专利寿命视角下的核心专利识别模型构建

Building Core Patent Identification Models from the Lifespan Perspective of Patents

文章基于专利寿命视角构建核心专利识别模型,为核心专利的识别以及专利情报分析提供新视角及研究模型。从专利寿命视角出发,筛选影响专利寿命的因素;建立基于深度学习生存预测算法的专利寿命预测模型,通过预测得出专利生存概率;根据设置的生存概率阈值筛选核心专利。研究发现:5种深度学习模型在专利寿命预测中均具有较好表现,其中MTLR模型的整体性能最好。通过与前人研究进行交叉验证后,设置16年为核心专利的生存阈值,根据MTLR模型和设置的生存阈值对5G关键技术领域的核心专利进行筛选,最终选出141项核心专利;对核心专利的技术主题进行聚类后发现,其技术主题集中在无线接入技术以及软件定义网络方面;核心专利相较于非核心专利,保护范围更大、权利更稳定、技术影响力更强。

Based on the lifespan of a patent, core-patent identification models are built to present new perspectives for identifying core patents and patent information analysis. From the perspective of the life of a patent, this paper screens the factors affecting patents’ duration, and devises prediction models of patent lifespan based on deep learning and survival prediction algorithms,which project the survival probability of a patent. Then,by setting thresholds on survival probability,the core patents are sifted out. The study suggests that the five deep learning models have good performance in life prediction for patents,and MTLR model performs the best in general. After cross-validating with previous studies,the paper establishes a survival threshold of 16 years for core patents. On the basis of MTLR model and the survival threshold,141 core patents in the field of 5G key technology are identified. With a cluster analysis on the technical subjects of the core patents,it is found that the patents mainly focus on wireless access technology and software defined network. Compared with the non-core patents,the core patents are of broader protection,more stable rights and stronger technical influence.