Prediction of learners’ learning styles in online environments has several advantages, including steering learners on the proper path, motivating and engaging them while learning, and improving their learning results. It also helps instructors in the formation of personalized resource recommendations. As a result, predicting learners learning styles is necessary to aid in the personalization process. Existing approaches use either conventional or automatic approaches for learning style identifications. However, the large volume of data stored in online platforms has become a challenge in analyzing the behavior of learners and predicting their learning styles in the real world. Also, most of the existing approaches rely on a particular learning platform and can not be used in other platforms without technical assistance. In this paper, we propose GNN-LS, a new approach to identify and predict learners learning styles using a graph neural network. First, the graph embedding technique is used to capture the representation of learners and resources as a bipartite graph and encode them into low-dimensional representation. The encoded L-R sequences were given as input to the K-means clustering algorithm to identify and obtain labels as per FSLSM dimensions. Then, Graph neural network is trained to predict the learner’s learning style in the real world. The GNN-LS technique can be applied in a variety of educational systems and adapted to fit a variety of learning style models. Extensive experiments are run using the 2015 KDD Cup public available dataset to demonstrate the capabilities of GNN-LS. 5.31-15.68% improvements are achieved across all four FSLSM dimensions in accuracy.

In the context of the information age, the rapid growth and increasing diversity of learning resources underscore the urgency of personalized learning, while learning style is the most crucial factor to consider in personalized learning as it significantly influences students’ academic achievements and learning experiences. Traditional methods of assessing learning styles, such as completing questionnaires, have many drawbacks, including subjectivity and time costs. Therefore, in recent years, researchers have been exploring automatic methods to identify learning styles by analyzing students’ interactive behaviors. Motivated by these limitations, we propose a learning style detection method using a graph attention network (GAT), named GAT-LS. We originally constructed a bipartite graph between learners and learning materials, utilizing node features to represent the students’ behavior. Subsequently, we employ GAT to obtain hidden vectors for the graph nodes. These hidden vectors encapsulate both the overall graph information and the importance of neighboring nodes. We employ a multi-head attention network to process student nodes and combine a dropout mechanism with a single-layer attention network to process learning material nodes. Finally, we map the obtained hidden node features to the Felder-Silverman learning style model (FSLSM) and use K-means clustering to detect learning styles. The proposed method can be integrated into various types of educational systems or online learning platforms, providing a better educational experience and learning resource recommendations for both teachers and students. Experiments on the real-world dataset, KDD CUP 2015, demonstrated the superiority of our method. Our proposed approach achieved outstanding results with average values of 0.9647 accuracy, 0.9478 precision, 0.9171 recall, and 0.9346 F1 score.

A growing number of students enrol in online education to improve their skills. However, students are faced with the challenge of finding courses that meet their individual needs. Recommender systems were introduced to help students choose the courses that best meet their needs. To learn better representations of students and courses for improved recommendation results, existing graph-based recommender systems utilize the high-order collaborative signals between set of students or set of courses from a bipartite graph. However, courses also have prerequisite dependency between them, which when utilized together with collaborative relations can improve recommendation results. On this basis, we propose a model that utilizes the high-order relation between set of courses, the prerequisite dependency between courses, as well as the direct relation between students and courses. Using meta-paths generated from the knowledge graph, our model extracts the prerequisite dependency between courses, which is then used to generate a course prerequisite graph. The course prerequisite graph and the student-course bipartite graph are used to learn the representation of the students and courses, jointly capturing the prerequisite dependency, high-order collaborative relations as well as direct relations. The learned representations are used for recommendation. The experiments on real-world dataset show the superiority of our proposed method, achieving 3.61% on F1@10 and 1.38% on Mrr@10.