Construction of a Learning Engagement Evaluation Model Based on Multi-modal Data Fusion

Learning engagement has become an important indicator affecting learning outcome in universities. It can not only reflect learners' learning process, but also fully resonate “learner-centered” educational concept. Accurate evaluation of learning engagement is an important task. Multi-modal data fusion can extract and fuse dynamic and multi-dimensional input, which is of great value in characterizing learners' learning process. But there are some problems in traditional evaluation methods, such as poor real-time evaluation, low evaluation effect of single modal data, and social approval response bias. Based on multi-modal data fusion, an evaluation model of learners' engagement was constructed and its predictive effect was verified. In this study, OpenCV region extraction method was applied and an automatic evaluation method was proposed based on multi-modal data fusion calculation. Questionnaires were collected to explore factors impacting learning engagement from university students in mainland China. Results showed (a) the evaluation model of learning engagement, i.e., behavioral engagement, cognitive engagement and emotional engagement has good reliability and validity; (b) It demonstrates that learner engagement has a significant predictive effect on academic achievement. The study finds that improving learning engagement will significantly improve learners' overall academic gains and the learning engagement evaluation model can accurately assess learning engagement.