A New Approach To Scientific Investigations Based On The Structure Of Citation Networks

Research evaluation as well as scientometric indicators are very important tools in the policy-making of scientific management. However, it is faced to the challenge that there are still many problems in the current research of scientometics, such as many indicators are all in one-dimension except the h-index, citation counting are just taking direct influence into account, few indicators for patentometrics, and reference have been overlooked in the citation network analysis.In the research, we develop a new approach of scientific investigation based on the structure of citation networks, combining the graph theory with citation analysis. Four topics were focused:the definitions of new indicators based on a new structural framework of "TSU-CN+E-CN"; the use of these new indicators; the mechanism of scientific investigation based on an ego article citation network extending over several generations; the use of these new indicators in patent networks based on "TSU-CN+E-CN". The main contents and achievements are listed as follows:First, a new structural framework of TSU-CN (Targeted node-subnetwork-universal network) is proposed based the graph theory of ego citation networks (E-CN), which is very suitable for indicating the scientific evolution in the environment of "targeted individual-small science-big science (universal science)". The results suggest that the new framework of "TSU-CN +E-CN" is very useful to reflect the mechanism of series indicators proposed in our study, such as the influence of citations generations, the structural indicators of article, the structural indicators of patent.Second, four citation generation types of Gs, Gm, Hs, Hm were proposed based on the theories of set and multiset, combined with graph theory, citation analysis as well as visual technique, they remove confusion when studying citation and reference generations. Based on that, the calculations of four types of influence of three forward and backward citation generations were also proposed using the exponential generating function, taking direct and indirect into account.Third, a new multi-layered approach to measure the technological value of patents based on "TSU-CN + E-CN" is developed in this research. Four structural indicators for patent were introduced, following Rousseau's structural indicators of article. The results of the real world samples as well as Monte Carlo Simulation suggest that the structural indicators developed in our study reflect a patent's performance in its environment of "TSU-CN+ E-CN". In this way a multi-dimensional picture of the intrinsic characteristics of the technological value of a patent is provided.And fourth, a new application of structure study of citation analysis was proposed. Considering the outgrow index as a time series, we study relative changes within "TSU-CN+E-CN" and determine when and if the citing article outgrows the ones on which is it based, by proposing three types of time series: two synchronous approaches, including a cumulative form, and a diachronous one. Combining two dimensions leads to an outgrow matrix. The results of real examples suggest that the time series of outgrow indices proposed by our research not only for comparing among articles, but also for sets of articles.In this research, a new approach of scientific investigation based on the structure of citation networks was developed, including the framework of TSU-CN, the definitions of four types citation generations Gs, Gm, Hs, Hm, the calculations of four types of influence of three forward and backward citation generations, the definitions of new types of structural indicators, the time series of outgrow indices, a new multi-layered approach to measure the technological value of patents based on the patent structural indicators. It provides a new, and we think useful, approach to citation analysis as well as scientometrics.