Analysis of multifactorial social unrest events with spatio-temporal k-dimensional tree-based DBSCAN

Sudeep Basnet, Leen-Kiat Soh, Ashok K Samal, Deepti Joshi

Research output: Chapter in Book/Report/Conference proceedingConference contribution

Abstract

Clustering geospatial event data requires defining a distance function between events as well as representing neighborhood characteristics where an event occurred in numerical or categorical values. For events such as social unrest events, in addition to the geospatial coordinates and time stamps, other factors needed to understand how they evolve include socioeconomic factors that fuel, say, the emergence of a social unrest event, and infrastructural factors that facilitate, say, the propagation of an event to nearby regions. In this paper, we focus on addressing two main challenges in spatiotemporal clustering of such event data: (1) how to derive a numeric representation of nearby geospatial objects in a neighborhood for an event, and (2) how to improve the clustering process to scale well for very large datasets. To address the first challenge, we propose two metrics—proximity and density—of geospatial objects, and incorporate them into the definition of a distance function between events. To address the second challenge, we propose a novel Spatio-Temporal k-Dimensional Tree-based DBSCAN (ST-KDT-DBSCAN) clustering approach that restricts the search radius for each event during clustering by first organizing the dataset into a k-dimensional tree structure, subsequently creating a Fixed-Radius Near Neighbor (FRNN) object for each event, and then carrying out DBSCAN considering only each event’s FRNN object when computing reachability. We have applied the solutions to 29,371 unrest events with socioeconomic and infrastructural factors recorded for the year 2014 in India, to identify event episodes in order to analyze how social unrest evolves. Our results show the viability and scalability of our solutions.

Original languageEnglish (US)
Title of host publicationProceedings of the 2nd ACM SIGSPATIAL International Workshop on Analytics for Local Events and News, LENS 2018
EditorsAmr Magdy, Liang Zhao, Xun Zhou, Yan Huang
PublisherAssociation for Computing Machinery, Inc
ISBN (Electronic)9781450360357
DOIs
StatePublished - Nov 6 2018
Event2nd ACM SIGSPATIAL International Workshop on Analytics for Local Events and News, LENS 2018 - Seattle, United States
Duration: Nov 6 2018Nov 6 2018

Publication series

NameProceedings of the 2nd ACM SIGSPATIAL International Workshop on Analytics for Local Events and News, LENS 2018

Other

Other2nd ACM SIGSPATIAL International Workshop on Analytics for Local Events and News, LENS 2018
CountryUnited States
CitySeattle
Period11/6/1811/6/18

Fingerprint

Scalability

Keywords

  • Geographic Information Systems
  • Social Unrest Analysis
  • Spatial-Temporal Clustering

ASJC Scopus subject areas

  • Computer Graphics and Computer-Aided Design
  • Computer Vision and Pattern Recognition

Cite this

Basnet, S., Soh, L-K., Samal, A. K., & Joshi, D. (2018). Analysis of multifactorial social unrest events with spatio-temporal k-dimensional tree-based DBSCAN. In A. Magdy, L. Zhao, X. Zhou, & Y. Huang (Eds.), Proceedings of the 2nd ACM SIGSPATIAL International Workshop on Analytics for Local Events and News, LENS 2018 [2] (Proceedings of the 2nd ACM SIGSPATIAL International Workshop on Analytics for Local Events and News, LENS 2018). Association for Computing Machinery, Inc. https://doi.org/10.1145/3282866.3282870

Analysis of multifactorial social unrest events with spatio-temporal k-dimensional tree-based DBSCAN. / Basnet, Sudeep; Soh, Leen-Kiat; Samal, Ashok K; Joshi, Deepti.

Proceedings of the 2nd ACM SIGSPATIAL International Workshop on Analytics for Local Events and News, LENS 2018. ed. / Amr Magdy; Liang Zhao; Xun Zhou; Yan Huang. Association for Computing Machinery, Inc, 2018. 2 (Proceedings of the 2nd ACM SIGSPATIAL International Workshop on Analytics for Local Events and News, LENS 2018).

Research output: Chapter in Book/Report/Conference proceedingConference contribution

Basnet, S, Soh, L-K, Samal, AK & Joshi, D 2018, Analysis of multifactorial social unrest events with spatio-temporal k-dimensional tree-based DBSCAN. in A Magdy, L Zhao, X Zhou & Y Huang (eds), Proceedings of the 2nd ACM SIGSPATIAL International Workshop on Analytics for Local Events and News, LENS 2018., 2, Proceedings of the 2nd ACM SIGSPATIAL International Workshop on Analytics for Local Events and News, LENS 2018, Association for Computing Machinery, Inc, 2nd ACM SIGSPATIAL International Workshop on Analytics for Local Events and News, LENS 2018, Seattle, United States, 11/6/18. https://doi.org/10.1145/3282866.3282870
Basnet S, Soh L-K, Samal AK, Joshi D. Analysis of multifactorial social unrest events with spatio-temporal k-dimensional tree-based DBSCAN. In Magdy A, Zhao L, Zhou X, Huang Y, editors, Proceedings of the 2nd ACM SIGSPATIAL International Workshop on Analytics for Local Events and News, LENS 2018. Association for Computing Machinery, Inc. 2018. 2. (Proceedings of the 2nd ACM SIGSPATIAL International Workshop on Analytics for Local Events and News, LENS 2018). https://doi.org/10.1145/3282866.3282870
Basnet, Sudeep ; Soh, Leen-Kiat ; Samal, Ashok K ; Joshi, Deepti. / Analysis of multifactorial social unrest events with spatio-temporal k-dimensional tree-based DBSCAN. Proceedings of the 2nd ACM SIGSPATIAL International Workshop on Analytics for Local Events and News, LENS 2018. editor / Amr Magdy ; Liang Zhao ; Xun Zhou ; Yan Huang. Association for Computing Machinery, Inc, 2018. (Proceedings of the 2nd ACM SIGSPATIAL International Workshop on Analytics for Local Events and News, LENS 2018).
@inproceedings{e27484dae39a428f9b6deeaf487c7b84,
title = "Analysis of multifactorial social unrest events with spatio-temporal k-dimensional tree-based DBSCAN",
abstract = "Clustering geospatial event data requires defining a distance function between events as well as representing neighborhood characteristics where an event occurred in numerical or categorical values. For events such as social unrest events, in addition to the geospatial coordinates and time stamps, other factors needed to understand how they evolve include socioeconomic factors that fuel, say, the emergence of a social unrest event, and infrastructural factors that facilitate, say, the propagation of an event to nearby regions. In this paper, we focus on addressing two main challenges in spatiotemporal clustering of such event data: (1) how to derive a numeric representation of nearby geospatial objects in a neighborhood for an event, and (2) how to improve the clustering process to scale well for very large datasets. To address the first challenge, we propose two metrics—proximity and density—of geospatial objects, and incorporate them into the definition of a distance function between events. To address the second challenge, we propose a novel Spatio-Temporal k-Dimensional Tree-based DBSCAN (ST-KDT-DBSCAN) clustering approach that restricts the search radius for each event during clustering by first organizing the dataset into a k-dimensional tree structure, subsequently creating a Fixed-Radius Near Neighbor (FRNN) object for each event, and then carrying out DBSCAN considering only each event’s FRNN object when computing reachability. We have applied the solutions to 29,371 unrest events with socioeconomic and infrastructural factors recorded for the year 2014 in India, to identify event episodes in order to analyze how social unrest evolves. Our results show the viability and scalability of our solutions.",
keywords = "Geographic Information Systems, Social Unrest Analysis, Spatial-Temporal Clustering",
author = "Sudeep Basnet and Leen-Kiat Soh and Samal, {Ashok K} and Deepti Joshi",
year = "2018",
month = "11",
day = "6",
doi = "10.1145/3282866.3282870",
language = "English (US)",
series = "Proceedings of the 2nd ACM SIGSPATIAL International Workshop on Analytics for Local Events and News, LENS 2018",
publisher = "Association for Computing Machinery, Inc",
editor = "Amr Magdy and Liang Zhao and Xun Zhou and Yan Huang",
booktitle = "Proceedings of the 2nd ACM SIGSPATIAL International Workshop on Analytics for Local Events and News, LENS 2018",

}

TY - GEN

T1 - Analysis of multifactorial social unrest events with spatio-temporal k-dimensional tree-based DBSCAN

AU - Basnet, Sudeep

AU - Soh, Leen-Kiat

AU - Samal, Ashok K

AU - Joshi, Deepti

PY - 2018/11/6

Y1 - 2018/11/6

N2 - Clustering geospatial event data requires defining a distance function between events as well as representing neighborhood characteristics where an event occurred in numerical or categorical values. For events such as social unrest events, in addition to the geospatial coordinates and time stamps, other factors needed to understand how they evolve include socioeconomic factors that fuel, say, the emergence of a social unrest event, and infrastructural factors that facilitate, say, the propagation of an event to nearby regions. In this paper, we focus on addressing two main challenges in spatiotemporal clustering of such event data: (1) how to derive a numeric representation of nearby geospatial objects in a neighborhood for an event, and (2) how to improve the clustering process to scale well for very large datasets. To address the first challenge, we propose two metrics—proximity and density—of geospatial objects, and incorporate them into the definition of a distance function between events. To address the second challenge, we propose a novel Spatio-Temporal k-Dimensional Tree-based DBSCAN (ST-KDT-DBSCAN) clustering approach that restricts the search radius for each event during clustering by first organizing the dataset into a k-dimensional tree structure, subsequently creating a Fixed-Radius Near Neighbor (FRNN) object for each event, and then carrying out DBSCAN considering only each event’s FRNN object when computing reachability. We have applied the solutions to 29,371 unrest events with socioeconomic and infrastructural factors recorded for the year 2014 in India, to identify event episodes in order to analyze how social unrest evolves. Our results show the viability and scalability of our solutions.

AB - Clustering geospatial event data requires defining a distance function between events as well as representing neighborhood characteristics where an event occurred in numerical or categorical values. For events such as social unrest events, in addition to the geospatial coordinates and time stamps, other factors needed to understand how they evolve include socioeconomic factors that fuel, say, the emergence of a social unrest event, and infrastructural factors that facilitate, say, the propagation of an event to nearby regions. In this paper, we focus on addressing two main challenges in spatiotemporal clustering of such event data: (1) how to derive a numeric representation of nearby geospatial objects in a neighborhood for an event, and (2) how to improve the clustering process to scale well for very large datasets. To address the first challenge, we propose two metrics—proximity and density—of geospatial objects, and incorporate them into the definition of a distance function between events. To address the second challenge, we propose a novel Spatio-Temporal k-Dimensional Tree-based DBSCAN (ST-KDT-DBSCAN) clustering approach that restricts the search radius for each event during clustering by first organizing the dataset into a k-dimensional tree structure, subsequently creating a Fixed-Radius Near Neighbor (FRNN) object for each event, and then carrying out DBSCAN considering only each event’s FRNN object when computing reachability. We have applied the solutions to 29,371 unrest events with socioeconomic and infrastructural factors recorded for the year 2014 in India, to identify event episodes in order to analyze how social unrest evolves. Our results show the viability and scalability of our solutions.

KW - Geographic Information Systems

KW - Social Unrest Analysis

KW - Spatial-Temporal Clustering

UR - http://www.scopus.com/inward/record.url?scp=85059004080&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=85059004080&partnerID=8YFLogxK

U2 - 10.1145/3282866.3282870

DO - 10.1145/3282866.3282870

M3 - Conference contribution

AN - SCOPUS:85059004080

T3 - Proceedings of the 2nd ACM SIGSPATIAL International Workshop on Analytics for Local Events and News, LENS 2018

BT - Proceedings of the 2nd ACM SIGSPATIAL International Workshop on Analytics for Local Events and News, LENS 2018

A2 - Magdy, Amr

A2 - Zhao, Liang

A2 - Zhou, Xun

A2 - Huang, Yan

PB - Association for Computing Machinery, Inc

ER -