Artificial Intelligence Solutions for Reliable Epidemic Forecasting
Infectious diseases, such as seasonal influenza, Zika, Ebola, and the ongoing COVID-19, can be spread, directly or indirectly, from one person to another leading to an outbreak, an epidemic, or a pandemic. Infectious diseases place a heavy social and economic burden on our society. Producing timely, well-informed, and reliable spatio-temporal forecasts of the epidemic dynamics can help inform policymakers on how to provision limited healthcare resources, develop effective interventions, rapidly control outbreaks, and ensure the safety of the general public. Traditional approaches are mainly based on theory-based mechanistic models such as an agent-based SEIR model and statistical time series models like autoregressive models. Recent advances in deep learning have significantly improved the state of the art in computer vision, natural language processing, and many other fields. Although deep learning-based predictive models have gained increased prominence in epidemic forecasting, they are far from being well explored. One challenge is the lack of sufficiently good quality of training data, particularly during new emerging epidemics. Another challenge is that they rarely consider epidemiological context as prior. Models in both cases are prone to be overfitting and are unlikely to provide explanatory power for the underlying phenomena due to the black box nature. Furthermore, a new challenge that is often overlooked in the academic setting involves the large variability that can exist across multiple training runs of the same model, so called inconsistency. Inconsistency behavior can result in severe consequences in AI systems. Given the challenges, my research focuses on deep learning-based methods that incorporate theory-based causal models, spatio-temporal features, multi-source data analysis, and ensembling techniques for a better understanding of disease spreading and improving forecasting accuracy, explainability, and consistency. Through interdisciplinary collaborative efforts, the key contributions of my Ph.D. research are 1) improving epidemic forecasting accuracy by proposing graph neural network-based (GNN) frameworks that consider temporal and spatial signals, 2) improving explainability and accuracy of deep learning-based forecasting models by combining deep learning models with theory-based causal models to incorporate epidemiological context, and 3) improving consistency and accuracy of deep learning-based forecasting models using new ensemble algorithms with theoretical proofs and empirical demonstrations. Through experiments on forecasting influenza and COVID-19 dynamics, we demonstrate that the proposed methods can boost model performance compared with a broad range of baselines.
- Anil Vullikanti, Committee Chair, (CS/UVA)
- Madhav Marathe, Co-Advisor, (CS/UVA)
- Jiangzhuo Chen, Co-Advisor, (NSSAC/BI/UVA)
- Jundong Li (CS, ECE/SEAS, SDS/UVA)
- Stephen Eubank, (PHS/GSAS/UVA)
- Adam Sadilek (Google)