Klasifikasi Tulisan Tangan Aksara Batak dengan Metode Hybrid CNN-SVM Berbasis Website

Abstract

Batak script is one of Indonesia's cultural heritages originating from North Sumatra which is now endangered due to the lack of use and understanding of this script. This is also influenced by its use in ancient times only limited to certain people. This research aims to develop a classification system for Batak script handwriting using the Hybrid CNN-SVM method, money can recognize five types of Batak script: Toba, Simalungun, Karo, Pakpak, and Mandailing. The CNN-SVM hybrid method works by using CNN combined with resnet-50 architecture as a feature extractor and svm is used for classification. PCA is also used after the features have been extracted from CNN, in order to reduce the dimensionality of the extracted features before entering SVM classification. Tests were carried out with five data sharing scenarios, and the best results were obtained in the fifth scenario, namely 80:10:10 with an average accuracy of 93.64% training and 95.06% testing. This model is implemented on the website, and is expected to help preserve the Batak script.