PERAMALAN DERET WAKTU MENGGUNAKAN MODEL JARINGAN SYARAF FUNGSI BASIS RADIAL (RBFNN) DAN AUTO REGRESSIVE INTEGRATED MOVING AVERAGE (ARIMA)

The accuracy of time series forecasting is being the subject of the decision making process. Time series using a quantitative approach to the data of the past that made reference to forecast the future. Some of research that have been doing research on time series, such as using statistics, neural networks, wavelets, and fuzzy systems. These methods have different advantages and disadvantages.But often the problem in the real world is a complex problem where just a single model maybe cannot overcome that problem well. The reason to combining these two models (ARIMA and RBF) is the assumption that a single model can not completely identify all the characteristics of time series. This research will make a forecasting for data of Wholesale Price Index (WPI) and inflation of Indonesian commodity. Both of data in the range of 2006 to several months in 2012, and each of the data has 6 variables. The results of ARIMA-RBF forecasting method will be compared with ARIMA method and RBF method individually. The result of analysis show that the combination method of ARIMA and RBF give accurate results than the ARIMA model or RBF model only. The result can be seen using the visual plot, MAPE, and MSE of all the variables in the two trial data.