KNN 简单现实

knn是什么，大家可以百度百科一下，这里就不做介绍，其作为机器学习的最基本的算法，这里有简单的实现。

# -*- coding: utf-8 -*-
"""
Created on Sat Jun  9 18:27:11 2018

@author: Qunxing
"""

import numpy as np

DEBUG_FLAG = 0 #1 表示调试标志

def Normalize_Data(dataSet):
    maxValue = dataSet.max(axis = 0);
    minValue = dataSet.min(axis = 0);
    
    ranges= maxValue - minValue;
    dataSet = (dataSet - minValue)/ranges
    return dataSet, minValue, ranges;


def KNN(train_data, labels, test_data, kV):
    distSquaArray = (train_data - test_data)**2
    distSquaCount = distSquaArray.sum(axis = 1)   #进行操作后会变成一维数组
    if DEBUG_FLAG:
        print(distSquaCount)
        
    dist = np.sqrt(distSquaCount)
    if DEBUG_FLAG:
        print(dist)
        
    sortedIndices = dist.argsort() #默认按升序排列，返回索引值
    indices = sortedIndices[:kV]   # 取最小的k个
    
    labelCount = {}
    for index in indices:
        label = labels[index]
        labelCount[label] = labelCount.get(label, 0) + 1
        
    #这里注意是字典的排序，先将字典转化为元组
    labelRet = sorted(labelCount.items(), key = lambda x: x[1], reverse = True) #降序排列
    return labelRet[0][0]


if __name__ =="__main__":
    dataSet = np.array([[10, 3.5], [12, 7.8], [19, 4], [11, 8]])
    labels = ['a', 'b', 'a', 'c']
    testSet = np.array([14, 6])
    dataSet, minValue, ranges = Normalize_Data(dataSet)
    
    testSet = (testSet - minValue)/ranges
    print(KNN(dataSet, labels, testSet, 2))