Research On The Application Of Enterprise Illegal Fund-raising Identification Based On XGBoost And LR Integrated Model

Maintaining the health and stability of the financial market is the core of national economic construction,and in recent years,illegal fund-raising cases emerge one after another,for the country’s economic activities and financial market may have a bad impact,regulate illegal fund-raising behavior,to ensure the safety of funds is the general trend.Research methods in the field of enterprise financing at home and abroad yet into the system,in raising funds illegally judgement more for human experience to judge,no more accurately and system identification methods,this paper,the innovative application of machine learning algorithms to enterprise financing risk assessment,the research method is novel,and obtain good recognition result.This paper mainly studies the use of machine learning model to predict the risk of illegal fund-raising.Firstly,the desensitized public data set provided by CCF BDCI was used to preprocess various enterprise information,including missing value processing,tag coding,and constructing new features by feature sorting method.Secondly,in order to avoid over-fitting,feature selection based on Pearson correlation coefficient and L1 normal form was adopted to eliminate features with high pin-correlation.Use SMOTE algorithm to avoid negative influence of unbalanced sample;Then the LR model,Decision Tree Model,Random Forest model,XGBoost model,and XGBoost+LR model are used for model training,and the risk of illegal fund-raising is predicted in the test set.In terms of recall rate,precision rate and AUC value,the prediction effects of each model are compared,which proves that the XGBoost and LR integrated models are better than other models in predicting illegal fund-raising.In addition,the importance of features under each model is output respectively to provide a reference for the subsequent effective prevention of illegal fund-raising risks of enterprises.