KNUST Team Wins Best Presenter Award for EduTrace Early‑Warning System at IURC 2026

Key Takeaways

• Five KNUST computer‑science students earned the Best Presenter Award (Parallel Session) at the 10th International Undergraduate Research Conference.
• Their project, EduTrace, delivers dropout‑risk alerts via 160‑character SMS, bypassing the need for smartphones or internet.
• Testing on 248 unseen records identified more true dropout cases than an untuned baseline, though ranking performance was not statistically superior to chance.
• The system is positioned as a decision‑support tool for resource‑constrained junior‑high schools in rural Ghana.

A team from Kwame Nkrumah University of Science and Technology’s Computer Science Department captured international attention at the virtual 10th International Undergraduate Research Conference hosted by Universiti Teknologi Malaysia. Their presentation of EduTrace, an explainable early‑warning system for school dropout, secured the Best Presenter Award in the parallel session track.

The accolade underscores the growing relevance of low‑tech, data‑driven interventions in Ghana’s rural education sector, where limited connectivity has historically hampered the deployment of sophisticated machine‑learning solutions.

Background & Context

School dropout rates in Ghana’s rural junior‑high schools remain a persistent challenge, driven by factors such as economic pressure, limited infrastructure, and irregular attendance tracking. Traditional early‑warning systems rely on digital dashboards and constant internet access, prerequisites that many remote schools cannot meet.

The KNUST team, operating within the CAN‑DO Virtual Lab and supervised by Dr. Eric Opoku Osei, sought to bridge this gap by designing a solution that leverages existing paper‑based registers while remaining operable on basic 2G networks.

System Design and Findings

EduTrace extracts six variables from ordinary school registers—attendance, grades, and other readily recorded metrics—and feeds them into a machine‑learning model that predicts dropout risk. The novel “SHAP‑to‑SMS” component translates the model’s explanatory output into a concise 160‑character text message, enabling alerts to be dispatched without internet reliance.

During development, the researchers evaluated 428 student‑year records from 180 pupils spanning 2024‑2026. A subsequent test on 248 unseen records, which included seven confirmed dropout cases, demonstrated that EduTrace flagged more genuine dropouts than an untuned baseline while keeping every alert within a single SMS.

Despite these gains, statistical analysis revealed that EduTrace’s ability to rank at‑risk pupils above others did not differ significantly from random chance, prompting the authors to frame the tool as supplementary decision support rather than a definitive predictor.

Implications for Rural Education

The system’s reliance on SMS and paper data addresses two critical barriers: lack of reliable broadband and limited device availability. By delivering actionable insights directly to teachers or school administrators via basic mobile phones, EduTrace can prompt timely interventions such as counseling, financial aid, or community outreach.

Adoption of such low‑tech solutions could reshape how Ghana’s Ministry of Education monitors student retention in underserved regions, potentially informing policy adjustments and resource allocation without the overhead of costly digital infrastructure.

Looking Ahead

Future work will focus on expanding the variable set, refining the ranking algorithm, and conducting longitudinal field trials across multiple districts. Partnerships with local education authorities could facilitate scaling while preserving the system’s low‑resource ethos.

Should EduTrace demonstrate consistent predictive value in broader deployments, it may serve as a template for analogous early‑warning frameworks in other sectors where connectivity constraints limit the use of advanced analytics.