基于装袋GEP分类器集成的信用评估

Credit Evaluation Based on Bagging GEP Classifier Integration

为提高信用评估的预测精度,提出一种基于装袋的基因表达式编程(GEP)多分类器集成算法。该算法采用Bagging方法将GEP产生的多个差异基分类器进行集成。在德国信用数据库真实数据集上的实验及性能分析表明,该算法较SVM算法的预测精度提高约2.7%;较KNN(K=17)算法的预测精度提高约7.93%;较单GEP分类算法的预测精度提高约1.1%。

To improve the prediction precision in credit evaluation,a novel classifier ensemble algorithm based on Gene Expressiong Programming（GEP） with Bagging,called BGEP-CREDIT,is proposed.The algorithm uses Bagging to combine the several GEP classifiers generated from GEP.Experiments and performance analysis on Germany credit database are given.The results show that compared with SVM algorithm,KNN（K=17） algorithm and a single GEP classifier,the prediction precision is increased by 2.7%,7.93% and 1.1% respectively by using BGEP-CREDIT.