Research On Risk Assessment Of Online Loan Project Using Scoring Model

In recent years,the scale of domestic financial credit has continued to grow steadily,and the problems accompanied by credit risk have emerged endlessly.Among them,the issue of "early repayment" is particularly prominent in the post-epidemic era.Since 2022,due to the fluctuations in the economic cycle,the trend of "baying loans in advance" of central banks has gradually appeared,which has harmed the banking industry revenue.Some domestic banks have begun to take multiple measures to delay customers’ "early loans" behaviour,but such efforts have a negligible effect on the behaviour of "repayment in advance".Banking institutions bring reputation risks.It can be seen that financial institutions such as banks need to evaluate and manage such risks.The online loan accounts for all "early loans" risks are relatively high.Therefore,evaluating the tendency of such loan repayment in online loan projects has a significant reference effect on responding to the overall "early loan repayment" risk.This article takes the user portrait and behavioural data provided by a domestic Internet bank as the research object and establishes a logical return model based on logical regression,digging a single feature variable path of the borrower and further designing the relationship between the characteristic variable path and the "baying loan" behaviour in advance.Risk management strategy.First,this article completes the selection of feature variables through methods such as variable pre-screening,deck boxes,WOE coding,and gradual regression.Based on a logical regression,a scoring model is established.To optimize the generalization of the model;secondly,this article defines positive and negative samples based on the scoring results and whether to repay in advance.Create a CART decision tree model.It uses the Gini coefficient to select the optimal division attribute to dig out the two user characteristics variables Path;Finally,based on the principle of completing the scoring model and risk management strategy,this article formulates the management priority sorting standard for management variables.According to the comparative analysis of the standard and the two paths,the Risk management strategy of "repayment".From the research results in this article,we can see that in online loan projects,financial institutions such as banks can more accurately predict the user’s behaviour tendency by constructing a credit scoring model.The value of the credit score model constructed in this article is 0.36,and the value of AUC is 0.74.Both model evaluation indicators have reached the performance requirements of the predictive model in the field of risk control;In the process of transitioning from the high evaluation partition to the low rating interval,the corresponding proportion of the corresponding early repayment increased from 16.15% to90.65%,which shows that the scoring ability is better for sample risk levels.In summary,the research results of this thesis provided the risk recognition and management methods for the borrower’s "early repayment" in the online loan project.