Research On Financial Risk Forecast Algorithm Based On Big Data

With the rapid development of Internet finance,banks are paying more and more attention to credit risk management,reducing the rate of non-performing loans,and it is particularly important to identify loan customers with greater default risk.Therefore,the prediction of customer credit overdue risks has always been the financial industry An important research direction.In recent years,the customer's unreliability has made the banks' non-performing loan rate very unoptimistic in recent years.Accurately assessing and predicting the credit risk of corporate customers is a risk management function that Chinese commercial banks urgently need to master.In the context of the arrival of the era of big data,this paper implements feature extraction and data weighting processing based on user basic attribute data and numerous data of downloaded APP types,and then uses punitive linear regression to build predictive models to improve debt insolvency.The accuracy of the customer's default judgment has been partially optimized to improve the risk management and control of the customer's commercial bank loans,thereby greatly reducing the risk of commercial bank loans.According to the characteristics of the collected sample data,the most suitable penalty linear regression prediction algorithm is selected and the experimental analysis is carried out in order to improve the bank's risk management level.The main research contents and results of this topic are as follows:(1)Through research and analysis of the mature technology of big data application in the Internet industry,it is determined to use pandas and numpy to read and process preliminary data,and clean and organize the original data.(2)For multi-faceted data integration,choose to use TF-IFD method to extract the features of customer downloaded APP data.After the extraction is completed,merge the APP data with the customer's basic attribute data to complete the data integration work.(3)Aiming at the problem of binary classification on unbalanced data sets,a weighted processing of ordinary penalty linear regression algorithm is proposed,and a comparison test is carried out for four combinations of different penalty coefficients and whether they are weighted.Finally,the best choice is made for the four situations through confusion matrix,accuracy rate,recall rate and accuracy rate.