K-近邻算法

 1 #!/usr/bin/env python
 2 # -*- coding: utf-8 -*-
 3 from numpy import *
 4 
 5 def file2matrix(filename):
 6     ‘‘‘
 7     将文本格式转化为Numpy
 8     :param filename: 需要处理的文本文件
 9     :return: 转化格式后的文本和标签
10     ‘‘‘
11     fr = open(filename)
12     numberOfLines = len(fr.readlines())
13     returnMat = zeros((numberOfLines, 3))  # 数据的前三列用于分类
14     classLabelVector = []
15     fr = open(filename)
16     index = 0
17     for line in fr.readlines():
18         line = line.strip()  # line.strip()截取掉所有的回车符
19         listFromLine = line.split(‘\t‘)   # 将整行数据分割成一个元素列表
20         returnMat[index, :] = listFromLine[0:3]
21         classLabelVector.append(int(listFromLine[-1]))
22         index += 1
23     return returnMat, classLabelVector
24 
25 if __name__==‘__main__‘:
26     datingDataMat,datingLabels=file2matrix(‘datingTestSet2.txt‘)
27     print datingDataMat
28     print datingLabels

 1 #!/usr/bin/env python
 2 # -*- coding: utf-8 -*-
 3 from numpy import *
 4 import operator
 5 
 6 
 7 def file2matrix(filename):
 8     ‘‘‘
 9     将文本格式转化为Numpy
10     :param filename: 需要处理的文本文件
11     :return: 转化格式后的文本和标签
12     ‘‘‘
13     fr = open(filename)
14     numberOfLines = len(fr.readlines())
15     returnMat = zeros((numberOfLines, 3))  # 数据的前三列用于分类
16     classLabelVector = []
17     fr = open(filename)
18     index = 0
19     for line in fr.readlines():
20         line = line.strip()  # line.strip()截取掉所有的回车符
21         listFromLine = line.split(‘\t‘)   # 将整行数据分割成一个元素列表
22         returnMat[index, :] = listFromLine[0:3]
23         classLabelVector.append(int(listFromLine[-1]))
24         index += 1
25     return returnMat, classLabelVector
26 
27 
28 def autoNorm(dataSet):
29     ‘‘‘
30     归一化特征值
31     :param dataSet:
32     :return:
33     ‘‘‘
34     minVals = dataSet.min(0)   # 存放每列中的最小值，一行三列的形式
35     maxVals = dataSet.max(0)   # 存放每列中的最大值，一行三列的形式
36     ranges = maxVals - minVals
37     normDataSet = zeros(shape(dataSet))    # 1000行3列
38     m = dataSet.shape[0]   # m=1000
39     normDataSet = dataSet - tile(minVals, (m,1))   # tile将minVals复制成为1000行3列的形式
40     normDataSet = normDataSet/tile(ranges, (m,1))  #  归一化后的所有数据
41     return normDataSet, ranges, minVals
42 
43 
44 def classify0(inX, dataSet, labels, k):
45     ‘‘‘
46     K 近邻算法
47     :param inX: 用于分类的输入向量
48     :param dataSet: 输入的训练样本集
49     :param labels: 标签向量
50     :param k: 选择最近邻的数目
51     :return:
52     ‘‘‘
53     dataSetSize = dataSet.shape[0]
54     diffMat = tile(inX, (dataSetSize,1)) - dataSet  # 测试数据转化成训练数据的格式后相减
55     sqDiffMat = diffMat**2
56     sqDistances = sqDiffMat.sum(axis=1)
57     distances = sqDistances**0.5
58     sortedDistIndicies = distances.argsort()  #argsort()根据元素的值从小到大对元素进行排序，返回下标
59     classCount={}
60     for i in range(k):   #　将字典分解为元组列表
61         voteIlabel = labels[sortedDistIndicies[i]]   # 获取前K个最小距离元素的下标作为主要分类
62         classCount[voteIlabel] = classCount.get(voteIlabel,0) + 1
63     sortedClassCount = sorted(classCount.iteritems(), key=operator.itemgetter(1), reverse=True)
64     # operator.itemgetter() 获取对象的第几个域的特征
65     return sortedClassCount[0][0]
66 
67 
68 def datingClassTest():
69     ‘‘‘
70     分类器对于约会网站的测试代码
71     :return:
72     ‘‘‘
73     hoRatio = 0.1  #hold out 10%
74     datingDataMat,datingLabels = file2matrix(‘datingTestSet2.txt‘)
75     normMat, ranges, minVals = autoNorm(datingDataMat)     # norMat归一化后的特征向量，range值域，minvals每列的最小值
76     m = normMat.shape[0]    # 数据的行数
77     numTestVecs = int(m*hoRatio)           # 作为测试数据的个数
78     errorCount = 0.0
79     for i in range(numTestVecs):
80         classifierResult = classify0(normMat[i,:],normMat[numTestVecs:m,:],datingLabels[numTestVecs:m],3)
81         print ("分类器返回的标签为: %d, 实际的标签为: %d" % (classifierResult, datingLabels[i]))
82         if (classifierResult != datingLabels[i]): errorCount += 1.0
83     print ("整体的错误率为: %f" % (errorCount/float(numTestVecs)))
84     print (errorCount)
85 
86 
87 
88 if __name__==‘__main__‘:
89     datingClassTest();

 1 #!/usr/bin/env python
 2 # -*- coding: utf-8 -*-
 3 from numpy import *
 4 import operator
 5 from os import listdir   # 列出给定目录的文件名
 6 
 7 def img2vector(filename):
 8     ‘‘‘
 9     把一个32X32的图像矩阵转化为一个1X1024的向量
10     :param filename:
11     :return:
12     ‘‘‘
13     returnVect = zeros((1,1024))
14     fr = open(filename)
15     for i in range(32):
16         lineStr = fr.readline()
17         for j in range(32):
18             returnVect[0,32*i+j] = int(lineStr[j])
19     return returnVect
20 
21 
22 def autoNorm(dataSet):
23     ‘‘‘
24     归一化特征值
25     :param dataSet:
26     :return:
27     ‘‘‘
28     minVals = dataSet.min(0)   # 存放每列中的最小值，一行三列的形式
29     maxVals = dataSet.max(0)   # 存放每列中的最大值，一行三列的形式
30     ranges = maxVals - minVals
31     normDataSet = zeros(shape(dataSet))    # 1000行3列
32     m = dataSet.shape[0]   # m=1000
33     normDataSet = dataSet - tile(minVals, (m,1))   # tile将minVals复制成为1000行3列的形式
34     normDataSet = normDataSet/tile(ranges, (m,1))  #  归一化后的所有数据
35     return normDataSet, ranges, minVals
36 
37 
38 def classify0(inX, dataSet, labels, k):
39     ‘‘‘
40     K 近邻算法
41     :param inX: 用于分类的输入向量
42     :param dataSet: 输入的训练样本集
43     :param labels: 标签向量
44     :param k: 选择最近邻的数目
45     :return:
46     ‘‘‘
47     dataSetSize = dataSet.shape[0]
48     diffMat = tile(inX, (dataSetSize,1)) - dataSet  # 测试数据转化成训练数据的格式后相减
49     sqDiffMat = diffMat**2
50     sqDistances = sqDiffMat.sum(axis=1)
51     distances = sqDistances**0.5
52     sortedDistIndicies = distances.argsort()  #argsort()根据元素的值从小到大对元素进行排序，返回下标
53     classCount={}
54     for i in range(k):   #　将字典分解为元组列表
55         voteIlabel = labels[sortedDistIndicies[i]]   # 获取前K个最小距离元素的下标作为主要分类
56         classCount[voteIlabel] = classCount.get(voteIlabel,0) + 1
57     sortedClassCount = sorted(classCount.iteritems(), key=operator.itemgetter(1), reverse=True)
58     # operator.itemgetter() 获取对象的第几个域的特征
59     return sortedClassCount[0][0]
60 
61 
62 def handwritingClassTest():
63     hwLabels = []
64     trainingFileList = listdir(‘trainingDigits‘)
65     m = len(trainingFileList)
66     trainingMat = zeros((m,1024))
67     for i in range(m):
68         fileNameStr = trainingFileList[i]
69         fileStr = fileNameStr.split(‘.‘)[0]     #take off .txt
70         classNumStr = int(fileStr.split(‘_‘)[0])
71         hwLabels.append(classNumStr)
72         trainingMat[i,:] = img2vector(‘trainingDigits/%s‘ % fileNameStr)
73     testFileList = listdir(‘testDigits‘)
74     errorCount = 0.0
75     mTest = len(testFileList)
76     for i in range(mTest):
77         fileNameStr = testFileList[i]
78         fileStr = fileNameStr.split(‘.‘)[0]     
79         classNumStr = int(fileStr.split(‘_‘)[0])
80         vectorUnderTest = img2vector(‘testDigits/%s‘ % fileNameStr)
81         classifierResult = classify0(vectorUnderTest, trainingMat, hwLabels, 3)
82         print "the classifier came back with: %d, the real answer is: %d" % (classifierResult, classNumStr)
83         if (classifierResult != classNumStr): errorCount += 1.0
84     print "\nthe total number of errors is: %d" % errorCount
85     print "\nthe total error rate is: %f" % (errorCount/float(mTest))
86 
87 
88 if __name__==‘__main__‘:
89     handwritingClassTest()