Kamarudin, Nur Azulia and Ismail, Suzilah (2016) Model selection approaches of water quality index data. Global Journal of Pure and Applied Mathematics, 12 (2). pp. 1821-1829. ISSN 0973-1768

PDF Restricted to Registered users only Download (870kB) | Request a copy

Abstract

Automatic model selection by using algorithm can avoid huge variability in model specification process compared to manual selection.With the employment of algorithm, the right model selected is then also used for forecasting purposes. In order to select the best model, it is vital to ensure that proper estimation method is chosen in the modelling process.Different estimators have been proposed for the estimation of parameters of a model, including the least square and iterative estimators.This study aims to evaluate the forecasting performances of two algorithms on water quality index (WQI) of a river in Malaysia based on root mean square error (RMSE) and geometric root mean square error (GRMSE).Feasible generalised least squares (FGLS) and iterative maximum likelihood (ML) estimation methods are used in the algorithms, respectively.The results showed that SUREMLE-Autometrics has surpassed SURE-Autometrics; another simultaneous selection procedure of multipleequation models.Two individual selections, namely Autometrics-SUREMLE and Autometrics-SURE, though showed consistency only for GRMSE.All in all, ML estimation is a more appropriate method to be employed in this seemingly unrelated regression equations (SURE) model selection.

Item Type:	Article
Uncontrolled Keywords:	Automated model selection, multiple equations, maximum likelihood estimation.
Subjects:	Q Science > QA Mathematics
Divisions:	School of Quantitative Sciences
Depositing User:	Mdm. Nur Azulia Kamarudin
Date Deposited:	05 Apr 2017 08:42
Last Modified:	05 Apr 2017 08:42
URI:	https://repo.uum.edu.my/id/eprint/21525

Actions (login required)

View Item