机器学习算法( 二、K - 近邻算法)

  1 ‘‘‘
  2     Created on Sep 16, 2010
  3     kNN: k Nearest Neighbors
  4     
  5     Input:      inX: vector to compare to existing dataset (1xN)
  6     dataSet: size m data set of known vectors (NxM)
  7     labels: data set labels (1xM vector)
  8     k: number of neighbors to use for comparison (should be an odd number)
  9     
 10     Output:     the most popular class label
 11     
 12     @author: pbharrin
 13     ‘‘‘
 14 from numpy import *
 15 import operator
 16 from os import listdir
 17 import matplotlib
 18 import matplotlib.pyplot as plt
 19 def show(d,l):
 20     #d,l=kNN.file2matrix(‘datingTestSet2.txt‘)
 21     fig=plt.figure()
 22     ax=fig.add_subplot(111)
 23     ax.scatter(d[:,0],d[:,1],15*array(l),15*array(l))
 24     plt.show()
 25 def show2():
 26     datingDataMat,datingLabels=file2matrix(‘datingTestSet2.txt‘)
 27     fig = plt.figure()
 28     ax = fig.add_subplot(111)
 29     l=datingDataMat.shape[0]
 30     X1=[]
 31     Y1=[]
 32     X2=[]
 33     Y2=[]
 34     X3=[]
 35     Y3=[]
 36     for i in range(l):
 37         if datingLabels[i]==1:
 38             X1.append(datingDataMat[i,0]);Y1.append(datingDataMat[i,1])
 39         elif datingLabels[i]==2:
 40             X2.append(datingDataMat[i,0]);Y2.append(datingDataMat[i,1])
 41         else:
 42             X3.append(datingDataMat[i,0]);Y3.append(datingDataMat[i,1])
 43     type1=ax.scatter(X1,Y1,c=‘red‘)
 44     type2=ax.scatter(X2,Y2,c=‘green‘)
 45     type3=ax.scatter(X3,Y3,c=‘blue‘)
 46     #ax.axis([-2,25,-0.2,2.0])
 47     ax.legend([type1, type2, type3], ["Did Not Like", "Liked in Small Doses", "Liked in Large Doses"], loc=2)
 48     plt.xlabel(‘Percentage of Time Spent Playing Video Games‘)
 49     plt.ylabel(‘Liters of Ice Cream Consumed Per Week‘)
 50     plt.show()
 51 
 52 def classify0(inX, dataSet, labels, k):
 53     dataSetSize = dataSet.shape[0]
 54     diffMat = tile(inX, (dataSetSize,1)) - dataSet
 55     sqDiffMat = diffMat**2
 56     sqDistances = sqDiffMat.sum(axis=1)
 57     distances = sqDistances**0.5
 58     sortedDistIndicies = distances.argsort()
 59     classCount={}
 60     for i in range(k):
 61         voteIlabel = labels[sortedDistIndicies[i]]
 62         classCount[voteIlabel] = classCount.get(voteIlabel,0) + 1
 63     sortedClassCount = sorted(classCount.iteritems(), key=operator.itemgetter(1), reverse=True)
 64     return sortedClassCount[0][0]
 65 
 66 def createDataSet():
 67     group = array([[1.0,1.1],[1.0,1.0],[0,0],[0,0.1]])
 68     labels = [‘A‘,‘A‘,‘B‘,‘B‘]
 69     return group, labels
 70 
 71 def file2matrix(filename):
 72     fr = open(filename)
 73     numberOfLines = len(fr.readlines())         #get the number of lines in the file
 74     returnMat = zeros((numberOfLines,3))        #prepare matrix to return
 75     classLabelVector = []                       #prepare labels return
 76     fr = open(filename)
 77     index = 0
 78     for line in fr.readlines():
 79         line = line.strip()
 80         listFromLine = line.split(‘\t‘)
 81         returnMat[index,:] = listFromLine[0:3]
 82         classLabelVector.append(int(listFromLine[-1]))
 83         index += 1
 84     return returnMat,classLabelVector
 85 
 86 def autoNorm(dataSet):
 87     minVals = dataSet.min(0)
 88     maxVals = dataSet.max(0)
 89     ranges = maxVals - minVals
 90     normDataSet = zeros(shape(dataSet))
 91     m = dataSet.shape[0]
 92     normDataSet = dataSet - tile(minVals, (m,1))
 93     normDataSet = normDataSet/tile(ranges, (m,1))   #element wise divide
 94     return normDataSet, ranges, minVals
 95 
 96 def datingClassTest():
 97     hoRatio = 0.50      #hold out 10%
 98     datingDataMat,datingLabels = file2matrix(‘datingTestSet2.txt‘)       #load data setfrom file
 99     normMat, ranges, minVals = autoNorm(datingDataMat)
100     m = normMat.shape[0]
101     numTestVecs = int(m*hoRatio)
102     errorCount = 0.0
103     for i in range(numTestVecs):
104         classifierResult = classify0(normMat[i,:],normMat[numTestVecs:m,:],datingLabels[numTestVecs:m],3)
105         print "the classifier came back with: %d, the real answer is: %d" % (classifierResult, datingLabels[i])
106         if (classifierResult != datingLabels[i]): errorCount += 1.0
107     print "the total error rate is: %f" % (errorCount/float(numTestVecs))
108     print errorCount
109 
110 def img2vector(filename):
111     returnVect = zeros((1,1024))
112     fr = open(filename)
113     for i in range(32):
114         lineStr = fr.readline()
115         for j in range(32):
116             returnVect[0,32*i+j] = int(lineStr[j])
117     return returnVect
118 
119 def handwritingClassTest():
120     hwLabels = []
121     trainingFileList = listdir(‘trainingDigits‘)           #load the training set
122     m = len(trainingFileList)
123     trainingMat = zeros((m,1024))
124     for i in range(m):
125         fileNameStr = trainingFileList[i]
126         fileStr = fileNameStr.split(‘.‘)[0]     #take off .txt
127         classNumStr = int(fileStr.split(‘_‘)[0])
128         hwLabels.append(classNumStr)
129         trainingMat[i,:] = img2vector(‘trainingDigits/%s‘ % fileNameStr)
130     testFileList = listdir(‘testDigits‘)        #iterate through the test set
131     errorCount = 0.0
132     mTest = len(testFileList)
133     for i in range(mTest):
134         fileNameStr = testFileList[i]
135         fileStr = fileNameStr.split(‘.‘)[0]     #take off .txt
136         classNumStr = int(fileStr.split(‘_‘)[0])
137         vectorUnderTest = img2vector(‘testDigits/%s‘ % fileNameStr)
138         classifierResult = classify0(vectorUnderTest, trainingMat, hwLabels, 3)
139         print "the classifier came back with: %d, the real answer is: %d" % (classifierResult, classNumStr)
140         if (classifierResult != classNumStr): errorCount += 1.0
141     print "\nthe total number of errors is: %d" % errorCount
142     print "\nthe total error rate is: %f" % (errorCount/float(mTest))