A major challenge in higher education is retaining students in academic programs. Too often, at-risk students drop out and do not complete their degrees. North Carolina State University (NCSU) is committed to harnessing technology to improve student learning and provide ready access to quality education for everyone. DELTA is the department within NCSU spearheading this initiative; they targeted student retention as an area that could benefit from a technology-based solution.
NCSU’s academic programs — both on campus and distance learning — needed to be able to identify at-risk students early on in the semester. This would allow faculty to intervene with students who were struggling and provide remediation designed to keep them on track towards successfully completing their courses, programs, and degrees.
DELTA decided to implement an institution-wide solution for learning analytics technology that would give faculty access to predictive analytics tools. These were their requirements:
- Implement an enterprise-class learning analytics solution using open-source components that could acquire data from LMS, SIS, and other potential source systems
- Develop an ‘open’ analytics model that could be customized/trained to identify potentially at-risk students very early in any given semester
- Provide easy-to-understand dashboard views into the data
- Design the solution to scale easily to accommodate NCSU’s large enrollment numbers
DELTA’s solution called for an open-source learning data warehouse (Open LRW), the Apache Hadoop distributed processing framework, and end-user facing technologies such as OpenDashboard. DELTA needed a partner to help them put this all together.
For a similar project, Unicon partnered with Marist College to operationalize an open source predictive learning analytics model that could be easily adopted by various institutions regardless of their size and demographic makeup. This model was a good fit for NCSU’s needs.
In addition, Unicon’s extensive experience with integrations, data warehousing, scalable infrastructure, and open-source applications such as OpenDashboard and OpenLRW made us the best candidate to deliver the solution.
Unicon set up integrations between OpenLRW and the source data systems to ingest NCSU student data into the learning data warehouse. We also deployed and configured the Apache Hadoop framework, integrated the Marist predictive learning analytics model, and engaged with data scientists at Marist to train the predictive model based on the student data. Finally, we customized OpenDashboard and other end-user facing open-source components to show useful views to instructors.
Using these new predictive modeling tools, NCSU faculty can now identify struggling students at any stage of the semester and take appropriate remediation steps. Because NCSU’s solution is an open model, the university can continue to customize it for their student population. The infrastructure put in place can easily scale to meet the needs of the entire university.
NCSU ran a small pilot in Spring 2019, and plans further pilots in Fall of 2019 and Spring 2020. Feedback from the pilots will be used to plan further customizations, and improvements, and model training and refinement. Once pilot participants are satisfied with and have confidence in the model's predictions, the solution will be offered campus-wide.