An official website of the United States government
Demand for Machine Learning (ML) courses remains high, and educators face open questions about which prerequisites are important for student success in upper-year ML courses. Prior work has shown that instructors and students in ML courses believe that the math prerequisites and their relative recency are barriers to success, but this relationship has not been demonstrated quantitatively. In this paper, we study the link between prerequisite grades and performance in an upper-year ML course at two sites. We use linear models to study the extent to which student grades in prerequisite courses in calculus, linear algebra, statistics, and software design are predictive of student performance in the ML course. We consider the effect of additional factors like gender, first-in-family status, prior experience, comfort with mathematics, and comfort with academic English. Like prior work in many domains, and consistent with ML instructor and student perspectives, we find that prerequisite grades are predictive of ML performance. However, different combinations of prerequisites are important at different sites. Also, we find that cumulative grade point average (cGPA) in past technical and non-technical courses are as predictive of ML grade, if not more. Moreover, recency in prerequisite courses is not predictive of ML course grades in our setting. These findings suggest that general academic preparation may be as robust a predictor of ML course performance as specific math prerequisites, challenging assumptions about the role of mathematical recency and preparedness—at least as measured by grades.
more »
« less
Instructor Perspectives on Prerequisite Courses in Computing
Recent research in computing has shown that student performance on prerequisite course content varies widely, even when students continue to progress further through the computing curriculum. Our work investigates instructors' perspectives on the purpose of prerequisite courses and whether that purpose is being fulfilled. In order to identify the range of instructor views, we interviewed twenty-one computer science instructors, at two institutions, that teach a variety of courses in their respective departments. We conducted a phenomenographic analysis on the interview transcripts, which revealed a wide variety of views on prerequisite courses. The responses shed light on various issues with prerequisite course knowledge, as well as issues around responsibility and conflicting pressures on instructors. These issues arise at the department level, as well as with individual course offerings.
more »
« less
- PAR ID:
- 10428488
- Date Published:
- Journal Name:
- Proceedings of the 54th ACM Technical Symposium on Computer Science Education
- Volume:
- 1
- Page Range / eLocation ID:
- 277 to 283
- Format(s):
- Medium: X
- Sponsoring Org:
- National Science Foundation
More Like this
-
-
With increasingly technology-driven workplaces and high data volumes, instructors across STEM+C disciplines are integrating more data science topics into their course learning objectives. However, instructors face significant challenges in integrating additional data science concepts into their already full course schedules. Streamlined instructional modules that are integrated with course content, and cover relevant data science topics, such as data collection, uncertainty in data, visualization, and analysis using statistical and machine learning methods can benefit instructors across multiple disciplines. As part of a cross-university research program, we designed a systematic structural approach–based on shared instructional and assessment principles–to construct modules that are tailored to meet the needs of multiple instructional disciplines, academic levels, and pedagogies. Adopting a research-practice partnership approach, we have collectively developed twelve modules working closely with instructors and their teaching assistants for six undergraduate courses. We identified and coded primary data science concepts in the modules into five common themes: 1) data acquisition, 2) data quality issues, 3) data use and visualization, 4) advanced machine learning techniques, and 5) miscellaneous topics that may be unique to a particular discipline (e.g., how to analyze data streams collected by a special sensor). These themes were further subdivided to make it easier for instructors to contextualize the data science concepts in discipline-specific work. In this paper, we present as a case study the design and analysis of four of the modules, primarily so we can compare and contrast pairs of similar courses that were taught at different levels or at different universities. Preliminary analyses show the wide distribution of data science topics that are common among a number of environmental science and engineering courses. We identified commonalities and differences in the integration of data science instruction (through modules) into these courses. This analysis informs the development of a set of key considerations for integrating data science concepts into a variety of STEM + C courses.more » « less
-
With increasingly technology-driven workplaces and high data volumes, instructors across STEM+C disciplines are integrating more data science topics into their course learning objectives. However, instructors face significant challenges in integrating additional data science concepts into their already full course schedules. Streamlined instructional modules that are integrated with course content, and cover relevant data science topics, such as data collection, uncertainty in data, visualization, and analysis using statistical and machine learning methods can benefit instructors across multiple disciplines. As part of a cross-university research program, we designed a systematic structural approach–based on shared instructional and assessment principles–to construct modules that are tailored to meet the needs of multiple instructional disciplines, academic levels, and pedagogies. Adopting a research-practice partnership approach, we have collectively developed twelve modules working closely with instructors and their teaching assistants for six undergraduate courses. We identified and coded primary data science concepts in the modules into five common themes: 1) data acquisition, 2) data quality issues, 3) data use and visualization, 4) advanced machine learning techniques, and 5) miscellaneous topics that may be unique to a particular discipline (e.g., how to analyze data streams collected by a special sensor). These themes were further subdivided to make it easier for instructors to contextualize the data science concepts in discipline-specific work. In this paper, we present as a case study the design and analysis of four of the modules, primarily so we can compare and contrast pairs of similar courses that were taught at different levels or at different universities. Preliminary analyses show the wide distribution of data science topics that are common among a number of environmental science and engineering courses. We identified commonalities and differences in the integration of data science instruction (through modules) into these courses. This analysis informs the development of a set of key considerations for integrating data science concepts into a variety of STEM + C courses.more » « less
-
Previous work in computing has shown that Black, Latinx, Native American and Pacific islander (BLNPI), women, first-generation, and transfer students tend to have worse outcomes during their time in university compared to their majority counterparts. Previous work has also found that students' incoming prerequisite course proficiency is positively correlated with their outcomes in a course. In this work, we investigate the role that prerequisite course proficiency has on outcomes between these groups of students. Specifically, we examine incoming prerequisite course proficiency in an Advanced Data Structures course. When comparing incoming prerequisite course proficiency between demographic pairs, we only see small differences for gender or by first-generation status. There is a sizeable difference by BLNPI status, although this difference is not statistically significant, possibly due to the small number of BLNPI students. In addition, we find that transfer students have sizeable and statistically significantly lower prerequisite course proficiency when compared to non-transfer students. For BLNPI and transfer students, we find that they also have lower grades in the prerequisite courses, which may partially explain their lower prerequisite course proficiency. These findings suggest that institutions need to find ways to better serve BLNPI and transfer students.more » « less
-
Developing online courses is a complex and time-consuming process that involves organizing a course into a sequence of topics and allocating the appropriate learning content within each topic. This task is especially difficult in complex domains like programming, due to the incremental nature of programming knowledge, where new topics extensively build upon domain concepts that were introduced in earlier lessons. In this paper, we propose a course-adaptive content-based recommender system that assists course authors and instructors in selecting the most relevant learning material for each course topic. The recommender system adapts to the deep prerequisite structure of the course as envisioned by a specific instructor, while unobtrusively deducing that structure from problem-solving examples that the instructor uses to present course concepts. We assessed the quality of recommendations and examined several aspects of the recommendation process by using three datasets collected from two different courses.While the presented recommender system was built for the domain of introductory programming, our course-adaptive recommendation approach could be used in a variety of other domains.more » « less
- Free Publicly Accessible Full Text
-
Accepted Manuscript1.0
- Conference Paper:
- https://doi.org/10.1145/3545945.3569787
