大数据驱动的气象环境健康风险评价研究

The impact of changes in the meteorological environment on health is increasingly valued by the public. However, because of the differences in geographical areas, data, and research methods, the study of meteorological environmental health risk assessment has always been limited. In recent years, the extensive application of big data technology has provided powerful theoretical and technical support for solving this interdisciplinary issue. In this project, firstly, we use web crawlers and other acquisition methods to collect multi-source heterogeneous data related to meteorological environment and health, and then use clustering, inversion, and other technologies to preprocess the data and store it on the Hadoop-based big data platform. Second, through intelligent algorithms such as deep learning, text mining, and sentiment analysis, as well as mature computing models, human influences are reduced as much as possible, and the impact of meteorological environmental changes on public health is explored in depth from multiple perspectives based on data. Finally, through the analysis of big data on the meteorological environment and health, a big data-driven risk assessment model was constructed to explore the extent of damage caused by changes in the meteorological environment to public health, and the results of the research were displayed using technologies such as GIS and mobile internet development. The research results will further deepen the understanding of the impact of meteorological environment on health, and provide a theoretical basis for the later use of the big data platform to provide services for the public and reduce public health losses.

气象环境的变化对健康的影响越来越受到公众的重视，然而由于地域、数据和研究方法的差异，气象环境健康风险评价的研究始终带有局限性。近年来，大数据技术的广泛应用，对解决该交叉学科问题提供了有力的理论及技术支持。本项目中，首先利用网络爬虫等采集方法收集与气象环境、健康相关的多源异构数据，进而使用聚类、反演法等技术进行数据的预处理并存储在基于Hadoop的大数据平台上；其次通过深度学习、文本挖掘、情感分析等智能算法以及成熟的运算模型，尽可能的减少人为的影响，基于数据从多角度来深入的探讨气象环境变化对公众健康的影响；最后，通过对气象环境与健康的大数据分析，进而构建大数据驱动的风险评价模型，探究气象环境的变化对公众健康造成的损失程度，并用GIS、移动互联网开发等技术将研究结果进行展示。研究成果将进一步加深气象环境对健康影响的认识，为后期利用该大数据平台为公众提供服务、减少公众健康损失提供理论基础。

本项目《大数据驱动的气象环境健康风险评价研究》以研究气象环境的变化对易感人群发病风险为主要内容，第一，完成了研究区域中13个城市的气象环境数据和医疗数据的预处理及整合，形成数据驱动的基础库；第二，完成数据驱动的气象环境发病风险的研究。利用基础库的数据为驱动对象，通过分布非线性滞后模型（DLNM）结合广义相加模型（GLM）对气象环境指标和天气敏感性疾病的暴露风险进行分析研究，发现在不同的地区高低温、24小时温差和大气污染物等因素对相关疾病的发病风险均存在不同的滞后效应，同时还对不同地域的相同指标，同一地域的不同气象环境指标的发病风险进行比较研究；第三，完成了对天气敏感性疾病的预测。利用面板平滑转换回归（PSTR）模型和NeuralProphet模型在预测确诊病例和住院病例数高准确度的预测。第四，完成了对我过大气污染治理政策的文本分析，对政策文本的主题、关键词、效力等内容进行研究分析。公开发表论文14篇，专著1部，4项软件著作权。研究中选择的气象环境暴露指标包括表观温度、气温日较差，大气污染物浓度等，这些指标能更加真实客观地反映人体的感觉，既丰富了分布滞后非线性模型在气象环境与心血管疾病方面的研究，也为当地居民和公共卫生部门制定防治措施提供依据。使用预测模型对心血管疾病住院人数进行预测，并将体感温度作为外生变量纳入模型进行研究，丰富了疾病预测模型的研究，并为农村地区医疗机构的人力资源优化以及地区医保资金的配置有一定的参考价值。