Publicação
Self-Organizing Maps Trading Strategy: Brazilian Stock Market
| datacite.subject.fos | Ciências Naturais::Ciências da Computação e da Informação | pt_PT |
| dc.contributor.advisor | Bação, Fernando José Ferreira Lucas | |
| dc.contributor.author | Ferreira, Rodrigo Arend Vaz | |
| dc.date.accessioned | 2025-07-04T13:38:07Z | |
| dc.date.available | 2025-07-04T13:38:07Z | |
| dc.date.issued | 2025-06-23 | |
| dc.description | Dissertation presented as the partial requirement for obtaining a Master's degree in Information Management, specialization in Business Intelligence | pt_PT |
| dc.description.abstract | This thesis presents a novel trading strategy for the Brazilian stock market, leveraging SelfOrganizing Maps (SOMs) to uncover hidden patterns in the daily behavior of the Ibovespa index from 2000 to 2023 and tested in data from 2024 to March 2025. By transforming raw price and technical indicators data into a set of Perceptually Important Points (PIPs), the model captures essential market movements while discarding less relevant fluctuations. A comprehensive time-series cross-validation approach is employed, ensuring rigorous performance evaluation on unseen data. The proposed method automatically clusters similar market conditions and assigns trading signals based on the expected future return within each cluster. Experimental results show that this SOM-based strategy outperforms both a simple buy-and-hold benchmark and other widely used technical analysis approaches in terms of overall returns and risk-adjusted performance. Notably, the inclusion of technical indicators provides a richer description of market trends, volatility, and momentum, enhancing the SOM’s ability to distinguish profitable patterns. The framework also incorporates risk management by evaluating the Sharpe ratio and other metrics, highlighting the robustness of the strategy across diverse market regimes. These findings suggest that a well-tuned unsupervised learning approach, combined with carefully selected financial features, can systematically exploit market inefficiencies. Moreover, the methodology is flexible and can be extended to other stock markets or asset classes with minimal adjustments, underlining the versatility and practical value of the presented solution. | pt_PT |
| dc.identifier.tid | 203969170 | |
| dc.identifier.uri | http://hdl.handle.net/10362/184803 | |
| dc.language.iso | eng | pt_PT |
| dc.rights.uri | http://creativecommons.org/licenses/by/4.0/ | pt_PT |
| dc.subject | Self-Organizing Maps | pt_PT |
| dc.subject | Trading Strategies | pt_PT |
| dc.subject | Brazilian Stock Market | pt_PT |
| dc.subject | Technical Analysis | pt_PT |
| dc.subject | Perceptually Important Points | pt_PT |
| dc.title | Self-Organizing Maps Trading Strategy: Brazilian Stock Market | pt_PT |
| dc.type | master thesis | |
| dspace.entity.type | Publication | |
| rcaap.rights | openAccess | pt_PT |
| rcaap.type | masterThesis | pt_PT |
| thesis.degree.name | Mestrado em Gestão de Informação, especialização em Inteligência de Negócio | pt_PT |