Employee turnover intention - Mapping profiles under a decision tree perspective

Abstract

This work aims to map some profiles having more propensity to quit prematurely a company. The analysis is important because affects the productivity of employees and it represents a high cost for companies around the world. The research applies a decision tree model in a study database of public domain with 1470 records, where it is possible to group profiles under 38 different variables to understand what can influence more the turnover. The result is a model with 81% of accuracy which has identified employees working overtime and new hires in the sales executive position with a higher risk of quitting prematurely the company. In some modeling approaches it is necessary focusing more on interpretability over performance. As the goal of this research is to map and understand key factors of turnover, the decision tree model is ideal. However, the model has a recall of 27%, which means that can predict about 1/3 of turnover cases. This paper contributes with a true modeling application towards People Analytics, sharing openly the model performance and discussing the features related to turnover. Companies can adapt this study in their databases in order to trace employees in turnover risk groups.

Keywords

• People Analytics
• HR Analytics
• Turnover
• Decision Trees