Data-Driven Early Academic Intervention: Harnessing AI for Students Achievement

Abstract

In the dynamic landscape of higher education, the timely identification and mitigation of factors contributing to academic failure among university students are paramount for fostering academic success and student well-being. This research follows a quantitative research method using machine learning algorithms and strategically designed features extracted from students’ laboratory practices and questionnaires, to predict students’ academic performance. The primary motivation driving this research is to develop a model capable of identifying students at potential academic risk at mid-course, thereby enabling timely intervention strategies. Changes in the evaluation of laboratory practices are introduced to enhance the model’s predictive accuracy. Results demonstrate the model’s effectiveness in predicting final exam outcomes, achieving over 90% accuracy at the end of the course. A mid-course identification experiment shows the feasibility of predicting student outcomes with an accuracy exceeding 85%. The findings suggest the potential for early intervention strategies to improve student success.