前期准备的基础理论,熵的计算,按照给定特征划分数据集,找出划分数据集的最优特征,但有些时候,数据集处理了所有属性,不过得到的类标签依然不是唯一的,书上说用之前KNN里面涉及的频次统计方法,遵从少数服从多数给该叶子节点分类,
import operator
def majorKind(dataList):
majorFeat = {}
for featVec in dataList.keys():
if majorFeat[featVec] not in majorFeat:
majorFeat[featVec] = 0
majorFeat[featVec] += 1
sortedMajorFeat = sorted(majorFeat.items(),key=operator.itemgetter(1),reverse=True)
return sortedMajorFeat[0][0]
把这些肥料组合一下,开始种树:
def createMytree(dataSet,Labels):
classList = [example[-1] for example in dataSet] #判断待分类数据集是否为同一类
if len(dataSet) == classList.count(classList[0]):
return classList[0]
if len(dataSet[0]) == 1: #判断是否已无其他属性可分类
return majorKind(classList) #选取频次最高的类标签最为该叶子节点的分类
bestFeat = findbestfeature(dataSet) #划分数据集的最优特征
bestFeatLabel = Labels[bestFeat]
myTree = {bestFeatLabel:{}}
del(Labels[bestFeat])
subfeat = [example[bestFeat] for example in dataSet] #选取最优分类的属性做之后的分类特征
uniqfeat = set(subfeat)
for value in uniqfeat: #根据特征连续划分,得到的值插入myTree中
subLabels = Labels[:]
myTree[bestFeatLabel][value] = createMytree(splitdataset(dataSet,bestFeat,value),
subLabels)
return Mytree
一脸懵逼,哎