PERAMALAN LALU-LINTAS JARINGAN UGM HOTSPOT MENGGUNAKAN METODE SEASONAL AUTOREGRESSIVE INTEGRATED MOVING AVERAGE

ISNA ALFI BUSTONI, Adhistya Erna Permanasari, S.T., M.T., Ph.D.

2014 | Skripsi | TEKNOLOGI INFORMASI

Manajemen bandwidth pada jaringan Wireless UGM-Hotspot di lingkungan Jurusan Teknik Elektro dan Teknologi Informasi UGM masih dikelola secara manual sehingga pembagian bandwidth kurang efektif. Untuk mengatasi hal tersebut diperlukan automasi. Namun, sebelum itu diperlukan analisis untuk menemukan model peramalan lalu-lintas jaringan UGM-Hotspot. Salah satu metode peramalan yang banyak digunakan adalah metode ARIMA (Autoregressive Integrated Moving Average). Namun metode ini kurang akurat untuk memodelkan lalu-lintas jaringan UGM-Hotspot yang mempunyai tren musiman. Sehingga, penelitian ini menggunakan metode Seasonal ARIMA(SARIMA) dengan penambahan deteksi outlier(pencilan) agar hasil lebih akurat. Berdasarkan hasil, diperoleh data MAPE(Mean Absolute Percentage Error) untuk model SARIMA dengan deteksi outlier (14%) lebih baik dibandingkan model SARIMA tanpa deteksi outlier (32%).

Bandwidth management for UGM - Hotspot Wireless network in the Department of Electrical Engineering and Information Technology UGM is still managed manually by looking at traffic data at one time then share it manually. It make bandwidth sharing becomes less effective. It is necessary to automatically adjust bandwidth allocation based on the everyday use. However, before reaching the automation process, it requires an analysis to find the appropriate forecasting model that can be applied in the development of automation. Different types of methods are used to produce accurate forecasting of network bandwidth, one of them is ARIMA (Autoregressive Integrated Moving Average) method. However, this method is less accurate for modeling the UGM - Hotspot network traffic data because its seasonal trends. Thus, this study using Seasonal ARIMA( SARIMA ) with the addition of outlier detection so that the result becomes more accurate . Based on the result, MAPE(Mean Absolute Percentage Error) for SARIMA model with outlier detection (14 %) is better than SARIMA model without outlier detection (32 %).

Kata Kunci : Peramalan, Bandwidth, Wireless, SARIMA, Outlier