Forecasting the price of nonrenewable commodity such as crude oil is a critical task and requires careful attention. Due to the vital role of nonrenewable commodity in the economics of an organization, forecasting its price has attracted both researchers and practitioners. In this paper, a relatively new Swarm Intelligence (SI) technique, namely Grey Wolf Optimization (GWO) is utilized as a method for short term time series forecasting. Classified as a nature inspired metaheuristic algorithm, the GWO emerged from the observation of leadership hierarchy and hunting mechanism of grey wolves in the nature. It incorporates of four groups which each of them belong to different level of hierarchy. The GWO has been proven to be comparable to existing optimization algorithms, thus carries a great potential for the said time series data. Realized in crude oil price time series data, the efficiency of GWO is measured based on statistical metric viz. Mean Absoluter Percentage Error (MAPE) and is compared against two Evolutionary Computation (EC) algorithms namely Artificial Bee Colony (ABC) and Differential Evolution (DE). Based on the obtained findings, it is noted that the GWO produces the lowest MAPE which is at 5.4779% while the ABC obtained a similar reading at 5.4170%. However, the performance of DE is not at the same par as it produces 11.9320% error rate. Thus, it can be concluded that the GWO may become a competitor in the domain of time series forecasting.
Zuriani Mustaffa, Universiti Utara Malaysia, Malaysia
Yuhanis Yusof, Universiti Utara Malaysia, Malaysia
Siti Sakira Kamaruddin, Universiti Utara Malaysia, Malaysia
Stream: Technology and Society – Green Computing: ICT
This paper is part of the ACSET2014 Conference Proceedings (View)
View / Download the full paper in a new tab/window