A. CEBRIAN-HERNANDEZ, E. JIMÉNEZ-RODRÍGUEZ, A. Díaz
This study aims to utilize Machine Learning techniques to analyze and predict Bitcoin volatility. Unsupervised learning methods like K-Means, DBSCAN, and HDBSCAN are employed to cluster behavioral patterns. The main goal is to determine the relationship between Bitcoin volatility and various exogenous financial variables such as RIOT Blockchain, NVIDIA, KBR, VISA, the EUR/USD exchange rate, and the S&P500 index. These variables serve as exogenous inputs for clustering. The data will be grouped by similar characteristics to assess the impact on cryptocurrency price movements. Preliminary findings suggest that DBSCAN outperforms in terms of cluster separation and noise resistance. However, all models face limitations due to Bitcoin's volatile and unpredictable nature, indicating a need for additional approaches to enhance forecasting accuracy.
Keywords: Bitcoin, Volatility, Machine Learning, Clustering Algorithms
Scheduled
AR1 Risk analysis II
June 12, 2025 3:30 PM
MR 1