Evolution monitoring for innovation sources using patent cluster analysis

Abstract

According to the increasing importance of advanced technologies for economy growth and the incremental complexity of research and development management, a novel methodology is proposed in this paper to monitor the evolution trace of innovation sources. This approach focuses on the knowledge-transfer among technologies using patent cluster analysis. More specifically, a citation network model, consisting of patents in "Coherent Light Generators" classification, is established with the data collected from the United States Patent and Trademark Office. In addition, dynamical topological structure is investigated to probe into the overview properties and identify key milestones for the expanding citation network from 1976 to 2014. Next, a novel framework for patent clustering is developed to find out knowledge chunks of which internal knowledge-flows are dense while cut edges are sparse. Community detection algorithms are compared with different assessment indices based on citation network and the selected solution is improved using optimization objectives of cluster analysis. Then, the dynamical structure of the detected knowledge chunks is investigated and the evolution of innovation sources, identified by k-core decomposition, is monitored to unveil the technology development trace. Finally, analysis results are discussed and related conclusions are summarized. This article improves approaches for patent cluster analysis and develops a new follow-up investigation methodology for detected knowledge chunks. It is discovered there are not only scale increases, but also the integration for knowledge chunks during the focal period. Identifying the knowledge chunks which obtain rapid growth in both cluster scale and innovation source is useful to detect technology development opportunities.