python如何实现决策树分类算法

#创建数据集def createdataset():    dataSet=[[0,2,0,0,'N'],            [0,2,0,1,'N'],            [1,2,0,0,'Y'],            [2,1,0,0,'Y'],            [2,0,1,0,'Y'],            [2,0,1,1,'N'],            [1,0,1,1,'Y'],            [0,1,0,0,'N'],            [0,0,1,0,'Y'],            [2,1,1,0,'Y'],            [0,1,1,1,'Y'],            [1,1,0,1,'Y'],            [1,2,1,0,'Y'],            [2,1,0,1,'N'],]    labels=['age','income','job','credit']    return dataSet,labels

ds1,lab = createdataset()print(ds1)print(lab)

def calcShannonEnt(dataSet):    numEntries = len(dataSet)    labelCounts = {}    for featVec in dataSet:        currentLabel = featVec[-1]        if currentLabel not in labelCounts.keys():            labelCounts[currentLabel] = 0                labelCounts[currentLabel] += 1                        shannonEnt = 0.0    for key in labelCounts:        prob = float(labelCounts[key])/numEntries        shannonEnt -= prob*log(prob,2)        return shannonEnt

shan = calcShannonEnt(ds1)print(shan)

def chooseBestFeatureToSplit(dataSet):    numFeatures = len(dataSet[0])-1    baseEntropy = calcShannonEnt(dataSet)    bestInfoGain = 0.0;bestFeature = -1    for i in range(numFeatures):        featList = [example[i] for example in dataSet]        uniqueVals = set(featList)        newEntroy = 0.0        for value in uniqueVals:            subDataSet = splitDataSet(dataSet, i, value)            prop = len(subDataSet)/float(len(dataSet))            newEntroy += prop * calcShannonEnt(subDataSet)        infoGain = baseEntropy - newEntroy        if(infoGain > bestInfoGain):            bestInfoGain = infoGain            bestFeature = i        return bestFeature

col = chooseBestFeatureToSplit(ds1)col

def majorityCnt(classList):    classCount = {}    for vote in classList:        if vote not in classCount.keys():classCount[vote] = 0        classCount[vote] += 1    sortedClassCount = sorted(classList.iteritems(),key=operator.itemgetter(1),reverse=True)#利用operator操作键值排序字典    return sortedClassCount[0][0]#创建树的函数    def createTree(dataSet,labels):    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)    bestFeatLabel = labels[bestFeat]    myTree = {bestFeatLabel:{}}    del(labels[bestFeat])    featValues = [example[bestFeat] for example in dataSet]    uniqueVals = set(featValues)    for value in uniqueVals:        subLabels = labels[:]        myTree[bestFeatLabel][value] = createTree(splitDataSet(dataSet, bestFeat, value), subLabels)            return myTree

Tree = createTree(ds1, lab)print("样本数据决策树：")print(Tree)

def classify(inputtree,featlabels,testvec):    firststr = list(inputtree.keys())[0]    seconddict = inputtree[firststr]    featindex = featlabels.index(firststr)    for key in seconddict.keys():        if testvec[featindex]==key:            if type(seconddict[key]).__name__=='dict':                classlabel=classify(seconddict[key],featlabels,testvec)            else:                classlabel=seconddict[key]    return classlabel

labels=['age','income','job','credit']tsvec=[0,0,1,1]print('result:',classify(Tree,labels,tsvec))tsvec1=[0,2,0,1]print('result1:',classify(Tree,labels,tsvec1))

import matplotlib.pyplot as pltdecisionNode = dict(box, fc="0.8")leafNode = dict(box, fc="0.8")arrow_args = dict(arrow)#获取叶节点的数目def getNumLeafs(myTree):    numLeafs = 0    firstStr = list(myTree.keys())[0]    secondDict = myTree[firstStr]    for key in secondDict.keys():        if type(secondDict[key]).__name__=='dict':#测试节点的数据是否为字典，以此判断是否为叶节点            numLeafs += getNumLeafs(secondDict[key])        else:   numLeafs +=1    return numLeafs#获取树的层数def getTreeDepth(myTree):    maxDepth = 0    firstStr = list(myTree.keys())[0]    secondDict = myTree[firstStr]    for key in secondDict.keys():        if type(secondDict[key]).__name__=='dict':#测试节点的数据是否为字典，以此判断是否为叶节点            thisDepth = 1 + getTreeDepth(secondDict[key])        else:   thisDepth = 1        if thisDepth > maxDepth: maxDepth = thisDepth    return maxDepth#绘制节点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 )#绘制连接线  def plotMidText(cntrPt, parentPt, txtString):    xMid = (parentPt[0]-cntrPt[0])/2.0 + cntrPt[0]    yMid = (parentPt[1]-cntrPt[1])/2.0 + cntrPt[1]    createPlot.ax1.text(xMid, yMid, txtString, va="center", ha="center", rotation=30)#绘制树结构  def plotTree(myTree, parentPt, nodeTxt):#if the first key tells you what feat was split on    numLeafs = getNumLeafs(myTree)  #this determines the x width of this tree    depth = getTreeDepth(myTree)    firstStr = list(myTree.keys())[0]     #the text label for this node should be this    cntrPt = (plotTree.xOff + (1.0 + float(numLeafs))/2.0/plotTree.totalW, plotTree.yOff)    plotMidText(cntrPt, parentPt, nodeTxt)    plotNode(firstStr, cntrPt, parentPt, decisionNode)    secondDict = myTree[firstStr]    plotTree.yOff = plotTree.yOff - 1.0/plotTree.totalD    for key in secondDict.keys():        if type(secondDict[key]).__name__=='dict':#test to see if the nodes are dictonaires, if not they are leaf nodes               plotTree(secondDict[key],cntrPt,str(key))        #recursion        else:   #it's a leaf node print the leaf node            plotTree.xOff = plotTree.xOff + 1.0/plotTree.totalW            plotNode(secondDict[key], (plotTree.xOff, plotTree.yOff), cntrPt, leafNode)            plotMidText((plotTree.xOff, plotTree.yOff), cntrPt, str(key))    plotTree.yOff = plotTree.yOff + 1.0/plotTree.totalD#创建决策树图形    def createPlot(inTree):    fig = plt.figure(1, facecolor='white')    fig.clf()    axprops = dict(xticks=[], yticks=[])    createPlot.ax1 = plt.subplot(111, frameon=False, **axprops)    #no ticks    #createPlot.ax1 = plt.subplot(111, frameon=False) #ticks for demo puropses     plotTree.totalW = float(getNumLeafs(inTree))    plotTree.totalD = float(getTreeDepth(inTree))    plotTree.xOff = -0.5/plotTree.totalW; plotTree.yOff = 1.0;    plotTree(inTree, (0.5,1.0), '')    plt.savefig('决策树.png',dpi=300,bbox_inches='tight')    plt.show()