A new hybrid approach to detect and track learner’s engagement in e-learning

Learner engagement is a critical concept that can lead to satisfaction, motivation, and success in e-learning courses. It covers contextual, emotional, behavioral, cognitive, social aspects. The instructors have difficulties identifying who involved the courses lack of face-to-face interaction with learning resource act upon reduce dropout rate. This paper presents novel approach aims predict learner online quantify relationship between learners’ their engagement. For this purpose, we used traces gathered from 1 356 reactions during winters 2020, 2021, 2022, implement approach. To model engagement, variety features were considered such as total number posts made forums time spent on platform. study uses BiLSTM method FastText word embedding detect emotions forum discussions. Then, an unsupervised clustering technique based new dataset was cluster learners into groups according level. Several supervised classification algorithms trained performance evaluated using cross-validation techniques diverse precision metrics. findings indicate decision tree rule more relevant than others, accuracy 98% AUC score 0.97. conclusions research reveal most are observers there nonlinear correlation