Application of TabNet Interpretable Model in the Evaluation of Student Satisfaction and Value Formation in Textile Ideological and Political Courses

With the rapid advancement of intelligent information processing and data-driven decision-making technologies, interpretable artificial intelligence has become increasingly important for complex evaluation tasks in engineering education and future intelligent communication systems. In textile universities, accurately assessing student satisfaction and value formation in ideological and political courses remains challenging because of the nonlinear interactions among multiple educational factors and the limited interpretability of conventional models. To address these issues, this study proposes an interpretable evaluation framework based on the TabNet deep learning architecture. A multi-task learning strategy is employed to simultaneously predict student satisfaction and value formation while providing transparent attribution of influential factors through sequential attention mechanisms. Using questionnaire data collected from 832 textile students, the proposed model achieves a satisfaction prediction accuracy of 91.3% and a value prediction mean squared error of 0.653, outperforming conventional machine learning approaches. The analysis identifies textile ethics integration, craftsmanship guidance, and professional identity as the dominant factors affecting educational outcomes and further reveals significant nonlinear interaction effects and population heterogeneity across majors and grade levels. The proposed framework provides an accurate and explainable methodology for intelligent educational assessment and offers valuable references for data-driven decision support, adaptive information processing, and interpretable learning models in future smart sensing and communication-oriented engineering applications.