통계적 모델링을 통한 특허소송 위험요인 탐색에 관한 연구

The importance of technology independence is increasing due to the recent deterioration of the foreign trade situation and intensification of technological competition between countries. It is essential to secure patents including core technologies for effective technology independence. Global companies' patent portfolio strategies and the emergence of NPEs promote litigation against patents that play a key role in specific technology fields. Patent litigation hinders technology independence by adversely affecting product or service supply and R&D. Companies use a variety of methods to identify and prevent patent litigation in advance. Among them, the quantitative litigation prediction model ensures objectivity for the results derived based on patent big data. In addition, time and cost can be saved, enabling efficient litigation prediction and R&D strategy establishment. Existing quantitative litigation prediction models used various information of patents, but models are constructed based on the time of data collection. Therefore, patents that may cause litigation after the point of collection, that is, Censored data, cannot be reflected in model construction. This study includes censored data through survival analysis and quantitatively explores factors that influence patent litigation. As an experimental method, a Cox propotional hazards model is constructed from the collected patent big data to search for litigation hazard factors. The data is registered with the USPTO among patents related to artificial intelligence technology. As a result of the experiment, the number of IPC codes, the number of family countries, the number of registered claims, technology transfer status, and time required from application to registration were derived as factors that increase the litigation hazard ratio.