Identifying The Main Path Based On The Influence Flow In Citation Network

With the increase of knowledge as the time goes on,the citation network has evolved into a huge network system.How to find the key documents from the huge citation network is an important issue to analyze the changes in specific fields.The main path analysis method effectively avoids the high-cited nodes selected only by considering the indegree value of nodes but ignoring the problem that the"key documents"may not be connected strongly,and it can also better describe the path development process in a target field.The traditional algorithm weighting the path based on the traversal counts approach have been applied in various field research,but the traversal counts will cause some shortcomings such as the deviation the same initial traversal count of each edge will bring,the incomplete search path and the singularity of the algorithm and so on.In this paper,the underlying problems of SPX algorithm are modified to some extent.According to the core idea of eigenvector centrality,the concept of influence flow in citation network is summarized,and the algorithms to weight the path based on equally-distributed and unequally-distributed influence flow are proposed.Firstly,main paths in desalination field and information security field through classical traversal counts SPLC algorithm are obtained as a comparative experiment of the main path weighted by equally-distributed influenced flow algorithms.The results show that the main paths based on SPLC algorithm exists some shortcomings such as not all nodes have high cited times and there are review literature occurred in the scientific main path.Secondly,this paper proposed the influence flow algorithms based on the PageRank and single traversal from to perspective to divide influence equally to get the corresponding main path results in desalination field and information security field respectively as well.The main paths obtained by each algorithm are compared to get the similarities and differences between the SPLC algorithm from the aspects of shape,node content,cited times of node and theme evolution.Comparisons show that the proposed two algorithms supplement some key nodes in the main path derived from the SPLC algorithm to some extent.Then,this paper proposed another influence flow algorithm based on the coupling strength that citing nodes transmit unequal influence to cited nodes because the roles played by different documents in the process of knowledge diffusion are different,the influence of the transmission between the nodes is also different but according to the closeness between nodes.This algorithm is applied to the existing DNA network and compared the forms and content of the main path with the classical traversal counts algorithm.Finally,this paper summarizes main conclusions of the new algorithms proposed.Then we discuss the limitations of the research which are not considered and prospects for future research directions on main path analysis.