@phdthesis{digilib18546,
           month = {September},
           title = {PENGENALAN POLA HANDWRITINGCITRA ANGKA ARAB
MENGGUNAKAN JST BACKPROPAGATION},
          school = {UIN SUNAN KALIJAGA YOGYAKARTA},
          author = {NIM. 10651068 ANDI HAMDIANAH},
            year = {2015},
            note = {Dr. Shofwatul ?Uyun, S.T., M.Kom.},
        keywords = {ANN, Anova, Arabic numbers, Backpropagation, feature extraction,
Handwriting, PCA.},
             url = {https://digilib.uin-suka.ac.id/id/eprint/18546/},
        abstract = {Research on pattern recognition of handwriting is often done by various
methods. Methods most commonly used in the classification that is
backpropagation. This method is believed to be able to classify each class well,
but the accuracy is highly influential in the pre-processing and feature extraction.
As we know handwritten Arabic numerals have many variations writing.
Pattern recognition handwritten Arabic numerals requires 207 sample
image captured using a digital camera then stored in a computer with a file
format jpg size 4320x3240 pixels. 207 samples were split into three parts, namely,
136 training data, validation data 36 and 36 test data. The whole sample through
praprosesing and feature extraction using PCA. Identification using ANN
Backpropagation to calculate the success rate in identifying the image.
The test results ANN Backpropagation ie, 100\% on the training data, 31\%
of the data validation, and 28\% in the test data. Optimal architecture used two
input nodes, three hidden layer, and output layer 4. The optimal parameters used
are MSE 0.0001; learning rate of 0.1; 0.3 momentum; and epoch 100000.
However, the results are lacking in recognizing handwriting Arabic numerals.
From here the authors classify the data of feature extraction using ANOVA
statistical techniques. The results show 72\% have the same data variation. This
suggests that, PCA feature extraction less represent traits in pattern recognition
handwriting Arabic numerals.}
}