Early identification of breakthrough technologies: Insights from science-driven innovations

Abstract

Identifying breakthrough technologies is crucial for advancing technological innovation and, in this sense, the innovation patterns driven by science are considered to be key pathways for forming breakthrough technologies. Building on this premise, this paper presents a framework for identifying breakthrough technologies that starts with these signals of scientific innovation. The first step in the method is to construct a science-technology knowledge network based on papers and patents. Then a two-stage selection method funnels the scientific innovation signals, filtering out those with the potential to trigger technological breakthroughs. Next, a machine learning-based link prediction model, integrating three types of features, identifies new links between science-driven signals and existing technologies. A community detection algorithm then identifies sub-networks of technologies formed around these new links. Finally, a structural entropy index is used to evaluate these sub-networks to determine potential breakthrough technologies. By systematically characterizing the content and core features of scientific innovation signals, this study reveals the driving sources of technological breakthroughs and sheds light on the absorption and diffusion processes of scientific innovation. We validated the method through a use case in the field of artificial intelligence. Those who manage technological innovation should find the insights of this research particularly valuable.