Abstract

This study investigates the foundation of an intelligent system using Artificial Intelligent (AI) technologies to assist decision makers in a specific business problem, namely business forecasting. In time series and macroeconomic modelling, there are many assumptions being imposed on the behavior and functional relationship of the underlying variables. In addition, one may face the complexity in the estimation of these models. This study uses Artificial Neural Network (ANN) and other AI technologies in an effective forecasting system in order to overcome the restrictions of traditional modelling and estimation methods. An ANN has been shown to be a universal function approximator (Cybenko, 1989; Hornik et al., 1989). It requires no prior assumptions on the behavior and functional form of the related variables but it is still able to capture the underlying dynamic and nonlinear relationships among variables in the problem space, ie. a macroeconomic model in this context. This study integrates the powerful ability of an ANN into an efficient framework incorporating Recurrent Algorithms (Jordan, 1986), Genetic Algorithms (Holland, 1975) in a Mixture-of-experts Architecture (Jacobs et al., 1991) to obtain accurate estimation and forecasts. As such, this study addresses the ability of a versatile intelligent technology to solve a general economic forecasting problem involving temporal and non-temporal variables. Using the contexts provided in the Klein Model I of the US interwar economy in 1921-1941 and the Klein-Goldberger Model of the US economy in 1929-1952, this study investigates the relative performance of the proposed system and traditional methods in modelling and forecasting a mix of economic variables. It extends these frameworks into the future to forecast with more recent data. The study specifies the conditions that will make the implementation of ANN more successful in estimation and forecasting. This study provides evidence on the effectiveness and efficiency of the proposed system. It asserts empirically the ability of the integrated ANN and GA in estimation and forecasting. The findings should contribute positively to the development of theory, methodology, and practice of using AI tools, particularly ANN and GA, to build intelligent forecasting systems.