LightGCN-LSC: Lightweight Graph Convolution Network for Learning Style Classification in Online Education

How to understand learners have always been the core goal of educational information science and technology, and will greatly lead the practical process of future education. Learning style recognition can largely reflect the learning status and internal cognitive structure of learners, then achieve personalized guidance and precise recommendations. However, the existing collaborative or automatic methods are limited by a large amount of data and have poor portability due to being limited to specific platforms. In this paper, we propose lightGCN-LSC, a novel learning style classfication approach using lightweight graph neural network. First, we build different embedding technique to extract hidden features of S-M bipartite graph, then we employ the K-means clustering algorithm to cluster learners and obtain labels utilizing the FSLSM model. Finally, lightweight graph neural network suitable for S-M graph data size and structure is trained to classify student’s learning style. The superiority of our method was demonstrated through experiments on dataset KDD CUP 2015. Our proposed lightweight models achieved outstanding results (about $1.18-8.24 \%$ improvement) over Graph Convolutional Networks (GCN), and the best result with improvment values of $2.05 \%$ accuracy, $8.24 \%$ precision, $2.05 \%$ recall, and $4.05 \%$ F1 score.