Research On Scientific Data Evaluation Based On Citation Network

With the advent of the era of big data,academic research has entered a new paradigm,the most prominent feature of which is the intensive application of data.In order to standardize the application of scientific data,so as to improve the availability and reusability of scientific data,more and more institutions are committed to improving the citation specification of scientific data.As a result,policies and practices related to scientific data citation have gradually developed.Scientific data is one of the important outputs of scientific research.With the increasingly mature data citation,how to evaluate the influence of scientific data has become an urgent problem.In order to solve the above problems,previous studies mainly used statistical analysis of data citation,taking the number of data cited as the main index to evaluate the influence of scientific data.However,the perspective of this evaluation method is relatively single.In order to expand the evaluation dimension of scientific data,this study uses scientometrics and social network analysis theory to conduct citation analysis and social network analysis on PLoS citation data,and evaluates scientific data by constructing and analyzing 2-mode data citation network and 1-mode data co-occurrence network based on CO citation.The main conclusions are as follows:1.In general,the data reference behavior of PLoS is becoming more and more common.Data set publishers have obvious Matthew effect,that is,a small number of data set publishers publish the vast majority of data sets.Moreover,the larger the amount of data published by the publisher,the higher the citation frequency of the publisher's dataset.Among 142 data set publishers,GBIF is the core publisher.3.It is necessary and effective to evaluate scientific data from the perspective of network structure: the data sets highlighted by 2-mode data citation network are not in the top ten cited times.4.ICPSR,the publisher of data sets,has a high degree of intermediary centrality and topic relevance.Nodes with high degree of intermediary centrality have a strong ability to control information flow.The above conclusion shows that the application of social network analysis method to the evaluation of scientific data can present the complex network structure of data reference,so as to tap the deep value of data reference.The evaluation of scientific data based on citation network not only expands the traditional statistical analysis perspective,but also helps to improve the scientific data evaluation system,so as to improve the scientific data citation standards,encourage data producers and data management institutions to share more high-quality and reliable data,and promote the development of scientific research.