A hierarchical attention neural network with multi-view fusion for online course recommendation

The ever-evolving landscape of online education has spawned numerous course recommendation methods. Yet, these methods have neglected learners’ preferences for multi-view course data, including the course content, concepts, instructors, institutions hosting the courses, and their relations. Additionally, they have overlooked the influence of various elements within different views on learners’ course selection behavior. Consequently, the integration of multi-view course data and distinct elements within each view into a comprehensive recommendation method remains a complex challenge. To tackle this issue, we propose a novel solution—a hierarchical attention neural network (HANN) for effective online course recommendation. HANN incorporates a course encoder, a user encoder, and a recommendation generator. The course encoder leverages two levels of attention to integrate multiple views of information, thereby generating comprehensive course representations. The user encoder employs bidirectional long short-term memory and another level of attention mechanism to acquire user representations based on their learning records. Lastly, the recommendation generator calculates recommendation probabilities by feeding user and course representations into a multi-layer perceptron. Experimentation with real-world data validates the effectiveness of HANN. Our results show the multi-view course data and attention mechanisms to be useful in online course recommendation. © The Author(s), under exclusive licence to Springer-Verlag London Ltd., part of Springer Nature 2025.