Abstract:

In the landscape of STEM education, understanding and addressing the factors that lead to student failure is critical to academic success and retention. Learning analytics, which uses data-driven insights from students' interactions with educational platforms, is emerging as a promising tool for early identification of at-risk students.

This paper focuses on identifying key points within learning analytics frameworks that signal potential student failure in STEM courses. The paper draws on a comprehensive analysis of various data modalities, including engagement metrics, assessment performance, and demographic variables, and identifies critical markers that indicate academic difficulty. It also discusses the development and validation of predictive models using different algorithms to illustrate the intricate interplay between various predictors and student outcomes. Furthermore, the paper discusses the practical implications of these findings for education professionals and offers actionable strategies for targeted interventions and personalised support mechanisms. The paper aims to equip educators with actionable insights to proactively identify and support students at risk of failure, fostering a more inclusive and supportive learning environment in STEM education.

The research uses data from the e-learning platform for maths courses at the Faculty of Engineering, University of Rijeka. Various correlation methods are used to determine and quantify the predictors. As a result, the ranked predictors of student failure are presented as a function of time, i.e. the week of class attendance.

Keywords:

Student failure, learning analytics, STEM education, supportive learning environment, prediction.