AI models predict at-risk students using digital learning traces

Researchers have investigated the generalizability of predictive models designed to identify at-risk students in higher education using digital learning traces. By analyzing data from undergraduate computer science courses at two universities, the study found that behaviors related to self-regulated learning, such as time management and sustained engagement, were strong predictors of student risk. While models could predict at-risk students early within courses, their performance decreased when applied across different institutions, especially when base rates of at-risk students varied. AI

IMPACT Highlights the challenges in generalizing educational AI models across different institutional contexts, suggesting caution is needed when applying predictive analytics.

RANK_REASON This is a research paper published on arXiv concerning predictive modeling in education.

Read on arXiv cs.LG →

• arXiv
• Elastic Net
• Random Forest
• XGBoost

• paper
• other

AI-generated summary · Google Gemini · from 1 sources. How we write summaries →