Title: Event detection in educational records: an application of big data approaches
Authors: Alan D. Smith
Addresses: Department of Marketing, Robert Morris University, 6001 University Blvd., Moon Township, PA 15108, USA
Abstract: As the exponential growth of predictive analytics applied to educational institutions continues, with various goals in mind, an event detection system could serve as the foundation for more sophisticated predictive systems. What is envisioned is a system that flags educational records that have a sudden change (i.e., event detection in education records). The basic algorithm prosed would retrieve the current week's data for each student, calculate Euclidian and Mahalanobis distances among current week's data, and historical student data, sort all student data based on these distances, and update historical data with current data. The educational event detection system uses a query-based, R-tree structure, big data analytical approach. A simulated experiment determined that the event detection algorithm was able to detect relatively small changes in simulated data. Although, the units of the two distance metrics do not coincide, so distance measurements are not directly comparable, the event detection algorithm with the two different distance metrics have comparable CPU run times. Research and suggestions for constructing predictive analytics on student learners' attrition/retention approaches were documented.
Keywords: big data; clustering; data mining; educational data clustering; EDC; Euclidian distance; Mahalanobis distance; event detection; network intrusion detection; predictive analytics; student success; systems.
International Journal of Business and Systems Research, 2021 Vol.15 No.3, pp.271 - 291
Received: 14 Sep 2019
Accepted: 24 Sep 2019
Published online: 12 May 2021 *