Penggunaan Decision Tree Dengan ID3 Algorithm Untuk Mengenali Dokumen Beraksara Jawa

Abstract

This thesis’ goal is to let computer recognize Javanese letters. Before a computer is able to recognize Javanese letters, segmentation and feature extraction process is needed. After the features of Java letters that have been segmented obtained, then the features will be processed using computer learning method so that the computer can recognize the letters.

Input is in the form of .CSV file that contains features of the Javanese document image that had previously segmented. Output text is a text file that contains unicode representation letter from Java font and a .CSV file that contains the test and data classification results. This application is created with C# programming language and Microsoft Visual Studio 2013 IDE.

As for the process that is done is as follows: prepare the .CSV file resulting from the extraction of features to be used as learning materials or trainer for computer learning network. After the trainer file ready, then the network will be trained. After the training is complete, the network can receive other data input, and then classify it according to what has been learned.

Probabilistic neural network and ID3 used as computer learning methods, which after data testing, it is found that PNN give higher accuracy result than ID3.