Student-Driven Projects: Learning Analytics in a Data-Driven Age

Analytics education is undergoing major reform. As data becomes more prevalent in industry, one of the most important questions for higher-learning institutions is how analytics should be taught in the classroom.

A conventional teaching approach involves lectures, problem sets and exams. These provide students with technical training, but that training is often insufficient preparation for real-world analytics. In contrast, course projects train students to think holistically about specific problems, which allows them to learn the intricacies of deploying real-world analytics solutions.

A common alternative to conventional teaching approaches is client-driven projects. These projects allow students to work with industry partners and provide insights for real-world problems via analytics. These projects yield positive learning outcomes, but they also come with challenges that make scaling them to large classes difficult (e.g., establishing industry connections, identifying appropriate projects with appropriate learning outcomes) [1].

In part to avoid the challenges of client-driven projects, our course adopted student-driven projects. These projects ask students to choose their own analytics problem and provide a holistic solution under the guidance of a project advisor. Given the rapid increase in demand for analytics education and the abundance of public data sets, we believe that student-driven projects provide a promising alternative or complement to client-driven projects.

Over the past few years, we served as project advisors in an undergraduate analytics course at the University of Toronto. Collectively, we advised more than 200 students and over 50 student-driven projects that spanned a diverse set of methodologies and applications. We observed students overcoming common challenges to deploy analytics from scratch to solve real-world problems. In this article, we briefly summarize our experiences and the lessons we learned as project advisors. More details can be found in our paper in INFORMS Transactions on Education [2].

Learning Outcomes in Student-Driven Projects

Students are responsible for all stages of a student-driven project, which includes defining a project topic, implementing appropriate analytics models, analyzing the performance of the proposed solution and communicating the impact. The primary learning outcomes of this project are the analytical skills to identify high-impact problems, solve high-impact problems and communicate progress.

Identifying high-impact problems. This course was often the first time that students were asked to define a topic for their own project. They used the opportunity to pursue topics related to personal interests, and we had the privilege of advising a breadth of diverse and creative projects, as evident from Table 1. The project topics were typically defined via an iterative process of searching for data sources and refining the topic based on data availability. This is an important process in real-world analytics, and, as one former student notes, “the iterative approach to projects has been extremely helpful at work.”

Table 1. Summary of projects and percentage of each category and subcategory

Solving high-impact problems. Students used analytics to solve their proposed project topic, which required a thorough understanding of existing data sets and methodologies. Many students researched and implemented models not covered in our course materials, or designed new models that were tailored to their specific problem topic. Additionally, implementing these models enabled students to refine coding and analysis skills.

Communicating progress. Students regularly met with their teams and project advisors to discuss current challenges and future plans. These meetings helped students learn to communicate with people who understood their project to different degrees. Communication is also important to iteratively refine ideas, models and insights. As one former student writes, this iterative exercise helped build “a really strong analytical thought process, patience with doing research and searching for the right answer, thinking about the answer and if it makes sense, interpreting and understanding results then communicating them clearly.”

Addressing Common Challenges  

The open-ended nature of student-driven projects can pose challenges for students. Project advisors should help students overcome these challenges to maximize their ability to identify high-impact problems, solve these problems and communicate progress.

Identifying high-impact problems. Students occasionally gravitated toward problems and data sets that were “interesting” at a surface-level without considering how their analysis may impact downstream decision making. As a result, we encouraged teams to give us concrete examples of how a specific stakeholder could use their model. This exercise helped shift the focus away from data exploration and toward using data to understand or inform decisions, which can improve the real-world value of the project.

Solving high-impact problems. Identifying appropriate models was often a major challenge, given the abundance of potential models. We generally encouraged students to propose a few simple approaches and specific metrics to measure model quality to serve as a starting point for their project. Together, these steps helped teams establish a base for their project to iterate on and measure improvement.

Communicating progress. A student-driven project consists of many moving parts. Students often have trouble communicating the individual components of their project and establishing a consistent narrative. We often found that this challenge could be resolved by helping students organize their thoughts with high-level flow charts to capture each piece of their project and how they are linked together. This also helped us work with students to resolve technical issues without losing sight of the overall picture.

Considerations for Course Instructors, Students and Advisors

The demand for analytics education is rapidly rising. To effectively satisfy this demand, we believe that more attention should be placed on introducing student-driven analytics projects into curriculums. We conclude with some additional considerations for course instructors, students and project advisors.

Course Instructors: Student-driven analytics projects offer several advantages for instructors. They require relatively low levels of time commitment and setup costs, which makes them easy to implement at scale. Additionally, unlike in client-driven projects, instructors maintain control over standardized learning objectives, deliverables and grading rubrics.

Students: Many students highlighted that the experience prepared them for internships and jobs. Some commented specifically that the project helped them learn to apply analytical skills, work as a team, consult an advisor and communicate with people who are not directly involved in the project.

Project Advisors (Teaching Assistants): Advising student-driven projects is a relatively low-stakes opportunity for graduate students to develop skills that are necessary in academia, including mentoring students, advising multiple ongoing projects and interacting with students from diverse backgrounds.

Overall, we found this opportunity tremendously helpful in our own research, internships and job searches.

References

1. Gorman, M.F., 2018, “A survey of research in field-based education: A summary of process, best practices, and lessons learned,” INFORMS Transactions on Education, Vol. 18, No. 3, pp. 145-161.
2. Babier, A., Fernandes, C. and Zhu, I.Y., 2022, “Advising student-driven analytics projects: A summary of experiences and lessons learned,” INFORMS Transactions on Education, August 26, https://doi.org/10.1287/ited.2022.0275.