On-chain blockchain metrics combined with social sentiment can reliably classify cryptocurrency market emotion, with XGBoost and SHAP providing both accuracy and transparency about which features matter most.
This paper builds a machine learning classifier to understand Bitcoin market sentiment by combining blockchain transaction data, price history, and Twitter posts. Using XGBoost and SHAP for interpretability, the model achieves 84% F1-score and reveals which on-chain metrics best explain market emotion.