Determining the best forecasting model of cement price index in Malaysia

Abstract

Malaysia is aiming towards a developed country by the year 2020. Therefore, implementation of Private Financial Initiative (PFI) in Malaysia is needed as a procurement method to improve the delivery and quality of infrastructure facilities and public services in this country. The most essential aspect that needs to be fulfilled in this program is value for money (VFM) whereby maximum efficiency and effectiveness of every purchase is attained. Hence, at the preliminary stage of this study, estimating materials price index in Malaysia is the main objective. This particular paper aims to discover the best forecasting method to estimate cement price index by different regions in Malaysia since cement is the main material used in construction industry. Cement index data used were from year 2005 to 2011 monthly data of different regions in Peninsular Malaysia, and year 2003 to 2011 monthly data in both Sabah and Sarawak. It was found that Backpropagation Neural Network (BPNN) with linear transfer function produced the most accurate and reliable results for estimating cement price index in every region in Malaysia. The neural network models selection were based on the Root Mean Squared Errors (RMSE), where the values were approximately zero errors and highly significant at p<0.01. Therefore, artificial neural network is sufficient to forecast cement price index in Malaysia. The estimated price indexes of cement will contribute significantly to value for money in PFI and soon towards Malaysian economical growth.