Predictive analytics to enable proactive student support interventions

One key area where the predictive analytic tools in education sector fall behind is their adaptability across education institutions or across modules. This paper introduces a portable and scalable tool for predicting student academic performance, which can be adopted by different institutions working with diverse set of Learning Management Systems (LMSs). The tool can process variety of student data formats, automatically select the most relevant performance metrics (features), and evaluate multiple ML models to determine the best fit for each module’s dataset. The tool is designed using microservice architecture, which can allow institutions full control over deployment and data management. The tool is accessible as a web application with a user-friendly dashboard, enabling module owners to preprocess data, select optimal ML model, and generate predictions accurately. Experimental results demonstrate high prediction accuracy (F1 Score up to 97%) while working with different data formats and multiple features, validating the tool’s robustness and reliability. Results further showcase how early identification of at-risk students can allow for timely interventions, contributing to improved academic outcomes. The paper also discusses implementation challenges and solutions adopted for broad dissemination. Future work includes further validation across more diverse modules and institutions to enhance trust and public availability.