape[0]
    result = 0.0
    # for i in D.groupby(D.columns[-1]).size().index:
    #遍历每一个值
    for i in set(D[D.columns[-1]]):
        #获取该属性下的某个值的次数
        cnt = D.iloc[:,-1].value_counts()[i]
        result = result + (cnt/setCnt)*math.log(cnt/setCnt, 2)
    return (-1*result)
"""
@description  :计算增益
---------
@param  :输入为DataFrame数据源，然后是需要计算增益的属性值
-------
@Returns  :返回增益值,浮点型
-------
"""
def CalcZengYi(D:pd.DataFrame, attr:str):
    sumShang = CalcShang(D)
    setCnt = D.shape[0]
    result = 0.0
    valus = D.groupby(attr).size()
    for subVal in valus.index:
        result = result + (valus[subVal]/setCnt)*CalcShang(D[D[attr] == subVal])
    return sumShang - result
"""
@description  :选择最佳的属性
---------
@param  :输入为数据源，以及还剩下的属性列表
-------
@Returns  :返回最佳属性
-------
"""
def Choose(D:pd.DataFrame, A:list):
    result = 0.0
    resultAttr = ''
    for attr in A:
        tmpVal = CalcZengYi(D, attr)
        if(tmpVal > result):
            resultAttr = attr
            result = tmpVal
    A.remove(resultAttr)
    return resultAttr
"""
@description  :检查数据在每一个属性下面的值是否相同
---------
@param  :输入为DataFrame以及剩下的属性列表
-------
@Returns  :返回bool值，相同返回1，不同返回0
-------
"""
def Check(D:pd.DataFrame, A:list):
    for i in A:
        if(len(set(D[i])) != 1):
            return 0
    return 1
"""
@description  :生成树主函数
---------
@param  :数据源DataFrame以及所有类型
-------
@Returns  :返回生成的字典树
-------
"""
def TreeGen(D:pd.DataFrame, A:list):
    if(len(set(D.好瓜)) == 1):
        #标记返回 
        return D.好瓜.iloc[0]
    elif((len(A) == 0) or Check(D, A[:-1])):
        #选择D中结果最多的为标记
        cnt = D.groupby('好瓜').size()
        #找到结果最多的结果
        maxValue = cnt[cnt == cnt.max()].index[0]
        return maxValue
    else:
        A1 = copy.deepcopy(A)
        attr = Choose(D, A1[:-1])
        tree = {attr:{}}
        for value in set(D[attr]):
            tree[attr][value] = TreeGen(D[D[attr] == value], A1)
    return tree
"""
@description  :验证集
---------
@param  :输入为待验证的数据(最后一列为真实结果)以及决策树模型
-------
@Returns  :无
-------
"""
def Test(D:pd.DataFrame, model:dict):
    for i in range(D.shape[0]):
            data = D.iloc[i]
            subModel = model
            while(1):
                attr = list(subModel)[0]
                subModel = subModel[attr][data[attr]]
                if(type(subModel).__name__ != 'dict'):
                    print(subModel, end='')
                    break
    print('')
name = ['色泽', '根蒂', '敲声', '纹理', '脐部', '触感', '好瓜']
df = pd.read_csv('./savedata.txt', names=name)
# CalcZengYi(df, '色泽')
resultTree = TreeGen(df, name)
print(resultTree)
# print(df[name[:-1]])
Test(df[name[:-1]], resultTree)
treeplot.plot_model(resultTree,"resultTree.gv")

from graphviz import Digraph

def plot_model(tree, name):
    g = Digraph("G", filename=name, fORMat='png', strict=False)
    first_label = list(tree.keys())[0]
    g.node("0", first_label)
    _sub_plot(g, tree, "0")
    g.view()
root = "0"

def _sub_plot(g, tree, inc):
    global root

    first_label = list(tree.keys())[0]
    ts = tree[first_label]
    for i in ts.keys():
        if isinstance(tree[first_label][i], dict):
            root = str(int(root) + 1)
            g.node(root, list(tree[first_label][i].keys())[0])
            g.edge(inc, root, str(i))
            _sub_plot(g, tree[first_label][i], root)
        else:
            root = str(int(root) + 1)
            g.node(root, tree[first_label][i])
            g.edge(inc, root, str(i))