An Improved Economic Early Warning Based on Rough Set and Support Vector Machine

Abstract

Economic early warning (EEW) helps decision-making by judging the tendency of economic development. However, little research is considered about the noise problem commonly existing in the economic data. Traditional EEW method such as Bayesian model needs the feature independent assumption; artificial neural network suffers from the over-fitting problem. This paper proposes a new method of combining rough sets and support vector machine, where rough set is applied to overcome the noise problem and eliminate the redundant economic information; and support vector machine based on structural risk minimization principle is used to solve the over-fitting and small-scale sample problem. The experiment indicates that our method has achieved a satisfying performance: 87.5% in precision in binary EEW, which is a desirable precision in EEW