Dengue is a vector-borne disease transmitted by the Aedes genus mosquito. It causes financial burdens on public health systems and considerable morbidity and mortality. Tropical regions in the Americas and Asia are the areas most affected by the virus. Fortaleza is a city with approximately 2.6 million inhabitants in northeastern Brazil that, during the recent decades, has been suffering from endemic dengue transmission, interspersed with larger epidemics. The objective of this paper is to study the impact of human mobility in urban areas on the spread of the dengue virus, and to test whether human mobility data can be used to improve predictions of dengue virus transmission at the neighbourhood level. We present two distinct forecasting systems for dengue transmission in Fortaleza: the first using artificial neural network methods and the second developed using a mechanistic model of disease transmission. We then present enhanced versions of the two forecasting systems that incorporate bus transportation data cataloguing movement among 119 neighbourhoods in Fortaleza. Each forecasting system was used to perform retrospective forecasts for historical dengue outbreaks from 2007 to 2015. Results show that both artificial neural networks and mechanistic models can accurately forecast dengue cases, and that the inclusion of human mobility data substantially improves the performance of both forecasting systems. While the mechanistic models perform better in capturing seasons with large-scale outbreaks, the neural networks more accurately forecast outbreak peak timing, peak intensity and annual dengue time series. These results have two practical implications: they support the creation of public policies from the use of the models created here to combat the disease and help to understand the impact of urban mobility on the epidemic in large cities.

中文翻译：

---

利用城市地区的人员流动性预测社区层面的登革热疫情

登革热是一种由伊蚊属蚊子传播的媒介传播疾病。它给公共卫生系统带来经济负担，并导致相当大的发病率和死亡率。美洲和亚洲的热带地区是受病毒影响最严重的地区。福塔莱萨是巴西东北部一座拥有约 260 万居民的城市，近几十年来，该城市一直遭受地方性登革热传播，并夹杂着更大的流行病。本文的目的是研究城市地区人口流动对登革热病毒传播的影响，并测试人类流动数据是否可用于改进对社区层面登革热病毒传播的预测。我们提出了两种不同的登革热传播预测系统：第一个使用人工神经网络方法，第二个使用疾病传播的机械模型开发。然后，我们展示了两个预测系统的增强版本，其中包含福塔莱萨 119 个社区之间的公共汽车交通数据编目运动。每个预测系统都用于对 2007 年至 2015 年的历史登革热疫情进行回顾性预测。结果表明，人工神经网络和机械模型都可以准确预测登革热病例，并且包含人类流动性数据大大提高了两个预测系统的性能. 虽然机械模型在捕捉大规模暴发季节方面表现更好，但神经网络更准确地预测暴发高峰时间、高峰强度和年度登革热时间序列。