Identification of Potential At-Risk Students Through an Intelligent Multi-model Academic Analytics Platform

The identification of students potentially at risk of having learning difficulties is of utmost importance for the provision of timely support. This paper presents an intelligent academic analytics platform which features a multi-model prediction approach to identify students potentially at risk of failing a course. The platform aims to help instructors identify potential at-risk students at the beginning of a course so that appropriate learning support can be provided early. It employs three supervised machine learning algorithms – linear regression, XGBoost, and decision tree – to build predictive models. A student’s risk level is based on the predicted grade s/he will receive for a course, determined by integrating the prediction results of the three models using a majority vote method. The platform also provides supplementary information about students, such as past course grades and factors contributing to the predicted risk levels, for instructors’ reference to inform learning support strategies. Preliminary evaluation results show that the platform correctly identified about 39% of students who failed their courses. The results suggest that some students who were predicted as having a high risk of failing at the beginning of a course would eventually pass with the learning support provided after early identification. This demonstrates the effectiveness of the platform in enhancing student success through proactive support enabled by early prediction of academic risk.