Study of the volatility of bitcoin cryptocurrency using machine learning methods : an implementation in R

• Κύριος συγγραφέας: Αριστείδου, Χριστόφορος
• Άλλοι συγγραφείς: Aristeidou, Christoforos
• Γλώσσα: English
• Έκδοση: 2020
• Θέματα: Bitcoin; Cryptocurrency; Realized volatility; Prediction; Blockchain; RandomForest; Machine learning; Algorithm; Forecast; Time series; Πρόβλεψη; Μεταβλητότητα; Κρυπτονόμισμα; Μηχανική μάθηση; Αλγόριθμος; Χρονοσειρά

In this dissertation we forecast Bitcoin’s realized volatility using Blockchain information variables and machine learning methods. Based on the literature of Bitcoin, we used Blockchain information and the empirical results correspond with the literature that Blockchain information can explain the variability of Bitcoin’s price. We are trying to examine which variables are more important and driving Bitcoin’s price which is consider highly volatile. Also for a better prediction results we use machine learning that is considered by the literature that results in a higher accuracy. In our analysis we use Random Forest machine learning algorithm for the prediction of realized volatility. From the empirical results we conclude that the trade_volume variable is the most important variable included in the model. Also medium to high important variables are transaction_fees_usd, cost_per_transaction_percent, miners_revenue and output_volume. In the other hand the less important variables are n_transaction and n_transaction_per_block.