EURO-Online login
- New to EURO? Create an account
- I forgot my username and/or my password.
- Help with cookies
(important for IE8 users)
2311. BitMood: Analyzing Bitcoin Trends through Facebook Emotions with AI
Invited abstract in session WC-31: Innovations in Digital Assets - IDA, stream Analytics.
Wednesday, 12:30-14:00Room: 046 (building: 208)
Authors (first author is the speaker)
| 1. | Daniel Traian Pele
|
| Bucharest University of Economic Studies. Institute for Economic Forecasting, Romanian Academy | |
| 2. | Miruna Mazurencu Marinescu Pele
|
| Statistics and Econometrics, Bucharest University of Economic Studies | |
| 3. | Stefan Gaman
|
| Cybernetics and Economic Statistics, Bucharest University of Economic Studies | |
| 4. | Alexandra Conda
|
| Bucharest University of Economic Studies | |
| 5. | Raul Cristian Bag
|
| Bucharest University of Economic Studies | |
| 6. | Bogdan Paul Saftiuc
|
| Bucharest University of Economic Studies |
Abstract
In the rapidly evolving landscape of cryptocurrency markets, the predictive power of social media sentiment has emerged as a pivotal area of exploration. This study presents a comprehensive analysis aimed at uncovering the intricate relationship between public sentiment expressed on Facebook and the volatility of Bitcoin trading volumes. Leveraging an extensive dataset of 120,000 Facebook posts tagged with Bitcoin-related keywords, combined with granular financial data from CoinGecko, we employ a methodologically advanced approach integrating Long Short-Term Memory (LSTM) networks, sentiment analysis via FinBERT, and topic modeling techniques.
The LSTM model's ability to capture temporal dependencies in data makes it particularly suited for the volatility inherent in cryptocurrency markets, providing a more accurate forecasting tool compared to traditional time-series models. The findings from our analysis reveal significant correlations between the sentiment derived from social media and the fluctuations in Bitcoin market trends. These correlations not only validate the hypothesis that social media sentiment can serve as an effective predictor of market movements but also underscore the efficacy of LSTM networks in enhancing the precision of volume predictions.
Keywords
- Economic Modeling
- Analytics and Data Science
- Artificial Intelligence
Status: accepted
Back to the list of papers