Co-authorship Network Analysis For Scientific Papers

Nowadays, bioinformatics becomes one of the hottest interdisciplinary subjects. Largenumbers of research subfields in this subject are booming rapidly. Complex Network isanother booming research field, which is widely applied in different disciplines to studyand analyze specific networks. Among all of them, Social Network Analysis is a commonand well study field. The scientific collaboration Network is a subfield of it. In this paper,we study the co-authorship Network with respect to bioinformatics, and perform somefurther analysis.The co-authorship collaboration network is the common form of scientificcollaboration network, which is built according to the co-authorship relations in one ormore papers. If two authors’ names appear in a same paper, there will be an edge betweenthem. This can strongly support the analysis of structure of scientific collaborationNetwork.First of all, we introduce the foundation of complex network. In the meantime, wereview the background, development and present hot topics in bioinformatics. Therefore,according to compare, six popular bioinformatics topics are chosen. Based on somecondition, six networks are constructed with six keywords. The data sets are collected fromthe ‘‘Web of Science which interval ranges from2002to2012.Then, we review our methods of data collecting, data pre-processing, networktopology attribute and network construction and visualization. During data pre-processing,there is a critical problem, name recognition, for building a network. We review theprevious work of Newman in the problem, and propose another method to handle it. Theresults are fairly good.After processing raw data into network data, we explore those networks, estimate theirstatistical properties and topology properties, and then, interpret their significance. In themeantime, we study whether the data sets are plausible to fit power-law degree distributions.Also, we note that the distribution of author number of the papers seems to be subject tokind of power law distribution. Thus, we testify our hypothesis in this paper and compare itwith other alternative distribution. We also divide these datasets into some subsets withdifferent conditions including impact factor, cite times and publish time to do some further analysis.The study of this paper could help understand co-authorship network, especially in theformation mechanism of various fields of bioinformatics. We believe that the study ofbioinformatics co-authorship networks could intensively improve our understanding ofpatterns and trends of the current research. In the meantime, it can help understand theimpact of different factors, including impact factor and cite time in macroscopic prospect.Besides, the study of relative authors and organization in the network could also help tracethe main efforts and progress in specific fields.