Analysing Learner Behaviour in an Ontology-Based E-learning System: A Graph Neural Network Approach

Wynn, Adam; Wang, Jingyun; Sun, Zhongtian; Shimada, Atsushi

Abstract

Despite the prevalence of e-learning systems, there is a lack of support for learners to identify and compare new knowledge with existing cognitive structures. Therefore, an ontology-based visualization support system was previously introduced which offers two modes: cache-cache, where relations are initially hidden and the learners are encouraged to create those relations, and receptive, where learners can view expert-generated topic maps. In this study, we aim to analyse learner behaviour by representing user behaviour as graphs and utilising a heterogeneous graph convolutional network. Two graphs are constructed for each student to capture behaviour before and after system use. Results indicate significant differences in mean embeddings between learners in receptive and cache-cache modes. Further analysis, considering pre-test performance, shows no significant differences in the receptive and cache-cache groups but highlights a considerably smaller mean for high prior
performers in the cache-cache group.

Citation

Wynn, A., Wang, J., Sun, Z., & Shimada, A. (2024, March). Analysing Learner Behaviour in an Ontology-Based E-learning System: A Graph Neural Network Approach. Paper presented at LAK '24: The 14th Learning Analytics and Knowledge Conference, Kyoto, Japan

• Presentation Conference Type: Conference Paper (unpublished)
• Conference Name: LAK '24: The 14th Learning Analytics and Knowledge Conference
• Start Date: Mar 18, 2024
• End Date: Mar 22, 2024
• Publication Date: 2024-03
• Deposit Date: Apr 29, 2024
• Publicly Available Date: Apr 29, 2024
• Public URL: https://durham-repository.worktribe.com/output/2407749
• Publisher URL: https://dl.acm.org/doi/proceedings/10.1145/3636555