The Hybrid Best Worst Integrated Machine Learning Methodology to Clustering the Countries Based on the Happiness Index

Abstract

Happiness, a universal component that affects human life, is defined as personal well-being, encompassing a person's evaluation of their life satisfaction. Although the perception of happiness varies from person to person, it is the primary goal that societies strive to achieve. In this regard, studies on measuring happiness have continued from the past to the present. In this context, the World Happiness Report includes happiness index values created using eight criteria with equal weights for each country. This study assumes that the variables used to calculate the happiness index values have different weights. These weights are obtained using the Best-Worst method (BWM), one of the multi-criteria decision-making methods. The BWM results, derived from expert evaluations, indicate that 'Freedom to Make Life Choices' is the most influential factor, while 'Generosity' is the least. Then, using these weights, countries are clustered into three using the k-means method, one of the unsupervised learning techniques. When the clusters are examined, it is seen that there are countries with similar characteristics in the clusters. The clustering analysis reveals three distinct country groups, where countries in Cluster 1 show higher and more homogeneous well-being indicators. In contrast, countries in Cluster 3 are characterized by low and more diverse scores. Furthermore, scenario analyses demonstrate the sensitivity of the clustering outcomes to changes in specific variable weights. The results emphasize that incorporating weighted criteria offers a more realistic and intuitive classification of the happiness profiles of the countries.