Dependency evaluation of financial market returns for classifying and grouping stocks

Abstract

Following the globalization of the economy and the increasing significance of international trade investments, linkages among economic variables of different countries are becoming strikingly evident. There is a strong interest among researchers to capture presence and extent of such negative or positive correlations. In this paper, we embark a novel methodology to identify the correlated market by a modified clustering procedure and finding the optimal number of countries within the clusters. The proposed methodology mainly works with the k-means clustering method in which its performance is improved by particle swarm optimization algorithm (PSO). The integration of these methods aims at finding the best number of clusters (k) within the dataset with a distance-based index in order to achieve the most appropriate stock market assigned to each cluster. As a case study, an experiment on daily and monthly stock market returns of 50 counties has been evaluated.