Zhen Liu

Fivefold cross-validation showed that the proposed RoBERTa + IFPA + decision tree framework outperformed six classical machine-learning baselines and three neural architectures, achieving an overall accuracy of 92.36% and an F1-score of 0.91.