Determinants of Subjective Well-Being in Taiwan: A Machine Learning and SHAP Analysis

This study aimed to identify and rank key determinants of subjective well-being (SWB), examine threshold values, distinguish protective from motivational roles, assess non-linear associations, and evaluate the performance of machine learning models in predicting SWB. We conducted a cross-sectional machine learning study using data from 10,712 participants in Taiwan, derived from a nationally representative 2024 survey. SHAP analysis was applied to interpret model outputs. We identified five key determinants: family, interpersonal relationships, health, life goal clarity, and financial stability. Family relationship satisfaction, interpersonal relationship satisfaction, and health are protective factors, all of which show nonlinear associations with SWB, including pronounced threshold effects: when scores exceed 5-6 on a 10-point scale, SWB increases sharply, underscoring the importance of reaching sufficient intervention intensity. In contrast, financial safety and goal clarity function as motivational factors. Gradient Boosting demonstrated the strongest predictive performance. These findings highlight several actionable levers for enhancing SWB, including strengthening social relationships, improving health, clarifying life goals, and ensuring financial safety. Importantly, the presence of threshold effects suggests that benefits are not uniformly distributed but disproportionately realized among individuals with lower baseline conditions--often the most vulnerable.