A Revision Approach for Scalable Community Detection
碩士 === 國立中正大學 === 資訊工程研究所 === 105 === In the past decade, community detection is an important task in the eld of graph data mining. Most high quality methods have limitation on large scale network. Scalable Community Detection(SCD) based on maximize Weighted Community Clustering(WCC) is proposed to...
| Main Authors: | , |
|---|---|
| Other Authors: | |
| Format: | Others |
| Language: | en_US |
| Published: |
2017
|
| Online Access: | http://ndltd.ncl.edu.tw/handle/30683306251988193605 |
| Summary: | 碩士 === 國立中正大學 === 資訊工程研究所 === 105 === In the past decade, community detection is an important task in the eld of
graph data mining. Most high quality methods have limitation on large scale
network. Scalable Community Detection(SCD) based on maximize Weighted
Community Clustering(WCC) is proposed to solve the large scale community
detection problem. However it have been suered in the clustering result is
too fragment. In this paper we proposed SCD-Nu and SCD-Merge to improve
the clustering result. SCD-Nu is based on WCC-Nu which is the redenition
of cohesion. We adjust the metric which still based on triangl analyze, and
implements it on SCD. SCD-Nu nd much more bigger communities and
decreasing the number of communities. Take DBLP for example, the biggest
community size is growing up from about 1 hundred to about 2 thousand
nodes. SCD-Merge is trying to merge the communities of the SCD result.
The merged condition is based on the idea of conductance. But due to the
merge condition is not based on the metric dene, sometimes the algorithm
will forced the merged community to separate in order to get higher WCC
value.
|
|---|