Mining Research, Based On The Citation Network Of Scientific Research Groups

The development of sci-tech greatly depends on the research work of scientific researchers which is also the primary thrust to social progress. Research community mining could help researchers to understand the trend of science.It plays an effective role in promoting cooperation among researchers from different academic background.Further it can stimulate the development of interdisciplinary science and give birth to new research topics.Scientific literature is the main carrier and indicator of researchers' achievements and performance.Citation network constituted by scientific literature and citation relations reflects the content similarities and knowledge transfer among research papers.It is widely used to estimate the academic influence of research papers.Mining research community on the basis of citation network can help to identify potential research groups with relevant research interest and increase the chance of scientific cooperation.In order to find out the research communities with relevant research interests,the citation relationship among papers is taken into account.Citation analysis model based on citation paths is established,upon which the relevant indicators are built up to represent the relevancy of research direction.Then text comparing indicators are considered to remove the fake connects.Cosine theorem of text matching is used in body text,summary as well as the main reference papers.After that the communities are recognized with DBScan clustering algorithm.Finally,the prototype simulation is established to reflect the feasibility and effectiveness of the algorithm.The citation analysis model contains more information than the past models,so that the subtle groups could be found and the potential collaboration opportunities could be explored more easily.The construction of paper relevancy and author relevancy indicators turns the original problem into a clustering problem.It implements a more objective and creative community mining methods.Citation path based indicators is of stronger comparability comparing to the past indicators. Besides,the application of text matching as content relevancy makes indicators more reasonable so that the mining communities could be more convincing.