Spatiotemporal dengue fever hotspots associated with climatic factors in taiwan including outbreak predictions based on machine-learning

Sumiko Anno, Takeshi Hara, Hiroki Kai, Ming An Lee, Yi Chang, Kei Oyoshi, Yousei Mizukami, Takeo Tadono

Research output: Contribution to journalArticlepeer-review

4 Citations (Scopus)

Abstract

Early warning systems (EWS) have been proposed as a measure for controlling and preventing dengue fever outbreaks in countries where this infection is endemic. A vaccine is not available and has yet to reach the market due to the economic burden of development, introduction and safety concerns. Understanding how dengue spreads and identifying the risk factors will facilitate the development of a dengue EWS, for which a climate-based model is still needed. An analysis was conducted to examine emerging spatiotemporal hotspots of dengue fever at the township level in Taiwan, associated with climatic factors obtained from remotely sensed data in order to identify the risk factors. Machine-learning was applied to support the search for factors with a spatiotemporal correlation with dengue fever outbreaks. Three dengue fever hotspot categories were found in southwest Taiwan and shown to be spatiotemporally associated with five kinds of sea surface temperatures. Machine-learning, based on the deep AlexNet model trained by transfer learning, yielded an accuracy of 100% on an 8-fold cross-validation test dataset of longitude-time sea surface temperature images.

Original languageEnglish
Article number771
Pages (from-to)183-194
Number of pages12
JournalGeospatial Health
Volume14
Issue number2
DOIs
Publication statusPublished - 2019 Nov 12

All Science Journal Classification (ASJC) codes

  • Medicine (miscellaneous)
  • Health(social science)
  • Geography, Planning and Development
  • Health Policy

Fingerprint Dive into the research topics of 'Spatiotemporal dengue fever hotspots associated with climatic factors in taiwan including outbreak predictions based on machine-learning'. Together they form a unique fingerprint.

Cite this