《机器学习实战》第二章－－k近邻算法的完整代码如下：
1 from numpy import * 2 import operator 3 import matplotlib 4 import matplotlib.pyplot as plt 5 from os import listdir 6 7 8 def createDataSet(): 9 group = array([[1.0,1.1],[1.0,1.0],[0,0],[0,0.1]]) 10 labels = ['A', 'A', 'B', 'B'] 11 return group, labels 12 13 14 def classify0(inX, dataSet, labels, k): 15 ''' 16 k-近邻算法 17 :param inX:用于分类的输入向量 18 :param dataSet: 输入的训练样本集 19 :param labels: 标签向量 20 :param k: 用于选择最近邻居的数目 21 :return: 返回k个邻居中距离最近且数量最多的类别作为预测类别 22 ''' 23 dataSetSize = dataSet.shape[0] 24 diffMat = tile(inX, (dataSetSize, 1)) - dataSet 25 sqDiffMat = diffMat ** 2 26 sqDistances = sqDiffMat.sum(axis=1) 27 distances = sqDistances ** 0.5 28 # 以上为计算输入向量与已有标签样本的欧式距离 29 sortedDistIndicies = distances.argsort() # argsort函数返回的是数组值从小到大的索引值，距离需要从小到大排序 30 classCount = {} 31 for i in range(k): 32 voteIlabel = labels[sortedDistIndicies[i]] 33 classCount[voteIlabel] = classCount.get(voteIlabel,0) + 1 34 # Python 字典(Dictionary) get() 函数返回指定键的值，如果值不在字典中返回默认值。 35 # get(voteIlabel,0)表示当能查询到相匹配的字典时，就会显示相应key对应的value，如果不能的话，就会显示后面的这个参数。 36 sortedClassCount = sorted(classCount.items(),key=operator.itemgetter(1),reverse=True) 37 # 按照元祖中第２个值的大小降序排序 38 # python2中的iteritems()方法需改为items() 39 return sortedClassCount[0][0] 40 41 42 def file2matrix(filename): 43 # 将文本记录转换为NumPy的解析程序 44 fr = open(filename) 45 arrayOLines = fr.readlines() 46 numberOfLines = len(arrayOLines) 47 print(numberOfLines) 48 returnMat = zeros((numberOfLines,3)) # 存放３种特征 49 classLabelVector = [] # 存放标签 50 index = 0 51 for line in arrayOLines: 52 line = line.strip() # strip() 方法用于移除字符串头尾指定的字符（默认为空格或换行符）或字符序列。 53 listFromLine = line.split(' ') # split() 通过指定分隔符对字符串进行切片，组成列表 54 returnMat[index, :] = listFromLine[0:3] # 将当前列表的前３个值赋予returnMat的当前行 55 classLabelVector.append(int(listFromLine[-1])) # 将标签添加到classLabelVector中 56 index += 1 57 return returnMat, classLabelVector 58 59 60 def autoNum(dataSet): 61 minVals = dataSet.min(0) # A.min(0) : 返回A每一列最小值组成的一维数组； 62 maxVals = dataSet.max(0) # A.max(0)：返回A每一列最大值组成的一维数组； 63 # https://blog.csdn.net/qq_41800366/article/details/86313052 64 ranges = maxVals - minVals 65 normDataSet = zeros(shape(dataSet)) 66 m = dataSet.shape[0] 67 normDataSet = dataSet - tile(minVals, (m,1)) 68 # tile将minVals的行数乘以m次重复，列数乘以１次重复，每一行都减掉minVals 69 normDataSet = normDataSet/tile(ranges,(m,1)) 70 # 每一行都除以ranges以是实现数据归一化 71 return normDataSet,ranges, minVals 72 73 74 def datingClassTest(): 75 hoRatio = 0.10 # 测试集比重 76 m = normMat.shape[0] 77 numTestVecs = int(m*hoRatio) # 测试集数量 78 errorCount = 0.0 79 for i in range(numTestVecs): 80 classfierResult = classify0(normMat[i,:],normMat[numTestVecs:m,:],datingLabels[numTestVecs:m],3) 81 print("the classifierResult came back with: %d,the real answer is: %d"%(classfierResult,datingLabels[i])) 82 if(classfierResult != datingLabels[i]):errorCount += 1.0 83 print("the total error rate is: %f"%(errorCount/float(numTestVecs))) 84 85 def classifyPerson(): 86 resultList = ['not at all', 'in small doses', 'in large doses'] 87 percentTats = float(input("percentage of time spent playing video games?")) 88 # 在 Python3.x 中 raw_input( ) 和 input( ) 进行了整合，去除了 raw_input( )，仅保留了 input( ) 函数， 89 # 其接收任意任性输入，将所有输入默认为字符串处理，并返回字符串类型。 90 ffMiles = float(input("frequent flier miles earned per year?")) 91 iceCream = float(input("liters of ice cream consumed per year?")) 92 inArr = array([ffMiles, percentTats, iceCream]) # 输入测试向量 93 classifierResult = classify0((inArr-minVals)/ranges,normMat,datingLabels,3) # 得到分类结果 94 print("You will probably like this person:",resultList[classifierResult-1]) 95 96 97 def img2vector(filename): 98 returnVect = zeros((1, 1024)) # 创建１行１０２４列的零数组 99 fr = open(filename) 100 for i in range(32): 101 lineStr = fr.readline() 102 for j in range(32): 103 returnVect[0, 32*i+j] = int(lineStr[j]) 104 return returnVect 105 106 107 def handwritingClassTest(): 108 hwLabels = [] # 存放手写数字的类别 109 trainingFileList = listdir('../machinelearninginaction/Ch02/digits/trainingDigits') 110 m = len(trainingFileList) 111 trainingMat = zeros((m, 1024)) # 创建ｍ行１０２４列的零数组 112 for i in range(m): 113 fileNameStr = trainingFileList[i] 114 fileStr = fileNameStr.split('.')[0] # 文件名字 115 classNumStr = int(fileStr.split('_')[0]) # 数字类别 116 hwLabels.append(classNumStr) 117 trainingMat[i,:]= img2vector('../machinelearninginaction/Ch02/digits/trainingDigits/%s'%fileNameStr) 118 testFileList = listdir('../machinelearninginaction/Ch02/digits/testDigits') 119 errorCount = 0.0 120 mTest = len(testFileList) 121 for i in range(mTest): 122 fileNameStr = testFileList[i] 123 fileStr = fileNameStr.split('.')[0] 124 classNumStr = int(fileStr.split('_')[0]) 125 vectorUnderTest = img2vector('../machinelearninginaction/Ch02/digits/testDigits/%s'%fileNameStr) 126 classifierResult = classify0(vectorUnderTest, trainingMat,hwLabels,3) # 对测试数据集进行ｋｎｎ计算 127 print('the classifier came back with: %d, the real answer is :%d'%(classifierResult,classNumStr)) 128 if(classifierResult != classNumStr):errorCount += 1.0 129 print(' the total number of error is: %d'%errorCount) 130 print(' the total error rate is: %f'%(errorCount/float(mTest))) 131 132 133 if __name__ == "__main__": 134 ''' 135 group,labels = createDataSet() 136 result = classify0([0,0],group,labels,3) 137 print(result) 138 ''' 139 datingDataMat, datingLabels = file2matrix("./datingTestSet2.txt") # 数据转换 140 fig = plt.figure() 141 ax = fig.add_subplot() 142 ax.scatter(datingDataMat[:,0],datingDataMat[:,1],15.0*array(datingLabels),15.0*array(datingLabels)) 143 plt.show() 144 normMat, ranges, minVals = autoNum(datingDataMat) # 输入数据归一化 145 # datingClassTest() 146 # classifyPerson() # 分类 147 # testVector = img2vector("../machinelearninginaction/Ch02/digits/trainingDigits/0_1.txt") 148 # print(testVector[0,0:31]) 149 handwritingClassTest()
手写数字识别的运行结果如下：
the total number of error is: 12
the total error rate is: 0.012685