Predicting Happiness Index Using Machine Learning

Happiness in its subjective form is difficult but important to measure. Various happiness indicators are considered when attempting to quantify the level of happiness of countries in the world. The ability to predict the happiness index based on any combination of indicators will provide governments with the understanding for better decision-making. Countries are being ranked based on the happiness perspective of the citizens. This study employed Machine Learning (ML) to predict the happiness score of 156 countries aiming to find the model that performs with close to a hundred percent accuracy, The 2018 and 2019 World Happiness Report was combined, cleaned, and prepared for modeling. Random Forest, XGBoost, and Lasso Regressor were fitted on the dataset utilizing an 80-20 percent split. Performance was evaluated based on R-squared and Mean Square Error. Our study results show that XGBoost performed optimally with a r-squared of 85.03% and MSE of 0.0032. Random Forest achieved 83.68% and 0.0035; Lasso obtained 80.61% and 0.0041 in accuracy.