Abstract:

During the COVID-19 crisis, the use of learning management systems (LMS) has increased tremendously. There are many developed LMSs, including Moodle, Edmodo, and Blackboard. These systems produce a ton of information regarding student activity. Students' LMS activity is stored in log files. Log file data contains records of students' interaction with course materials through resources and activities. These data are an important source of knowledge for faculty management, teachers, and students. Raw data should be pre-processed and transformed to be prepared for knowledge discovery in the data process. Knowledge discovery in data enables the identification of patterns in students’ behaviour and leads to the improvement of teaching approaches and the enhancement of students’ success. To create descriptive models of student behaviour and success, this study uses descriptive learning analytics methodologies. Data from LMS Moodle is employed with the aim to examine online students' LMS activities. The study presented here was conducted at the University of Zagreb, Faculty of Organization and Informatics at the courses of two study programmes. Collected data consists of students' logs to the different types of files, forums, assignments and other resources. Variables that were extracted from logs at particular resources and activities were File, Forum, Student Report, Folder, Choice, File Submission, Overview Report, Page, System, Test, and Assignment.

The following research questions were examined:
(i) can we create good student clusters based on their usage of the LMS?
(ii) is there any correlation between students' clusters and students' success?

In the empirical study, a CRISP DM data mining standard is applied with a k-means clustering algorithm used in the modelling phase. An unsupervised machine learning approach is used in order to cluster students based on their behaviour at LMS. Descriptive model evaluation through the internal clustering measures shows that our approach produced a good quality of cluster groups.

The findings of the study revealed:
(i) student groups with similar behavioural patterns at the LMS and
(ii) student activities at the LMS that result in effective course completion.

These findings serve as guidance for teachers when creating courses and for students when signing up for the course. Results from this research help to personalize teaching and learning strategies, particularly in online environments. An innovative strategy in education that can benefit both teachers and students and enhance learning outcomes is knowledge discovery in data along with machine learning algorithms.

Keywords:

LMS, clustering, data mining, students' activity.