Strategic Human Resource Management and Employee Retention: An Explainable Machine Learning Analysis of Workforce Performance and Attrition

Abstract

Employee attrition is a major concern in strategic human resource management because it can influence workforce stability, organizational continuity, and overall performance. This study examined demographic, occupational, satisfaction, compensation, and career-related factors associated with employee attrition and applied explainable machine learning to support retention-oriented HR decision-making. A secondary cross-sectional analysis was conducted using 1,470 employee records from the IBM HR Analytics Employee Attrition & Performance source. Descriptive statistics, chi-square tests, independent-samples t-tests, and Pearson correlation analysis were applied, while Logistic Regression and Random Forest models were developed for attrition prediction. Model performance was assessed using accuracy, precision, recall, F1-score, and ROC-AUC, and SHAP analysis was used to interpret the better-performing model. Attrition was significantly associated with job role, job level, business travel, job satisfaction, environmental satisfaction, job involvement, work-life balance, overtime, and stock option level. Employees experiencing attrition were generally younger, had lower monthly income, fewer total working years, and shorter organizational tenure. Logistic Regression produced better recall and F1-score than Random Forest and was therefore selected for explainability analysis. SHAP identified years since last promotion, environmental satisfaction, job satisfaction, overtime, manageri al continuity, role tenure, and job involvement as important predictors. The findings show that explainable machine learning can provide practical and interpretable evidence for targeted employee-retention strategies.