Predicting Employee Attrition & Employee Retention Period using Supervised Learning

This paper employs machine learning (ML) to predict employee retention and attrition during recruitment. Predictions assist decision-makers in selecting the most suitable long-term candidates while reducing the likelihood of losing current employees unexpectedly. The study used a mid-scale apparel company's corporate human resource dataset, which included pre- and post-recruitment information. The algorithms “SVM,” “Decision Tree,” and “Random Forest” were chosen for model construction based on the literature. Relevant metrics from six experiments using different inputs were presented. The Random Forest (RF) classifier outperformed the others in five of the six experiments. The RF algorithm outperformed both models, with an accuracy of 94.5% for retention and 89.7% for attrition prediction. Additional metrics, including Balanced Accuracy, Recall, Precision, and F1 Score, confirmed that RF is the most effective algorithm. A user interface predicted employee retention. A dashboard of active employees was created using the attrition prediction model. Demographic and job factors are distinguished to determine employee turnover causes. Gender was the most important demographic factor, while salary was the most important job factor.