Abstract Dengue incidence has been increasing dramatically in last few years with nearly four hundred million annual cases worldwide. It has been postulated that the wide-spread of dengue be due to climate change and increased exposure following the increasing human population in the affected regions. Climate change impacts on ecosystem have also set a critical role in the unpredictability of vector breeding behavior. A compelling strategy in the modeling of dengue outbreak must therefore integrate climate factors inasmuch as they determinedly govern incidence patterns. The aim of this paper is to construct a clustering integrated multiple regression model for predicting dengue incidence rate based on incidence, rainfall, and humidity data, which renders early warning information. The data used were dengue incidence data in Jakarta obtained from Jakarta Health Office and meteorological data from Indonesian Agency for Meteorology, Climatology and Geophysics (BMKG) in the period 2008–2016, defined on weekly basis. Cross-correlation was used to determine the interrelationship between dengue, rainfall, and relative humidity in Jakarta. Further improvement of the model was done by instrumenting the accumulated preceding one-month dengue incidence as an additional correction term in the model. The best fittings in terms of outbreak catchment and minimal mean squared error were obtained from the model variants involving the accumulated original and logarithm of the incidence rates respectively. Both the historical incidence rate locale and centroids of the meteorological data related to the clustering as well as the accumulated incidence rate serve as the key determinant for the upcoming incidence rate. An optimal clustering was determined in a way that the mean squared error achieves its foremost minimum, which almost coincides with the division into tertiles. These clustering strategies can be utilized to provide a more accurate forecast of the ominous dengue incidence for a few weeks’ lead-time.

中文翻译：

---

通过聚类集成多元回归模型评估雅加达登革热发病率与天气之间的相互作用

摘要 登革热发病率在过去几年急剧增加，全世界每年有近 4 亿例病例。据推测，登革热的广泛传播是由于气候变化和受影响地区人口增加后暴露增加所致。气候变化对生态系统的影响也在媒介繁殖行为的不可预测性方面发挥了关键作用。因此，登革热爆发建模中的一项引人注目的策略必须整合气候因素，因为它们决定了发病率模式。本文的目的是构建一个聚类集成多元回归模型，用于基于发病率、降雨量和湿度数据预测登革热发病率，从而呈现预警信息。使用的数据是从雅加达卫生办公室获得的雅加达登革热发病率数据和印度尼西亚气象、气候和地球物理局 (BMKG) 2008-2016 年期间每周定义的气象数据。互相关用于确定雅加达登革热、降雨量和相对湿度之间的相互关系。通过将累积的前一个月登革热发病率作为模型中的附加校正项，对模型进行了进一步改进。爆发流域和最小均方误差方面的最佳拟合分别从涉及发病率累积原始值和对数的模型变体中获得。与聚类相关的气象数据的历史发生率地点和质心以及累积发生率都是即将到来的发生率的关键决定因素。以均方误差达到其最大最小值的方式确定最佳聚类，这几乎与划分为三分位数一致。这些聚类策略可用于在几周的准备时间内更准确地预测不祥的登革热发病率。