import numpy as np
from os import listdir
import operator

def img2vector(filename):
    fr = open(filename)
    lines = fr.readlines()
    returnVect = np.zeros((1,1024))
    for i in range(32):
        line = lines[i]
        for j in range(32):
            returnVect[0,32*i+j] = int(line[j])
    return returnVect

def classify(inX, dataset, labels, k):
    datasetSize = dataset.shape[0]
    diffMat = np.tile(inX,(datasetSize,1)) - dataset
    sqDiffMat = diffMat**2
    sqDistance = np.sum(sqDiffMat, axis=1)
    distances = sqDistance**0.5
    sortedDistIndicies = np.argsort(distances)
    classCount = {}
    for i in range(k):
        voteIlable = labels[sortedDistIndicies[i]]
        classCount[voteIlable] = classCount.get(voteIlable,0)+1
    sortedDistIndicies = sorted(classCount.items(),key=operator.itemgetter(1),reverse=True)
    return sortedDistIndicies[0][0]

def handwritingClassTest():
    hwLabels = []
    trainingFileList = listdir(‘trainingDigits‘)
    m = len(trainingFileList)
    trainingMat = np.zeros((m,1024))
    for i in range(m):
        fileNameStr = trainingFileList[i]
        filestr = fileNameStr.split(‘.‘)[0]
        classNumStr = filestr.split(‘_‘)[0]
        hwLabels.append(classNumStr)
        trainingMat[i,:] = img2vector(‘trainingDigits/%s‘ % fileNameStr)
    errorCount = 0.0
    testFileList = listdir(‘testDigits‘)
    mTest = len(testFileList)
    for i in range(mTest):
        fileNameStr = testFileList[i]
        filestr = fileNameStr.split(‘.‘)[0]
        classNumStr = filestr.split(‘_‘)[0]
        vectorunderTest = img2vector(‘testDigits/%s‘ % fileNameStr)
        classifierResult = classify(vectorunderTest, trainingMat, hwLabels, 3)
        print "the classfier came back with: %s, the real answer is: %s" % (classifierResult, classNumStr)
        if(classifierResult!=classNumStr):
            errorCount+=1
    print "the total error rate is: %f" % (errorCount/float(mTest))

kNN_handwriting