Numerical Statistic Approach For Expert System In Rainfall Prediction Based On Data Series

Numerical Statistic Approach For Expert System In Rainfall Prediction Based On Data Series

Numerical Statistic Approach For Expert System In Rainfall Prediction Based On Data Series

Vol.5 No.2 May 2013/International Journal of Engineering and Technology (IJET) (SCOPUS INDEX)

Indrabayu, A. Achmad, M.S. Pallu, N. Harun

Abstract

The potential of statistical approach in predicting rain fall is discussed in this paper. Two most implemented methods i.e. Auto-Regressive Integrated Moving Average (ARIMA) and Adaptive Splines Threshold Autoregressive (ASTAR) are compared in term of accuracy in prediction. Both methods are constructed to predict daily rainfall in the area of Makassar, Indonesia. Rain problem in Indonesia increasingly complex due to climate shifts that result in high intensity rainfall in the dry season so it is very influential on the development of many aspect of social-economy sector. A ten years daily data (2001-2010) obtained from BMKG (the Meteorology, Climatology and Geophysics). Several complementary data is also obtained from LAPAN (Government Space Agent). From various meteorological variables, four variables are selected for predicting rainfall- There are temperature, humidity, wind speed, and previous precipitation based on their high correlation to rain event.. These four variables are then input to the ARIMA and ASTAR. The accuracy of prediction is measured based on root mean square error (RMSE). ASTAR outperformed ARIMA with less RMSE which is 0.02 to 0.24.

Keywords: ARIMA, ASTAR, Expert System, Rain Prediction