基于结构化大数据深度挖掘的呼吸道症候群监测与早期预警机制研究

The existing disease surveillance systems, including the Chinese Centers for Disease Control and Prevention (Chinese CDC), rely on diagnosis results provided by hospitals or medical labs, which results in weeks of lag in case reporting. Our previous research in pre-diagnosed medical data suggested that the actual disease infection trend was strongly related with syndromic indicators like Internet search frequency; therefore, we propose to apply data mining approaches, especial the Deep Learning methods, to analyze respiratory syndromic data for real-time disease surveillance. This project plans to start from theoretic research in Deep Data Mining. After analyzing the multisource feature structural, the multiscaling space-time structure and the multilayer heterogeneous data structure of the syndromic surveillance data, the proposed project will build the syndromic indicator framework for the purpose of respiratory infectious disease surveillance. We will further study the surveillance models to predict actual disease infections in different districts and cities, and ultimately build the syndromic surveillance platform for monitoring respiratory infectious diseases. By providing data and application interfaces to the Chinese CDC, the respiratory syndromic surveillance platform will become an important mechanism for national infectious disease control and prevention. It will provide scientific and technical foundation for analyzing epidemic stages, detecting outbreaks and developing response strategies in early stages of potential epidemics.

针对现有传染病直报系统"监测结果"滞后于"传染病发病"的缺陷，课题组前期开展了"确诊前"呼吸道症候群数据挖掘研究。研究发现，甲型N1N1流感等呼吸道传染病的发病趋势与特定"呼吸道症候群指标"的变化趋势具有明显的相关性。本课题据此提出，以禽流感等法定呼吸道传染病疫情实时监测为目标，以结构化症候群大数据为研究对象，探索"深度挖掘"理论，分析症候群数据的多源特征结构、多尺度时空结构与层次化特异性结构，规划症候群监测指标体系，建立预测传染病发病率与疫情爆发风险的关键理论与算法，最终搭建呼吸道症候群实时监测网络平台，实现与政府疾控部门在数据层、应用层的对接。课题涉及海量异构数据特征分析、大数据深度挖掘模型等关键科学问题。研究成果将成为国家传染病防控机制中的重要环节，为我国非典、H7N9禽流感等法定呼吸道传染病疫情分析、早期预警和防控策略的制定提供科学支撑和技术服务，具有重要的科学和社会意义。

针对现有传染病直报系统"监测结果"滞后于"传染病发病"的缺陷，课题组开展了"确诊前"呼吸道症候群数据挖掘研究。研究发现，甲型N1N1流感等呼吸道传染病的发病趋势与特定"呼吸道症候群指标"的变化趋势具有明显的相关性。以禽流感等法定呼吸道传染病疫情实时监测为目标，以结构化症候群大数据为研究对象，基于机器学习理论，分析症候群数据的多源特征结构、多元摩尔科夫多尺度时空结构与层次化特异性结构，建立了预测传染病发病率与疫情爆发风险的理论与算法，搭建了DiseaseTrends.org呼吸道症候群实时监测网络平台。课题涉及海量异构数据特征分析、大数据深度挖掘模型等关键科学问题。研究成果将成为国家传染病防控机制中的重要环节，为我国非典、H7N9禽流感等法定呼吸道传染病疫情分析、早期预警和防控策略的制定提供科学支撑和技术服务，具有较为重要的科学和社会意义。