tt.zip_UJUN_python/digital/KNN_rockz2j

#!/usr/bin/env python # -*- coding: utf-8 -*- from Digit import Digit from optparse import OptionParser from variantKNN import notKNN import pickle import os import glob # NOQA N = 100 K = 1 def _initializeDigitsFromFile(filename): result = [] with open(filename) as trainFile: for line in trainFile: tempDigit = Digit(line) result.append(tempDigit) return result def trainTheModel(trainFile, trainObj=None): """The function will train from the file, if trainObject is None, will genernate new one, otherwise, reuse the exist trainObject and go on training. :trainFile: The file contained the data for training :trainObj: The trained object model :returns: The trained object model """ trainList = _initializeDigitsFromFile(trainFile) # digit1 = trainList[0] # print(digit1) # print(digit1.getPointsWithNumberList()) # print(digit1.getSortedPoints()) # print(digit1.getTarget()) # print(digit1.imagewithpredicition(5)) trainMatrix = [] for i in trainList: trainMatrix.append(i.getPointsWithNumberList()) # trainMatrix.append(i.getResamblePoints()) targets = [] for i in trainList: targets.append(i.getTarget()) trainedObj = notKNN() # trainedObj.training(trainMatrix[:2000], targets[:2000], N) trainedObj.training(trainMatrix, targets, N) return trainedObj def savetofile(predfile, points, predResults): """Save the prediction to the file predfile: The file to store the result points: The points list predResult: The list of prediction result :returns: None """ with open(predfile, 'w', encoding='utf8') as saveFile: for i in range(len(predResults)): saveFile.write(points[i].imagewithpredicition(predResults[i])) def prediction(predFile, trainedModelObj): """The function will prediction the data from the predFile. If the prediction file had target, will report the sucseful rate. If the prediction file had no target, will save the file with prediction numbers :predFile: The file contain the prediction digits. :returns: None """ predList = _initializeDigitsFromFile(predFile) predMatrix = [] needSaveResult = False for i in predList: predMatrix.append(i.getPointsWithNumberList()) # predMatrix.append(i.getResamblePoints()) targets = [] for i in predList: targets.append(i.getTarget()) if targets[0] == -1: needSaveResult = True # res = trainedModelObj.predict(predMatrix, targets) print("predmatrix len={0}".format(len(predMatrix))) if needSaveResult is True: res = trainedModelObj.predict(predMatrix, None, K) savetofile(predFile, predList, res) return # if the predction file with targets, will caculate the hit rate. res = trainedModelObj.predict(predMatrix, targets, K) count = 0 for i in range(len(res)): if res[i] == targets[i]: count += 1 accurate = float(count) / len(res) * 100.0 print("Test set total = {0}".format(len(res))) print("Test match = {0}".format(count)) print("Accuracy = {0}%".format(accurate)) pass def loadTrainedModel(filename): """Load the trainedObject from file, if file non-exsit, retrun None return None or saved tranedObject """ if not(os.path.exists(filename) and os.path.isfile(filename)): return None with open(filename, 'rb') as dbfile: trainedObj = pickle.load(dbfile) return trainedObj return None def saveTrainedModel(filename, trainedObj): """Save the trainedObject to file. :returns: None """ with open(filename, 'wb') as dbfile: pickle.dump(trainedObj, dbfile) pass def walkTrainFileInDir(folder, outputfilename): filelocation = folder filearray = [] output = [] for filename in glob.glob(filelocation + "*.tra"): filearray.append(filename) print(filearray) for i in range(len(filearray)): fname = filearray[i] count = 1 with open(fname) as trainFile: for line in trainFile: line = line.strip('\n') if len(line.strip()) <= 1: continue trainTarget = count % 10 linestr = line + ", " + str(trainTarget) + ('\n') output.append(linestr) with open(outputfilename, 'w') as outputfile: for i in output: outputfile.write(i) pass def main(): """The main function """ try: # Parse the input options opt = OptionParser() opt.add_option('-t', '--trainFile', dest='trainFile', type=str, help='the file stored the pendigits train data') opt.add_option('-p', '--predictionFile', dest='predictionFile', type=str, help='the file stored the pendigits test data') opt.add_option('-n', '--nosave', dest='nosave', action="store_true", default=False, help='if set --nosave, the program will not save the\ model which ahas been trained.') opt.add_option('-w', '--walkfolder', dest='walkfolder', type=str, help='the walk folder which locate the *.tra files') (options, args) = opt.parse_args() # first try to load the trained model, if no traied model, the # trainTheModel will create one, so -t is a must. if options.walkfolder: walkTrainFileInDir(options.walkfolder, "output") return saveTrainedFile = "TrainObj-pickle" trainedModelObj = loadTrainedModel(saveTrainedFile) if options.trainFile: trainedModelObj = trainTheModel(options.trainFile) if options.predictionFile: prediction(options.predictionFile, trainedModelObj) # trainedModelObj.image() if not options.nosave: saveTrainedModel(saveTrainedFile, trainedModelObj) except Exception as e: print("Check if TrainObj-pickle exsit, if none, train again") raise e if __name__ == "__main__": main()