Research On The Design Of Personal Credit Evaluation Model And Related Issues

In recent years,the users of Internet personal credit loans are increasing day by day,and the credit risks are constantly expanding.How to build an effective personal credit evaluation model and realize accurate personal credit evaluation has become an urgent problem for banks and other financial institutions.Generally,personal credit loan data sets have the attributes of unbalanced categories and high feature dimensions.The resulting personal credit evaluation models trained based on such data sets will be affected by unbalanced data issues and high-dimensional feature issues,leading to classification accuracy of the model is low and the recall ability of default users is low.In order to solve the problem of umbalanced data and high-dimensional features faced by the personal credit evaluation model,and build an effective personal credit evaluation model,this paper introduces the unbalanced data processing and feature selection into the personal credit evaluation model.Due to the insufficiency of the existing unbalanced data processing and feature selection,this paper aims at the unbalanced data problem faced by the personal credit evaluation model,combined with the inverse random under sampling strategy and the "Stacking+Bagging" integrated learning model,and proposes an unbalanced data processing method based on Stacking inverse random under sampling;in view of the high-dimensional feature problem faced by personal credit evaluation models,combined with simulated annealing,disturbance factor and adaptive multi-neighborhood improvements,a feature selection method based on improved multi-neighborhood iterative local search is proposed.In this paper,the proposed unbalanced data processing method based on Stacking inverse random under sampling and the feature selection method based on improved multi-neighborhood iterative local search are introduced into the construction of personal credit evaluation model,and an adaptive personal credit based on inverse random under sampling is proposed to achieve accurate personal credit evaluation.Based on the personal credit loan data set of Lending Club in the United States and a financial institution in Europe,comparative experiments are conducted to verify that the proposed unbalanced data processing method based on Stacking inverse random under sampling has strong discriminating ability and robustness,and can realize more effective disequilibrium data processing;the feature selection method based on improved multineighborhood iterative local search has a strong global optimization ability and can achieve good feature selection;the adaptive personal credit evaluation model based on inverse random under sampling can achieve accurate personal credit evaluation,and has good reference value for the current personal credit evaluation application of financial institutions.