Abstract: 【Purpose/significance】Patent early warning plays a significant role in alerting and actively preventing risks for patent hold⁃ ers during the technology development process. Exploring scientific and effective patent early warning schemes is crucial for helping relevant research entities to avoid the risks of patent infringement, clarify the trends of technological changes, and understand the de⁃ velopment status of competitors.【Method/process】This paper takes universities as the objects of early warning, focusing on the microlevel patent infringement risks, meso-level technological change risks, and macro-level potential competitive risks of university pat⁃ ents. By relying on topic modeling and Word2vec word vector technology, a patent infringement early warning process is formed. Com⁃ bined with Doc2vec sentence vector technology and outlier detection algorithms, a technological trend early warning process is estab⁃ lished. Utilizing patent mapping, a potential competitor early warning process is created. Based on the aforementioned content, an inte⁃ grated university patent early warning model of "patents, technology, and competition" is constructed.【Result/conclusion】Taking the field of "virtual reality" as an example, a specific university is selected to conduct an empirical study on the model proposed in this pa⁃ per to verify its effectiveness.【Innovation/limitation】The patent early warning model constructed in this paper provides innovative support for universities to guard against research and development risks and enhance their technological competitiveness. However, a quantitative validation method for the patent early warning results has not been established, and further exploration is needed to inte⁃ grate multi-source data into the patent early warning schemes for universities.