Course periodic behavior modelling and its application in LMS activity prediction

The use of technology in education has risen so rapidly that many of the e-learning tools have become a great source for data gathering. In response to the growth, learning analytics are developed to extract the meaning from extensively large datasets and optimize learning opportunities for learners. Understanding users' behaviors is one of the key factors that can help educational institutes improve their curriculum design or non-instructional intervention. This paper, therefore, explores LMS's user behaviors through the study of courses' periodic behaviors. It assumes that there is a weekly pattern in most college-level courses where classes meet on a weekly basis, and that the pattern is reflected through the number of daily activities the courses produce. Courses that share similar weekly patterns are grouped into the same cluster using an unsupervised clustering algorithm. The knowledge obtained from the clustering can be used to describe the pattern in other courses. This paper continues to demonstrate one of course periodic modelling's applications by proposing a method to predict the activities that occur in an LMS. As a result, the weekly pattern found in each course, when aggregated, can represent the weekly behavior of the overall system. It can also predict the future trend of activities with correct shape and accuracy within the range of 82-86 percent.