This paper proposes a method to improve the identification effect of technical Trajectory by adding ghost edges in the patent citation network, which includes calculating patent technology similarity, constructing ghost edge candidate set, adding the ghost edges by evaluating the utility measures, and using main path analysis to identify four technical trajectories. Taking US e-commerce data technology as an example, we find the following three points. (1) Adding a small amount of ghost edges in the patent citation network helps to increase the accuracy of technical trajectory identification, but adding a large number of ghost edges may cause destructive effects on the network structure and lead to identification bias. The experience value of this case is at most 10%. (2) Different construction methods of ghost edge candidate sets will have an important impact on the result of improving the trajectory recognition. No matter which candidate set is used, there is no deviation in the primary technical trajectory identification. However, there are differences in the subsequent technical trajectory identification. (3) The addition of the ghost edges further improves the network characteristics, especially the technical trajectory differences in subsequent locations which are identified.

提出一种通过在专利引证网络中添加鬼边来提高技术轨迹识别效果的方法，包括计算专利技术相似度、构建鬼边候选集、通过评估效用度量添加鬼边以及利用主路径分析以确定四种技术轨迹。以美国电商数据技术为例，我们发现以下三点。 (1)在专利引用网络中添加少量鬼边有助于提高技术轨迹识别的准确性，但添加大量鬼边可能会对网络结构造成破坏性影响，导致识别偏差。本案的经验值最多为10%。 (2)鬼边候选集的不同构建方法将对提高轨迹识别的结果产生重要影响。无论使用哪种候选集，初级技术轨迹识别都不会出现偏差。但后续技术轨迹识别存在差异。 (3)鬼边的加入进一步改善了网络特性，特别是后续识别位置的技术轨迹差异。