Abstract:
Employee attrition is a significant problem that impacts organizational stability, productivity, and financial performance. High employee turnover generates various far-reaching costs, such as recruitment expenses, training costs, and loss of institutional memory. Traditional methods for predicting employee exit, such as surveys and historical trend analysis, are not very effective at providing real-time, data-driven insights. In this research, the application of Artificial Intelligence (AI) and Machine Learning (ML) was looked into to predict change intentions and improve HR strategies. This study used various ML models like Logistic Regression, Decision Trees, Random Forest, XGBoost, and LGBM to determine their effectiveness at predicting employee attrition. Techniques like SMOTE, Borderline-SMOTE, and ADASYN were applied to address class imbalance, making for fairer model predictions with a lesser bias towards the majority class. For Feature Engineering Variance Inflation Factor (VIF) to remove high-correlated features, apply principal component analysis for dimensionality reduction and permutation importance to remove low-importance features. Model evaluation, including accuracy, precision, recall, F1 score, and AUCROC, never left a shroud of doubt when measuring the ensemble model’s potency. XGBoost and Random Forest model mechanisms are vastly different from traditional classification techniques. These models could single out the most critical factors contributing to employees’ fleeing the workplace, allowing HR professionals to act proactively on retention strategies.
