An intelligent autoregressive-distributed lag model: A climate-driven approach for predicting dengue fever incidence in Taiwan cities.

Abstract

Dengue fever lacks specific treatments and vaccines. Its prevalent areas globally are mainly in tropical and subtropical regions, with its spread being strongly influenced by meteorological factors such as temperature and rainfall. Given the numerous influencing variables, an accurate prediction model is highly desirable to support anti-dengue control strategies. This study developed a climate-driven dengue fever model to predict the areas susceptible to dengue fever spread in Taiwan under current and future climate change. The data were sourced from the CDC's open platform in Kaohsiung City and Tainan City, which have higher dengue fever incidence compared to other cities in Taiwan. Climate factors and Google Trends variables were utilized to forecast dengue fever incidence. A novel hybrid model, integrating an intelligent algorithm with an autoregressive-distributed lag model and incorporating a mechanism to include or exclude outbreak periods, was proposed to assess their influence on forecasting accuracy. The results indicated that the proposed model with support vector regression yielded the best results for Kaohsiung data, while the proposed model with gene expression programming showed the best performance for Tainan data. Additionally, the findings revealed that once dengue fever occurs, its duration is quite long, up to 10 weeks, and the lag periods of weather attributes contribute to the continued recurrence of dengue in Taiwan. Overall, the results of this study can serve as a reference for implementing sustainable prevention and control programs and for government agencies to prepare early responses to dengue fever.