Successfully reported this slideshow.
We use your LinkedIn profile and activity data to personalize ads and to show you more relevant ads. You can change your ad preferences anytime.

Situational awareness in social media: lessons learned using information entropy in flood risk management

98 views

Published on

Increasing situational awareness using social media data is still a problem for the surveillance of disaster-related events due to the amount of data. In order to address this problem, a number of studies have been conducted on the basis of the Tobler's first law of geography, in which social media messages nearest to events are more relevant than the more distant messages. However, these studies fail to take the explicit content of the messages in terms of quantitative measures into account. A quantitative measure is important to prioritize and rank social media messages using another criterion beyond the geographical distance. This paper conducts a case study in the city of S\~{a}o Paulo, Brazil, for assessing the relationship between the information entropy and the distance to flooded areas of rain- flood-related Twitter messages. The results provide evidence that the entropy measure of the tweets is not governed by the Tobler's law of geography. Nonetheless, our findings do not challenge the Tobler's law assumption, but put forward discussions in terms of the relevance of the social media's content in relation to distance to the affected areas by disasters.

Published in: Education
  • Be the first to comment

  • Be the first to like this

Situational awareness in social media: lessons learned using information entropy in flood risk management

  1. 1. Núcleo de Universos Virtuais, Entretenimento e Mobilidade – II Workshop @NUVEM II Workshop @NUVEM 29 e 30 de Novembro de 2018 Sala S-213-0 – Campus Santo André – UFABC Núcleo de Universos Virtuais, Entretenimento e Mobilidade Situational awareness in social media: lessons learned using information entropy in flood risk management Sidgley C de Andrade1 , Camilo Restrepo-Estrada2 , Thiago A. G. da Costa3 Jó Ueyama3 , Alexandre C. B. Delbem4 , João Porto de Albuquerque4 1 Federal University of Technology - Paraná, Toledo, Brazil 2 Faculty of Economic Sciences, University of Antioquia, Medellín, Colombia 3 University of São Paulo, São Carlos, Brazil 4 Warwick University, Coventry, UK Publication available at http://nuvem.ufabc.edu.br/certificados/ii-workshop/2-workshop-nuvem-ufabc_paper_9.pdf
  2. 2. Núcleo de Universos Virtuais, Entretenimento e Mobilidade – II Workshop @NUVEM Introduction • Increasing situational awareness using social media data is still a problem for the surveillance of disaster-related events due to:  the amount of data and  how the prioritization of the data is done. • Most studies use only the geographic distance as a prioritization criterion (in accordance with the Tobler’s first law of geography). • No quantitative measure to extract the relevance of the explicit content has been addressed.  Decision-maker might choose the data by the “informational power” contained in them and not only by the geographical distance criterion. 2
  3. 3. Núcleo de Universos Virtuais, Entretenimento e Mobilidade – II Workshop @NUVEM Purpose • Carry out a case study in the city of São Paulo for assessing the relationship between the “informational power” and the distance to flooded areas of rain- flood-related Twitter messages. 3
  4. 4. Núcleo de Universos Virtuais, Entretenimento e Mobilidade – II Workshop @NUVEM Methodology 4 To determine the flooded areas To calculate the entropy of the on-topic tweets and the distances of them to the flooded areas To perform the correlation of both measures 21 3
  5. 5. Núcleo de Universos Virtuais, Entretenimento e Mobilidade – II Workshop @NUVEM Flooded areas from flood points 5 1 DBScanFlood points Flooded areas Parameters = {MinPts = 1, Eps = 216.43} rangeSemivariogram
  6. 6. Núcleo de Universos Virtuais, Entretenimento e Mobilidade – II Workshop @NUVEM Reference events 6
  7. 7. Núcleo de Universos Virtuais, Entretenimento e Mobilidade – II Workshop @NUVEM Distance to the flooded areas 7 2
  8. 8. Núcleo de Universos Virtuais, Entretenimento e Mobilidade – II Workshop @NUVEM Entropy 8 2
  9. 9. Núcleo de Universos Virtuais, Entretenimento e Mobilidade – II Workshop @NUVEM Linear correlation 9 3
  10. 10. Núcleo de Universos Virtuais, Entretenimento e Mobilidade – II Workshop @NUVEM Results 10 • Hypothesis: there is a relationship between the entropy and the distance measures of the on-topic tweets.
  11. 11. Núcleo de Universos Virtuais, Entretenimento e Mobilidade – II Workshop @NUVEM Results 11 • Hypothesis: there is a relationship between the entropy and the distance measures of the on-topic tweets. REJECT
  12. 12. Núcleo de Universos Virtuais, Entretenimento e Mobilidade – II Workshop @NUVEM Discussions and Conclusions • The results are still incipient, however, provide evidences that:  the content of the tweets are not governed by the Tobler’s law of geography.  (in scenarios with high resolution and multiple occurrences of events) • Further in-depth investigations are necessary and future works should:  incorporate more scenarios,  alternative methods to assess the relevance of social media content, and  other type of contents should be taken into account (e.g. images and videos). 12
  13. 13. Núcleo de Universos Virtuais, Entretenimento e Mobilidade – II Workshop @NUVEM Reference Tobler W., (1970) “A computer movie simulating urban growth in the Detroit region”. Economic Geography, 46(Supplement): 234-240. Gray, R. M. (1990). “Entropy and Information Theory”. https://doi.org/10.1007/978- 1-4757-3982-4 IBGE “Censo Demográfico 2010”. Brazilian Institute of Geography and Statistics, Rio de Janeiro, 2010. 13
  14. 14. Thank you very much! Sidgley Camargo de Andrade Lecturer (tenured) at Federal University of Technology - Paraná sidgleyandrade@utfpr.edu.br http://pessoal.utfpr.edu.br/sidgleyandrade/ Acknowledgements/Funding Núcleo de Universos Virtuais, Entretenimento e Mobilidade – II Workshop @NUVEM

×