Feature Selection Optimization for Employee Retention Prediction: A Machine Learning Approach for Human Resource Management

Employee retention represents a critical challenge for organizations, with high turnover rates impacting operational continuity and financial stability. This research introduces an optimized feature selection framework for predicting employee attrition using machine learning techniques. The study employs a hybrid approach integrating filter, wrapper, and embedded methods to identify the most influential predictors while reducing dimensionality by 60%. A comprehensive dataset containing 1,470 employee records with 35 attributes undergoes preprocessing with MinMaxScaler normalization and SMOTE balancing to address class imbalance issues. Multiple classification algorithms are evaluated, with XGBoost demonstrating superior performance (87.33% accuracy, 0.684 F1-score) using the optimized feature subset. SHAP value analysis reveals overtime requirements, monthly income, and job involvement as the primary predictors of attrition, with significant interaction effects between compensation and workload variables. The proposed framework enhances model interpretability while maintaining predictive power, enabling HR practitioners to implement targeted retention strategies. The integration of advanced feature selection techniques with ensemble learning methods provides both theoretical contributions to HR analytics and practical applications for workforce management, supporting proactive intervention before attrition indicators emerge.