Author/Authors :
Zaharim, Azami Universiti Kebangsaan Malaysia - Faculty of Engineering - Department of Archtitecture, Malaysia , Ismail, Abdul Halim Universiti Kebangsaan Malaysia - Faculty of Engineering - Department of Archtitecture, Malaysia , Md Darus, Zuhairuse Universiti Kebangsaan Malaysia - Faculty of Engineering - Department of Archtitecture, Malaysia , Yahya, Mohd. Sahar university of malaya - Matriculation Centre, Malaysia , Mohamed, Ibrahim university of malaya - Institute of Mathematical Sciences, Malaysia , Mohd Nopiah, Zulkifli Universiti Kebangsaan Malaysia - Faculty of Engineering - Department of Mechanical and Material, Malaysia
Abstract :
Surveyed from 160 time series data used in scientific articles concluded that 10-13% of them were generated by the nonlinear process. Several nonlinear models are available but only the bilinear model will be considered here. The bilinear model is basically an extension of the linear model ARIMA. It was first introduced by control theorists in 1972 before it was documented by Granger and Anderson in 1978. The bilinear model is believed to be able to fit hydrological and meteorological data well. In this paper, the coefficients of a special case of bilinear model, BL(1,1,1,1), are estimated using nonlinear least squares method. Results of modeling the linear and bilinear models on water level data from the Kelantan river in Malaysia when compared and evaluated gave better fitting to the bilinear model
Keywords :
Bilinear , ARIMA models , nonlinear , water level
