Explainable Machine Learning Models for Employee Retention and Turnover Analysis

Abstract

Modern businesses face serious challenges in terms of financial and operational outflows due to employee turnover. This empirical research presents a reliable and comprehensible predictive model of employees' turnover, based on 1470 employees' records and 35 organizational variables. The framework combines state-of-the-art machine learning algorithms like Random Forest, XGBoost, and LightGBM to identify potential attrition issues. Furthermore, class imbalance and cost sensitive learning methods are incorporated as well as multiple evaluation metrics are included to evaluate the models. Furthermore, new and emerging AI technologies such as explainable AI (XAI) like Shapley Additive Explanations (SHAP) and Local Interpretable Model-agnostic Explanations (LIME) are integrated. These explainability techniques can translate the outputs of complex models into meaningful insights that can assist in effective retention initiatives and ensure organisations keep important workers.