Exploring happiness factors with explainable ensemble learning in a global pandemic.

Area of Science:

• Social Sciences
• Computer Science
• Data Science

Background:

• Happiness is a multifaceted state influencing overall well-being.
• The World Happiness Report (WHR) provides key metrics like GDP per capita, social support, and life expectancy.
• Understanding factors influencing happiness is crucial for societal progress.

Purpose of the Study:

• To predict happiness scores using Machine Learning (ML) and Deep Learning (DL) algorithms.
• To analyze the impact of individual variables on the happiness index.
• To assess the influence of the COVID-19 pandemic on happiness indicators.

Main Methods:

• Development of two ensemble ML/DL models: Blending RGMLL and Stacking LRGR.
• Utilizing techniques like Ridge Regression, Gradient Boosting, Multilayer Perceptron, Long Short-Term Memory, Linear Regression, and Random Forest.
• Application of Explainable Artificial Intelligence (XAI) for detailed analysis.

Main Results:

• The Blending RGMLL model achieved high predictive accuracy (R²=85%).
• 'GDP per capita' was identified as the primary driver of happiness scores.
• During the COVID-19 pandemic, 'social support' became the most significant factor, followed by 'healthy life expectancy' and 'GDP per capita'.

Conclusions:

• ML and DL models effectively predict happiness and its determinants.
• The pandemic highlighted the critical role of social support in maintaining happiness.
• Findings offer insights for enhancing happiness and resilience during crises.