[A-TT30] Machine Learning Techniques in Weather, Climate, Ocean, Hydrology and Disease Predictions

[ATT30-07] Forecasting Dengue Outbreaks in Vietnam: A Machine Learning Approach Utilizing Climate Data

*Patrick Martineau¹, Noboru Minakawa², Michael Teron², Trang Huynh³ (1.Japan Agency for Marine-Earth Science and Technology, 2.School of Tropical Medicine and Global Health, Univ. of Nagasaki, 3.Ho Chi Minh City Pasteur Institute, Vietnam)

Keywords:Infectious diseases (dengue), Climate, Machine learning, Predictions

Dengue outbreaks cause important socioeconomic impacts in Vietnam (70-370K cases per year, 2M USD or 300M JPY). Like other vector-borne diseases, outbreaks tend to be more severe when climatic conditions are favorable for vector reproduction. The risk of contracting dengue is typically higher from May to December when temperatures are warmer and rainfall is more abundant, conditions that favor the spread of aedes aegypti mosquitoes. Likewise, interannual climatic variations are known to influence the severity of dengue outbreaks. Here we use information about these interannual variations in the climate system to provide early warning predictions with the use of machine learning. More specifically, we use global sea surface temperature (SST) fluctuations as predictor variables to perform categorical predictions (more or less than usual) of dengue incidence for clusters of provinces that are grouped according to similarities in their temporal evolution. Among these clusters, we find evidence of predictive skill only for the southernmost cluster which is associated with SST variations in the Atlantic, Indian, and Pacific Oceans. Predictive skill at lead times of 3-5 months tends to be higher from July to November, indicating that these predictions are potentially useful to mitigate outbreaks during the high-risk season.