A Machine Learning Based Method for Automatic Identification of Disaster Related Information Using Twitter Data


Christidou A. N., Drakaki M., Linardos V.

4th International Conference on Intelligent and Fuzzy Systems (INFUS), Bornova, Turkey, 19 - 21 July 2022, vol.505, pp.70-76 identifier identifier

  • Publication Type: Conference Paper / Full Text
  • Volume: 505
  • Doi Number: 10.1007/978-3-031-09176-6_8
  • City: Bornova
  • Country: Turkey
  • Page Numbers: pp.70-76
  • Keywords: Machine learning, Twitter, Social media, Disaster management, Natural disaster
  • Istanbul Technical University Affiliated: No

Abstract

The impacts of natural disasters on communities are devastating including loss of human lives and severe damages to properties. Social media data including Twitter and Facebook data can play a critical role in various phases of disaster management, in particular in the phases of disaster response and disaster preparedness. Geospatial data available by Twitter provide valuable real time information for search and rescue operations of emergency response units, damage assessment and disaster monitoring. Online social networks are an integral part of disaster communication and response systems. Machine learning (ML) algorithms can leverage social media data to facilitate disaster management operations. In this paper an ML based method is proposed to swiftly recognize natural disaster events in order to enhance the performance of emergency services to cope with the events. Various machine learning algorithms were used to automatically identify tweets related to natural disasters. Best results were achieved with the Logistic Regression (LR) and Support Vector Machine (SVM) algorithms. The developed method provides promising results to enhance informed decision making during and after disaster events.