Research into Bank Loan Risk Based on UDM and Self-adaptive RBF Neural Network

At present, the application of neural network technology in our country´s bank loan risk evaluation is very limited. The reason lies in the difficulty to find high quality training samples for neural network self-learning. Therefore, we adopt uniform design method (UDM) to design representative, uniformity and large-scale samples. And then we use those samples to train the self-adaptive radial basis function neural network (RBFNN) which is applied to carry out the bank loan risk evaluation. The result of the experiment shows that the generalization ability of self-adaptive RBFNN combined with UDM is far better than that of traditional RBFNN based on Monte-Carlo method. The self-adaptive RBFNN combined with UDM not only realizes the self-adaptive ability and non-linear approaching ability, but also conquers the performance limitations of traditional RBFNN. And also it avoids the subjectivity and uncertainty of traditional evaluation.