Research On Risk Identification Of An Internet Credit Institution Lender Based On Machine Learning Method

Since 2022,China’s economic growth has faced various internal and external pressures.Against this backdrop,Premier Li Keqiang proposed that consumption is the ultimate demand,and expanding consumption is not only conducive to improving people’s living standards,but also can drive employment,investment,and industrial upgrading.At this time,the role of consumer credit is self-evident.First,traditional financial institutions such as banks do not cover the increasing demand for personal loans in the credit market in terms of quantity and scale,and their risk control requirements are relatively high,so internet credit institutions are necessary supplements to the credit market.Secondly,regulatory policies indicate that loan institutions should attach importance to identifying and judging loan applicant information,strengthen credit risk control for internet loan businesses,strengthen risk data and risk model management,establish a sound independent and controllable risk management operation mechanism and information system,and strictly adhere to customer access standards.Finally,the country currently attaches great importance to the development of financial technology,and using machine learning methods for risk identification is a branch of financial technology.The application and development of financial technology is the only way for the future development of the financial industry towards high-quality development.Based on this,this article’s use of machine learning for loan applicant risk identification is in line with national requirements for the development and regulation of the financial industry.This approach not only plays a role in empowering financial technology and promoting high-quality development but also represents an important attempt to explore the value of data.Building on existing research,this article uses real loan data from an internet credit institution to identify loan applicant risk.First,the raw data is analyzed,processed,and used to construct and select features.Next,representative models,including linear logistic regression models,and Bagging and Boosting algorithms from ensemble learning-Random Forest,XGBoost,and LightGBM-are selected.These models are trained using the real data of the credit institution and are compared based on Accuracy,Precision,Recall,F1＿score,and AUC.Based on the AUC value,the best model for this article’s data is determined to be the LightGBM model,which is used to rank the importance of risk features for loan applicants.Finally,using the analytic hierarchy process,this article proposes that the internet credit institution should identify loan applicant risk based on four important features: basic information,loan information,repayment information,and nonfinancial information.