Bitcoin high frequency trading with Transformers
Συναλλαγές υψηλής συχνότητας Bitcoin με νευρωνικά δίκτυα Transformer
View/
Keywords
Transformers ; Bitcoin ; High frequency trading ; Neural networks ; Price movement prediction ; Machine learning ; Deep learningAbstract
In recent years, the cryptocurrency market, with Bitcoin at its forefront, has emerged as a highly volatile and dynamic trading landscape, capturing the interest of traders, researchers, and investors alike. This thesis investigates the application of transformer models in high-frequency trading, focusing on their ability to predict Bitcoin price movements using Limit Order Book (LOB) data from the Binance exchange. Traditional forecasting methods often fall short in capturing the rapid fluctuations inherent in cryptocurrency markets. Thus, the research explores advanced machine learning techniques, specifically transformer architectures, which have shown promise in sequence modeling tasks across various domains.
The study detail the methodology employed in data collection and preprocessing to create a robust dataset sufficient for high-frequency trading analysis. Key challenges, such as model overfitting, data imbalance, and the intricacies of financial data characteristics like non-stationarity, are thoroughly examined. Through empirical testing, the thesis evaluates the performance of transformer models in predicting price dynamics and highlights their strengths over traditional deep learning approaches.
Despite the theoretical advantages of transformers, the findings reveal limitations in predictive success under specific market conditions. This thesis provides insights into the integration of domain-specific features, the importance of data preprocessing, and the need for tailored modeling strategies to enhance predictive capabilities in cryptocurrency trading. Future research directions are proposed, emphasizing the potential of stochastic modeling approaches and the incorporation of diverse data sources to improve trading strategies in the complex realm of Bitcoin high-frequency trading.
Postgraduate Studies Programme
Κυβερνοασφάλεια και Επιστήμη ΔεδομένωνDepartment
Σχολή Τεχνολογιών Πληροφορικής και Επικοινωνιών. Τμήμα ΠληροφορικήςNumber of pages
46Language
EnglishCollections
Except where otherwise noted, this item's license is described as
Αναφορά Δημιουργού-Όχι Παράγωγα Έργα 3.0 Ελλάδα
Αναφορά Δημιουργού-Όχι Παράγωγα Έργα 3.0 Ελλάδα