2024 ASEE Annual Conference & Exposition

This research paper describes the development and analysis of a longitudinal dataset of students’ academic performance using regression and machine learning modeling. This study is contextualized within The School of Electrical and Computer Engineering (ECE) at a large, public, research-intensive institution in the Southeast United States. Undergraduate students in this School have a diverse range of backgrounds, experiences, and needs. As such, program leaders must work to (1) provide effective, accurate, and personalized support; and (2) provide information and recommendations for curricular developments and resource management. Both efforts rely on a strong foundation of data to inform decision-making. As such, this paper describes the quantitative portion of a larger mixed-methods project, from which the authors identified initial baseline conditions of students’ academic performance and revealed potential influential factors in difficult ECE courses. This work suggests opportunities for programmatic improvements with the highest potential for success.

Many engineering educational researchers have worked with large-scale datasets of students’ academic records to understand influential factors on students’ performance. This paper contributes to those efforts by taking advantage of the size of the School to create a statistically powerful dataset while still being able to capture the nuances and legacies specific to the department. Further, the use of machine learning modeling can offer novel insights into underlying trends in the data.

At the time of submission, the project has identified a focal course to frame the analysis: [course number]. This course is a 4-credit hour, junior-level course on semiconductor architecture, and it is notorious within the department for its difficulty. This reputation will be investigated further through qualitative focus groups with former students and interviews with instructors. For the quantitative strand that is the focus of this paper, the data revealed notable baseline conditions. 1,225 students have taken the course between Fall 2016 and Spring 2023. From those students, 20.58% earned a final grade of a D, F, or W. Women made up 19.2 % of the course population and performed at the same level as their male peers, with the exception of the Spring and Fall 2022 semesters (after which their average performance recovered to parity). Students from minoritized racial or ethnic backgrounds made up 23.43% of the data. The data revealed that students from minoritized backgrounds had lower performance in Spring semesters compared to Fall semesters, a trend that is not explained by differences in instructors. Finally, students who had transferred into the institution made up 26.86% of the course population on average, although that number has fluctuated over time. Transfer students demonstrated lower average performance compared to students who had directly matriculated. Regression and machine learning modeling will be conducted and presented in the full paper.

We report these trends by student groups to present baseline conditions for this abstract, but we want to be intentional in avoiding any essentialist, deficit-based conclusions. The goal is to engage in deeper analysis to identify influential factors behind these symptoms, including possible insights at the intersection of different identities. This work has implications for developing data-driven insights to help program leadership make decisions about curricular developments and resource management.