Study on Commercial Bank Risk Early Warning System Based on UDM and Self-Adaptive RBFNN

摘要

By using scientific uniform design method,the representative and uniformity samples are designed. Thus,the multi-factors' and multi-levels' self- learning training is arranged using limited experiments times,and the self-adaptive RBFNN are adopted to realize the bank risk early warning diagnosis. Experiments show the results between self- adaptive RBFNN evaluation and experts fuzzy comprehensive evaluation (FCE) are very close,The generalization ability of self-adaptive RBFNN with UDM is far better than that of traditional RBFNN with Monte-Carlo method The self-adaptive RBFNN with UDM realizes non-linear approaching ability of evaluation,meantime conquers the capability limitation of traditional RBFNN and BP neural network,and avoids the subjectivity and uncertainty of traditional FCE.