An Empirical Study of Dynamic Financial Early Warning Based on Grey Correlation and BP Neural Network
Wang hui zhen1
COMPUTER MODELLING & NEW TECHNOLOGIES 2014 18(12C) 943-946
1School of Business , Xi’an University of Finance and Economics , Shaanxi, Xi’an, China, 710100
Current research on early warning of financial crises mainly focuses on financial early warning such as multivariate linear pre-warning and Bayes discrimination, whereas the research methods are inclined to mathematical statistics, so there are rigorous requirements for data and hypotheses. Nevertheless, corporate financial risks and predetermined indices for early warnings are possibly changeable. In consideration of ineffective control of crises by earning warning, crises were dynamically monitored with a BP neural network based on Grey Model (1, 1) from the perspective of risk and crisis forecast (the only measures available for financial crisis in modern enterprises). Besides, a three-tier BP neural network was constructed by transforming fitting accuracy of exponential functions. The results have suggested that the changing indices about corporate financial crises have direct impacts upon corresponding early warning results. All simulation trainings based on BP neural network have been validated and can be used to further verify dynamic grey correlations in the process of financial early warning. Furthermore, all ST enterprises were predicted to face crisis by the pre-warning mechanism based on grey model, BP neural network training and the analog control, while corresponding non-ST enterprises were forecasted to be sound. Hence, it is helpful for listed enterprises to effectively forecast their possible and potential financial crises.