机器学习实战笔记-K近邻算法3(手写识别系统)

Posted 2020-10-13 kevincong

1 准备数据：将图像转换为测试向量
这次数据集还是有两种，训练数据集和测试数据集，分别有2000个，900个。
我们将把一个32*32的二进制图像矩阵转换为1 x 1024的向量，这样前两节使用的分类器就可以处理数字图像信息了。
代码:

def img2vector(filename): 

    returnVect = zeros((1,1024)) 

    file = open(filename) 

    for i in range(32): 

        line = file.readline() 

        for j in range(32): 

            returnVect[0,i*32+j] = line[j] 

    return returnVect

 

效果截图:

测试算法
代码:

def handWritingTest(): 

    hwLabels = [] 

    trainingFileList = os.listdir(‘trainingDigits‘) 

    trainingFileLength = len(trainingFileList) 

    trainingMat = zeros((trainingFileLength,1024)) 

    for i in range(trainingFileLength): 

        fileNameStr = trainingFileList[i] 

        className = fileNameStr.split(‘_‘)[0] 

        hwLabels.append(int(className)) 

        fileVector = img2vector(‘trainingDigits/‘ + fileNameStr) 

        trainingMat[i,:] = fileVector 

    testFileList = os.listdir(‘testDigits‘) 

    testFileLength = len(testFileList) 

    errorCount = 0.0 

    for i in range(testFileLength): 

        fileNameStr = testFileList[i] 

        className = int(fileNameStr.split(‘_‘)[0]) 

        fileVector = img2vector(‘testDigits/‘ + fileNameStr) 

        testResult = classify0(fileVector,trainingMat,hwLabels,3) 

        print("the classifier came back with: %d, the real answer is: %d" % (testResult,className)) 

        if(testResult != className): 

            errorCount+=1.0 

    errorRate = errorCount/float(testFileLength) 

    print("the errorRate is : %f" % errorRate)

结果截图:

分别将k改为4,5:


可以发现错误率逐渐增高