Vector autoregressive (VAR) model for rain-fall forecasting in west java indonesia at the peak of the rainy season

Abstract

Rain is a natural event that occurs in every region. Rainfall intensity in several regions that are close together indicates the same rainfall pat-tern each year, especially during the peak of the rainy season. Time series modeling that uses more than one variable can be used to see the relationship between the rainfall patterns. Vector Autoregressive (VAR) is a multivariate time series method that is used to see the relation-ships between variables and to forecast. The data that is used in this study are monthly rainfall data at the peak of the rainy season (December, January, February) in three locations in West Java, namely Bandung City, Cimahi City, and West Bandung Regency with a period of time from December 1981 - February 2017. Based on the results of the analysis, it shows the similarity of rainfall patterns from the three locations based on plot data and high correlation values between locations. The appropriate VAR model is VAR(3) based on the smallest AIC value and has an average MAPE value 18,817 percent.