(dataSet)
| 107 | 2017-07-20 |
| 108 | """ |
| 109 | def chooseBestFeatureToSplit(dataSet): |
| 110 | numFeatures = len(dataSet[0]) - 1 #特征数量 |
| 111 | baseEntropy = calcShannonEnt(dataSet) #计算数据集的香农熵 |
| 112 | bestInfoGain = 0.0 #信息增益 |
| 113 | bestFeature = -1 #最优特征的索引值 |
| 114 | for i in range(numFeatures): #遍历所有特征 |
| 115 | #获取dataSet的第i个所有特征 |
| 116 | featList = [example[i] for example in dataSet] |
| 117 | uniqueVals = set(featList) #创建set集合{},元素不可重复 |
| 118 | newEntropy = 0.0 #经验条件熵 |
| 119 | for value in uniqueVals: #计算信息增益 |
| 120 | subDataSet = splitDataSet(dataSet, i, value) #subDataSet划分后的子集 |
| 121 | prob = len(subDataSet) / float(len(dataSet)) #计算子集的概率 |
| 122 | newEntropy += prob * calcShannonEnt(subDataSet) #根据公式计算经验条件熵 |
| 123 | infoGain = baseEntropy - newEntropy #信息增益 |
| 124 | # print("第%d个特征的增益为%.3f" % (i, infoGain)) #打印每个特征的信息增益 |
| 125 | if (infoGain > bestInfoGain): #计算信息增益 |
| 126 | bestInfoGain = infoGain #更新信息增益，找到最大的信息增益 |
| 127 | bestFeature = i #记录信息增益最大的特征的索引值 |
| 128 | return bestFeature #返回信息增益最大的特征的索引值 |
| 129 | |
| 130 | |
| 131 | """ |
no test coverage detected