Optimized Machine Learning Approaches for Predicting Work-Life Balance: An Indonesian Case Study

Abstract

This study explores the application of optimized machine learning techniques to predict work-life balance determinants in Indonesia. Utilizing data from the World Happiness Report (2005-2023), the research implements multiple advanced algorithms including Ordinary Least Squares with Recursive Feature Elimination (OLS_RFE), Ridge Regression, Random Forest, Gradient Boosting, and Ensemble methods. The methodology incorporates comprehensive feature engineering, hyperparameter optimization, and cross-validation techniques to enhance predictive accuracy. The analysis reveals that OLS_RFE achieved perfect predictive performance (R² = 1.000, RMSE = 0.000), followed by Ridge Regression (R² = 0.947, RMSE = 0.007) and Ensemble methods (R² = 0.722, RMSE = 0.017). Feature importance analysis identified social support systems, workplace flexibility measures, and economic-social factor interactions as the most significant determinants of work-life balance. The optimized models demonstrated substantial improvement over conventional approaches, with the ensemble method providing balanced performance between accuracy and generalization. These findings offer valuable insights for policymakers and organizational leaders in developing evidence-based strategies to enhance workforce well-being. The research contributes to the literature by demonstrating the efficacy of machine learning optimization in social science research, particularly in the context of developing economies. The study establishes a robust framework for predicting work-life balance outcomes that can be adapted to other socio-cultural contexts