Market prediction using machine learning based on social media specific features.

Area of Science:

• Computational Finance
• Natural Language Processing
• Machine Learning

Background:

• Social media sentiment significantly impacts financial markets, including virtual currencies.
• Predicting sudden price changes in cryptocurrencies remains a challenge.

Purpose of the Study:

• To investigate the efficacy of Twitter data and natural language expressions for predicting virtual currency market information.
• To develop a model for forecasting sudden price changes (drops, rises, or stability) in cryptocurrencies.

Main Methods:

• Feature extraction from tweets using Sentence-BERT.
• Training a LightGBM classifier with these linguistic features.
• Classification task with three labels: sudden drop, sudden rise, or no sudden change.

Main Results:

• Linguistic features derived from tweets enhance the prediction of cryptocurrency price changes.
• The proposed method demonstrates the advantage of using natural language expressions for market trend prediction.

Conclusions:

• Social media sentiment analysis is a valuable tool for understanding and predicting cryptocurrency market dynamics.
• Integrating linguistic features into machine learning models improves the accuracy of financial market forecasting.