A Network-Based Learning Analytics Approach to Measuring Student Engagement and Predicting STEM Achievement

• In Event: AERA Poster Session 6
  In Poster Session: Connecting Learning Systems Through Technology (San Diego)

Abstract

One challenge for learning analytics is to identify metrics that are useful to practitioners and stakeholders. We used a network-based approach to analyze students’ learning management system (LMS) events from log data, interpreted network metrics from an engagement perspective that is intuitive to instructors and aligned with educational psychological theory, and analyzed relationships with course grades. Data drawn from one in-person (n = 226; events = 257k) and one online (n = 364; events = 496k) semester and submitted to regression models predicted 28% of variance in students’ grades. Our results suggested three types of domain general engagement variables that can be derived from native LMS logs, i.e., breadth, depth, and patterned, which can be used to predict STEM performance.

Authors

• Jonathan C. Hilpert, University of Nevada Las Vegas

• Matthew L. Bernacki, University of North Carolina Chapel Hill

• MeganClaire Cogliano, University of Nevada - Las Vegas

• Christy Strong, University of Nevada - Las Vegas

• Jeff A. Greene, University of North Carolina - Chapel Hill