K-nearest neighbor actual handwritten numeral recognition

1, the leader packet

import numpy as np
import operator
from os import listdir
from sklearn.neighbors import KNeighborsClassifier as KNN

%config ZMQInteractiveShell.ast_node_interactivity='all'

 

2, the definition of a function to convert the image vectors

"" " 
Function Description: Converts a binary image 32x32 into 1x1024 vector of 

the Parameters: 
    filename - the file name 
Returns: 
    returnVect - 1x1024 vector return binary image 
", "" 
DEF img2vector (filename):
     # create 1x1024 zero vector 
    returnVect = np. zeros ((. 1, 1024 ))
     # open files 
    fr = open (filename)
     # rows read 
    for I in Range (32 ):
         # read data line 
        lineStr = fr.readline ()
         # the first 32 elements in each row a added to the returnVect 
        for J in Range (32 ): 
            returnVect [0,* I + J 32] = int (lineStr [J])
     # the return 1x1024 vector conversion 
    return returnVect

 

3, the definition of handwritten numeral recognition system function

 

"" " 
Function Description: Handwriting digital Category Test 

Parameters: 
    None 
Returns: 
    None 
." "" 
DEF handwritingClassTest ():
     # training set Labels 
    hwLabels = []
     # Returns the file name under trainingDigits directory 
    trainingFileList = listdir ( ' trainingDigits ' )
     # returns the file folder number 
    m = len (trainingFileList)
     # initialization Mat training matrix, the training set 
    trainingMat np.zeros = ((m, 1024 ))
     # from a centralized file parsed category training set 
    for I in Range (m):
         # get file name
        = fileNameStr trainingFileList [I]
         # obtained classification numbers 
        classNumber = int (fileNameStr.split ( ' _ ' ) [0])
         # Add category obtained in the hwLabels 
        hwLabels.append (classNumber)
         # The data of each file of 1x1024 stored trainingMat matrix 
        trainingMat [I,:] = img2vector ( ' trainingDigits / S% ' % (fileNameStr))
     # Construction KNN classifier 
    neigh = KNN (N_NEIGHBORS =. 3, algorithm = ' Auto ' )
     # fit model, trainingMat for the training matrix, hwLabels the corresponding label 
    neigh.fit (trainingMat, hwLabels)
     #Returns the file in testDigits directory listing 
    testFileList = the listdir ( ' testDigits ' )
     # error check count 
    errorCount = 0.0 # number of test data 
    MTEST = len (testFileList)
     # parses category test set and classify test from a file for I in Range (MTEST):
         # obtain a file name 
        fileNameStr = testFileList [I]
         # obtained classification numbers 
        classNumber = int (fileNameStr.split ( ' _ ' ) [0])
         # obtained test set 1x1024 vectors used for training 
        vectorUnderTest = img2vector ( ' testDigits / S%
    
    ' % (FileNameStr))
         # get predictions 
        classifierResult = neigh.predict (vectorUnderTest)
         # print 
        Print ( ' classification returns a value of% d \ t real results d% ' % (classifierResult, classNumber))
         IF (classifierResult! = ClassNumber ): 
            errorCount + = 1.0
     Print ( ' total wrong data% d \ n error rate %% F% ' % (errorCount, errorCount / MTEST * 100))

 

4 run results

 

if __name__ == "__main__":
    handwritingClassTest()

 

Dataset Address:

Link: https: //pan.baidu.com/s/1-F2LyVh63i4yjIwweTYjNg
extraction code: 3gsa

reference:

1, "machine learning real" books

2、https://github.com/apachecn/AiLearning

3、https://cuijiahua.com/blog/2017/11/ml_1_knn.html

4, eye depth of machine learning combat training camp Homework ( http://www.deepshare.net/ )