Answers are grounded in this paper and cite their sources.
Deep learning modeling study demonstrates high-accuracy adaptive learning recommendations via sentiment-performance fusion in students, indicating a viable pathway for educational AI adoption.
In this research study, our main aim is to investigate how the students’ perceptions and performance interact for adoption of intelligent tool for learning purposes. The proposed framework integrates various perspectives including students’ results prediction, sentiment analysis and adoptability measures. The users’ behavior features are explored for proposal of deep learning based dual-channel data fusion-based methods. The study covers four core aspects. First, in sentiment classification, the proposed model achieves the highest accuracy (0.91) and F1-score (0.89), demonstrating the strength of combining semantic and affective embeddings. Second, in learning outcome regression, the framework predicts improvement with the lowest error (MAE = 3.8) and the highest explanatory power (R 2 = 0.72). Third, in adoption intention prediction, the model records superior results with 0.88 accuracy and 0.90 AUC, showing that trust, ease of use, and satisfaction strongly influence adoption. Finally, in adaptive recommendation, the integration of an agentic AI layer delivers personalized interventions, achieving top ranking metrics such as Precision@5 = 0.78 and MRR = 0.80. All tasks operate under the same proposed mechanism, the Dual-Channel Fusion Performance Encoder, which integrates sentiment embeddings with performance features through cross-attention and contrastive alignment. In addition, SHAP-based feature analysis, counterfactual evaluation, and attention weight visualization ensure transparency and reliability in decision-making. The findings establish a comprehensive pathway for advancing AI adoption in education, enabling adaptive, trustworthy, and effective learning experiences for both students and teachers.
No takes yet. Share an insight, caveat, or question.
Wang et al. (2026) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: