Clustering of Financial Development Indices using AGNES and AGNES-PCA algorithms

Abstract

Global financial indices play a prominent role in the prior identification of future economic growth and shrinking trends, developing financial strategies, and reviewing financial and economic policies. Annual financial indices published by the International Monetary Fund [IMF] associated with the 2020 financial year gauging the development of financial institutions and markets in terms of depth, accessibility, and efficiency were considered. In this study, hierarchical clustering was carried out through two main approaches to identify recurring economic and financial patterns and development trends across countries and territories. Clustering was initially implemented with all six financial indices using the traditional agglomerative nesting (AGNES) algorithm. Subsequently, clustering was executed through AGNES after conducting dimensional reduction in financial indices through principal component analysis (PCA). Internal validation measures: connectivity, Silhouette and Dunn index were improved by 46.45%, 39.48% and 96.55%, respectively through AGNES with PCA (AGNES-PCA) compared to the AGNES without PCA. All stability measures, including average proportion of non-overlaps (APN), average distance (AD) between data points, average distance between means (ADM), and figure of merit (FOM) were increased by 91.55%, 230.81%, 26.78%, and 54.42% respectively during AGNES-PCA. Unstable economies like Bangladesh, Sri Lanka, El Salvador, Estonia, Honduras, Lebanon, Nigeria, Pakistan, Tunisia, Venezuela, and Zambia were clustered into a common group during AGNES-PCA. This study is helpful in effectively identifying the best clustering method for a group in stable and unstable economies around the globe. Moreover, clustering financial indices using AGNES-PCA could be used to identify future financial and economic crises.