A genetic-based hybrid approach to corporate failure prediction. (17th March 2008)

• Record Type: Journal Article
• Authors: Lin, Ping-Chen; Chen, Jiah-Shing
• Is Part Of: International journal of electronic finance. Volume 2:Number 2(2008)
• Page Start: 241
• Page End: 255
• Publication Date: 2008-03-17
• Subjects: corporate failure -- genetic algorithms -- GAs -- neural networks -- logistic regression -- discriminant analysis -- e-finance -- electronic finance -- failure prediction; Financial services industry -- Computer networks -- Periodicals; Electronic commerce -- Periodicals; 332.178
• Languages: English
• ISSNs: 1746-0069

Abstract

This paper proposes a genetic-based hybrid approach to predict the possibility of corporate failure. We use Genetic Algorithm (GA) to select the critical variables set and optimise the weight of each classifier for integrating the best features of several classification approaches (such as discriminant analysis, logistic regression and neural networks) in order to enhance prediction results. GA with nonlinear searching capabilities extracts more critical feature variables if compared with the Stepwise Method. This means that the undesirable variables for classification models will be cleaned out by GA. In addition, our experimental results show that this hybrid approach obtains better prediction performance than when using a single approach effectively.