Construction of Information Architecture for Ideological and Political Teaching Management Driven by Deep Learning from the Perspective of AI Empowerment

The digital transformation of engineering education requires teaching management systems that can process learning data, evaluate professional development, and support curriculum coordination. This need is particularly relevant for students in textile engineering, electronic information, antenna technology, and electromagnetic wave propagation, where interdisciplinary knowledge and engineering ethics are increasingly important. From the perspective of artificial intelligence empowerment, this study constructs a deep-learning-driven information architecture for ideological and political teaching management. Recent literature is reviewed to identify existing limitations in data integration, teaching evaluation, and management feedback. An information architecture is then designed with four layers: data, algorithm, application, and guarantee. Experiments compare deep learning with traditional machine learning methods. For student performance prediction, the Bi-LSTM model achieves an accuracy of 87.6%, which is 15.3% higher than random forest. For teaching resource classification, the improved Attention-CNN reaches an F1-score of 0.89, outperforming SVM. The results also show stronger robustness of deep learning models in long-sequence learning behavior analysis. This study provides a data-driven management framework for engineering education and offers support for cultivating students in technical fields that combine material engineering, intelligent manufacturing, and electromagnetic applications.