Predicting engagement using machine learning techniques

Abstract

A learner who is sufficiently involved in the learning process will surely show higher performance with respect to the objectives set for the process. With detailed planning of every miniscule unit of learning in place, focus is entirely on the learner's engagement for successful learning to happen. Engagement prediction has been investigated in various ways not only in education but in other domains also. Most of them that involve machine learning techniques either use invasive methods or narrowly focus on only one aspect of the learning environment particularly if they are using log based measurements. The aim of this work is to propose and validate a framework that will be suitable for detecting engagement at a unit level of sessions in an online environment. Using machine learning techniques, an engagement detector will then be created that will track the tasks under each of the component of this framework and depending upon depth of involvement would predict the learner's engagement level. It will not require any special wearable or external instruments like cameras to capture the data used for the detection. It will use only the logs from the Moodle platform for the activities set up by the faculty for the session and track the activities of the learner for them to calculate the engagement level. This calculated index of engagement can then help the instructors to take corrective actions so that the learner's involvement can be increased.