.png){kind=icon}
.png){kind=icon}
Return
Multimodal Cryptocurrency Volatility Prediction Analysis Based on LLM
.png){kind=icon}
OA
AI
DOI：10.1016/j.bcra.2026.100479.png)
.png){kind=icon}
.png){kind=icon}
Share.png){kind=icon}
Save
Abstract
En 中文
As the digital currency market rapidly evolves, predicting cryptocurrency volatility has become a critical area of financial research. However, traditional and machine learning approaches struggle to effectively capture the multi-dimensional market dynamics, abrupt fluctuations, and long-term dependencies, while also facing interpretability challenges. This study introduces a multimodal prediction approach leveraging large language models (LLMs) to enhance cryptocurrency volatility forecasting. We propose Multimodal Optimized Deep Embedding Learner (MODEL), a comprehensive framework that integrates historical volatility data with textual news information. MODEL incorporates a bimodal encoding module, an attention alignment mechanism, and a sequence prediction module to enhance predictive performance. Empirical analysis on main transaction data from Bitcoin (BTC), Ethereum (ETH), and Binance Coin (BNB) demonstrates the superior accuracy and robustness of our approach compared to existing techniques. By effectively capturing multimodal market signals and sentiment, MODEL improves prediction reliability. Ablation studies further validate the contributions of each module, underscoring the theoretical and practical significance of our research for quantitative trading strategies and related domains.
Keywords:
LLM
Multimodal Data
Market Sentiment Analysis
Cryptocurrency
Realized Volatility
AI Summary
Key information extracted from the uploaded paper, including a brief overview, abstract, background, key highlights, visual analysis, and future outlook.
Journal
B
IF:
0
Papers:
87
Citations:
0
View more organizations