Visualising Student Learning Patterns: A Deep Learning Approach to Learning Analytics

Understanding learning patterns and relevant teaching methodologies is essential for improving academic output and enabling custom-made learning experiences. This paper developed a deep-learning framework model by integrating multiple advanced NN architectures, like the Transfer model, graph neural network, and autoencoders, to analyse and visualise students' learning behaviour. The proposed deep learning framework uses multimodal datasets collected from different educational resources, like the log data of the learning management system, engagement metrics, and assessment scores, to make the DL framework a generalised model. It also helps to generate intuitive visual representations, such as knowledge graphs, heatmaps, and learning trajectories. The DL framework is developed by integrating a transformer model to extract sequential features and dependencies from the learning activity data. The GNN model extracts the relationship between the learning person and resources and the people interacting during the period. The autoencoders included in the DL framework help the teacher and learner to identify the people who are struggling in the earlier stage of the learning period and detect the anomalies in the learning patterns. The DL framework is implemented in Python with specific datasets to demonstrate its achievements. The output shows that the proposed DL framework uses hybrid DL models and obtained 97.6% accuracy with reduced dropouts by 38% compared with the conventional methods. It is also identified that the DL framework included potential advancements of deep learning with visualised methods in improving learning analytics, optimisation, and fostering data-based decision-making.