Forecasting Exchange Rates in the Short-Term: The Statistical and Machine Learning Approach
| Thesis title: | Forecasting Exchange Rates in the Short-Term: The Statistical and Machine Learning Approach |
|---|---|
| Author: | Feio Moreira da Silva, Gonçalo |
| Thesis type: | Diploma thesis |
| Supervisor: | Brůna, Karel |
| Opponents: | Časta, Martin |
| Thesis language: | English |
| Abstract: | We show that regime classification can be a beneficial addition to deep learning models when forecasting exchange rates, especially in volatile environments or highly information efficient currencies. We build on existing applications of Deep Learning technologies in the financial spectrum, and develop a combinational approach to regime classification utilizing know Regime Aware Multiscale Transformer methods and a Mixture of Experts, comparing its performance with econometric and Deep Learning benchmarks. Overall, the findings indicate that the effectiveness of regime-aware Deep Learning is conditional on the underlying market environment and characteristics. |
| Keywords: | Deep Learning; Financial Time Series; LSTM; Regime-Awareness; Exchange Rates; RAMuST; Mixture of Experts |
| Thesis title: | Forecasting Exchange Rates in the Short-Term: The Statistical and Machine Learning Approach |
|---|---|
| Author: | Feio Moreira da Silva, Gonçalo |
| Thesis type: | Diplomová práce |
| Supervisor: | Brůna, Karel |
| Opponents: | Časta, Martin |
| Thesis language: | English |
| Abstract: | We show that regime classification can be a beneficial addition to deep learning models when forecasting exchange rates, especially in volatile environments or highly information efficient currencies. We build on existing applications of Deep Learning technologies in the financial spectrum, and develop a combinational approach to regime classification utilizing know Regime Aware Multiscale Transformer methods and a Mixture of Experts, comparing its performance with econometric and Deep Learning benchmarks. Overall, the findings indicate that the effectiveness of regime-aware Deep Learning is conditional on the underlying market environment and characteristics. |
| Keywords: | Deep Learning; Exchange Rates; Regime-Awareness; Financial Time Series; LSTM; RAMuST; Mixture of Experts |
Information about study
| Study programme: | Finance and Accounting |
|---|---|
| Type of study programme: | Magisterský studijní program |
| Assigned degree: | Ing. |
| Institutions assigning academic degree: | Vysoká škola ekonomická v Praze |
| Faculty: | Faculty of Finance and Accounting |
| Department: | Department of Monetary Theory and Policy |
Information on submission and defense
| Date of assignment: | 15. 11. 2025 |
|---|---|
| Date of submission: | 10. 5. 2026 |
| Date of defense: | 9. 6. 2026 |
| Identifier in the InSIS system: | https://insis.vse.cz/zp/94503/podrobnosti |