| G. Vonitsanos, A. Kanavos, Ph. Mylonas |
| Context-Enriched Hybrid Modeling for Cryptocurrency Price Prediction: Integrating ARIMAX and Deep Learning Architectures |
| 21th International Conference on Web Information Systems and Technologies (WEBIST 2025), 21-23 October 2025, Marbella, Spain |
|
ABSTRACT
|
| The inherent volatility and nonlinear dynamics of cryptocurrency markets present significant challenges to accurate price forecasting. This study explores a hybrid modeling approach combining classical time series analysis and deep learning techniques to enhance prediction accuracy in the context of Bitcoin price movements. We comprehensively evaluate ARIMA, ARIMAX, Support Vector Machines (SVM), and Long Short-Term Memory (LSTM) networks using high-resolution historical market data from 2019 to 2024. Emphasis is placed on integrating exogenous variables such as market volume, capitalization, and moving averages to enrich model inputs. The experimental results demonstrate that hybrid models, particularly ARIMAX, outperform standalone statistical and machine learning methods in terms of Root Mean Squared Error (RMSE) and R2 score, achieving superior alignment with actual market trends. The findings underscore the utility of synergistic frameworks that leverage historical statistical regularities and deep learning¢s capacity to model nonlinear temporal dependencies. This research contributes to the advancement of robust, data-driven tools for financial forecasting in highly dynamic and speculative digital asset markets.
|
| 21 October , 2025 |
| G. Vonitsanos, A. Kanavos, Ph. Mylonas, "Context-Enriched Hybrid Modeling for Cryptocurrency Price Prediction: Integrating ARIMAX and Deep Learning Architectures", 21th International Conference on Web Information Systems and Technologies (WEBIST 2025), 21-23 October 2025, Marbella, Spain |
[  PDF] [
BibTex] [
Print] [
 Back] |
