一种面向个人信用评分数据的改进型SMOTE算法

打开文本图片集

中图分类号：TP301.6；TP181 文献标识码：A 文章编号：2096-4706（2025）23-0083-07

Abstract： In the training of personal credit scoring model，it is often faced with the problem of imbalanced sample size.Therefore，various samplingalgorithms atthedata levelare proposed.Among them，SMOTEis widelyusedasatypical minorityclasssample expansionalgorithm.Inorderto improve the performanceofSMOTEindealing withhigh-dimensional complex datasets and improve the diversityof synthetic samples，an improved SMOTEalgorithm combining least squares methodand Gausiannoiseisdesignedinthispaper.Firstly，theleastsquaresalgorithmisusedtoftthefeaturespaceofthe minorityclassamples.Secondly，newsamplesaresythesized inthefeature space.FinallyGausiannoise isadedtothe new samples toenhancethediversityof thesamples，soas to improvethequalityanddistributionrationalityofthe synthetic samples.Experimental results onfour real datasets andfive scoring modelsshowthatthe algorithmcan effectively improve the performance of the scoring model.

Keywords： personal credit scoring; imbalanced data; SMOTE; least squares method; data processing

0 引言

如今，个人信用贷款已成为现代经济中的重要组成部分。（剩余9436字）