This Diploma thesis dives into predicting cryptocurrency prices, with a particular emphasis on Bitcoin, due to its leading role in the market. Utilizing a comprehensive dataset that contains daily Bitcoin prices from September 24, 2019, to February 6, 2024, I developed and compared three distinct deep learning models: ARIMAX (AutoRegressive Integrated Moving Average with exogenous parameters), SVM (Support Vector Machine), and LSTM (Long Short-Term Memory).
- Notebook: ARIMAX Model Notebook
- Root Mean Squared Error:
508.450 - R-squared Value:
0.992
- Notebook: SVM Model Notebook
- Root Mean Squared Error:
793.323 - R^2 Score:
0.981
- Notebook: LSTM Model Notebook
- Root Mean Squared Error:
943.177 - R^2 Score:
0.973
The dataset comprises daily Bitcoin prices from September 24, 2019, to February 6, 2024, sourced from CoinMarketCap.
The analysis of these models provides valuable insights into the complexities of predicting cryptocurrency prices. The high R^2 scores indicate that all three models can closely track the real price movements of Bitcoin, offering promising directions for further research in financial market prediction.