Research And Prototype Implementation Of Credit Risk Control Warning Method Based On Machine Learning

The security of network credit fund is one of the important research contents of Internet finance.This thesis studies the warning method of credit risk based on the integration of machine learning and big data technology to construct an intelligent credit risk assessment system with the pre-perception and warning of credit risk for credit financial enterprises and financial institutions,which reduces the financial risk of enterprises and ensures the safety of funds.The main work is as follows:First,based on the study of pre-loan risk prevention and control methods,a variety of methods are adopted to deal with the data,including box sorting,coding and variable processing methods.The random forest method and LASSO method are combined to select variables,giving full play to its advantages in variable selection.At the same time,the weighted logistic regression method is used to construct the pre-loan risk warning model,so as to perceive the credit risk in advance and give the early warning.Secondly,this thesis studies the variable selection and parameter optimization method of GBDT algorithm,and designs a credit control method integrating GBDT algorithm and logistic regression.On top of the above credit prevention and control methods,L1 constraints are integrated into the logistic regression algorithm.Thus,the application effect of the model is better.Thirdly,based on the random forest algorithm and parameter optimization method,a post-loan risk warning models based on random forest algorithm is designed.The model parameters are optimized to make the risk warning model more accurate.Last,the distributed financial credit warning system architecture was designed.Based on the research of pre-loan,mid-loan and post-loan by using distributed big data technology.This system mainly includes financial data set module,data storage module,data analysis module,model storage module and data visualization module.Based on the realization of the system function module,the system is tested.