Research On Co-word Analysis Based On Community Detection Algorithm

The co-word network composed of the key words and their co-occurrence is a special kind of scientific knowledge network,which represents the structure of scientific cognition.However,among the co-word analysis clustering algorithms,the traditional co-word analysis clustering algorithm has the following shortcomings:(1)The clustering process needs to determine the number of clustering in advance;(2)keywords can only be grouped into a cluster.In view of the above shortcomings,this paper adopts complex network community detection algorithm to improve the co-word analysis method.On the basis of UEOC(unfold and extract overlapping communities)community detection algorithm,two improved community detection algorithms are proposed: CW＿UEOC(Co-occurrence Weighting and Extract Weighting),ATCW＿UEOC Community Detection Algorithm(Article Citation Weighting Unfold and extract overlapping communities),and build co-word network to carry out clustering experiment,hoping to solve the deficiency in traditional co-word analysis that clustering process needs to determine the number of clusters in advance and the keywords can only be grouped into one category.First of all,the paper mainly aims at the shortcoming that keywords can only be divided into one class in the clustering process based on co-word analysis,and uses UEOC community detection algorithm in complex network to weight word network.Secondly,the UEOC community detection algorithm based on complex network is improved,and two new community detection algorithms are proposed,and both of them use connectivity degree as the truncation criterion to provide objective basis for judging the number of clusters.Thirdly,the ATCW＿UEOC community detection algorithm uses the measure of timeliness of cited references and the measure of total cited references to identify them as literatures,so as to improve the problem of "same quantity but different quality" of keywords faced by the general co-word analysis method at the literature level.Finally,we use the method of strategic coordinate system to compare with CW＿UEOC community detection algorithm and ATCW＿UEOC community detection algorithm.The experiment shows that the proposed CW＿UEOC community detection algorithm and ATCW＿UEOC community detection algorithm have a good improvement on the clustering results of keywords.The same keyword in multiple fields can be divided into different research topics,and both of the two community detection algorithms use connectivity as a truncation criterion to provide objective basis for judging the number of clusters.In addition,ATCW＿UEOC community detection algorithm proposed a keyword feature selection method based on the measurement of timeliness of citation and the weight of total citation metrics.