Machine learning predicts stablecoin

Xiangjin Liu, Dehua Shen,
Machine learning predicts stablecoin,
Finance Research Letters,
Volume 86, Part E,
2025,
108444,
ISSN 1544-6123,
https://doi.org/10.1016/j.frl.2025.108444.
(https://www.sciencedirect.com/science/article/pii/S1544612325016988)
Abstract: This paper employs five types of machine learning methods, i.e., AdaBoost, RF, XGBoost, LightGBM, and LSTM, to predict stablecoin’s return. The empirical results mainly reveal that (1) investor attention proxies, e.g., extreme return, abnormal trading volume, and Google Trends, can effectively enhance predictive accuracy, and (2) SHapley Additive exPlanations (SHAP) indicates that both traditional attention proxies and direct attention proxies play an important role in predicting stablecoin.
Keywords: Stablecoin; Investor attention; Alternative data; Machine Learning; SHAP