Please use this identifier to cite or link to this item: http://hdl.handle.net/10609/152079
Title: Evaluating Discrete and Continuous Action Spaces in Soft Actor-Critic for Single-Asset Cryptocurrency Trading
Other Titles: Evaluación de espacios de acción discretos y continuos en el modelo Soft Actor Critic para el comercio de criptomonedas de un solo activo
Author: López Marzabal, Miguel
Director: Pérez Ibáñez, Rubén
Tutor: Benito Altamirano, Ismael
Abstract: This thesis explores the impact of action space design on the performance of Soft Actor-Critic (SAC) algorithms in single-asset cryptocurrency trading environments. Reinforcement learning (RL) has shown promise in financial markets, where decision-making under uncertainty is critical. SAC, with its adaptability to continuous and discrete actions, offers a unique opportunity to assess how action granularity influences trading performance. The study involves implementing SAC in two versions of a trading environment: one with discrete actions (e.g., buy, hold, sell) and another with continuous actions (e.g., varying position sizes). By analyzing key metrics such as profitability, risk-adjusted returns, and computational efficiency, this research aims to provide insights into the trade-offs between action space design choices in RL for financial applications. The findings are expected to contribute to the development of more effective RL-based trading systems.
Keywords: Deep Reinforcement Learning
Soft Actor-Critic (SAC)
Action Space Comparison
Cryptocurrency Trading
Single-Asset
Document type: info:eu-repo/semantics/masterThesis
Issue Date: 2025
Publication license: http://creativecommons.org/licenses/by-nc-nd/3.0/es/  
Appears in Collections:Bachelor thesis, research projects, etc.

Files in This Item:
File Description SizeFormat 
mlopezmarzTFM0225.pdf5,03 MBAdobe PDFThumbnail
View/Open
Share:
Export:
View statistics

Items in repository are protected by copyright, with all rights reserved, unless otherwise indicated.