The Evolution Of Science Of Science Data Of APS Based On Complex Networks

With the development and wide application of computer technology,many experts and scholars in various fields focus on computational social science.By applying modern science and technology to abundant data,they aim to get novel ideas and methods to explain human activities and solve social problems,and thus computational social science flourishes.As an important role to promote social progress,scientific research attracts considerable attention.Applying computer technology and network science theories to scientific data is a significant way to research further in scientometrics which is a part of science of science(SciSci).Data about papers of six journals from 1958 to 2015 is selected from data sets publicly provided by American Physical Society(APS),and three kinds of accumulative network models are constructed.These networks are analyzed in the thesis from two aspects,nodes and motifs,to reveal the underlying mechanisms of the evolution of citation networks and collaboration networks.1.Citation networks between papers are undirected and unweighted.With nodes,the growth rate of network nodes and edges accelerate over time,but slows down in recent years due to the decrease in journals publications.There is a strong positive correlation between the number of nodes and edges.No matter how the networks evolve,their indegree distributions follow the power-law distributions,and the out-degree distributions follow the exponential distributions.Four degree-degree correlation coefficients excluding r(out,out)are close to 0 in recent years.With motifs,four directed motifs of three nodes in the networks take into consideration.M10 which represents two papers without any reference between them cite the same paper has the largest number and the highest proportion,while M5 called the feed-forward loop is the fewest.The proportion of the four motifs in the network is stable.Papers with different motifs in similar positions have similar exponential distributions of motif year differences,but less heterogeneous in M5.However,the stable tail proportion of M5 indicates that more similar and incremental studies are more sustainable.2.Citation networks between scientists are directed and weighted.First of all,the scientists' publications distributions follow the power-law distributions.It's further confirmed that the number of papers published by scientists has a strong relationship with their teams size.With nodes,networks have a large number of self-loop edges,indicating that self-citations among scientists are very common.The in-degree and out-degree distributions,in-strength and out-strength distributions of networks follow the power-law distributions.Large-degree nodes are high-strength,which is indicated by the linear relationship between the degree and strength.The assortative coefficients and the weighted assortative coefficients decrease with time,which shows weak assortativity or dissortativity of networks.With motifs,since mutual references are common,there are a large number of motifs with bidirectional edges in the networks.M9 called chain reference and M10 accounted for the highest proportion.The high ratio ofM9 indicates that there are significant gaps in scientific research ability and research depth among authors of journals papers.3.Collaboration networks between scientists are undirected and unweighted.First of all,the average number of authors increases over time,indicating that scientific research activities are increasingly inclined to collaborations.With nodes,the degree distributions follow the power-law distributions.Although unconnected,giant components of these networks exhibit some typical features,such as high clustered and small average shortest paths.With undirected motifs,M1 which has two edges accounts for the highest proportion,and a few M1 are converted into the fully connected M2 later,which indicates that having the same collaborators promote the cooperation of scientists,but with limited influence.And about 30%of the researchers in M2 come from the same affiliations,which shows that the same affiliations have greater positive influence on the collaboration between scientists.