Paper Info
Title
Geo-located community detection in Twitter with enhanced fast-greedy optimization of modularity: the case study of typhoon Haiyan
Abstract
As they increase in popularity, social media are regarded as important sources of information on geographical phenomena. Studies have also shown that people rely on social media to communicate during disasters and emergency situation, and that the exchanged messages can be used to get an insight into the situation. Spatial data mining techniques are one way to extract relevant information from social media. In this article, our aim is to contribute to this field by investigating how graph clustering can be applied to support the detection of geo-located communities in Twitter in disaster situations. For this purpose, we have enhanced the fast-greedy optimization of modularity (FGM) clustering algorithm with semantic similarity so that it can deal with the complex social graphs extracted from Twitter. Then, we have coupled the enhanced FGM with the varied density-based spatial clustering of applications with noise spatial clustering algorithm to obtain spatial clusters at different temporal snapshots. The method was experimented with a case study on typhoon Haiyan in the Philippines, and Twitter's different interaction modes were compared to create the graph of users and to detect communities. The experiments show that communities that are relevant to identify areas where disaster-related incidents were reported can be extracted, and that the enhanced algorithm outperforms the generic one in this task.
Year
DOI
Venue
2015
10.1080/13658816.2014.964247
International Journal of Geographical Information Science
Keywords
Field
DocType
social media
Semantic similarity,Data mining,Social media,Typhoon,Computer science,Popularity,Artificial intelligence,Clustering coefficient,Cluster analysis,Snapshot (computer storage),Modularity,Machine learning
Journal
Volume
Issue
ISSN
29
2
1365-8816
Citations 
PageRank 
References 
17
0.87
31
Authors
3
Name
Order
Citations
PageRank
Mohamed Bakillah116111.55
Ren-Yu Li2170.87
Steve H. L. Liang324423.13