Institutional Repository of Vilnius University
Ekonominių veiksnių įtaka prekybos strategijose taikant rekurentinius neuroninius tinklus ir mokymąsi su pastiprinimu /
Abstract
dc:descriptionAs the financial market complexity and sensitivity for external factors expands, there is a growing need for innovative, adaptable trade strategies. Deep reinforcement learning methods appears to be more and more applicable for trading in finance markets, but external factors seem to be ignored. This Bachelor research aims to explore how the integration of economic factors via recurrent neural networks into the environment of a deep reinforcement learning agent, affects its trading decisions. The experimental study led to the development of a program, which automatically scans information about economic factors from publicly available websites. A trading agent based on the PPO algorithm has also been developed with additional GRU neural network layer. Comparisons were made between agent trading results, when agent accepts trade solution only based on price dynamics and technical indicators, and when information is integrated into the agent’s environment indicating the duration to the nearest economic event. The results of the experimental study showed that an agent with no information about economic factors, traded during these periods without considering market dynamics and usually suffered losses. It was also found that adding an attribute to the agent’s environment, indicating the time to the next economic event, made its trading strategy significantly more effective: increased trading efficiency and profitability, improved riskreturn ratios and significant losses were avoided during economic events.
Degree
thesis:*- Grantor dc:publisher
- Institutional Repository of Vilnius University
- Year dc:date
- 2025
Author and committee
dc:creator, dc:contributor.*- Author dc:creator
-
- Dragūnas, Domas,
Rights
dc:rights- Statement dc:rights
-
- info:eu-repo/semantics/openAccess
- Language dc:language
- lit
Identifiers
dc:identifier.*- Repository record dc:identifier
- https://repository.vu.lt/VU:ELABAETD239030229&prefLang=en_US