Vector Based Programming for Machine Learning

Diterbitkan 48 tahun 10 bulan yang lalu

Nowadays Machine Learning course is substantially taught in curriculum of computer science because some technical approaches in the course are urgently required to solve computational problems. The Department of Information and Computer in Electronic Engineering Polytechnic Institute of Surabaya (EEPIS) therefore decided to adopt the Machine Learning course and implement the course as one of main subjects for the students of D4 (Bachelor of Applied science) level. However, students sometimes face difficulties in practically understanding and implementing several machine learning’s approaches into computational code. This final project proposes a new tool for assisting the students by presenting a vector based programming library for Machine Learning Course in EEPIS. It is developed under Java and used for a specific-purpose for vector manipulation to make an easier-use of computational solves with matrix-operational functions. The proposed library consists of five layers, which are (1) metadata generation, (2) matrix operation, (3) vector manipulation, (4) technical approach, and (5) application layer. A series of experimental study is conducted for applicability testing, consists of nine issues: (1) readCSV(), (2) printMatrix(), (3) saveCSV(), (4) addition(), (5) transpose(), (6) sortMatrix(), (7) singleLinkage(), (8) averageLinkage(), and (9) kmeans(). All of them show good result in handling various data sizes. The proposed library is also tested by 31 EEPIS students. As the result 28 students agree that it is suitable to be used for assisting EEPIS students in Machine Learning course.



    2 + 6 =