Researchers develop machine learning system to identify and classify predictive statements about five major cryptocurrencies in social media tweets.

New two-stage classification framework analyzes cryptocurrency tweets for Cardano, Matic, Binance, Ripple, and Fantom, first identifying predictive statements, then categorizing them as incremental, decremental, or neutral

Study combines manual annotation with GPT-based labeling methods and uses SenticNet to extract emotion features corresponding to each prediction category

Addresses class imbalance through GPT-generated paraphrasing for data augmentation to improve model robustness

Research aims to decode market sentiment and emotional signals in cryptocurrency social media discussions to understand speculative activity trends