基于IAPSO-RELM的信用卡用户风险评估

Research of credit card user risk assessment based on IAPSO-LSSVM

为了提高正则极限学习机(RELM)分类模型的精度,准确的将银行信用卡客户按照信用等级进行分类,利用改进的自适应粒子群算法(IAPSO)的全局搜索能力,寻找RELM的最优输入权值(IW)和隐藏层阈值(B),提出一种IAPSO-RELM分类算法。在分析影响信用卡用户信用等级的个人状况、工作状况和与银行关系的基础上,建立基于IAPSO-RELM的信用等级分类模型,并以FX银行信用卡中心的数据进行仿真实验。结果表明:与RELM、PSO-RELM算法相比,该方法分类效果更好。

In order to improve the classification accuracy of Regularized Extreme Learning Machine（RELM） model,the global search ability of improved adaptive particle swarm optimization（IAPSO）is used to find out the most optimal IW and B,a IAPSO-RELM classification algorithm is put forward.According to the factors affecting the credit card user credit rating,personal status,working status,and bank relationship,the credit rating classification model is established based on lAPSO-LSSVM and the simulation experiment with the data of FX Bank Credit Card Center is conducted.The results show that the proposed method is better than the ELM and RELM method.