python实现kaggle中的数字识别

#-*-coding:utf-8-*- import sys,os import csv from numpy import * import operator path=sys.path[0] def loadTrainData(): trdata=list() with open(path+'/train.csv') as file: lines=csv.reader(file) for line in lines:#839*785 trdata.append(line) trdata.remove(trdata[0]) trdata=array(trdata) row,col=len(trdata),len(trdata[0]) print 'train',row,col label=[] data=zeros((row,col-1)) for i in range(0,row): label.append(trdata[i][0]) for j in range(1,col): data[i][j-1]=trdata[i][j] label=array(label).reshape(len(label),1) data=array(data).reshape(len(label),len(trdata[0])-1) label=toint(label) data=toint(data) return nomalizing(data),label def loadTestdata(): tedata=list() with open(path+'/test.csv') as file: i=1 lines=csv.reader(file) for line in lines: tedata.append(line) tedata.remove(tedata[0]) row,col=len(tedata),len(tedata[0]) print 'test',row,col data=zeros((row,col)) for i in range(0,row): for j in range(0,col): data[i][j]=tedata[i][j] data=toint(data) return nomalizing(data) def loadTestResult(): #28001*2 terdata=list() label=list() with open(path+'/sample_submission.csv') as file: lines=csv.reader(file) for line in lines: terdata.append(line) terdata.remove(terdata[0]) row,col=len(terdata),len(terdata[0]) print 'test',row,col for i in range(0,row): label.append(terdata[i][1]) label=array(label).reshape(len(label),1) return toint(label) def nomalizing(array): m,n=shape(array) for i in xrange(m): for j in xrange(n): if array[i][j]!=0: array[i][j]=1 return array def toint(array): m,n=shape(array) print m,n newdata=zeros((m,n)) for i in xrange(m): for j in xrange(n): newdata[i][j]=int(array[i][j]) return newdata def classify(inX, dataSet, labels, k): inX=mat(inX)#1*28000 dataSet=mat(dataSet)#42000*784 labels=mat(labels)#1*42000 dataSetSize = dataSet.shape[0] #42000 diffMat = tile(inX, (dataSetSize,1)) - dataSet # 将数组inX作为元素构成42000*784的矩阵，即将inX重复42000遍，并与dataSet相减，求出inX与每一个train向量之间的距离 sqDiffMat = array(diffMat)**2 #对矩阵中的每个元素求平方 distances = sqDiffMat.sum(axis=1)#对矩阵中的每一行求和，得到42000*1的矩阵 #distances = sqDistances**0.5 sortedDistIndicies = distances.argsort()#对距离矩阵由小到大排序，得到排序矩阵，其中存储每个值的索引 classCount={} for i in range(k): voteIlabel = labels[0,sortedDistIndicies[i]] classCount[voteIlabel] = classCount.get(voteIlabel,0) + 1 sortedClassCount = sorted(classCount.iteritems(), key=operator.itemgetter(1), reverse=True) return sortedClassCount[0][0] def saveResult(result): l=array(result).reshape(len(result),1) row,col= shape(l) l1=[] for i in range(row): for j in range(col): l1.append(int(float(l[i][j]))) print l1 j=1 with open('result3.csv','wb') as file: myW=csv.writer(file) tmp=[] tmp.append('ImageId') tmp.append('Label') myW.writerow(tmp) for i in l1: tmp=[] tmp.append(str(j)) j+=1 tmp.append(i) myW.writerow(tmp) #with open(path+'/result.csv') as file: # myW=csv.writer(file) # myW.writerow(result) if __name__=='__main__': trainData,trainLabel=loadTrainData()#42000*784,42000*1 testData=loadTestdata()#28000*784 testLabel=loadTestResult()#28000*1 m,n=shape(testData)#28000,784 errorCount=0# resultList=list() for i in range(m): testD=testData[i,:]#1*784 classifyResults=classify(testD,trainData,trainLabel.transpose(),9) resultList.append(classifyResults) #print "the classifier came back with: %d, the real answer is: %d" % (classifierResult, testLabel[0,i]) #if (classifyResults != testLabel[0,i]): errorCount += 1 #print "\nthe total number of errors is: %d" % errorCount #print "\nthe total error rate is: %f" % (errorCount/float(m)) saveResult(resultList)