机器学习实战：knn实现手写数字识别

import os.path
from sklearn.neighbors import KNeighborsClassifier as knn
import numpy as np
from os import listdir

"""
函数说明：将32*32的图像转化为（1，1024）的向量
Parameters:
    filename - 文件名
returns:
    returnVector - 返回的（1，1024）的向量
"""
def img2vector(filename):
    returnVector = np.zeros(1024)
    with open(filename) as fr:
        for i in range(32):
            line = fr.readline()
            for j in range(32):
                returnVector[32*i+j] = int(line[j])
    return returnVector

"""
函数说明：返回数据矩阵与label列表
Parameters:
    filename
returns：
    returnMetrix,labels
"""
def datalabel(filename):
    filelist = listdir(filename)
    m = len(filelist)
    returnMetrix = np.zeros((m,1024))
    labels = []
    for i in range(m):
        filenamestr = filelist[i]
        filepath = os.path.join(filename,filenamestr)
        returnMetrix[i,:] = img2vector(filepath)
        labels.append(filenamestr.split('_')[0])
    return returnMetrix,labels


"""
函数说明：knn分类器
Parameters:
    k - knn选择的最近数据个数
returns：
    无
"""
def classfiy(k):
    # 训练数据集和标签
    traindata,trainlabels = datalabel('trainingDigits')
    # 测试数据和真实标签
    testdata,testlabels = datalabel('testDigits')
    clf = knn(n_neighbors=k)
    clf.fit(traindata,trainlabels)
    errorcount = 0.0
    testlen = len(testlabels)
    for i in range(testlen):
        label_predict = clf.predict(testdata[i].reshape(-1,1024))
        if label_predict != (testlabels[i]):
            errorcount += 1
        print("预测:%d\t真实:%s"%(label_predict,testlabels[i]))
    print('Error:%f%%'%(errorcount/len(testlabels)*100))

from 实现 import *

if __name__ == '__main__':
    classfiy(5)