Knn的python实现

python实现knn
运行的结果为

实现代码如下：

import numpy as np
import pickle as pic
import heapq as hp
from collections import Counter
class Solution:
    path_train=''
    train_data=[]
    train_label=[]
    test_data=[]
    test_label=[]
    test_record={}
    def readData(self,file):
        with open(file,'rb') as fo:
            dict=pic.load(fo,encoding='bytes')
        return dict

    def Init(self,path1,path2):
        self.path_train=path1
        self.test_record=self.readData(path2)#读取测试数据
        self.test_data=np.array(self.test_record[b'data'])
        self.test_label=np.array(self.test_record[b'labels'])
        for i in range(1,6):#读取训练数据
            path_temp=self.path_train+str(i)
            dict_temp=self.readData(path_temp)
            if(i==1):
                self.train_data=dict_temp[b'data']
                self.train_label=dict_temp[b'labels']
            else:
                self.train_data=np.append(self.train_data,dict_temp[b'data'],axis=0)
                self.train_label+=dict_temp[b'labels']
        self.train_data=np.array(self.train_data)
        self.train_label=np.array(self.train_label)

    def finalTest(self,k):#距离计算L1 norm 使用的是knn
        total=np.size(self.test_label)
        res_l=[]
        for i in range(0,total):
            data_t=self.test_data[i]
            temp=np.sum(abs(self.train_data-data_t),axis=1)
            ind_temp=hp.nsmallest(k,range(len(temp)), temp.take)
            lab_temp=self.train_label[ind_temp]
            counts = Counter(lab_temp)
            t=counts.most_common(k)
            res_l.append(t[0][0])
        print("acurracy: "+str(np.sum(res_l==self.test_label)/total))

a=Solution()
a.Init("D:\\data\\cifar-10-batches-py\\data_batch_","D:\\data\\cifar-10-batches-py\\test_batch")
for i in range(1,10,3):
    a.finalTest(i)