COMPARISON OF K-NEAREST NEIGHBOR, NAÏVE BAYES AND SUPPORT VECTOR MACHINE FOR CHARACTER RECOGNITION IN MOBILE APPLICATION; PERBANDINGAN K-NEAREST NEIGHBOR, NAÏVE BAYES DAN SUPPORT VECTOR MACHINE PADA PENGENALAN KARAKTER DALAM APLIKASI PIRANTI BERGERAK

LIANTARA, CHRISTINA CARASWATI, Reza M.I. Pulungan

2016 | Skripsi | FMIPA

Due to its variations of techniques and applications, optical character recognition is still the most popular research in digital image processing. As the increasing number of mobile devices and it becomes the future operating system, this application is built on mobile based. With this application, user can easily recognized the characters by capturing image using mobile phone camera. Open Computer Vision library is used to build the application as this library is the most popular one. Besides, machine learning algorithms are used for the recognition process in the application. There are k-nearest neighbor, naïve Bayes classifier and support vector machine algorithms which are used in this research. The performances of these three algorithms are calculated including recognition accuracy, memory usage and time consumed. From testing process, k-nearest neighbor show the best performance. It has the highest character recognition accuracy at 94% and the least average memory usage and time consumed in the training and prediction process. The result shows that it consumed average memory of 650300.7 KB in training and 650409.1 KB in the prediction and consumed average time of 128.5 microseconds in training and 520.6 microseconds in prediction.

Kata Kunci : mobile application, character recognition, OpenCV library, machine learning