An application of intelligent neural network to time series business fluctuation prediction

Economy is a dynamic system that inherits nonlinearity through long term trends, seasonal patterns, cyclical movements, and irregular factors. Time series prediction of business cycle indicators plays a critical role in managing an economy. Artificial neural network models have been successfully applied with univariate time series, multivariate datasets, classification techniques, and integrated techniques that incorporate various methods for economy prediction. Flexible intelligent systems for soft computing (FISSC) allow dynamic multi-level reasoning and inference using a fuzzy neural multicriteria group decision making framework. This paper focus on the application of the FISSC intelligent neural module to time series economy prediction. The neural module performance is evaluated in terms of convergence, generalization, scalability, sensitivity, and structural stability using the USA business cyclical indicator data set. This data set includes the lead, lag, and coincidental indicators that span over 420 months. The results are presented and the authors conclude with a discussion on their ongoing research direction