| With the continuous improvement of people’s living standards,the concept of consumption has also gradually changed.The number of bank loans,credit card consumption and online loans has increased significantly,and over-consumption has become a common way of life.To address this situation,it is essential for financial institutions to assess the credit risk of borrowers’ credit.A good credit classification model will help financial institutions control risks more accurately,reduce bad debt losses,and improve business efficiency and profitability.For the entire financial market,credit classification is also conducive to protecting consumer rights and promoting the stable development of the financial market.Aiming at the problems of low accuracy and low AUC value of the existing credit risk assessment models,we propose the BSCB-ELM credit classification model.Relying on W-BS algorithm to complete data balancing and weighting processing,and then fusing Convolutional Neural Networks and Bagging algorithms into Extreme Learning Machines to construct a strong classifiers.The combined efforts improve the accuracy and AUC values of the BSCBELM credit classification model and other indicators.First,during the data balancing weighting processing phase,most models improve the data imbalance problem by adding minority class samples.However,the difference between the authenticity of the newly generated data and the original data is not considered,and the two types of data are directly mixed and input into the training model.In order to improve the quality of data preprocessing,W-BS algorithm is proposed.We propose to distinguish the original data from the newly generated data and give them different weights,so as to reduce the influence of the newly generated data on model prediction.Second,in the credit classification stage,we propose a strong classifier construction method that fuse Convolutional Neural Networks,Bagging algorithms and Extreme Learning Machines.Aiming to solve the problems such as inaccurate prediction results of single classification model.Finally,the BSCB-ELM model is verified on the standard credit data set,and the prediction accuracy,error rate,and AUC value are used as model evaluation indicators.Through comparative experiments,we can see that the BSCB-ELM model performs well in many indicators such as accuracy.Thus,the superiority of the BSCB-ELM model in dealing with credit classification problems is verified. |
