Pendugaan Curah Hujan Dengan Bayesian Network Studi Kasus: Curah Hujan di Daerah Indramayu

Abstract

The objective of this study was to built the probabilistic modeling and to compare the Bayesian Networks (BNs) method which considers the spatial dependencies among variables for predict rainfall to Auto Regressive Integrated Moving Average (ARIMA) method which assumes the spatial independencies among variables. This research use rainfall data of 14 stations in period 1979 to 2001 in Indramayu, West Java. A network is generated by K2 algorithm, and this networks is used to built probabilistic model using multinomial BNs. The rainfall prediction is based on the combination of Gaussian BNs and ARIMA (BNARIMA). The result shows that BNARIMA is more effective to represent the spatial dependencies among the stations and to predict rainfall than ARIMA.