python实现C4.5决策树算法

时间:2022-09-04 17:40:32

C4.5算法使用信息增益率来代替ID3的信息增益进行特征的选择,克服了信息增益选择特征时偏向于特征值个数较多的不足。信息增益率的定义如下:

python实现C4.5决策树算法

?
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
# -*- coding: utf-8 -*-
 
 
from numpy import *
import math
import copy
import cPickle as pickle
 
 
class C45DTree(object):
 def __init__(self): # 构造方法
  self.tree = {} # 生成树
  self.dataSet = [] # 数据集
  self.labels = [] # 标签集
 
 
 # 数据导入函数
 def loadDataSet(self, path, labels):
  recordList = []
  fp = open(path, "rb") # 读取文件内容
  content = fp.read()
  fp.close()
  rowList = content.splitlines() # 按行转换为一维表
  recordList = [row.split("\t") for row in rowList if row.strip()] # strip()函数删除空格、Tab等
  self.dataSet = recordList
  self.labels = labels
 
 
 # 执行决策树函数
 def train(self):
  labels = copy.deepcopy(self.labels)
  self.tree = self.buildTree(self.dataSet, labels)
 
 
 # 构件决策树:穿件决策树主程序
 def buildTree(self, dataSet, lables):
  cateList = [data[-1] for data in dataSet] # 抽取源数据集中的决策标签列
  # 程序终止条件1:如果classList只有一种决策标签,停止划分,返回这个决策标签
  if cateList.count(cateList[0]) == len(cateList):
   return cateList[0]
  # 程序终止条件2:如果数据集的第一个决策标签只有一个,返回这个标签
  if len(dataSet[0]) == 1:
   return self.maxCate(cateList)
  # 核心部分
  bestFeat, featValueList= self.getBestFeat(dataSet) # 返回数据集的最优特征轴
  bestFeatLabel = lables[bestFeat]
  tree = {bestFeatLabel: {}}
  del (lables[bestFeat])
  for value in featValueList: # 决策树递归生长
   subLables = lables[:] # 将删除后的特征类别集建立子类别集
   # 按最优特征列和值分隔数据集
   splitDataset = self.splitDataSet(dataSet, bestFeat, value)
   subTree = self.buildTree(splitDataset, subLables) # 构建子树
   tree[bestFeatLabel][value] = subTree
  return tree
 
 
 # 计算出现次数最多的类别标签
 def maxCate(self, cateList):
  items = dict([(cateList.count(i), i) for i in cateList])
  return items[max(items.keys())]
 
 
 # 计算最优特征
 def getBestFeat(self, dataSet):
  Num_Feats = len(dataSet[0][:-1])
  totality = len(dataSet)
  BaseEntropy = self.computeEntropy(dataSet)
  ConditionEntropy = []  # 初始化条件熵
  slpitInfo = [] # for C4.5,caculate gain ratio
  allFeatVList = []
  for f in xrange(Num_Feats):
   featList = [example[f] for example in dataSet]
   [splitI, featureValueList] = self.computeSplitInfo(featList)
   allFeatVList.append(featureValueList)
   slpitInfo.append(splitI)
   resultGain = 0.0
   for value in featureValueList:
    subSet = self.splitDataSet(dataSet, f, value)
    appearNum = float(len(subSet))
    subEntropy = self.computeEntropy(subSet)
    resultGain += (appearNum/totality)*subEntropy
   ConditionEntropy.append(resultGain) # 总条件熵
  infoGainArray = BaseEntropy*ones(Num_Feats)-array(ConditionEntropy)
  infoGainRatio = infoGainArray/array(slpitInfo) # C4.5信息增益的计算
  bestFeatureIndex = argsort(-infoGainRatio)[0]
  return bestFeatureIndex, allFeatVList[bestFeatureIndex]
 
 # 计算划分信息
 def computeSplitInfo(self, featureVList):
  numEntries = len(featureVList)
  featureVauleSetList = list(set(featureVList))
  valueCounts = [featureVList.count(featVec) for featVec in featureVauleSetList]
  pList = [float(item)/numEntries for item in valueCounts]
  lList = [item*math.log(item, 2) for item in pList]
  splitInfo = -sum(lList)
  return splitInfo, featureVauleSetList
 
 
 # 计算信息熵
 # @staticmethod
 def computeEntropy(self, dataSet):
  dataLen = float(len(dataSet))
  cateList = [data[-1] for data in dataSet] # 从数据集中得到类别标签
  # 得到类别为key、 出现次数value的字典
  items = dict([(i, cateList.count(i)) for i in cateList])
  infoEntropy = 0.0
  for key in items: # 香农熵: = -p*log2(p) --infoEntropy = -prob * log(prob, 2)
   prob = float(items[key]) / dataLen
   infoEntropy -= prob * math.log(prob, 2)
  return infoEntropy
 
 
 # 划分数据集: 分割数据集; 删除特征轴所在的数据列,返回剩余的数据集
 # dataSet : 数据集; axis: 特征轴; value: 特征轴的取值
 def splitDataSet(self, dataSet, axis, value):
  rtnList = []
  for featVec in dataSet:
   if featVec[axis] == value:
    rFeatVec = featVec[:axis] # list操作:提取0~(axis-1)的元素
    rFeatVec.extend(featVec[axis + 1:]) # 将特征轴之后的元素加回
    rtnList.append(rFeatVec)
  return rtnList
 
 # 存取树到文件
 def storetree(self, inputTree, filename):
  fw = open(filename,'w')
  pickle.dump(inputTree, fw)
  fw.close()
 
 # 从文件抓取树
 def grabTree(self, filename):
  fr = open(filename)
  return pickle.load(fr)

调用代码

?
1
2
3
4
5
6
7
8
9
10
11
12
# -*- coding: utf-8 -*-
 
from numpy import *
from C45DTree import *
 
dtree = C45DTree()
dtree.loadDataSet("dataset.dat",["age", "revenue", "student", "credit"])
dtree.train()
 
dtree.storetree(dtree.tree, "data.tree")
mytree = dtree.grabTree("data.tree")
print mytree

以上就是本文的全部内容,希望对大家的学习有所帮助,也希望大家多多支持服务器之家。

原文链接:https://blog.csdn.net/yjIvan/article/details/71272968