python实现KNN，识别手写数字

Posted 2020-09-30

写了识别手写数字的KNN算法，如下图所示。参考链接http://blog.csdn.net/april_newnew/article/details/44176059。

# -*- coding: utf-8 -*-

import numpy as np
import pandas as pd
import os
def readtxt(filename):
    text=[]
    f = open(filename,‘r‘,encoding=‘utf-8‘)
    for line in f.readlines():
        text.append(line)
    txt = list(text)
    txt=np.array(txt,dtype=‘float‘)
    txt = txt.tolist()
    return txt

def readdata(rootfile):
    data = []
    label = []
    for root,dirs,files in os.walk(rootfile):
        for name in files:
            filename = root +‘\\\\‘+name
            txt = readtxt(filename)
            data.append(txt)
            label1 = name.split(‘_‘)[0]
            label.append(label1)
    data = pd.DataFrame(data)
    return data,label

def KNN(traindata,trainlabel,testdatai,K):
    length = len(traindata)
    newtest = np.tile(testdatai, (length,1))
    newtest = pd.DataFrame(newtest)
    diff = newtest - traindata
    diff = diff**2
    cha = diff.sum(axis=1)
    cha = cha**0.5
    result = pd.DataFrame({‘label‘:trainlabel,
                       ‘cha‘:cha})
    labels = result.sort_values(by=‘cha‘)[:K]
    frequent =labels.groupby(labels[‘label‘]).size()
    labely = frequent.argmax()
    return labely
        
def test(trainfile,testfile,K):
    result = []
    traindata, trainlabel= readdata(trainfile)
    testdata, testlabel = readdata(testfile)
    for i in range(len(testdata)):
        labely = KNN(traindata,trainlabel,testdata.loc[i,:],K)
        result.append(labely)
    tongji  = pd.DataFrame({‘result‘:result,‘testlabel‘:testlabel})
    accuary = len(tongji[tongji[‘result‘]==tongji[‘testlabel‘]])/len(result)
    return result,accuary
    
trainfile=r‘E:\\trainingDigits‘
testfile=r‘E:\\testDigits‘
K=3    
result, accuary= test(trainfile,testfile,K)
            

注：训练数据集有2,210条记录，测试数据有670条。准确率并不高，只有0.45。目前不知道为什么，以后多学习，争取优化代码。