Tugas Akhir Informatika
PENGENALAN POLA TULISAN TANGAN AKSARA ARAB MENGGUNAKAN CONVOLU Untuk memenuhi sebagian persyaratan mencapai derajat Sarjana S PROGRAM STUDI TEKNIK INFORMATIKA i TUGAS AKHIR PENGENALAN POLA TULISAN TANGAN AKSARA ARAB MENGGUNAKAN METODE CONVOLUTIONAL NEURAL NETWORK
Arabic is the language found in the holy book Al-Qur'an. Learning Arabic with the concept of letterforms is a very effective
method. The recognition of Arabic script handwriting patterns is one of the previous studies, where the accurate results
can vary according to the research method used. This study aims to build a machine learning model and the resulting
accuracy of handwriting Arabic script recognition using convolution neural network, and to correct some of the
deficiencies in previous research. Convolution neural network is a classification method by providing labels when learning
or classified as supervised learning. The data used for this research is data that comes from handwriting on A4 HVS paper
using markers with two categories, namely ages 5 to 20 years and ages 20 years and over in order to get handwriting
variations. Model testing is carried out on research data and previous research data. The research produced an
architectural model with 3 convolution layers each with 128 filters, 5x5 kernel size, 1HL each layer with 128 neurons,
30% dropout weight, 0.001 learning rate, 64x64 image size, normalization with ReLU activation function, and 1 input
image dimension, comparison of testing data and training 70:30 and accuracy is 78.10%.
Tidak ada salinan data
Tidak tersedia versi lain