From ranking and clustering of evolving networks to patent citation analysis

摘要

The network of patents connected by citations is an evolving graph that represents the innovation process of society. A patent citing another implies that the cited patent contains a piece of previously existing knowledge that the citing patent is building upon. Understanding the development of the patent citation network contributes to the discovery of the rules that govern its growth. By adopting a citation-based recursive ranking method for patents, the evolution of new fields of technology can be traced. Specifically, a reinforcement learning based ranking algorithm was adopted and found more appropriate than the now classical PageRank algorithm. The temporal evolution of patent classes and the eventual interaction among them were studied by combining regression and clustering methods. While some patterns for the network dynamics have clearly been identified, more work is needed to see the details and to be able to make predictions for the emerging fields of technologies.