RAPID: Real-time Forecasting of COVID-19 risk in the USA

为了支持政府和个人做出与COVID-19大流行相关的决策，研究人员将开发一套流行病预测模型，以准确评估COVID-19在美国县、州和国家层面带来的风险。这些模型将建立在CSSE团队公开的COVID-19跟踪地图的流行病学数据基础上，沿着匿名移动的电话数据、人口和社会经济信息、气候和季节性因素以及各种健康和行为指标。建模框架将是灵活的，因此能够为各种政策需求和缓解战略提供决策支持。 该团队将共同努力，最大限度地提高模型对决策者的有用性，并确保将建模结果成功转化为有用的行动。在短期内，该团队生成的模型输出将通过COVID-19 Forecast Hub为CDC的COVID-19国家预测工作做出贡献。从长远来看，这项研究工作的系统工程方法将有助于建立一套强大的、经过审查的工具，可用于在下一次大流行之前和期间进行流行病预测。该项目还将支持研究生的培训。预测模型将利用经验机器学习方法，将不同的数据输入组合成一个有意义的预测模型，使用原始数据和内部生成的新指标作为输入。研究团队将探索，评估和比较不同统计方法的性能，以满足不同的建模目标，此外还将开发新技术以进一步提高预测能力，如集成方法和输入聚类。各种方法和研究目标的组合将被考虑和优化，以找到最佳的配对。该团队将共同努力，根据观察到的数据不断验证模型，并作为回应，继续完善模型，以提高预测的准确性并推断推动疫情爆发的最重要因素，从而提高我们对COVID-19传播风险的总体理解。 拟议的建模工作将同时建立在研究团队正在进行的数据收集工作的基础上，该工作支持JHU CSSE COVID-19仪表板和数据集，从而使团队能够进一步提高数据的质量，并改善数据集的沟通，文档和管理，已成为全球COVID-19病例和死亡数据的权威来源，是数十个研究团队进行的国家和地方COVID-19建模的基础，该奖项反映了NSF的法定使命，并通过使用基金会的知识价值和更广泛的影响审查标准进行评估，被认为值得支持。

In an effort to support decision making by governments and individuals related to the COVID-19 pandemic, the researchers will develop a set of epidemic forecast models to accurately assess the risk presented by COVID-19 in the United States at county, state, and national levels. The models will build on epidemiological data from the CSSE team’s publicly available COVID-19 tracking map, along with anonymized mobile phone data, demographic and socioeconomic information, climate and seasonality factors, and various health and behavioral metrics. The modeling framework will be flexible, and thus able to provide decision support for various policy needs and mitigation strategies. The team will make concerted efforts to maximize the model’s usefulness to decision-makers and ensure the successful translation of modeling outcomes into useful actions. In the short term, the model outputs generated by the team will contribute to the CDC’s COVID-19 national forecasting efforts through the COVID-19 Forecast Hub. In the long term, the systems engineering approach to this research effort will contribute to the establishment of a robust, vetted set of tools that can be used for epidemic forecasting, prior to and during the next pandemic. This project will also support the training of graduate students. The forecasting model will utilize an empirical machine learning approach that combines disparate data inputs into a meaningful predictive model using a combination of raw data and novel metrics generated in-house as inputs. The research team will explore, evaluate and compare the performance of different statistical methodologies for answering different proposed modeling objectives, in addition to developing new techniques to further improve predictive capabilities such as ensemble approaches and input clustering. Various combinations of methodologies and research objectives will be considered and optimized to find the best pairing. The team will make a concerted effort to continually validate the model based on observed data, and in response, continue to refine the model to both increase the accuracy of the predictions and infer the most important factors driving the outbreak, thus improving our general understanding of COVID-19 transmission risk. The proposed modeling effort will simultaneously build on the research team’s ongoing data collection effort that supports the JHU CSSE COVID-19 Dashboard and data set, and thus enable the team to further improve the quality of the data, as well as improve the communication, documentation, and management of the dataset, which has become the authoritative source of COVID-19 case and death data globally serves as the foundation for national and local level COVID-19 modeling conducted by dozens of research teams, governmental organizations and public health agencies around the world.This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.