On the Performance of an Intelligent Forecasting System using Modular Neural Networks in Building and Forecasting an Economic Model

摘要

An intelligent forecasting system using modular Neural Networks is presented in this study. In the instrumental stage, the system has a recurrent module to handle time series data and a standard module to handle other non-temporal variables. In the final stage, a gateway network is used to aggregate instrumental estimations into final estimations/forecasts. The system uses Genetic Algorithms for network building and variable selection. Klein-Goldberger Model of US Economy is used to represent a general business forecasting context. Findings show that the proposed system outperformed the traditional estimations and forecasts.