Research On Credit Risk Analysis Based On XGBoost

Financial industry credit risk analysis is a global problem.Banks of tens of millions of customers will generate a large amount of personal information,transactions,savings and other data,and combined with data mining methods to process hundreds of millions of credit data of banks,Sexual and hierarchical credit risk prevention and control measures are of great significance to the banking credit risk control.This paper first introduces the research background of credit risk analysis in financial industry,the research status and research significance of credit risk analysis by combining different algorithms at home and abroad,and introduces the principle of common algorithm of credit risk analysis and analyzes its defects.The computational complexity is high and the adaptability is weak.Try to introduce the XGBoost algorithm into credit risk analysis to verify its feasibility.XGBoost is a machine learning method that can perform multi-thread parallel computing.It adds regular items to the objective function to find the optimal solution,balances the drop of the objective function and the complexity of the model,avoids over-fitting,and has fast running speed and good classification effect.Support for custom loss functions.Although the XGBoost algorithm has good processing speed and processing precision for medium and low-dimensional data,for the credit risk big data with more variables,the XGBoost algorithm may deal with risks such as reduced accuracy,and for random values and continuous values.Parameters,in the face of various uncertain factors that may arise,have an impact on classification accuracy and accuracy.In view of the above-mentioned XGBoost-based credit risk analysis problems,combined with the integrated learning idea and the elastic network algorithm(EEN),the data set is selected,and the variable with better classification effect is selected as the reference data set;The distance-based variable step size firefly algorithm CSFA optimizes some important parameters such as random values and continuous values of the XGBoost algorithm,selects the optimal parameters to classify and analyze the reference data set,and finally proposes the EENFA-XGBoost algorithm based on UCI public data.The comparison with other algorithms verifies the high performance and high accuracy of EENFA-XGBoost for credit risk analysis.Finally,the EENFA-XGBoost algorithm is used to analyze the application of credit risk big data.