SAPP: Student Academic Performance Predictor

Post COVID-19 pandemic, the education system has significantly shifted towards blended and hybrid learning approaches. In most cases, course delivery and materials are now being hosted online through various Learning Management Systems (LMS). In the recent years, researchers in the education paradigm have proposed methodologies to find key performance indicators from the LMS data and used Machine Learning (ML) models to identify students who require an early intervention to improve their academic performance. Although these ML-based solutions show high accuracy, these solutions are often restricted to a single course/module belonging to a specific department or university. Moreover, they evaluate a fixed set of ML models to determine the best performer. To address this, we propose a dynamic Student Academic Performance Predictor (SAPP) tool which can work for different types of modules having diverse student records. The tool can predict poorly performing students to make early intervention and improve their academic performance. As proof of concept, the tool has been designed as a Python application which can take LMS data from different modules and predict a list of students requiring early intervention. The application uses the best performing ML model out of 18 ML models. Initial results using LMS data collected from two different modules running at a UK university give us early indication that the SAPP tool can accurately predict student performance for various modules with diverse range of students.