MFB

MFB: A Generalized Multimodal Fusion Approach for Bitcoin Price Prediction Using Time-Lagged Sentiment and Indicator Features

Environment

Install Python IDE, PyCharm from here:

https://www.jetbrains.com/pycharm/download/?section=windows

Datasets

• Tweets: https://www.kaggle.com/datasets/hiraddolatzadeh/bitcoin-tweets-2021-2022
• News: https://figshare.com/articles/dataset/MarketData_for_MarketPredict_RESTFul_API_including_News_and_Market_Data/14754966
• Bitcoin price: https://coinmarketcap.com/

Experiment steps

• Data Collection: Download Bitcoin price, financial tweets and news data respectively from the links given above.
• Preprocessing: run Data Standardization.ipynb, preprocesses the collected data to obtain structured data.
• Lagged Data Correlation: run Correlation Analysis.ipynb, firstly applies VADER to extract sentiment scores and then correlates the results with Bitcoin price to identify temporal relationships.
• Feature Optimization: run Data Standardization.ipynb, uses a combination of multiple methods to optimize features.
• Model Training and Performance Evaluation: run Split_data.ipynb, divides the test and training set; run MFB and baseline on news.ipynb and run MFB and baseline on tweets.ipynb, applies performance metrics to compare the effectiveness of MFB and baseline models on news and tweets.
• Bitcoin Price Forecasting: run MFB Optimizer Comparison.ipynb and run MFB Price Prediction on Bitcoin.ipynb, optimizes and deploys the trained model to predict Bitcoin for the next hour, and compares the results with other related methods.

License

EDS is licensed under the GNU General Public License; for more information, read the LICENSE file or refer to:

http://www.gnu.org/licenses/

Citation

A related paper is submitted to the SCI journal.