基于ID3决策树算法的实现(Python版)

时间:2022-05-19 03:00:56

实例如下:

?
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
# -*- coding:utf-8 -*-
 
from numpy import *
import numpy as np
import pandas as pd
from math import log
import operator
 
#计算数据集的香农熵
def calcShannonEnt(dataSet):
  numEntries=len(dataSet)
  labelCounts={}
  #给所有可能分类创建字典
  for featVec in dataSet:
    currentLabel=featVec[-1]
    if currentLabel not in labelCounts.keys():
      labelCounts[currentLabel]=0
    labelCounts[currentLabel]+=1
  shannonEnt=0.0
  #以2为底数计算香农熵
  for key in labelCounts:
    prob = float(labelCounts[key])/numEntries
    shannonEnt-=prob*log(prob,2)
  return shannonEnt
 
 
#对离散变量划分数据集,取出该特征取值为value的所有样本
def splitDataSet(dataSet,axis,value):
  retDataSet=[]
  for featVec in dataSet:
    if featVec[axis]==value:
      reducedFeatVec=featVec[:axis]
      reducedFeatVec.extend(featVec[axis+1:])
      retDataSet.append(reducedFeatVec)
  return retDataSet
 
#对连续变量划分数据集,direction规定划分的方向,
#决定是划分出小于value的数据样本还是大于value的数据样本集
def splitContinuousDataSet(dataSet,axis,value,direction):
  retDataSet=[]
  for featVec in dataSet:
    if direction==0:
      if featVec[axis]>value:
        reducedFeatVec=featVec[:axis]
        reducedFeatVec.extend(featVec[axis+1:])
        retDataSet.append(reducedFeatVec)
    else:
      if featVec[axis]<=value:
        reducedFeatVec=featVec[:axis]
        reducedFeatVec.extend(featVec[axis+1:])
        retDataSet.append(reducedFeatVec)
  return retDataSet
 
#选择最好的数据集划分方式
def chooseBestFeatureToSplit(dataSet,labels):
  numFeatures=len(dataSet[0])-1
  baseEntropy=calcShannonEnt(dataSet)
  bestInfoGain=0.0
  bestFeature=-1
  bestSplitDict={}
  for i in range(numFeatures):
    featList=[example[i] for example in dataSet]
    #对连续型特征进行处理
    if type(featList[0]).__name__=='float' or type(featList[0]).__name__=='int':
      #产生n-1个候选划分点
      sortfeatList=sorted(featList)
      splitList=[]
      for j in range(len(sortfeatList)-1):
        splitList.append((sortfeatList[j]+sortfeatList[j+1])/2.0)
 
      bestSplitEntropy=10000
      slen=len(splitList)
      #求用第j个候选划分点划分时,得到的信息熵,并记录最佳划分点
      for j in range(slen):
        value=splitList[j]
        newEntropy=0.0
        subDataSet0=splitContinuousDataSet(dataSet,i,value,0)
        subDataSet1=splitContinuousDataSet(dataSet,i,value,1)
        prob0=len(subDataSet0)/float(len(dataSet))
        newEntropy+=prob0*calcShannonEnt(subDataSet0)
        prob1=len(subDataSet1)/float(len(dataSet))
        newEntropy+=prob1*calcShannonEnt(subDataSet1)
        if newEntropy<bestSplitEntropy:
          bestSplitEntropy=newEntropy
          bestSplit=j
      #用字典记录当前特征的最佳划分点
      bestSplitDict[labels[i]]=splitList[bestSplit]
      infoGain=baseEntropy-bestSplitEntropy
    #对离散型特征进行处理
    else:
      uniqueVals=set(featList)
      newEntropy=0.0
      #计算该特征下每种划分的信息熵
      for value in uniqueVals:
        subDataSet=splitDataSet(dataSet,i,value)
        prob=len(subDataSet)/float(len(dataSet))
        newEntropy+=prob*calcShannonEnt(subDataSet)
      infoGain=baseEntropy-newEntropy
    if infoGain>bestInfoGain:
      bestInfoGain=infoGain
      bestFeature=i
  #若当前节点的最佳划分特征为连续特征,则将其以之前记录的划分点为界进行二值化处理
  #即是否小于等于bestSplitValue
  if type(dataSet[0][bestFeature]).__name__=='float' or type(dataSet[0][bestFeature]).__name__=='int':
    bestSplitValue=bestSplitDict[labels[bestFeature]]
    labels[bestFeature]=labels[bestFeature]+'<='+str(bestSplitValue)
    for i in range(shape(dataSet)[0]):
      if dataSet[i][bestFeature]<=bestSplitValue:
        dataSet[i][bestFeature]=1
      else:
        dataSet[i][bestFeature]=0
  return bestFeature
 
#特征若已经划分完,节点下的样本还没有统一取值,则需要进行投票
def majorityCnt(classList):
  classCount={}
  for vote in classList:
    if vote not in classCount.keys():
      classCount[vote]=0
    classCount[vote]+=1
  return max(classCount)
 
#主程序,递归产生决策树
def createTree(dataSet,labels,data_full,labels_full):
  classList=[example[-1] for example in dataSet]
  if classList.count(classList[0])==len(classList):
    return classList[0]
  if len(dataSet[0])==1:
    return majorityCnt(classList)
  bestFeat=chooseBestFeatureToSplit(dataSet,labels)
  bestFeatLabel=labels[bestFeat]
  myTree={bestFeatLabel:{}}
  featValues=[example[bestFeat] for example in dataSet]
  uniqueVals=set(featValues)
  if type(dataSet[0][bestFeat]).__name__=='str':
    currentlabel=labels_full.index(labels[bestFeat])
    featValuesFull=[example[currentlabel] for example in data_full]
    uniqueValsFull=set(featValuesFull)
  del(labels[bestFeat])
  #针对bestFeat的每个取值,划分出一个子树。
  for value in uniqueVals:
    subLabels=labels[:]
    if type(dataSet[0][bestFeat]).__name__=='str':
      uniqueValsFull.remove(value)
    myTree[bestFeatLabel][value]=createTree(splitDataSet\
     (dataSet,bestFeat,value),subLabels,data_full,labels_full)
  if type(dataSet[0][bestFeat]).__name__=='str':
    for value in uniqueValsFull:
      myTree[bestFeatLabel][value]=majorityCnt(classList)
  return myTree
 
import matplotlib.pyplot as plt
decisionNode=dict(boxstyle="sawtooth",fc="0.8")
leafNode=dict(boxstyle="round4",fc="0.8")
arrow_args=dict(arrowstyle="<-")
 
 
#计算树的叶子节点数量
def getNumLeafs(myTree):
  numLeafs=0
  firstSides = list(myTree.keys())
  firstStr=firstSides[0]
  secondDict=myTree[firstStr]
  for key in secondDict.keys():
    if type(secondDict[key]).__name__=='dict':
      numLeafs+=getNumLeafs(secondDict[key])
    else: numLeafs+=1
  return numLeafs
 
#计算树的最大深度
def getTreeDepth(myTree):
  maxDepth=0
  firstSides = list(myTree.keys())
  firstStr=firstSides[0]
  secondDict=myTree[firstStr]
  for key in secondDict.keys():
    if type(secondDict[key]).__name__=='dict':
      thisDepth=1+getTreeDepth(secondDict[key])
    else: thisDepth=1
    if thisDepth>maxDepth:
      maxDepth=thisDepth
  return maxDepth
 
#画节点
def plotNode(nodeTxt,centerPt,parentPt,nodeType):
  createPlot.ax1.annotate(nodeTxt,xy=parentPt,xycoords='axes fraction',\
  xytext=centerPt,textcoords='axes fraction',va="center", ha="center",\
  bbox=nodeType,arrowprops=arrow_args)
 
#画箭头上的文字
def plotMidText(cntrPt,parentPt,txtString):
  lens=len(txtString)
  xMid=(parentPt[0]+cntrPt[0])/2.0-lens*0.002
  yMid=(parentPt[1]+cntrPt[1])/2.0
  createPlot.ax1.text(xMid,yMid,txtString)
 
def plotTree(myTree,parentPt,nodeTxt):
  numLeafs=getNumLeafs(myTree)
  depth=getTreeDepth(myTree)
  firstSides = list(myTree.keys())
  firstStr=firstSides[0]
  cntrPt=(plotTree.x0ff+(1.0+float(numLeafs))/2.0/plotTree.totalW,plotTree.y0ff)
  plotMidText(cntrPt,parentPt,nodeTxt)
  plotNode(firstStr,cntrPt,parentPt,decisionNode)
  secondDict=myTree[firstStr]
  plotTree.y0ff=plotTree.y0ff-1.0/plotTree.totalD
  for key in secondDict.keys():
    if type(secondDict[key]).__name__=='dict':
      plotTree(secondDict[key],cntrPt,str(key))
    else:
      plotTree.x0ff=plotTree.x0ff+1.0/plotTree.totalW
      plotNode(secondDict[key],(plotTree.x0ff,plotTree.y0ff),cntrPt,leafNode)
      plotMidText((plotTree.x0ff,plotTree.y0ff),cntrPt,str(key))
  plotTree.y0ff=plotTree.y0ff+1.0/plotTree.totalD
 
def createPlot(inTree):
  fig=plt.figure(1,facecolor='white')
  fig.clf()
  axprops=dict(xticks=[],yticks=[])
  createPlot.ax1=plt.subplot(111,frameon=False,**axprops)
  plotTree.totalW=float(getNumLeafs(inTree))
  plotTree.totalD=float(getTreeDepth(inTree))
  plotTree.x0ff=-0.5/plotTree.totalW
  plotTree.y0ff=1.0
  plotTree(inTree,(0.5,1.0),'')
  plt.show()
 
df=pd.read_csv('watermelon_4_3.csv')
data=df.values[:,1:].tolist()
data_full=data[:]
labels=df.columns.values[1:-1].tolist()
labels_full=labels[:]
myTree=createTree(data,labels,data_full,labels_full)
print(myTree)
createPlot(myTree)

最终结果如下:

{'texture': {'blur': 0, 'little_blur': {'touch': {'soft_stick': 1, 'hard_smooth': 0}}, 'distinct': {'density<=0.38149999999999995': {0: 1, 1: 0}}}}

得到的决策树如下:

基于ID3决策树算法的实现(Python版)

参考资料:

《机器学习实战》

《机器学习》周志华著

以上这篇基于ID3决策树算法的实现(Python版)就是小编分享给大家的全部内容了,希望能给大家一个参考,也希望大家多多支持服务器之家。