Prediction of Rainfall Using ARIMA Mixed Models

Abstract

The average rainfall in Aceh Barat every year is different pattern and it is influenced by several factors. In this paper we used rainfall dataset, which is changing time to time. The change is caused by an element of fluctuate and volatility in the data. The purpose of this study was to find the best ARIMA mixed models as combination with ARCH and GARCH models. The data used in this study are rainfall data and the number of rainy days in Aceh Barat district from the period January 2008 to December 2017. The results showed that stationary rainfall in the transformation results of Zt 0.27 and the first differencing (d=1) and test results Lagrange multiplier-ARCH for rainfall data and the number of rainy days shows significant lag 4. The best model for predicting rainfall uses the ARIMA(2,1,0)- ARCH(3) model and for the number of rainy day using the ARIMA(2,0.2) model. The calculation results obtained prediction accuracy value for rainfall using ARIMA(2,1,0)- GARCH(1,3) model with MAD, RMSE, MAE, and MASE values of 1,175, 1.163, 0.941 and 0.720 respectively and for the number of rainy days using ARIMA(2,0,2) model were accuracy
value respectively 4.448, 3.849, 3.189 and 0.737.