Internal dynamics of patent reference networks using the Bray–Curtis dissimilarity measure

Abstract

Abstract Background Patents are indicators of technological developments. The science & technology categories, to which they are assigned to, form a directed, weighted network where the links are the references between patents belonging to the respective categories. This network can be conceived as a kind of intellectual ecology, lending itself to mathematical analyses analogous to those carried out in numerical ecology. The non-metric Bray–Curtis dissimilarity, commonly used in quantitative ecology, can be used to describe the internal dynamics of this network. Results While the degree-distribution of the network remained stable during the studied years, that of the sub-networks of with at least k links showed that k = 5 is a critical number of citations: this many are needed that the bias towards already highly cited works come into effect (preferential attachment). Using the dij Bay-Curtis dissimilarity between nodes i and j, a surprising pattern emerged: the log-probability of a change in dij during a quarter of year depended linearly, with a negative coefficient, on the magnitude of the change itself. Conclusions The developed methodology could be useful to detect emerging technological developments, to aid decisions, for example, on resource allocation. The pattern found on the internal dynamics of the system depends on the categorisation of the patents, therefore it can serve as an indicator when comparing different categorisation methods. Graphical Abstract