Temporal Community Analysis Using Centrality Measures

This research explores temporal community detection to identify patterns in course grading and offerings at The University of Illinois from Spring 2010 to Spring 2020. Using a dataset that includes course grades, subject codes, instructors, and student enrolment, the study applies four community detection algorithms—Greedy Modularity, Louvain, Leiden, and Walktrap—to examine how courses group based on grading similarities over time. The data is pre-processed, split into time slices, and represented as graphs where nodes are courses and edges are weighted by cosine similarity of grading patterns. Key metrics such as modularity, community size, edge density, overlap, and persistence are used to evaluate algorithm performance. The results provide insights into the evolution of grading structures, course groupings, and instructional trends, offering a deeper understanding of the dynamics in educational data. This analysis contributes to improved decision-making in academic program development and resource allocation.