Journal Title
Title of Journal:
|
|
|
|
|
|
|
|
|
|
Authors: BingJie Sun HuaWei Shen XueQi Cheng
Publish Date: 2015/8/4
Volume: , Issue: , Pages: 104-115
Abstract
Clustering is one fundamental task in network analysis A widelyused clustering method is kmeans clustering where clustering is iteratively refined by minimizing the distance between each data point and its cluster center For kmeans clustering one key issue is initialization which heavily affects its accuracy and computational cost This issue is particularly critical when applying kmeans clustering to graph data where nodes are not embedded in a metric space In this paper we propose to use diversified ranking method to initialize kmeans clustering ie finding a set of seed nodes In diversified ranking seed nodes are figured out by considering their centrality and diversity in a unified manner With seed nodes as starting points kmeans clustering is used to cluster nodes into groups We apply the proposed method to detect communities in synthetic network and realworld network Results indicate that the proposed method exhibits high effectiveness and efficiency
Keywords:
.
|
Other Papers In This Journal:
|