Early Warning Of Enterprise Talent Loss Risk Based On XGBoost Model Fusion

Talent is the first resource.The loss of talent in the enterprise may lead to a series of chain problems,traditional human resources decisions often rely on intuition or management experience,there may be a large deviation in the selection of employees,bringing huge losses to the enterprise.It can be seen that scientific and reasonable analysis of the characteristics of departing employees and the establishment of enterprise talent loss early warning model are very important for enterprises and individuals.This paper takes IBM employees as the research object,first conducts a descriptive analysis,and explores the factors affecting employee turnover.The K-means clustering algorithm was used to divide the departing employees into three categories,and the characteristics of various departing employees were explored and analyzed.Then,six single models,Naive Bayes,SVM,KNN,Random Forest,XGBoost and Light GBM,were used to predict the departing employees of enterprises,and it was found that the three integrated models of random forest,XGBoost and Light GBM had good prediction effects,and the AUC values were greater than 0.9,of which the AUC value of XGBoost was0.947,which had the best prediction effect.In order to explore a more effective enterprise talent drain early warning model,this paper uses voting methods(including hard voting and soft voting)to fuse random forest,XGBoost and Light GBM,and uses random forest,XGBoost and Light GBM as Stacking base learners,and logistic regression as Stacking meta-learners.The results show that both model fusion methods effectively improve the performance of employee turnover prediction,and the stacking model fusion effect is the best,and the AUC value reaches 0.957.Since XGBoost has good prediction effect whether it is predicted alone or after model fusion,this paper ranks the feature importance based on the model,and obtains that the top ten features of feature importance are: monthly salary,monthly rate,daily rate,hourly rate,age,distance from home of the company,number of years of work with the current manager,work environment satisfaction,job satisfaction and interpersonal relationship satisfaction at work.In view of the characteristics of departing personnel,this paper finally puts forward several suggestions for enterprises through analysis to help enterprises retain talents.