Abstract
dc:descriptionThis thesis explores the applications of high-frequency and intraday financial data in the Bitcoin market, focusing on its unique characteristics of high volatility and extreme jumpy behavior. Bitcoin's price is known for significant fluctuations within short periods, attributed to speculative trading, regulatory news, technological developments, and macroeconomic trends. Understanding and predicting these price movements is crucial for traders, investors, and policymakers. Besides regular volatility, the Bitcoin market often exhibits sudden and large price changes, referred to as “jumps”, triggered by market news, large trades, or shifts in market sentiment. Capturing and analyzing these jumps can provide valuable insights into market dynamics and aid in developing robust trading strategies. In particular, the thesis presents a novel four-step methodological approach for predicting volatility estimators derived from Bitcoin prices. First, open, high, low, and close prices are transformed into volatility estimators using Brownian motion assumptions and logarithmic transformations. Second, the optimal number of time-series lags is determined using random forest regression, optimizing lags based on the Mean Decrease in Impurity (MDI) criterion. Third, Elastic Net Regression (ENR) is fitted to the volatility estimator dataset. Finally, the predictive accuracy of ENR is compared to decision tree (DTR), random forest (RFR), and support vector regression (SVR). Results show ENR's superior predictive accuracy for open and close prices, while SVR prevails for high and low prices due to their susceptibility to transient news and intra-day market sentiments. Additionally, the thesis investigates Bitcoin volatility patterns using R/S analysis, simple moving average (SMA), and the relative strength index (RSI). These methodologies capture Bitcoin's unpredictable volatility patterns, revealing underlying trends. Estimated Hurst exponents indicate persistence in these patterns, suggesting potential momentum-based trading strategies. Finally, the thesis introduces a dual approach using the Lévy-Khintchine process and the Wasserstein Generative Adversarial Network (WGAN-GP) for a detailed analysis of Bitcoin's price behavior under various market conditions. The Lévy-Khintchine process captures extreme price changes, while WGAN-GP tests the robustness of these predictions, showing high accuracy in extreme conditions. In summary, this thesis underscores the essential role of high-frequency and intraday data in understanding and predicting Bitcoin's market behavior, providing a foundation for more accurate predictive models, deeper insights into the market microstructure, and improved trading strategies and risk management practices in this volatile and dynamic market.
Degree
thesis:*- Grantor dc:publisher
- Ελληνικό Μεσογειακό Πανεπιστήμιο
- Year dc:date
- 2024
Author and committee
dc:creator, dc:contributor.*- Authors dc:creator
-
- Zournatzidou, Georgia
- Ζουρνατζίδου, Γεωργία
Subjects
dc:subject × 30- Ανάλυση κινδύνου
- Μεταβλητότητα
- Δείκτες κινδύνου
- Άλματα
- Δεδομένα υψηλής συχνότητας
- Οικονομικά σοκ
- Πολυμεταβλητή θεωρία αλμάτων
- Χρηματοπιστωτική κρίση
- Risk analysis
- Volatility
- Risk metrics
- Jumps
- High frequency data
- Economic shocks
- Multivariate jump theory
- Financial crisis
- Μαθηματικά
- Φυσικές Επιστήμες
- Εφαρμοσμένα μαθηματικά
- Οικονομικά και Επιχειρήσεις
- Κοινωνικές Επιστήμες
- Χρηματοοικονομικά
- Στατιστικές, Πιθανότητες και Αβεβαιότητα
- Mathematics
- Natural Sciences
- Applied Mathematics
- Economics and Business
- Social Sciences
- Finance
- Statistics, Probability and Uncertainty
Rights
- Language dc:language
- eng
Identifiers
dc:identifier.*- Identifier
- 10.12681/eadd/58103