Workforce Analytics for Manufacturing: Predicting Employee Job Satisfaction via Explainable Machine Learning and SHAP

Abstract

This study employs machine learning to investigate Job Satisfaction (JS) drivers in the manufacturing sector, analyzing demographic attributes and four composite variables. Utilizing a 201-point dataset, three algorithms were evaluated via StandardScaler normalization, Leave-One-Out Cross-Validation, and GridSearchCV. Results indicate the optimized XGBoost model achieved superior regression accuracy (R² = 0.8211, MSE = 0.4108) and classification performance (AUC = 0.9754). By integrating SHAP (Shapley Additive exPlanations), the research provides an interpretable framework for feature importance. Findings suggest XGBoost and Random Forest offer robust predictive capabilities, providing manufacturing executives with a data-driven roadmap to optimize organizational health and employee engagement.