Co-ranking Algorithm For Journals In A Heterogeneous Network

Since Eugene Garfield put forward the Impact Factor,the citation analysis of the journal is already one of the most popular bibliometric methods.In this paper,we analyze the principles of the three ranking methods that are PageRank,HITS and Random Co-Rankin model.On this basis,combined with the background of journal rankings,a co-ranking algorithm in multi-layer heterogeneous networks is proposed.In this algorithm,we learn from HITS,build local attribute index PB-index to filter the sub-network of the macroscopic network,and obtain the corresponding citation sub-network,which is also known as the local network.In the macroscopic global network,the algorithm uses PageRank to rank the journals.While in the microscopic local network,the PR value got from the macroscopic network is used to correct the probability transfer matrix and the convergence value is obtained to modify the macroscopic information.Thus,we achieve the interaction of macro and micro information.To study the nature of the algorithm,we conduct a lot of data experiments on the random networks,and run to the conclusion that the co-ranking algorithm is efficient in the scale-free network.Besides,we also study the effect of different selection path of root sets.Although different root set routing on the same network will get different convergence results,the difference is almost negligible.In addition,when some nodes in the macro network are cheating,studies show that the co-ranking algorithm has a significant advantage over the PageRank.As for the high-energy physics citation network,we analyze the degree distribution and find that that network is a typical scale-free network.Therefore,we use the co-ranking algorithm to rank journals and get to know that the ranking results of co-ranking algorithm and PageRank are almost consistent.However,when it comes to local ranking results,the results of the co-ranking algorithm are more consistent with our knowledge about high energy physics.