PEMODELAN AUTOREGRESSIVE FRACTIONALLY INTEGRATED MOVING AVERAGE (ARFIMA) UNTUK AKTIVITAS CURAH HUJAN DI KOTA MEDAN

Abstract

The Autoregressive Fractionally Integrated Moving Average (ARFIMA) model is a development of the ARIMA model with the differencing values ​​being fractional numbers. This study aims to model ARFIMA on rainfall activity data in the city of Medan which consists of three variables including air temperature, humidity and previous rainfall. In this study, a stationary test of the three data was carried out, after conducting several tests, it can be concluded that the air temperature and rainfall transformation data are stationary with respect to the mean, but the air humidity transformation data shows that the data is not stationary in the mean shown from the slow decreasing ACF plot value. and p-value is more than =0.05. After the data of air temperature, rainfall and humidity are stationary, the estimation of model parameters, diasgonic test of the model, and determination of the best model will be carried out using ARIMA. The ARIMA model equation (1,0,0)(1,0,0).The temperature is asfollows: ARIMA model equation (2,0,3)(1,0,0),Rainfall ARIMA model equation (1,1,1)(1,1,0)12 HumidityAfter the three variables get the equation from ARIMA, what is done is the data on air humidity is tested again using ARFIMA, in order to compare the suitability of the air humidity data, whether the data is more suitable for ARIMA or ARFIMA. After doing some testing and then getting the ARFIMA results(0,d,[1,2])with isRMSE 19,83846907, MAE 3,786387874, dan MAPE 3,786387874, So it can be concluded that the ARIMA method is better used on air humidity data

Keywords

• r/s
• gph
• arfima
• arima
• long memory