Name
Abstract | ||
|---|---|---|
Coupling social media with geographic location has boosted the worth of understanding the real-world situations. In particular, event detection based on clustering algorithms or bursty detection aims to find more specific topics that represent real-world events from geo-tagged social media. However, it is also necessary to identify unusual and seemingly inconsistent patterns in data, namely outliers. For example, it is difficult to obtain social media posted by residents of the places where a disaster is happening for quite some while. In this paper, we focus on a problem in partitioning a space to find a meaningful local outlier pattern by using a genetic algorithm (GA). We first describe a model of local patterns based on spatio-temporal neighbors and a normal distribution test. Then we propose our optimization process to maximize the number of patterns. Finally, we show results of the performance simulation with a real dataset related to a landslide disaster. |
Year | DOI | Venue |
|---|---|---|
2017 | 10.1007/978-3-319-59072-1_13 | ADVANCES IN NEURAL NETWORKS, PT I |
Keywords | Field | DocType |
Geo-social media,Spatio-temporal analysis,Outlier detection,KD-tree partitioning,Genetic algorithm | Anomaly detection,Local outlier factor,Normal distribution,Social media,Spatio-Temporal Analysis,Computer science,Outlier,Artificial intelligence,Cluster analysis,Genetic algorithm,Machine learning | Conference |
Volume | ISSN | Citations |
10261 | 0302-9743 | 0 |
PageRank | References | Authors |
0.34 | 8 | 4 |
Name | Order | Citations | PageRank |
|---|---|---|---|
| Teerawat Kumrai | 1 | 21 | 2.72 |
| Kyoung-Sook Kim | 2 | 24 | 14.07 |
| Mianxiong Dong | 3 | 2018 | 152.73 |
| Hirotaka Ogawa | 4 | 196 | 23.58 |