Educational Concept Discovery Using Nonnegative Matrix Factorization in Curriculum Mining Systems

Curriculum mining has emerged as a process capable of uncovering implicit patterns, relationships, and educational concepts in substantial university datasets with respect to technology use and learning outcomes. As a result, recent reflection accounts have been more strongly oriented to the quantitative processes for automated qualitative analysis of course structure, learning resources, and student-related outcomes towards better curriculum design and delivery. Unfortunately, many of the commercially-based approaches have struggled to generate educationally interpretable concepts in high-dimensional curriculum data. Traditional clustering and associated methods have either overlooked or failed to discover latent educational structures and relationships to implicit discipline-specific learning outcomes. These obstacles have further complicated the legal to establish trustworthiness and usefulness of the concept discoveries in authentic settings. This presentation will present Non-Negative Matrix Factorization-Based Educational Concept Discovery (NMFECD) as a method to address these compromises. NMF-ECD involves the enchancement use of non-negative matrix factorization to identify latent concept structures from curricular data, while still maintaining at least one level of interpretability worth pursuing. NMF-ECD can include the use of semantic similarity methods with flexible weighting for better alignment with educational intentions, and record curricular gaps, redundancies and progression behaviour. The NMF-ECD will equally emphasize accuracy and interpretability to provide a more trustworthy method for discovering actionable knowledge regarding education. Evaluative data found that NMF-ECD improved concept coherence, correctly identified curriculum gaps, and systematically mapped knowledge in cultivate knowledge categories compared to the standard approaches. Yet, results showed the method delineated the capacity for large data analytics and provided fine-grain evaluative differentiating factors among curriculum approaches. The structure relates directly to personalized learning systems and adaptive pathways that include the discovery of concepts and the formation of curriculum.

Hali tarjima qilinmagan