Research On Employee Turnover Prediction Of Company A Based On Data Mining

With the rapid development of industry 4.0、industrial big data、artificial intelligence and industrial Internet technology,the fourth industrial revolution began to bring the manufacturing industry into an era facing great changes.Manufacturing enterprises have a large number of employees and a high turnover rate,and the frequent turnover of excellent employees will inevitably bring about the loss of the company’s economic benefits.Therefore,grasping the trend of employees’ turnover and understanding the reasons for employees’ turnover will play a positive role in formulating talent retention measures and improving the rational allocation and management of human resources.With the rapid development of data mining technology,it has been widely used in various fields such as finance,bioengineering,industrial engineering,etc.Using data mining technology to deeply explore enterprise human resources information,mining the potential reasons of employee turnover and predicting employee turnover intention can provide reference suggestions for enterprise managers to make decisions.In this paper,based on the employee turnover data of Company A,a variety of single and integrated machine learning algorithms are applied to establish classification prediction models with the theme of predicting employee turnover probability.Classical decision tree model,Naive Bayesian model and Support Vector Machine model are selected as the representative of single prediction model,and random forest model and XGBoost model are selected as the representative of integrated model.Through the visual display of the prediction results of the model,we can understand the decision path of employee turnover and the ranking of important influencing factors,so as to provide targeted policy suggestions for enterprise human resources management.When using machine learning algorithm to predict and model practical problems,the accuracy and credibility of the prediction results are directly determined by the advantages and disadvantages of the models.In order to provide comprehensive reference suggestions for enterprise decision-making in combination with the actual situation,this paper evaluates the prediction effects of the above three single models and their integrated models in a comprehensive dimension,hoping to screen out the most suitable model for enterprise employee turnover prediction.In the evaluation of the model,it is different from the single evaluation index of accuracy used in previous studies,and also comprehensively considers the Kappa coefficient,root mean square error,relative absolute error,ROC curve area and running time of the model.The multi-attribute decision-making theory of preference is innovatively introduced,and the algorithm evaluation problem is modeled as a multi-attribute decision-making problem.According to the preference information of decision makers,AHP is used to give different weights to each index.Then TOPSIS method is used to evaluate and sort the comprehensive performance of the model,and the algorithm model with the best performance in the data set of employee turnover problem is screened out.The research in this paper provides a new idea for the analysis and early warning of employee turnover and brain drain,and has certain theoretical and applied value.There are 27 figures,16 tables and 46 references.