The Forecast Model Of Employees' Turnover

The loss of talents,especially the loss of key talents,will have a great impact on companies.The key technologies,management systems,and customer resources of enterprises are often in the hands of key personnel.Therefore,their departure will cause huge short-term business and production problems.Losses also add extra costs.Therefore,it is extremely important to guide employees who have a tendency to leave and avoid brain drain.Using machine learning algorithms and statistics related knowledge to calculate the probability of employee turnover,to explore what factors are the decisive factors leading to resignation and the corresponding impact relationship,so that you can infer which people are the most inclined to leave the company,companies can respond to strategies in advance,Try to avoid the occurrence of such things,while adjusting the company's employment system and policies,if the employees leave is purely personal reasons,the company should prepare reserve personnel in advance,so that at any time can take the post,the company will not be difficult due to the loss of individual talent Loss of remedy.Precaution is forbidden in all cases,seeing deeper levels from the data,and digging out internal problems,so that early adoption of measures can avoid further losses.This paper aims at predicting and analyzing the employee turnover,using the machine learning Gradient Boosting Classification Tree(GBDT)algorithm to build employee turnover prediction model,predicting whether employees will leave the company,and analyzing and summarizing several important factors that affect employee turnover,such as salary.Levels,ages,etc.,provide early warnings for companies,assist HR teams in key interventions,and let management guide what factors influence the retention of people,and in turn promote companies to “select people”,“educate people” and “employ people.".This article selected data shared on the IBM Watson Analytics analytics platform for empirical analysis.Before modeling,we performed related preprocessing on the raw data,including processing dirty data,data normalization,and unique thermal encoding of attribute variables.