Analysis of clustering algorithms for group discovery in a web-based intelligent tutoring system
D. Bunic, Igor Jugo, Božidar Kovačić · 2019
DITUS is a web-based intelligent tutoring system developed and used at our institution as an additional learning platform. To make the system even more adaptive we expanded its architecture with several modules that perform educational data mining tasks such as clustering, to discover groups of students that use the system in a similar manner, and high-utility sequential pattern mining to discover efficient learning paths through the knowledge domain. The results of these modules enable the system to offer hints to students on which knowledge units to learn before or after the currently selected unit. One of the main pre-conditions of the quality of hints is the clustering phase in which we discover groups of students that are using the system in a similar manner in terms of learning activity and efficiency. In this paper, we analyze the results of several well-known clustering algorithms on our datasets, to determine which one is best suited for the needs of our system.