Abstract
Commercial banks are of great value to social and economic development. Therefore, how to accurately evaluate their credit risk and establish a credit risk prevention system has important theoretical and practical significance. This paper combines BP neural network with a mutation genetic algorithm, focuses on the credit risk assessment of commercial banks, applies neural network as the main modeling tool of the credit risk assessment of commercial banks, and uses the mutation genetic algorithm to optimize the main parameter combination of neural network, so as to give better play to the efficiency of neural network. After verification of various evaluation models, the accuracy of the evaluation model designed in this paper is more than 65%, while the acceptability of the evaluation results optimized by the mutation genetic algorithm is more than 85%. Compared with the accuracy of about 50% of the traditional credit scoring method, the accuracy of the credit risk evaluation using neural network technology is improved by more than 10%. It is proved that the performance of the optimized algorithm is better than that of the traditional neural network algorithm. It has important theoretical and practical significance for the establishment of the credit risk prevention system of commercial banks.
Highlights
Erefore, it is necessary to explore the index system, evaluation method, and practical application of comprehensive evaluation of credit risk. erefore, it will be of great significance to accelerate the research work of the credit risk assessment, find the credit risk assessment method and model suitable for the actual situation of commercial banks as soon as possible, provide the basis for loan decision making of commercial banks, and improve the competitiveness of commercial banks [2]
Some statistics-based methods divide the interval with each attribute value and create larger intervals by merging adjacent intervals similar to those based on statistical tests. e outlier analysis function of data mining is used to find the samples containing outlier data in the training samples, and discard or flatten these samples so that the neural network model can more truly reflect the mapping relationship between credit indicators and credit ratings [6]
A backpropagation network includes input layer, output layer, and several hidden layers. e principle of the network for the credit risk assessment of commercial banks is to take the index information used to measure the financial and nonfinancial status of loan enterprises as the input vector of God and network, the classification result is used as the output vector of neural network, and the network is trained with training samples to obtain different output values
Summary
For the credit risk assessment model, its structure can be expressed by the number of input and output nodes, the number of hidden layers, and the number of nodes in each hidden layer of backpropagation network. It is generally believed that increasing the number of hidden layers can reduce the network error and improve the classification accuracy of the evaluation model, but it makes the model complex, the training time is too long, and it tends to “overfit.”. When the neural network model enters the later stage of training, the more stable is the connection weight between neurons; at this time, the learning rate should tend to be smaller, because the larger learning rate is easy to cause oscillation in the modification process of weight W. E momentum adding method is suitable for batch learning In this way, the error is the sum of the output errors of all training samples, that is,. Error backpropagation is that the output error is transmitted back to the input layer by layer through the hidden layer in some form, and the error is allocated to all units of each layer, so as to obtain the error signal of each layer, which is used as the basis for correcting the weight of the unit. e credit risk assessment model optimized adopts the mutation genetic algorithm
Sign-in/Register to view highlights & summary Sign-in/Register to access full text options 
Free) 
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call
