Research Of Personal Credit Risk Evaluation Based On Hybrid Ensemble Algorithm

With the mainstream of the times,credit consumption is gradually forming,and the means of credit economy have been widely used in many aspects,and the scale and mode of transaction between each subject is increasing.Related data show that in the financial institutions such as commercial banks in China,the scale of enterprise loans is gradually smaller than the number of personal credit consumption loans.With the increase of individual consumer credit capacity in our country,in many commercial banks and other financial institutions,the individual loan regulations are steadily rising,gradually surpassing.The loan scale of enterprises brings credit risk more and more.Therefore,in the process of credit exchange and credit consumption among the various credit subjects,the effective tools and means of effective credit risk control and control are urgently needed to avoid these hidden risks.It is an urgent problem for banks and other organizations to effectively avoid the risk of default by the loan applicants.How to establish a strict personal credit risk assessment system to guarantee the profit and effectively control the credit risk is an urgent problem for all banks and other financial institutions.Accurate assessment of personal credit risk is of great value both in academia and in industry.At present,one of the main research directions in the field of personal credit risk assessment is ensemble learning.The advantage of ensemble learning is that it can make decisions together through multiple base classifiers,which can improve the accuracy of the prediction model.This paper systematically discusses the research status of the index system and evaluation model of personal credit risk assessment,and finds that the selection of the existing research index system can not meet the needs of the present.Based on the idea of integrated model,the evaluation model of hybrid ensemble algorithm is set up by constructing a diversified base classifier.Finally,comparing the model by using a credit data set from UCI and a credit data set from Kaggle,compare the models presented in this paper by comparing several other existing ensemble models.The work of this article mainly includes the following two aspects:Combing the related research on the index system of personal credit risk assessment,we found that the existing index system may not meet the standards of the present times,and introduced a method of generating the feature combination in He in 2014.Through the experiment,it is proved that the combination feature method can explore the potential effective characteristics through the combination of features.Compared with the same model,the combined feature method can get better results.The Hybrid Ensemble Model of the base classifier is constructed,and several training subsamples are constructed by sampling,and the parameters of the Gradient Boosting Decision Tree(GBDT)model and the number of the input features are mixed together.A variety of base classifier is constructed,and the GBDT-LR and GBDT-SVM models are added.The classification of multiple base classifiers is the final prediction result by the way of voting.To realize the effective prediction of the unknown samples.Finally,we validate the model through two sets of data,and the experimental results prove that the model has the better performance compare to the other ensemble models.