Research And Application Of Personal Credit Evaluation Method Based On Flexible Neural Tree Model

With the rapid development of domestic economy and the substantial improvement of people’s consumption ability,more and more credit consumer products appear in our side.Credit consumption has become the key business of domestic commercial banks,and with the continuous development of the Internet,more and more Internet credit consumer products have been introduced to the market.However,commercial banks and Internet financial institutions have certain deficiencies in credit risk management,and there is no unified credit evaluation method in China,so there are many acts of breach of trust,which has caused great losses to financial institutions.Therefore,it is of great practical value and research significance to study and construct a complete and effective personal credit evaluation model to solve the problems in domestic credit evaluation.At present,financial institutions mainly use borrowers’ historical credit consumption records,assets and basic personal information as credit evaluation data.These data have disadvantages such as high dimension and high noise,which will cause problems such as long time of customer information collection,too many evaluation indicators and too long time of audit.By analyzing the current development status of personal credit risk assessment,this paper studies the selection of personal credit risk assessment indicators and the assessment classification model.Firstly,the flexible neural tree model fused with BP is applied to personal credit risk assessment,and then the flexible neural tree model based on pseudo inverse learning is constructed and used to solve the problem of personal credit risk assessment.The experimental results show that the Pseudo-inverse learning based flexible neural tree can be used to screen the important evaluation indicators and improve the interpretation of the evaluation results.In addition,it also improves the classification accuracy of bank credit risk assessment and provides suggestions for the development of subsequent credit risk assessment.The main research contents of this paper are as follows:(1)The flexible neural tree model fused with BP is applied to personal credit risk assessment.Firstly,genetic programming and particle swarm optimization were used to generate a flexible neural tree.The tree structure was constructed from the selected important input attributes.Then the neural network is constructed on the basis of the obtained tree structure,and the parameters are optimized by using BP algorithm,so as to achieve better classification effect while obtaining important indexes.Through experiments,this model is better than support vector machine,logistic regression,decision tree and other algorithms in classification accuracy.(2)A flexible neural tree model based on pseudo inverse learning is constructed and applied to personal credit risk assessment.This model only uses the optimal part of tree structure design in the flexible neural tree to get the tree structure based on the data set.The depth of the tree structure is limited to three layers,and then the neural network is constructed.Pseudo inverse learning algorithm is used to accurately learn the neural network.Through experiments,this model can achieve better classification effect compared with support vector machine,logistic regression,decision tree,flexible neural tree and other classification algorithms on the condition that the running time of the flexible neural tree model with BP fusion is much reduced.(3)A credit risk assessment software based on the flexible neural tree model combined with BP is implemented.The software is modeled based on German credit data,and the flexible neural tree model integrated with BP is used to predict the personal credit risk assessment data.The software interface,database and system logic are designed according to the business process and their functions are realized to ensure the practicability and effectiveness of the software.