A Research On Employee Turnover Prediction Based On Data Mining

The success of an enterprise not only depends on fortuitous luck,but also relies on the efforts of the teams and the employees' struggle for the company.It is these factors that make the company stronger and bigger.From this point of view,the importance of employees to the enterprise is self-evident.Although the phenomenon of "old people go,new people come" is very common in enterprises,the mobility of employees within a certain range does not have much impact on the survival and development of the enterprises,and may make enterprises full of vitality and enhance the vitality of employees,a high proportion of staff turnover will not only increase the financial burden of companies,but also a series of economic losses caused by the inefficiency caused by new employees' unfamiliarity with the company's business.Seriously,it may even cause the companies' core secrets to be leaked,thereby putting the company in trouble.If this kind of problem is not effectively controlled,it will eventually affect the sustainable and healthy development of enterprises,and may even cause enterprises to crash.Therefore,it is particularly important to help enterprises establish a reasonable and efficient prediction model of employee turnover,to help enterprises lock in employees with high turnover tendency and reduce losses.This article is based on 4,410 pieces of human resources employee data provided by XYZ company.First,use SPSS Statistics 20.0 software for data type conversion,deletion of missing values and other pre-processing.Then use SPSS Modeler 18.0software to remove irrelevant variables.Then the univariate churn prediction ability test is performed on the removed variables.The final forecasting modeling indicators are determined using the normal distribution test and two independent sample nonparametric tests.Finally,a random forest algorithm,support vector machine algorithm,C5.0 decision tree algorithm and Naive Bayes algorithm are used to build four models for the problem of employee turnover.At the same time,using the prediction results of the four models established above,a Lagrange method is used to establish a combined prediction model.The five models were compared and analyzed according to the evaluation criteria of accuracy,recall,F-Score,accuracy,ROC curve and AUC value.By comparing the above five models,the combined prediction model has the highest accuracy,recall,F-Score,accuracy,ROC curve and AUC value compared to other models,indicating that the prediction effect of the combined prediction model is best than other models.And because the combined prediction model can make full useof the effective information in the data than the single model,it avoids unnecessary waste of information as much as possible,considers the problem more systematically,and has strong practicality.Therefore,the combined prediction model was finally selected to solve the problem of employee turnover.