As the core basic course of computer science, the computer network course has long been faced with teaching dilemmas such as many and fragmented knowledge points, abstract and difficult to understand protocol levels, and a disconnection between theory and practice. Aiming at these problems, this study constructs a knowledge framework ontology covering all levels of TCP/IP protocol stack, designs a hierarchical and multidimensional visual teaching system, and carries out quasi-experimental research in university classrooms to verify its teaching effect. The research adopts the methodology combining the design research method and quasi-experiment method, and develops an interactive visualization platform based on knowledge graphs, which supports the dynamic demonstration of protocol flow, layer-by-layer analysis of packet structure, interactive exploration of network topology, and other functions. The experimental results show that the visual teaching group is significantly better than the traditional teaching group in the dimensions of agreement process understanding and cross-layer concept association, especially for learners with weak spatial ability. This study provides a reusable theoretical framework and practical paradigm for the knowledge visualization teaching of computer courses.