資料github地址。

import numpy as np
import pandas as pd
import matplotlib.pyplot as plt
import copy

def autoNorm(x):
    """
    最大值最小值歸一化
    :param x: 需要歸一化的特徵向量
    :return: 新的陣列、極差、最小值
    """
    assert isinstance(x,np.ndarray)

    minVals=x.min(axis=0)
    maxVals=x.max(axis=0)
    ranges=
 
maxVals-minVals

    x_new=(x-minVals)/ranges # 廣播

    return x_new,ranges,minVals

def getdata_normal():
    """
    讀取原始文字資料
    :return:
    """
    fp="datingTestSet.txt"

    f=open(fp,mode='r')

    data=[line.strip().split('\t') for line in f.readlines()]
    data=np.array(data)

    x=data[:
 
,:-1].astype(np.float)
    y=data[:,-1]

    f.close()

    # 視覺化特徵
    # labels=copy.deepcopy(y)
    # dic={"largeDoses": "r", "smallDoses": 'g', "didntLike": 'b'}
    # for k, v in dic.items():
    #     labels[labels==k]=v
    # plt.scatter(x[:, 0], x[:, 1], 10,labels)
    # plt.title(dic)
    # plt.show()
 


    return x,y

def knnClaffify(testItem,trainX,trainY,k):
    """
    knn分類演算法，單條資料測試
    :param testItem: 測試的單條資料
    :param trainX: 訓練集特徵
    :param trainY: 訓練集標籤
    :param k: 鄰居個數
    :return: 分類類別
    """
    distances=np.sqrt(np.sum((trainX-testItem)**2,axis=1))

    ind=np.argsort(distances)

    classCount={}
    for i in range(k):
        vote=trainY[ind[i]]
        classCount[vote]=classCount.get(vote,0)+1
    classCount=sorted(classCount.items(),key=lambda x:x[0])

    return classCount[0][0]

def knnTest():
    """
    測試演算法
    :return:
    """
    x, y = getdata_normal()
    x, _, _ = autoNorm(x)

    total=len(x)
    splittest=int(0.9*total) # 分割訓練集和測試集,訓練集佔比0.8
    print("分割位置：{}，總數：{}".format(splittest,total))

    trueCount=0
    for i in range(splittest,total):
        result = knnClaffify(x[i], x[:splittest],y[:splittest],k=3)
        trueCount+=(y[i]==result)

    print("正確率：{2}({0}/{1})".format(trueCount,total-splittest,trueCount/(total-splittest)))

def knnForPerson():

    x, y = getdata_normal()
    x,ranges,minVals = autoNorm(x)

    t1 = float(input("每年旅行距離："))
    t2 = float(input("玩遊戲時間佔比："))
    t3 = float(input("每週吃的冰激凌："))

    item=(np.array([t1,t2,t3])-minVals)/ranges
    result=knnClaffify(item,x,y,k=3)
    print("predict: ",result)

if __name__ == '__main__':

    knnTest()
    knnForPerson()