Research On Personal Loan Credit Risk Prediction Based On Deep Learnin

With the development of the socialist market economy with Chinese characteristics,China’s economic system and model are becoming increasingly modern,and a credit economy model that relies on credit endorsement or potential assets and resources to exchange for currently available cash has gradually emerged.According to statistics from the People’s Bank of China,the proportion of credit loans in the total business amount of state-owned financial institutions is increasing year by year;The relevant central ministries and commissions have also proposed to conduct in-depth research on financial health issues and promote high-quality development of inclusive finance.Against the backdrop of China’s increasing economic output and loose policies,credit loan business is increasingly sought after,with a large number of newly established banks and financial institutions emerging.On the one hand,the credit loan business,due to its objective benefits,increases the willingness of banks and financial institutions to lend,and once the lender defaults,it will inevitably cause serious losses to the lender,thereby affecting its operational ability;On the other hand,banks and financial institutions often use raising loan interest rates as a risk hedge when facing such risks,making it more difficult to promote financial inclusion.Therefore,how to improve the ability of banks and financial institutions to resist credit risks has become one of the current research hotspots.This article first conducted a research on the published research results on personal credit risk prediction both domestically and internationally.The results showed that credit scoring cards are currently the most mature risk prediction model and are widely popular due to their strong interpretability.However,research has shown that the performance of credit scoring cards has fallen behind more modern models,and there are few cases where deep learning is applied to this.Therefore,this paper attempts to use the Tab Net deep learning model proposed in recent years to assess personal credit risk,test its progressiveness performance,and at the same time contribute to the research in this field.This article uses CCF＿BDCI＿The 2021 competition dataset was used as a starting point,and four different models were constructed using comparative research methods to comprehensively evaluate the model performance using indicators such as accuracy,recall,and AUC.After performing conventional data science processes,it was found that Tab Net’s model performs well,with slightly better overall performance than the other three models.It also performs well in interpretability,which can compensate for the shortcomings of deep learning models in interpretability.The comparative experimental results of this study indicate that the Tab Net model has slightly better overall performance than other models,but significantly better time consumption than other models.Therefore,it has a significant improvement effect in identifying personal credit risks.