How to Manage Portfolios in Different Clusters: An Unsupervised Machine Learning Approach

International Review of Economics & Finance 2026 Open access

Read full text
Authors: Mohammad Mahbubur Rahman, Zeeshan Syed, Cynthia Abiemwense Akiotu, Rasol Eskandari
DOI: 10.1016/j.iref.2026.105309
Cited by: 0
References: 25

Abstract

Using daily data from 2015 to 2025 on global indices for the prices of 34 heterogeneous stocks, this study examines traditional portfolio theory by applying an unsupervised machine learning method, known as K-means clustering. Based on correlation coefficients of daily stock returns, this clustering method identifies four clusters, representing a specialised portfolio strategy in each cluster. These clusters display clear differences in risk-return characteristics and internal correlations. The results suggest that cluster-based portfolio construction offers flexible investment approaches, which may be particularly beneficial during periods of market volatility.

Fields of Study & Subjects

Expert finding and Q&A systems AI and HR Technologies

Publication details

Journal / source
International Review of Economics & Finance
Publisher
Elsevier BV
Volume
108
Pages
105309–105309
Publication date
24 April 2026
Language
EN