Exploration of the online learners’ actions: A sequence mining approach

Abstract

This paper presents the exploration of the learners’ learning engagement in a self-paced massive open online course (MOOC). Research often claims that engagement contributes to learning success. However, there is still limited understanding of engagement and its characteristics. This research aims to fulfil this gap by exploring how different patterns detected based on the density levels of engagement contribute to learning performance. A total number of 159,804 records of trace data from 971 learners who enrolled in a self-paced MOOC were used in this study. The sequence mining technique was used to formulate the sequence of learning engagement. Hierarchical clustering was then used to automate the pattern recognition of the formulated sequences. As a result, four groups of learners were detected based on a similar pattern of engagement levels. Sequence mining was then used to examine the learning engagement pattern. The Kruskal-Wallis test was used to examine the statistically significant differences in terms of final scores among the detected groups. The results revealed two successful groups of learners with different patterns of engagement and two unsuccessful groups. Successful learners are intensively engaged in learning activities in the short and long run, whereas unsuccessful groups tend to be less engaged. This paper extends the previous exploration of the engagement. That is, the level identified based on the density of interactive engagement as recorded in the system can be used to determine the learning patterns, consequently, reflective of individual’s learning profiles. It has a significant association with academic performance.