Abstract:
"Employee attrition is a problem that affects many organizations and companies. Many companies
are actively strategizing and investing in employee retention strategies and experts in the field in
order to retain their valuable employees. This project proposes an employee attrition prediction
system, AttritionPro, that is able to utilize deep learning models applied in ensemble techniques to
produce reliable and accurate predictions. The purpose of the system is to serve as a preemptive
measure for dealing with employee attrition before it happens and allow managements to develop
retention plans and make tactical decisions using forecasts of employee resignations.
The proposed functionalities of AttritionPro will allow HR departments to preemptively forecast
employee resignations, the attrition risk level of an employee and generate a breakdown of the
features contributing to that employee’s attrition. This project tackles several aspects of the
problem domain of Employee Attrition and contributes valuable research and insights into the
problem and research domain. Ensemble methods such as stacking, voting, and simple averaging
are used to combine various deep learning methods, including convolutional neural networks
(CNN) and feedforward neural networks (FNN) and Wide and Deep models to achieve the best
results. This study demonstrates the effectiveness of deep learning in identifying risk factors and
recommending retention programs through evaluation and analysis. The results of the study
indicate that Stacking gives the best accuracy and performance of these models. These findings
contribute to the advancement of HR analytics and talent management practices, providing insights
for organizations looking to reduce employee turnover."
