Hybrid Deep Learning for Predicting Student Engagement in Open Distance Education

Abstract

Open Distance Learning is a flexible learning approach whereby learners are exposed to learning material from a distance without the constraints of physical classrooms. In this research, student satisfaction with distance learning is investigated using a hybrid deep learning model consisting of a Bi-Stacked Gated Recurrent Unit (GRU) and ResNet. Data was collected from an online survey, cleaned to eliminate inconsistencies, and optimized by Ant Colony Optimization (ACO) feature selection. The Bi-Stacked GRU successfully learned sequential learning patterns, whereas ResNet learned deep features for improved classification performance. The combined model gave a combined view of student performance and engagement. Experimental results showed better prediction accuracy and interpretability, which proved the efficiency of the proposed method in assessing Open Distance Education.