Abstract: [Purpose/significance] From the perspective of industry chain, this paper takes virtual reality technology as an example, constructs VR patent industry chain corpus, and explores the technical theme, research and development hotspot and future development trend of China VR patent. [Method/process] First of all, this paper used Python to crawl the patent text in VR field and got effective corpus through data cleaning. Secondly, combining IPC classification number and K-means clustering algorithm, this paper constructed and validates VR patent industry chain. In addition, based on TF-IDF algorithm and LDA theme model, we identified the core technology themes and their comprehensive strength, technology research and development hotspots and future trends of China VR patents from the perspective of production chain. [Result/conclusion] At present, the proportion of patents in each link of China VR industry chain is unbalanced. The upstream link is the most popular, followed by the downstream link, and the weakest link is the midstream link. In terms of theme mining, the upstream hot spot is software development, the midstream hot spot is film and television production, and the downstream hot spot is medical, educational and entertainment applications. In terms of future trends, the upstream of the industrial chain will be dominated by technologies such as electronic digital data processing, optical components, image communication, etc., the midstream will be dominated by technologies such as vehicle components, power devices, damping devices, etc., and the downstream will be dominated by technologies such as indoor games, medical diagnosis, identification, etc..
Keywords: K-means clustering algorithm; LDA theme model; technology theme evolution; text mining; VR