Publicação
Enhancing Buy and Hold Strategy with Deep Reinforcement Learning Ensemble for Drawdown Risk Mitigation in Cryptocurrency Markets
| datacite.subject.fos | Ciências Naturais::Ciências da Computação e da Informação | pt_PT |
| dc.contributor.advisor | Scott, Ian James | |
| dc.contributor.advisor | Alpalhão, Nuno Tiago Falcão | |
| dc.contributor.author | Slesarev, Stanislav | |
| dc.date.accessioned | 2024-12-30T10:10:37Z | |
| dc.date.available | 2024-12-30T10:10:37Z | |
| dc.date.issued | 2024-10-30 | |
| dc.description | Dissertation presented as the partial requirement for obtaining a Master's degree in Data Science and Advanced Analytics, specialization in Data Science | pt_PT |
| dc.description.abstract | It is known that the average retail investor finds the financial markets quite complex, unpredictable and volatile, which can lead an investor to emotional mistakes and significant financial losses. Buy and Hold investment strategy is known to be the best for retail investors, but it is susceptible to severe drawdowns in the phases of declining market. To address this issue, Buy and Hold can be enhanced by some risk management approach (e.g. trend following strategy). Another way to control risk can be the application of Machine Learning in Finance, namely Deep Reinforcement Learning, which proved to be successful in multiple fields including e-commerce, energy sector, gaming etc. This thesis is dedicated to researching the application of Reinforcement Learning in the form of an ensemble of Deep Reinforcement Learning Agents to mitigate drawdown risks and prevent investors from significant financial losses. The ensemble of agents is trained on past cryptocurrency market data, namely Bitcoin market, while a part of that data is used to evaluate how well the ensemble is generalized and to measure its performance. The findings of this study show the ensemble of trading agents can succeed at risk control and reduction compared to Buy and Hold passive investment strategy but still lacks the ability to achieve the same returns. | pt_PT |
| dc.identifier.tid | 203784871 | pt_PT |
| dc.identifier.uri | http://hdl.handle.net/10362/176840 | |
| dc.language.iso | eng | pt_PT |
| dc.rights.uri | http://creativecommons.org/licenses/by/4.0/ | pt_PT |
| dc.subject | Reinforcement Learning | pt_PT |
| dc.subject | Q-Learning | pt_PT |
| dc.subject | Double Dueling Deep Q Network | pt_PT |
| dc.subject | Ensemble of DQN Agents | pt_PT |
| dc.subject | Bitcoin | pt_PT |
| dc.subject | Trading | pt_PT |
| dc.subject | Risk management | pt_PT |
| dc.title | Enhancing Buy and Hold Strategy with Deep Reinforcement Learning Ensemble for Drawdown Risk Mitigation in Cryptocurrency Markets | 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 Ciência de Dados e Métodos Analíticos Avançados, especialização em Ciência de Dados | pt_PT |