Enhanced Prediction of Happiness Scores Using Explainable AI and Ensemble Models with Socioeconomic Data

By making use of a dataset that includes annual records of happiness scores, alcohol consumption (beer, wine, and spirits per capita), and important economic indicators like GDP per capita and the Human Development Index (HDI), this study investigates the use of various machine learning and explainable AI (EAI) models to predict happiness levels across nations. The dataset provides a multi-dimensional picture of the elements impacting happiness with parameters such as Region, Country, Year, HappinessScore, HDI, GDP_PerCapita, Beer_PerCapita, Spirit_PerCapita, and Wine_PerCapita. Each of the models-Linear Regression, Decision Tree, Random Forest, LightGBM, Explainable Boosting Machine (EBM), and Ridge Regression-is assessed using metrics like variance, R-squared, accuracy, Mean Absolute Percentage Error (MAPE), Mean Squared Error (MSE), and Mean Absolute Error (MAE). The results indicate that ensemble models, particularly LightGBM and Random Forest, outperform simpler models in predictive accuracy, achieving accuracies of 86.76% and 82.66%, respectively. The findings underscore the importance of integrating economic indicators and lifestyle factors like alcohol consumption in happiness prediction models, revealing that economic prosperity and moderate consumption patterns tend to correlate with higher happiness scores. This research contributes to the understanding of happiness prediction, offering insights into the roles of socioeconomic and lifestyle factors and presenting a comparative evaluation of models that prioritize interpretability alongside predictive power. Future work could incorporate additional regional factors, longitudinal data, and more advanced EAI methods to deepen insights into the global determinants of happiness.