Thousands of deaths associated with air pollution each year could be prevented by forecasting the behavior of factors that pose risks to people’s health and their geographical distribution. Proximity to pollution sources, degree of urbanization, and population density are some of the factors whose spatial distribution enables the identification of possible influence on the presence of respiratory diseases (RD). Currently, Bogotá is among the cities with the poorest air quality in Latin America. Specifically, the locality of Kennedy is one of the zones in the city with the highest recorded concentration levels of local pollutants over the last 10 years. From 2009 to 2016, there were 8619 deaths associated with respiratory and cardiovascular diseases in the locality. Given these characteristics, this study set out to identify and analyze the areas in which the primary socioeconomic and environmental conditions contribute to the presence of symptoms associated with RD. To this end, information collected in field by performing georeferenced surveys was analyzed through geostatistical and machine learning tools which carried out cluster and pattern analyses. Random forests and AdaBoost were applied to establish hot spots where RD could occur, given the conjugation of predictor variables in the micro-territory. It was found that random forests outperformed AdaBoost with 0.63 AUC. In particular, this study’s approach applies to densely populated municipalities with high levels of air pollution. In using these tools, municipalities can anticipate environmental health situations and reduce the cost of respiratory disease treatments.

中文翻译：

---

空气质量对人体健康的影响分析：以波哥大肯尼迪市污染物浓度与呼吸系统疾病关系为例

通过预测对人们健康及其地理分布构成风险的因素的行为，可以避免每年与空气污染相关的数千人死亡。靠近污染源、城市化程度和人口密度是一些因素，其空间分布能够识别对呼吸系统疾病 (RD) 存在的可能影响。目前，波哥大是拉丁美洲空气质量最差的城市之一。具体来说，肯尼迪地区是该市过去 10 年中当地污染物浓度水平最高的地区之一。2009年至2016年，当地与呼吸系统和心血管疾病相关的死亡人数为8619人。鉴于这些特点，本研究旨在确定和分析主要社会经济和环境条件导致 RD 相关症状出现的领域。为此，通过执行地理参考调查在现场收集的信息通过地质统计和机器学习工具进行了分析，这些工具进行了聚类和模式分析。考虑到微区域中预测变量的共轭，随机森林和 AdaBoost 被应用于建立可能发生 RD 的热点。结果发现，随机森林的 AUC 为 0.63，优于 AdaBoost。特别是，本研究的方法适用于空气污染严重的人口稠密城市。通过使用这些工具，市政当局可以预测环境健康状况并降低呼吸系统疾病治疗的成本。