Business Cycle Forecasting Model using Fuzzy Interactive Naive Bayesian Network

Abstract

This study presents a business cycle forecasting model using economic indicators based on Fuzzy Interactive Naive Bayesian (FINB) network. FINB classifier is a modified model to enhance the classification capacity by weakening the conditional independence of naive Bayesian network. In particular, it facilitates the construction of an interaction network map consisting of leading indexes, thereby clarifying the degree of influence of a certain index to and from other indexes. The experimental results of the final interaction network map provide valuable information hidden in the prediction mechanism. Particularly, under the changing economic environment in the wake of the global financial crisis of 2008, future economic situations are hard to predict owing to the complexity of financial systems in which increased variables and the heightened interdependencies between them are interwoven within the system. These factors in turn aggravated the fragility of the financial system as a whole, resulting in increased uncertainty in predicting the outlook of the world economy. The proposed interaction network map described here will provide new insight into how the financial mechanism operates, new information pinpointing the main factors of the business cycle.