Portfolio optimization based on the pre-selection of stocks by the Support Vector Machine model

Abstract

This study aims to analyze the performance of an investment portfolio using the Markowitz model, which maximizes the Sharpe ratio from a set of assets preselected through the Support Vector Machine (SVM) model using fundamental indicators in the Brazilian stock market. With an accuracy of 61% for the SVM model, the results indicate that preselecting assets based on fundamental indicators and subsequently optimizing them by maximizing the Sharpe ratio showed a superior return and faster recovery after drawdown periods compared to the benchmark or SVM (1/n) strategy. These results suggest the relevance of including the SVM in the optimization portfolio process.