PERAMALAN RUNTUN WAKTU MUSIMAN DENGAN MENGGUNAKAN METODE WAVELET

Abstract

ABSTRACT

Forecasting is one of important things in making decision. Forecasting’s part had covered on many fields, such as physics, environment, economics, engineering and health. A method which had been used in forecasting is time series analysis whose present values based on same kind of values on the past. Time series method that often used in seasonal data is known as SARIMA. This research proposes wavelet method, viz. MODWT (Maximal Overlap Discrete Transform), which will be used to analyze the data. Modeling’ procedure done with wavelet until scale and wavelet coefficient reached by using MODWT transformation to the time series data. Based on this transformation’s result, time series will be modeled and used to forecast one period ahead by assuming each scale and wavelet coefficient as an auto regression process which called as MAR(Multiscale Autoregressive). MAR method is used in this research and compared to SARIMA method. Based on the empiric result for two analyzed data, it shows thattwo data with wavelet method have forecasting error for testing data smaller than SARIMA method

Keywords : MODWT, MAR, SARIMA, Forecast