This research introduces a two-stage dynamic topic analysis framework that enhances AI's ability to track how academic topics evolve over time, with superior consistency and interpretability.

• Employs convex non-negative matrix factorization to improve topic consistency across time periods
• Creates more interpretable and sparse topic representations compared to traditional methods
• Analyzes large-scale IEEE journal data to identify key topic trends across disciplines
• Enables better knowledge organization and discovery in academic research

For education stakeholders, this framework offers powerful tools to map curriculum relevance against emerging research trends, identify cross-disciplinary opportunities, and guide strategic research investment decisions.

Dynamic Topic Analysis in Academic Journals using Convex Non-negative Matrix Factorization Method