A data-driven approach to establishing a patent strategy by generating a patent map based on generative topographic mapping

As competition among companies intensifies through patents, the need for the strategic utilization and visualization of these patents is growing. However, establishing a patent strategy often relies on subjective insights from experts, which presents a significant limitation. Accordingly, this study aims to develop an analytical methodology that identifies the competitive landscape in technology and business, visualizes patent strategies, and helps in formulating future patent strategies with a focus on technical feature information. Initially, the methodology involves extracting the subject–action–object (SAO) structure from patent data, followed by the visualization of a patent map using generative topographic mapping (GTM). K-means clustering is then applied to further segment sub-technical areas. Subsequently, technology nodes on the GTM map are characterized from the perspective of companies. This process helps in deriving patent strategy patterns that reflect both technological competition and strategic intentions. Future patent strategies are established by scoring these patterns based on predictions of company occupancy using GTM-based classification (GTC) model-based vacuum nodes and other strategic quantitative indicators. This methodology particularly highlights the intersections between technological advancement and corporate competitiveness. An empirical study focusing on the autonomous vehicle industry validates the effectiveness of this methodology in providing insights about leveraging patent strategies for technological leadership. The significant contribution of this study lies in its proposition of a patent map enriched with detailed technical information from patents and the quantification and visualization of patent strategies, guiding the direction for future patent strategizing.