Paper: arXiv 2405.00522

Authors: Yihang Fu, Mingyu Zhou, Luyao Zhang

Abstract

In the distributed systems landscape, Blockchain has catalyzed the rise of cryptocurrencies, merging enhanced security and decentralization with significant investment opportunities. Despite their potential, current research on cryptocurrency trend forecasting often falls short by simplistically merging sentiment data without fully considering the nuanced interplay between financial market dynamics and external sentiment influences. This paper presents a novel Dual Attention Mechanism (DAM) for forecasting cryptocurrency trends using multimodal time-series data. Our approach, which integrates critical cryptocurrency metrics with sentiment data from news and social media analyzed through CryptoBERT, addresses the inherent volatility and prediction challenges in cryptocurrency markets. By combining elements of distributed systems, natural language processing, and financial forecasting, our method outperforms conventional models like LSTM and Transformer by up to 20% in prediction accuracy. This advancement deepens the understanding of distributed systems and has practical implications in financial markets, benefiting stakeholders in cryptocurrency and blockchain technologies. Moreover, our enhanced forecasting approach can significantly support decentralized science (DeSci) by facilitating strategic planning and the efficient adoption of blockchain technologies, improving operational efficiency and financial risk management in the rapidly evolving digital asset domain, thus ensuring optimal resource allocation.

Complexity vs Empirical Score

  • Math Complexity: 7.0/10
  • Empirical Rigor: 5.0/10
  • Quadrant: Holy Grail — high math complexity, high empirical rigor

Why this score: The paper employs advanced deep learning architectures like attention mechanisms and LSTMs, suggesting significant mathematical complexity, but also includes concrete empirical components like data sourcing, model comparison, and a GitHub link, indicating moderate to high rigor.

Research Flowchart

  flowchart TD
  Start(["Research Goal:<br>Cryptocurrency Trend Forecasting<br>with Sentiment & Financial Data"]) --> Inputs
  
  subgraph Inputs ["Data/Inputs"]
      A["Financial Time-Series<br>(Price, Volume)"]
      B["Sentiment Data<br>News & Social Media"]
      C["CryptoBERT<br>Pre-trained Model"]
  end
  
  Inputs --> Methodology
  
  subgraph Methodology ["Key Methodology Steps"]
      D["Dual Attention Mechanism<br>Integrates Multimodal Data"]
      E["Model Architecture<br>LSTM/Transformer Baseline"]
  end
  
  Methodology --> Process
  
  subgraph Process ["Computational Processes"]
      F["Training & Optimization<br>Outperforms Baselines by 20%"]
  end
  
  Process --> Outcomes(["Key Findings & Outcomes<br>Enhanced Forecasting Accuracy<br>Supports DeSci & Risk Management"])

Evidence Card

What the authors claim, on what data, and who has checked it. Read from the paper's own text by a language model: every quote is verbatim, nothing was re-run. How to read this card

Proposes

The paper proposes a Dual Attention Mechanism (DAM) for multimodal cryptocurrency time‑series forecasting that fuses financial and sentiment data.

Source sentence
We designed DAM, a universal Dual Attention Mechanism for multimodal financial time-series data fusion.

Assumes 4

  • The target variable for forecasting is the daily closing price of Bitcoin. data
    Source sentence
    The target variable is the daily closing price.
  • The last 70 days of data are used as a validation set. data
    Source sentence
    We split the last 70‑day data as the validation data.
  • Sentiment from news and social media influences cryptocurrency price movements. market
    Source sentence
    news or social media information on other stocks or cryptocurrencies could cause the co‑movements of the cryptocurrency.
  • Variables with high correlation (high, low, open price, volume) are relevant for the model. statistical
    Source sentence
    the variables high, low, open price, and volume show a high correlation with t…

Data

  • Bitcoin price data · CryptoCompare API market-prices public
    2020-01-01 to 2023-06-28 · Bitcoin
    Source sentence
    We queried cryptocurrency data (Bitcoin) ranges on a daily level from 2020-01-01 to 2023-06-28 by using CryptoCompare API.
  • Bitcoin news sentiment · Nasdaq text-news public
    Bitcoin
    Source sentence
    The crypto news information is queried from Nasdaq.
  • Bitcoin social media posts · Kaggle public dataset alternative public
    Bitcoin
    Source sentence
    The social media posts are collected from Kaggle public dataset.

Recognised sources: Kaggle

Costs

Not applicable The paper reports no trading or execution result.

Finds 5

  • The Dual Attention Mechanism improves prediction accuracy by up to 20 % compared with conventional models. performance
    Source sentence
    Our dual attention mechanism increases a 20 percent improvement.
  • DAM‑LSTM outperforms NeuralProphet and other baselines, showing a 20 % improvement over simple concatenation fusion. predictive-accuracy
    Source sentence
    Our Dual Attention Mechanism‑LSTM (DAM‑LSTM) model outperforms the NeuralProphet model and other selected models and shows a 20 percent improvement on the same LSTM architecture but with the simple concatenation fusion.
  • Integrating sentiment data raises TFT model prediction performance by about 20 % and LSTM model by about 5 %. predictive-accuracy
    Source sentence
    The application of multimodal data confers an elevation of prediction by approximately 20 percent for the TFT model and 5 percent for the LSTM model.
  • Stationary data input improves the LSTM‑based model by nearly 50 % in MAE. statistical
    Source sentence
    the stationary data input improves the LSTM‑based model by nearly 50% percent.
  • DAM‑LSTM achieves MAE 719.82 and MAPE 0.0297 on the multimodal test set. performance
    Source sentence
    DAM‑LSTM 719.82 431.86 0.0297 Yes

Checked by others

Qualified by

  • Understanding the Impact of News Articles on the Movement of Market Index: A Case on Nifty 50 (2024)
    On the question: Does adding sentiment data improve prediction accuracy for financial time series?
    Compare the findings

    This paper: Integrating sentiment data raises TFT model prediction performance by about 20 % and LSTM model by about 5 %.

    That paper: Using VADER sentiment scores with up to 3 lags marginally improves RMSE from 134.85 to 134.15 compared to a model without sentiment.

    A reports ~20% gain for TFT and ~5% for LSTM; B shows RMSE drop from 134.85 to 134.15 (~0.5%). Both find sentiment beneficial, though magnitude and metric differ.

Cited by 2 papers in the archive

Code

No repository by the authors found in the paper.

Extracted 2026-11-14 · same card as JSON · A language model chose these items: check the source sentence before relying on one.

Cited References (83)

  1. Replication Data for: ``DAM: A Universal Dual Attention Mechanism for Multimodal Timeseries Cryptocurrency Trend Forecasting’’ (2024) — link
  2. Temporal fusion transformers for interpretable multi-horizon time series forecasting (2021)
  3. Towards an understanding of cryptocurrency: a comparative analysis of cryptocurrency, foreign exchange, and stock (2019)
  4. A multimodal event-driven lstm model for stock prediction using online news (2020)
  5. The predictive power of public Twitter sentiment for forecasting cryptocurrency prices (2020)
  6. Cryptocurrency price prediction using tweet volumes and sentiment analysis (2018)
  7. News sentiment in the cryptocurrency market: An empirical comparison with Forex (2020)
  8. Ai ethics on blockchain: Topic analysis on twitter data for blockchain security — → 2212.06951
and 75 more
  1. Machine Learning based Cryptocurrency Price Prediction using historical data and Social Media Sentiment (2023)
  2. Multimodal deep learning for short-term stock volatility prediction (2018) — link
  3. Improving stock market prediction via heterogeneous information fusion (2018)
  4. Stock price prediction using generative adversarial networks (2021)
  5. Do US stock markets typically overreact to corporate news stories? (2006)
  6. A deep multimodal reinforcement learning system combined with CNN and LSTM for stock trading (2019)
  7. A systematic literature review of investor behavior in the cryptocurrency markets (2023)
  8. Vader: A parsimonious rule-based model for sentiment analysis of social media text (2014)
  9. Bitcoin price change and trend prediction through twitter sentiment and data volume (2022)
  10. Multisource financial sentiment analysis for detecting Bitcoin price change indications using deep learning (2022)
  11. Does Twitter happiness sentiment predict cryptocurrency? (2021)
  12. A hybrid data analytics framework with sentiment convergence and multi-feature fusion for stock trend prediction (2022)
  13. A hybrid model for bitcoin prices prediction using hidden Markov models and optimized LSTM networks (2019)
  14. Exploring the predictability of cryptocurrencies via Bayesian hidden Markov models (2022)
  15. Show me your NFT and I tell you how it will perform: Multimodal representation learning for NFT selling price prediction (2023)
  16. Forecasting Method of Stock Market Volatility Based on Multidimensional Data Fusion (2022)
  17. Big data fusion model for heterogeneous financial market data (findf) (2019)
  18. Predicting cryptocurrency price movements based on Social Media (2019)
  19. Forecasting stock prices with a feature fusion LSTM-CNN model using different representations of the same data (2019)
  20. Deep learning for event-driven stock prediction (2015)
  21. Crypto-currency price prediction using decision tree and regression techniques (2019)
  22. Bitcoin price forecasting method based on CNN-LSTM hybrid neural network model (2020)
  23. Sentiment-driven price prediction of the bitcoin based on statistical and deep learning approaches (2020)
  24. Machine learning models comparison for bitcoin price prediction (2018)
  25. Stock price prediction using the ARIMA model (2014)
  26. CryptoBERT
  27. Are transformers effective for time series forecasting? (2023)
  28. Transformers in time series: A survey (2022) — link
  29. Seasonality in cryptocurrencies (2019)
  30. Fast RobustSTL: Efficient and robust seasonal-trend decomposition for time series with complex patterns (2020)
  31. Transformer-based deep learning model for stock price prediction: A case study on Bangladesh stock market (2023)
  32. Accurate multivariate stock movement prediction via data-axis transformer with multi-level contexts (2021)
  33. Multi-head self-attention transformer for dogecoin price prediction (2021)
  34. Comprehensive analysis of non-fungible tokens valuation and accounting under IFRS: Challenges and artificial intelligence implications (2023) — link
  35. Ether price prediction using advanced deep learning models (2021)
  36. Fisher transformation — Wikipedia , The Free Encyclopedia (2023)
  37. Bitcoin network mechanics: Forecasting the btc closing price using vector auto-regression models based on endogenous and exogenous feature variables (2020)
  38. Bitcoin is not the New Gold – A comparison of volatility, correlation, and portfolio performance (2018)
  39. Testing for linear and nonlinear Granger causality in the stock price-volume relation (1994)
  40. Forecasting economic time series (2014)
  41. Deep learning applications and challenges in big data analytics (2015)
  42. Self-adaptive forecasting for improved deep learning on non-stationary time-series (2022) — link
  43. Discrepancy-based theory and algorithms for forecasting non-stationary time series (2020)
  44. Stock price forecasting with deep learning: A comparative study (2020)
  45. Ensemble and multimodal approach for forecasting cryptocurrency price (2022) — link
  46. Multimodal deep learning for biomedical data fusion: a review (2022)
  47. Foundations and recent trends in multimodal machine learning: Principles, challenges, and open questions (2022) — link
  48. Information fusion in attention networks using adaptive and multi-level factorized bilinear pooling for audio-visual emotion recognition (2021)
  49. A cross-modal fusion network based on self-attention and residual structure for multimodal emotion recognition (2021) — link
  50. Temporal pattern attention for multivariate time series forecasting (2019)
  51. A temporal fusion transformer for short-term freeway traffic speed multistep prediction (2022)
  52. On the properties of neural machine translation: Encoder-decoder approaches (2014) — link
  53. A dual-stage attention-based recurrent neural network for time series prediction (2017) — link
  54. Poster: Stock Price Prediction Using Machine Learning (2023)
  55. Recent advances in natural language processing via large pre-trained language models: A survey (2023)
  56. LSTM-MSNet: Leveraging forecasts on sets of related time series with multiple seasonal patterns (2020)
  57. Flashattention: Fast and memory-efficient exact attention with io-awareness (2022)
  58. NeuralProphet: Explainable Forecasting at Scale (2021) — link
  59. Frequency distribution of the values of the correlation coefficient in samples from an indefinitely large population (1915)
  60. Twitter sentiment data analysis of user behavior on cryptocurrencies: Bitcoin and ethereum (2021)
  61. Bitcoin Price Prediction Considering Sentiment Analysis on Twitter and Google News (2023)
  62. Another look at measures of forecast accuracy (2006)
  63. DL-GuesS: Deep learning and sentiment analysis-based cryptocurrency price prediction (2022)
  64. Does sentiment impact cryptocurrency? (2023)
  65. News-based sentiment and bitcoin volatility (2022)
  66. Predicting bitcoin price movements using sentiment analysis: a machine learning approach (2022)
  67. Real-time prediction of BITCOIN price using machine learning techniques and public sentiment analysis (2020) — link
  68. Blockchain challenges and opportunities: A survey (2018)
  69. Transform blockchain into distributed parallel computing architecture for precision medicine (2018)
  70. A Data Flow Framework with High Throughput and Low Latency for Permissioned Blockchains (2023)
  71. LDSP: shopping with cryptocurrency privately and quickly under leadership (2021)
  72. Cryptocurrency, a successful application of blockchain technology (2020)
  73. Blockchains from a distributed computing perspective (2019)
  74. Policy, ethical, social, and environmental considerations of Web3 and the metaverse (2022)
  75. DeSci based on Web3 and DAO: A comprehensive overview and reference model (2022)

1 of 83 cited works are in the corpus; resolved via arXiv source bundles, Semantic Scholar and OpenAlex. Corpus-wide: 2596 linked papers.