Year: 2026
Type: article
Abstract: The article discusses the application of clustering methods to form an optimal investment portfolio that allows the investor to achieve an effective risk-reward ratio. Three popular clustering methods, K-Means, MeanShift, and DBSCAN, are examined. The article focuses on the DBSCAN clustering method and highlights its advantages over other clustering methods in the context of financial data analysis. DBSCAN is particularly useful for identifying clusters of arbitrary shapes, being resistant to noise and eliminating the need to pre-define the number of clusters. The article presents a comprehensive approach to forming an optimal portfolio. The first step involves preparing the data and clustering it based on historical data on returns, volatility, and correlations using a programming language. The second stage involves further ranking by assigning integral scores that take into account a variety of criteria and allow for the identification of the stocks most attractive for investment within each selected cluster. After completing these stages, an optimal portfolio is formed with the highest Sharpe ratio, Sortino ratio, and other metrics that outperform the weighted average portfolio and the MOEX index. To validate the results, Monte Carlo simulations are used to assess the portfolio’s resilience in various market scenarios, including periods of volatility and crises. The study fills a gap in the study of the application of clustering methods for optimizing an investment portfolio, proposing a practical algorithm that can be adapted for individual and institutional investors. The findings highlight the potential of applying clustering methods to form an optimal investment portfolio.
Source: Financial Journal
Authors: N. I. Lysenok, Yu. E. Markova
Language: English
Field: Decision Sciences
Domain: Social Sciences
Open Access status: diamond
OpenAlex ID: W7204589709
