Predicting employee retention based on socio-economic factors : a case of an Sri Lankan apparel organization

Abstract

Employee attrition is considered as one of the biggest challenges faced by the apparel manufacturing organizations in Sri Lanka and it’s an industry which highly depends on its human workforce. Nearly 60% of the terminated employees are terminated during the first six months of their time at the organization considered in this study which makes it difficult to ensure a smooth production process. Currently a scientific method is not in place to identify the first six months termination risk associated with the employees in advance and if so, the management can take necessary actions to mitigate those risk factors. This study is aimed to build a predictive model for one of the leading apparel manufacturing organizations in Sri Lanka to identify the first six months retention risk of the newly recruits one month after recruitment based on socio-demographic factors and their earnings. Data obtained through the employee interview process and payroll process were used for this study and a scalable end-to-end machine learning tree boosting algorithm called XG-Boost was used to build the predictive model. The built model went through a parameter tuning process to improve its accuracy and after that the accuracy of the pruned model is around 86% and the model suggested that employee earnings, BMI (Body Mass Index), Education level, Age, Family opinion on the job are the most prominent factors affecting employee decision to stay in the organization for more than six months or not. The developed model has been implemented at one of the manufacturing facilities in the organization, and insights given by the model on each employee’s retention status being shared with all stakeholders through a dashboard.