Research On The Method Of Consumer Financial Default Prediction Based On LOMST-VS Model

The rapid development of information technology on the Internet and the increasing consumption level of people have created conditions for the booming development of consumer finance.Compared with traditional lending methods,more and more consumers are more inclined to choose among consumption channels such as the Internet and e-commerce.More convenient and faster consumer finance lending.However,the difference between consumer finance lending and traditional lending methods makes it impossible for consumer finance companies to assess the default risk of consumer finance relatively objectively based on the existing traditional risk assessment models,which may result in the loss of the company's interests,and inaccurate models reject consumers.Borrowing also violates the rights and interests of consumers.Therefore,it is imperative to build a default detection model in the consumer finance field with good performance.This dissertation proposes a LOMST-VS two-layer composite model for consumer finance loan default prediction based on hybrid sampling and multi-model fusion.The main research contents are as follows:(1)Training feature engineering: Based on the data in the consumer finance field,the feature engineering model is used to extract and integrate multi-dimensional feature attributes with the characteristics of default prediction in the consumer finance field.(2)Constructing a mixed sampling model for unbalanced consumer financial data: In view of the imbalance of the amount of default data in the experimental data and the influence of noise points on the prediction model,this dissertation proposes an improved model based on the SMOTE algorithm—LOMST hybrid The sampling model reduces the difference ratio between positive and negative sample data,which is more conducive to subsequent model training.(3)Constructing the LOMST-VS consumer finance default detection model: In the modeling process of LOMST mixed sampling + single model,there is a problem that the boundary point blur affects the model effect.Based on this,this dissertation further proposes the LOMST-VS double layer The compound model predicts the default of consumer finance loans.(4)Model experiment comparison: In order to explore a good-performance consumer finance default detection model,three sets of comparative experiments are carried out in this dissertation.One is the comparison between the SMOTE model and the MSW-SMOTE sampling processing experiment;the second is the LOMST mixed sampling model and other imbalance processing The third is the experimental comparison between the LOMST-VS two-layer composite consumer finance loan default model proposed in this dissertation and the base model LightGBM.The comparative analysis of the three sets of experiments finally verifies the effectiveness of the LOMST-VS double-layer consumer finance loan default prediction model proposed in this dissertation.