Dissertation defense (October 02, 2023): Leon Victor Medeiros de Lima
Student: Leon Victor Medeiros de Lima
Title: Portfólio Diversificado com Múltiplas Restrições de Cardinalidade
Advisor: Pedro Henrique González Silva
Day: October 02, 2023
Abstract: The Financial markets and investments are crucial to the economy. Furthermore, advances in technology and open services are impacting the integration of financial markets and institutions. This has made financial decision-making processes even more complex, as precise methods are applied to deal with specific constraints. The portfolio selection problem (PSP) is one of the principal and most relevant problems in the finance area. The objective is to choose a set of assets that maximize or minimize return or risk, respectively. This work presents a new formulation of the portfolio selection problem to single-objective with cardinality constraints for minimizing the portfolio risk with multiple categories of assets. Its objective is to compose a multi-asset portfolio with a fixed number of three diverse asset categories (Brazilian stocks and American stocks and REITs) to minimize portfolio risk given an expected return. Additionally, this work uses Recurrent Neural Networks of the Long Short-Term Memory (LSTM) architecture and Support Vector Regressor (SVR) algorithms to predict the expected asset return. It also presents mathematical models for the portfolio composition process, along with computational experiments that validate the proposed approach. Our main results validated the proposed model, showing that it is possible to obtain diversified portfolios given a level of return, outperforming market benchmarks.
