A Data-Driven Analysis for the World Happiness

This study presents a data-driven investigation into the determinants of national happinessusing data from the World Happiness Report. Leveraging a structured workflow ofexploratory data analysis and predictive modeling, the study assesses how socio-economic,health and governance indicators influence subjective well-being. Four regression models,Linear Regression, Random Forest, Gradient Boosting and XGBoost, were trained andevaluated using a time-based validation split, with robustness assessed over 5 randomizedseeds per model. Gradient Boosting outperformed the others, achieving the highest average R2 of 0.83 and the lowest mean squared error (MSE) of 0.24. Model interpretability wasaddressed using SHapley Additive exPlanations (SHAP), which identified GDP per capita,healthy life expectancy, and social support as the most influential predictors of happiness.These findings underscore the multifaceted nature of well-being and highlight the importanceof combining economic development with social support and good governance. The resultssuggest policy implications that prioritize investments in health, education, institutionaltrust, and social infrastructure alongside economic growth.