In the context of the continued expansion of online learning resources in higher education, learners face greater burdens in content filtering and learning-path decision making during course learning and review. Traditional organizational approaches dominated by catalogs or lists struggle to support non-linear, exploratory learning processes. Although course knowledge graphs can explicitly visualize relationships among knowledge points and provide structured entry points, structural visibility does not necessarily translate into actionable learning paths; learners may still experience disorientation, repeated backtracking, and stage-wise stagnation during graph-based exploration.

　　To address this issue, this study takes learning interest as a key cue for personalized, process-oriented support and proposes and implements an interest-oriented knowledge graph recommendation learning system for knowledge-graph exploration in online university courses. At the framework level, the study operationalizes an interest-development perspective at the node granularity of the knowledge graph and constructs a relational framework of “node-level interest development—exploration behavioral cues—support strategies.” At the mechanism level, the system characterizes learners’ interest states and stage features based on node-level interaction behaviors on the graph, thereby driving the selection of recommendation sources and stage-adaptive recommendation strategies. At the interaction level, aligned with the continuous learning process of exploration, comparison, planning, and consolidation, recommendations are embedded into knowledge-graph exploration as process support that is perceivable and explainable. Guided by user research on information needs and strategy preferences, the system forms information presentation and interaction-feedback approaches matched to different exploration stages.

　　Experimental evaluation and qualitative feedback indicate that, compared with the control condition, interest-oriented node-level identification and embedded recommendations are more conducive to enhancing learners’ proactiveness and engagement in graph-based exploration, supporting sustained exploration, expanding exploration scope, and improving the organization and understanding of knowledge structures and learning paths. The recommendation mechanism functions more as a supplementary entry point within the exploration process, triggering subsequent learning actions and supporting extended exploration without replacing learners’ main decision making. This study offers actionable interaction strategies and system implementation references for integrated designs that combine graph-based representation, process-oriented support, and personalized recommendation in online learning platforms in higher education.