PERBANDINGAN METODE FUZZY TIME SERIES SONG-CHISSOM DAN METODE FUZZY TIME SERIES SINGH UNTUK PREDIKSI KEBUTUHAN BANDWIDTH PADA JARINGAN KOMPUTER

Abstract

The prediction of bandwidth need on computer network in higher education institutions is required to identify the bandwidth need that will arise, both on LAN network and internet conection, thus it can enhance the academic service quality and determine how much to pay the bandwidth rent. This research implemented the fuzzy time series Song-Chissom and fuzzy time series Singh methods to predict bandwidth utilizers’ historical data in terms of having courses, having examinations, and having off-days adjusted to the academic calender. The predicting system by using fuzzy time series captured a pattern of the previous data, then used to project the further data. Form the results of the tests obtained by using Song-Chissom method, for the term of having courses had an average error 7.669%, and Singh method had the average error 8.523%. For the term of having examinations, the Song-Chissom obtained the average error 12.456%, and Singh method had the average error 35.781%. As for the term of having off-days, Song-Chissom method obtained the average error 7.376% and Singh method had the average error 68.774%. Generally, it can be stated that fuzzy time series Song-Chissom method gained the better prediction than the fuzzy time series Singh method, which its prediction results were close in the factual data. From the results of bandwidth need predictions, it is recommended to Politeknik Negeri Sriwijaya that 16 Mbps is for the term of having courses, 4 Mbps is for the term of having examinations, and 6 Mbps is for the term of having off-days. Key Words: Prediction, Bandwidth, Fuzzy time series