简单决策树的python实现

是 否 +2 论坛币

k人 参与回答

经管之家送您一份

应届毕业生专属福利!

求职就业群

赵安豆老师微信：zhaoandou666

经管之家联合CDA

送您一个全额奖学金名额~ !

立即领取

感谢您参与论坛问题回答

经管之家送您两个论坛币！

+2 论坛币

1. # coding: utf-8
   
2. 
   
3. # In[28]:
   
4. 
   
5. 
   
6. #####香农熵
   
7. from math import log
   
8. import operator
   
9. def calcShannonEnt(dataSet):
   
10.     numEntries=len(dataSet)
    
11.     labelCounts={}
    
12.     for featVec in dataSet:
    
13.         currentLabel = featVec[-1]
    
14.         if currentLabel not in labelCounts.keys():
    
15.             labelCounts[currentLabel] = 0
    
16.         labelCounts[currentLabel] += 1
    
17.     shannonEnt = 0.0
    
18.     for key in labelCounts:
    
19.         prob = float(labelCounts[key])/numEntries
    
20.         shannonEnt -= prob * log(prob,2)
    
21.     return shannonEnt
    
22. 
    
23. 
    
24. # In[6]:
    
25. 
    
26. 
    
27. def createDataSet():
    
28.     dataSet = [[1,1,'yes'],[1,1,'yes'],[1,0,'no'],[0,1,'no'],[0,1,'no']]
    
29.     labels=['no surfacing','flippers']
    
30.     return dataSet, labels
    
31. 
    
32. 
    
33. # In[9]:
    
34. 
    
35. 
    
36. #############划分数据集
    
37. def splitDataSet(dataSet,axis,value):
    
38.     retDataSet = []
    
39.     for featVec in dataSet:
    
40.         if featVec[axis] == value:
    
41.             reducedFeatVec = featVec[:axis]
    
42.             reducedFeatVec.extend(featVec[axis+1:])
    
43.             retDataSet.append(reducedFeatVec)
    
44.     return retDataSet
    
45. 
    
46. 
    
47. # In[26]:
    
48. 
    
49. 
    
50. #######选择最佳划分
    
51. def chooseBestFeatureToSplit(dataSet):
    
52.     numFeatures = len(dataSet[0]) - 1
    
53.     baseEntropy = calcShannonEnt(dataSet)
    
54.     bestInfoGain = 0.0;bestFeature = -1
    
55.     for i in range(numFeatures):
    
56.         featList = [example[i] for example in dataSet]
    
57.         uniqueVals = set(featList)
    
58.         newEntropy = 0.0
    
59.         for value in uniqueVals:
    
60.             subDataSet = splitDataSet(dataSet,i,value)
    
61.             prob =len(subDataSet)/float(len(dataSet))
    
62.             newEntropy += prob* calcShannonEnt(subDataSet)
    
63.         infoGain = baseEntropy - newEntropy
    
64.         if (infoGain > bestInfoGain):
    
65.             bestInfoGain = infoGain
    
66.             bestFeature = i
    
67.     return bestFeature
    
68. 
    
69. 
    
70. # In[29]:
    
71. 
    
72. 
    
73. ####投票决定标签
    
74. def majorityCnt(classList):
    
75.     classCount = {}
    
76.     for vote in classList:
    
77.         if vote not in classCount.keys(): classCount[vote] = 0
    
78.         classCount[vote] +=1
    
79.     sortedClassCount = sorted(classCount.iteritems(),                             key=operator.itemgetter(1),reverse = True)
    
80.     return sortedClassCount[0][0]
    
81. 
    
82. 
    
83. # In[32]:
    
84. 
    
85. 
    
86. #####递归构造决策树
    
87. def createTree(dataSet,labels):
    
88.     classList = [example[-1] for example in dataSet]
    
89.     if classList.count(classList[0]) == len(classList):
    
90.         return classList[0]
    
91.     if len(dataSet[0]) == 1:
    
92.         return majorityCnt(classList)
    
93.     bestFeat = chooseBestFeatureToSplit(dataSet)
    
94.     bestFeatLabel = labels[bestFeat]
    
95.     myTree = {bestFeatLabel:{}}
    
96.     del(labels[bestFeat])
    
97.     featValues = [example[bestFeat] for example in dataSet]
    
98.     uniqueVals = set(featValues)
    
99.     for value in uniqueVals:
    
100.         subLabels = labels[:]
     
101.         myTree[bestFeatLabel][value] = createTree(splitDataSet(dataSet,bestFeat,value),subLabels)
     
102.     return myTree
     
103. 
     
104. 
     
105. # In[33]:
     
106. 
     
107. 
     
108. ####测试
     
109. myDat,labels =createDataSet()
     
110. print(myDat)
     
111. splitDataSet(myDat,0,0)
     
112. chooseBestFeatureToSplit(myDat)
     
113. myTree = createTree(myDat,labels)
     
114. myTree

复制代码

扫码加我 拉你入群

请注明：姓名-公司-职位

以便审核进群资格，未注明则拒绝

分享0 收藏3 回帖

关键词：python 决策树 Majority operator features python python核心编程 决策树算法

代码能跑通 如果有可视化就最好了 希望添加

不懂