Introducing a cluster based concept to portfolio selection from colombo stock exchange (CSE)

Abstract

Portfolio management is a compelling concept in the financial sector for strategizing investments resulting maximum return with minimum risk. Normal procedure of selecting companies to the portfolios out of the registered companies in Colombo Stock Exchange (CSE) depends on the experience as well as the gut-feeling of the investor or the broker. Thereby, this selection depends on qualitative measures without much quantitative justification. This research objective is to introduce a quantitative criterion to select companies in an investment portfolio that result higher return compared to the selection made by qualitative measures, using clustering based algorithm. As per literature or the knowledge of collaborators, this technique was never used. Here, the daily share prices of 215 companies were collected for the period of five years, starting from 1st January 2012 and K-means clustering algorithm was used in clustering with Norm of the Variance of the Quarterly Share Prices (VQSP) as clustering variable. Based on the VQPS values, all companies were sorted, and the analysis was carried out by considering first 100%, 95%, 90%, 80%, 70% selected companies separately. There after the selected companies were clustered into 5 clusters and those clusters were used to build different portfolios by introducing a quantitative measure on the selection criterion. The return gained by the portfolios that were built-up based on company clusters was higher than that gained by the traditional selection criterion. Therefore, the research exhibits that the portfolio which are having higher return can be constructed by the selections from the clusters suitably.