Title: A design of experiment based procedure for real-time forecasting

Authors: Shankar Vinay Arul, Angappa Gunasekaran, SP. Nachiappan, V. Naganathan

Addresses: Department of Mechatronics, Thiagarajar College of Engineering, Madurai, India. ' Charlton College of Business, University of Massachusetts, USA. ' Department of Mechanical Engineering, Thiagarajar College of Engineering, Madurai, India. ' Department of Mechanical Engineering, Thiagarajar College of Engineering, Madurai, India

Abstract: Time series methods base their forecasts on extrapolations from past patterns and inter-relationships. Consequently, they work well only when the future is similar to the past or when changes (by chance) happen to cancel out; they are also quite handicapped when it comes to the consideration of environmental factors. In a turbulent environment with high uncertainty, the need for accurate forecasts is paramount. Hence, this paper proposes a design-approach, which incorporates the principles of Design of Experiments (DOE) into the real-time forecasting model, such as causal methods, to minimise the standard error. DOE is employed to select the most significant factors while forming the causal equation and further to perfect the coefficients of these factors. A pilot study data has been used to validate the proposed model.

Keywords: causal methods; real-time forecasting; time series; design of experiments; DOE; uncertainty.

DOI: 10.1504/IJISE.2007.011436

International Journal of Industrial and Systems Engineering, 2007 Vol.2 No.1, pp.61 - 78

Published online: 30 Nov 2006 *

Full-text access for editors Full-text access for subscribers Purchase this article Comment on this article