def normData(dataSet):
 maxVals = dataSet.max(axis=0)#按列获取最大值，并返回数组
 minVals = dataSet.min(axis=0)
 ranges = maxVals - minVals
 retData = (dataSet - minVals) / ranges
 return retData, ranges, minVals

#！／user/bin/env python
#-*- coding:utf-8 -*-
import numpy as np
import operator as opt

def normData(dataSet):#标准化训练集数据
 maxVals = dataSet.max(axis=0)
 minVals = dataSet.min(axis=0)
 ranges = maxVals - minVals
 retData = (dataSet - minVals) / ranges
 return retData, ranges, minVals

def kNN(dataSet, labels, testData, k):
 distSquareMat = (dataSet - testData) ** 2 # 计算差值的平方
 distSquareSums = distSquareMat.sum(axis=1) # 求每一行的差值平方和，axis=0则按列计算
 distances = distSquareSums ** 0.5 # 开根号，得出每个样本到测试点的距离
 sortedIndices = distances.argsort() # 排序，得到排序后的下标
 indices = sortedIndices[:k] # 取最小的k个
 labelCount = {} # 存储每个label的出现次数，出现次数最多的就是我们要选择的类别
 for i in indices:
 label = labels[i]
 labelCount[label] = labelCount.get(label, 0) + 1 # 次数加一，使用字典的get方法，第一次出现时默认值是0
 sortedCount = sorted(labelCount.items(), key=opt.itemgetter(1), reverse=True) # 对label出现的次数从大到小进行排序
 return sortedCount[0][0] # 返回出现次数最大的label

if __name__ == "__main__":#测试程序
 dataSet = np.array([[2, 3], [6, 8]])#训练集
 normDataSet, ranges, minVals = normData(dataSet)
 labels = ['a', 'b']#训练集分别为a和b类
 testData = np.array([3.9, 5.5])#测试数据
 normTestData = (testData - minVals) / ranges#同样需要将测试数据标准化
 result = kNN(normDataSet, labels, normTestData, 1)#k=1
 print(result)