skip to main content


The NSF Public Access Repository (NSF-PAR) system and access will be unavailable from 11:00 PM ET on Thursday, May 23 until 2:00 AM ET on Friday, May 24 due to maintenance. We apologize for the inconvenience.

This content will become publicly available on June 27, 2024

Title: Phase-Informed Bayesian Ensemble Models Improve Performance of COVID-19 Forecasts

Despite hundreds of methods published in the literature, forecasting epidemic dynamics remains challenging yet important. The challenges stem from multiple sources, including: the need for timely data, co-evolution of epidemic dynamics with behavioral and immunological adaptations, and the evolution of new pathogen strains. The ongoing COVID-19 pandemic highlighted these challenges; in an important article, Reich et al. did a comprehensive analysis highlighting many of these challenges.In this paper, we take another step in critically evaluating existing epidemic forecasting methods. Our methods are based on a simple yet crucial observation - epidemic dynamics go through a number of phases (waves). Armed with this understanding, we propose a modification to our deployed Bayesian ensembling case time series forecasting framework. We show that ensembling methods employing the phase information and using different weighting schemes for each phase can produce improved forecasts. We evaluate our proposed method with both the currently deployed model and the COVID-19 forecasthub models. The overall performance of the proposed model is consistent across the pandemic but more importantly, it is ranked third and first during two critical rapid growth phases in cases, regimes where the performance of most models from the CDC forecasting hub dropped significantly.

more » « less
Award ID(s):
1918656 2142997 1633028 2028004 2027541
Author(s) / Creator(s):
; ; ; ; ; ; ;
Publisher / Repository:
Date Published:
Journal Name:
Proceedings of the AAAI Conference on Artificial Intelligence
Page Range / eLocation ID:
15647 to 15653
Medium: X
Sponsoring Org:
National Science Foundation
More Like this
  1. Real-time forecasting of non-stationary time series is a challenging problem, especially when the time series evolves rapidly. For such cases, it has been observed that ensemble models consisting of a diverse set of model classes can perform consistently better than individual models. In order to account for the nonstationarity of the data and the lack of availability of training examples, the models are retrained in real-time using the most recent observed data samples. Motivated by the robust performance properties of ensemble models, we developed a Bayesian model averaging ensemble technique consisting of statistical, deep learning, and compartmental models for fore-casting epidemiological signals, specifically, COVID-19 signals. We observed the epidemic dynamics go through several phases (waves). In our ensemble model, we observed that different model classes performed differently during the various phases. Armed with this understanding, in this paper, we propose a modification to the ensembling method to employ this phase information and use different weighting schemes for each phase to produce improved forecasts. However, predicting the phases of such time series is a significant challenge, especially when behavioral and immunological adaptations govern the evolution of the time series. We explore multiple datasets that can serve as leading indicators of trend changes and employ transfer entropy techniques to capture the relevant indicator. We propose a phase prediction algorithm to estimate the phases using the leading indicators. Using the knowledge of the estimated phase, we selectively sample the training data from similar phases. We evaluate our proposed methodology on our currently deployed COVID-19 forecasting model and the COVID-19 ForecastHub models. The overall performance of the proposed model is consistent across the pandemic. More importantly, it is ranked second during two critical rapid growth phases in cases, regimes where the performance of most models from the ForecastHub dropped significantly. 
    more » « less
  2. The COVID-19 pandemic represents the most significant public health disaster since the 1918 influenza pandemic. During pandemics such as COVID-19, timely and reliable spatiotemporal forecasting of epidemic dynamics is crucial. Deep learning-based time series models for forecasting have recently gained popularity and have been successfully used for epidemic forecasting. Here we focus on the design and analysis of deep learning-based models for COVID-19 forecasting. We implement multiple recurrent neural network-based deep learning models and combine them using the stacking ensemble technique. In order to incorporate the effects of multiple factors in COVID-19 spread, we consider multiple sources such as COVID-19 confirmed and death case count data and testing data for better predictions. To overcome the sparsity of training data and to address the dynamic correlation of the disease, we propose clustering-based training for high-resolution forecasting. The methods help us to identify the similar trends of certain groups of regions due to various spatio-temporal effects. We examine the proposed method for forecasting weekly COVID-19 new confirmed cases at county-, state-, and country-level. A comprehensive comparison between different time series models in COVID-19 context is conducted and analyzed. The results show that simple deep learning models can achieve comparable or better performance when compared with more complicated models. We are currently integrating our methods as a part of our weekly forecasts that we provide state and federal authorities. 
    more » « less
  3. Abstract Human mobility plays an important role in the dynamics of infectious disease spread. Evidence from the initial nationwide lockdowns for COVID− 19 indicates that restricting human mobility is an effective strategy to contain the spread. While a direct correlation was observed early on, it is not known how mobility impacted COVID− 19 infection growth rates once lockdowns are lifted, primarily due to modulation by other factors such as face masks, social distancing, and the non-linear patterns of both mobility and infection growth. This paper introduces a piece-wise approach to better explore the phase-wise association between state-level COVID− 19 incidence data and anonymized mobile phone data for various states in the United States. Prior literature analyzed the linear correlation between mobility and the number of cases during the early stages of the pandemic. However, it is important to capture the non-linear dynamics of case growth and mobility to be usable for both tracking and forecasting COVID− 19 infections, which is accomplished by the piece-wise approach. The associations between mobility and case growth rate varied widely for various phases of the epidemic curve when the stay-at-home orders were lifted. The mobility growth patterns had a strong positive association of 0.7 with the growth in the number of cases, with a lag of 5 to 7 weeks, for the fast-growth phase of the pandemic, for only 20 states that had a peak between July 1st and September 30, 2020. Overall though, mobility cannot be used to predict the rise in the number of cases after initial lockdowns have been lifted. Our analysis explores the gradual diminishing value of mobility associations in the later stage of the outbreak. Our analysis indicates that the relationship between mobility and the increase in the number of cases, once lockdowns have been lifted, is tenuous at best and there is no strong relationship between these signals. But we identify the remnants of the last associations in specific phases of the growth curve. 
    more » « less
  4. It is increasingly common to encounter prediction tasks in the biomedical sciences for which multiple datasets are available for model training. Common approaches such as pooling datasets before model fitting can produce poor out‐of‐study prediction performance when datasets are heterogeneous. Theoretical and applied work has shownmultistudy ensemblingto be a viable alternative that leverages the variability across datasets in a manner that promotes model generalizability. Multistudy ensembling uses a two‐stagestackingstrategy which fits study‐specific models and estimates ensemble weights separately. This approach ignores, however, the ensemble properties at the model‐fitting stage, potentially resulting in performance losses. Motivated by challenges in the estimation of COVID‐attributable mortality, we proposeoptimal ensemble construction, an approach to multistudy stacking whereby we jointly estimate ensemble weights and parameters associated with study‐specific models. We prove that limiting cases of our approach yield existing methods such as multistudy stacking and pooling datasets before model fitting. We propose an efficient block coordinate descent algorithm to optimize the loss function. We use our method to perform multicountry COVID‐19 baseline mortality prediction. We show that when little data is available for a country before the onset of the pandemic, leveraging data from other countries can substantially improve prediction accuracy. We further compare and characterize the method's performance in data‐driven simulations and other numerical experiments. Our method remains competitive with or outperforms multistudy stacking and other earlier methods in the COVID‐19 data application and in a range of simulation settings.

    more » « less
  5. Abstract Background

    No versatile web app exists that allows epidemiologists and managers around the world to comprehensively analyze the impacts of COVID-19 mitigation. The app presented here fills this gap.


    Our web app uses a model that explicitly identifies susceptible, contact, latent, asymptomatic, symptomatic and recovered classes of individuals, and a parallel set of response classes, subject to lower pathogen-contact rates. The user inputs a CSV file of incidence and, if of interest, mortality rate data. A default set of parameters is available that can be overwritten through input or online entry, and a user-selected subset of these can be fitted to the model using maximum-likelihood estimation (MLE). Model fitting and forecasting intervals are specifiable and changes to parameters allow counterfactual and forecasting scenarios. Confidence or credible intervals can be generated using stochastic simulations, based on MLE values, or on an inputted CSV file containing Markov chain Monte Carlo (MCMC) estimates of one or more parameters.


    We illustrate the use of our web app in extracting social distancing, social relaxation, surveillance or virulence switching functions (i.e., time varying drivers) from the incidence and mortality rates of COVID-19 epidemics in Israel, South Africa, and England. The Israeli outbreak exhibits four distinct phases: initial outbreak, social distancing, social relaxation, and a second wave mitigation phase. An MCMC projection of this latter phase suggests the Israeli epidemic will continue to produce into late November an average of around 1500 new case per day, unless the population practices social-relaxation measures at least 5-fold below the level in August, which itself is 4-fold below the level at the start of July. Our analysis of the relatively late South African outbreak that became the world’s fifth largest COVID-19 epidemic in July revealed that the decline through late July and early August was characterised by a social distancing driver operating at more than twice the per-capita applicable-disease-class (pc-adc) rate of the social relaxation driver. Our analysis of the relatively early English outbreak, identified a more than 2-fold improvement in surveillance over the course of the epidemic. It also identified a pc-adc social distancing rate in early August that, though nearly four times the pc-adc social relaxation rate, appeared to barely contain a second wave that would break out if social distancing was further relaxed.


    Our web app provides policy makers and health officers who have no epidemiological modelling or computer coding expertise with an invaluable tool for assessing the impacts of different outbreak mitigation policies and measures. This includes an ability to generate an epidemic-suppression or curve-flattening index that measures the intensity with which behavioural responses suppress or flatten the epidemic curve in the region under consideration.

    more » « less