決策樹,decision的pyton程式碼和註釋(機器學習實戰)
阿新 • • 發佈:2019-01-06
Decison Tree的註釋:畫圖部分不給註釋了
from math import log import numpy def calcShannonEnt(dataSet): numEntries = len(dataSet) labelCounts = {}
#這個是字典,{a:1,b:2}其中a,b是key,1,2是對應的value for featVec in dataSet: currentLabel = featVec[-1]
#-1代表最後一行,也就是類標 if currentLabel not in labelCounts.keys(): labelCounts[currentLabel] = 0labelCounts[currentLabel] += 1 shannonEnt = 0.0 for key in labelCounts: prob = float(labelCounts[key])/numEntries shannonEnt -= prob * log(prob,2) return shannonEnt def createDataSet(): dataSet=[[1,1,'yes'], [1,1,'yes'], [1,0,'no'], [0,1,'yes'], [0,1,'no']] labels=['no surfacing','flippers'] return dataSet,labels #依據特徵劃分資料集 axis代表第幾個特徵 value代表該特徵所對應的值 返回的是劃分後的資料集def splitDataSet(dataSet, axis, value): retDataSet = [] for featVec in dataSet: if featVec[axis] == value: reducedFeatVec = featVec[:axis]
#這裡的featVec[:axis],是指從第1(就是下標0)個數到第axis個,不包含 reducedFeatVec.extend(featVec[axis+1:])
#同上,這裡的[axis+1,:]就是從最後到axis+1
retDataSet.append(reducedFeatVec)
#extend,append都是擴充套件用的,a=[1,2],b=[3,4],a.append(b)=[1,2,[3,4]],a.extend(b)=[1,2,3,4] return retDataSet
#選擇最好的資料集(特徵)劃分方式 返回最佳特徵下標
def chooseBestFeatureToSplit(dataSet):
numFeatures = len(dataSet[0]) - 1 #特徵個數
baseEntropy = calcShannonEnt(dataSet)
bestInfoGain = 0.0; bestFeature = -1
for i in range(numFeatures): #遍歷特徵 第i個
featureSet = set([example[i] for example in dataSet]) #第i個特徵取值集合
#這一部分程式碼沒啥難度,跟matalb差不多,唯一就是這個set
newEntropy= 0.0
for value in featureSet:
subDataSet = splitDataSet(dataSet, i, value)
prob = len(subDataSet)/float(len(dataSet))
newEntropy += prob * calcShannonEnt(subDataSet) #該特徵劃分所對應的entropy
infoGain = baseEntropy - newEntropy
if infoGain > bestInfoGain:
bestInfoGain = infoGain
bestFeature = i
return bestFeature
#建立樹的函式程式碼 python中用字典型別來儲存樹的結構 返回的結果是myTree-字典 def createTree(dataSet, labels): classList = [example[-1] for example in dataSet] if classList.count(classList[0]) == len(classList): #類別完全相同則停止繼續劃分 返回類標籤-葉子節點 return classList[0] if len(dataSet[0]) == 1: return majorityCnt(classList) #遍歷完所有的特徵時返回出現次數最多的 bestFeat = chooseBestFeatureToSplit(dataSet) bestFeatLabel = labels[bestFeat] myTree = {bestFeatLabel:{}} del(labels[bestFeat]) featValues = [example[bestFeat] for example in dataSet] #得到的列表包含所有的屬性值 uniqueVals = set(featValues) for value in uniqueVals: subLabels = labels[:] myTree[bestFeatLabel][value] = createTree(splitDataSet(dataSet, bestFeat, value), subLabels) return myTree
#多數表決的方法決定葉子節點的分類 ---- 當所有的特徵全部用完時仍屬於多類 def majorityCnt(classList): classCount = {} for vote in classList: if vote not in classCount.key(): classCount[vote] = 0; classCount[vote] += 1 sortedClassCount = sorted(classCount.iteritems(), key = operator.itemgetter(1), reverse = True)
#排序函式,至於怎麼用,help就好,裡面引數設定有詳細例子
return sortedClassCount[0][0]
建立樹的函式程式碼 其實這一步應該放在上一步前面 def createTree(dataSet, labels): classList = [example[-1] for example in dataSet] if classList.count(classList[0]) == len(classList): #類別完全相同則停止繼續劃分 返回類標籤-葉子節點 return classList[0]
#count是數數目的函式,a=[1,1,2] a.count[1]=2 len相當於matalb裡的length if len(dataSet[0]) == 1: return majorityCnt(classList) #遍歷完所有的特徵時返回出現次數最多的 bestFeat = chooseBestFeatureToSplit(dataSet) bestFeatLabel = labels[bestFeat] myTree = {bestFeatLabel:{}} del(labels[bestFeat]) featValues = [example[bestFeat] for example in dataSet] #得到的列表包含所有的屬性值 uniqueVals = set(featValues) for value in uniqueVals: subLabels = labels[:] myTree[bestFeatLabel][value] = createTree(splitDataSet(dataSet, bestFeat, value), subLabels)
#這一步creteTree裡面又用了creatTree,遞迴呼叫,直到len(dataSet[0]) == 1: return myTree