skip to main content
Explore Scholarly Publications and Datasets in the NSF-PAR

Title: Inter-Series Attention Model for COVID-19 Forecasting
COVID-19 pandemic has an unprecedented impact all over the world since early 2020. During this public health crisis, reliable forecasting of the disease becomes critical for resource allocation and administrative planning. The results from compartmental models such as SIR and SEIR are popularly referred by CDC and news media. With more and more COVID-19 data becoming available, we examine the following question: Can a direct data-driven approach without modeling the disease spreading dynamics outperform the well referred compartmental models and their variants? In this paper, we show the possibility. It is observed that as COVID-19 spreads at different speed and scale in different geographic regions, it is highly likely that similar progression patterns are shared among these regions within different time periods. This intuition lead us to develop a new neural forecasting model, called Attention Crossing Time Series (ACTS), that makes forecasts via comparing patterns across time series obtained from multiple regions. The attention mechanism originally developed for natural language processing can be leveraged and generalized to materialize this idea. Among 13 out of 18 testings including forecasting newly con rmed cases, hospitalizations and deaths, ACTS outperforms all the leading COVID-19 forecasters highlighted by CDC.
Authors:
; ;
Award ID(s):
2029626
Publication Date:
NSF-PAR ID:
10232366
Journal Name:
SIAM Data Mining
Sponsoring Org:
National Science Foundation
More Like this
  1. ; ; ; ; ; ( , ArXivorg)
    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-19more »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.« less
  2. ; ; ( , BMC Infectious Diseases)
    Abstract Background The COVID-19 outbreak in Wuhan started in December 2019 and was under control by the end of March 2020 with a total of 50,006 confirmed cases by the implementation of a series of nonpharmaceutical interventions (NPIs) including unprecedented lockdown of the city. This study analyzes the complete outbreak data from Wuhan, assesses the impact of these public health interventions, and estimates the asymptomatic, undetected and total cases for the COVID-19 outbreak in Wuhan. Methods By taking different stages of the outbreak into account, we developed a time-dependent compartmental model to describe the dynamics of disease transmission and casemore »detection and reporting. Model coefficients were parameterized by using the reported cases and following key events and escalated control strategies. Then the model was used to calibrate the complete outbreak data by using the Monte Carlo Markov Chain (MCMC) method. Finally we used the model to estimate asymptomatic and undetected cases and approximate the overall antibody prevalence level. Results We found that the transmission rate between Jan 24 and Feb 1, 2020, was twice as large as that before the lockdown on Jan 23 and 67.6 % (95% CI [0.584,0.759]) of detectable infections occurred during this period. Based on the reported estimates that around 20% of infections were asymptomatic and their transmission ability was about 70% of symptomatic ones, we estimated that there were about 14,448 asymptomatic and undetected cases (95% CI [12,364,23,254]), which yields an estimate of a total of 64,454 infected cases (95% CI [62,370,73,260]), and the overall antibody prevalence level in the population of Wuhan was 0.745% (95% CI [0.693 % ,0.814 % ]) by March 31, 2020. Conclusions We conclude that the control of the COVID-19 outbreak in Wuhan was achieved via the enforcement of a combination of multiple NPIs: the lockdown on Jan 23, the stay-at-home order on Feb 2, the massive isolation of all symptomatic individuals via newly constructed special shelter hospitals on Feb 6, and the large scale screening process on Feb 18. Our results indicate that the population in Wuhan is far away from establishing herd immunity and provide insights for other affected countries and regions in designing control strategies and planing vaccination programs.« less
  3. ; ; ; ; ; ; ; ; ; ; et al ( , medRxiv)
    The COVID-19 pandemic brought to the forefront an unprecedented need for experts, as well as citizens, to visualize spatio-temporal disease surveillance data. Web application dashboards were quickly developed to fill this gap, including those built by JHU, WHO, and CDC, but all of these dashboards supported a particular niche view of the pandemic (ie, current status or specific regions). In this paper1, we describe our work developing our own COVID-19 Surveillance Dashboard, available at https://nssac.bii.virginia.edu/covid19/dashboard/, which offers a universal view of the pandemic while also allowing users to focus on the details that interest them. From the beginning, our goalmore »was to provide a simple visual way to compare, organize, and track near-real-time surveillance data as the pandemic progresses. Our dashboard includes a number of advanced features for zooming, filtering, categorizing and visualizing multiple time series on a single canvas. In developing this dashboard, we have also identified 6 key metrics we call the 6Cs standard which we propose as a standard for the design and evaluation of real-time epidemic science dashboards. Our dashboard was one of the first released to the public, and remains one of the most visited and highly used. Our group uses it to support federal, state and local public health authorities, and it is used by people worldwide to track the pandemic evolution, build their own dashboards, and support their organizations as they plan their responses to the pandemic. We illustrate the utility of our dashboard by describing how it can be used to support data story-telling – an important emerging area in data science.« less
  4. ; ( , Transportation Research Record: Journal of the Transportation Research Board)
    Social media can be a significant tool for transportation and transit agencies providing passengers with real-time information on traffic events. Moreover, COVID-19 and other limitations have compelled the agencies to engage with travelers online to promote public knowledge about COVID-related issues. It is, therefore, important to understand the agencies’ communication patterns. In this original study, the Twitter communication patterns of different transportation actors—types of message, communication sufficiency, consistency, and coordination—were examined using a social media data-driven approach applying text mining techniques and dynamic network analysis. A total of 850,000 tweets from 395 different transportation and transit agencies, starting in 2018more »and the periods before, during and after the pandemic, were studied. Transportation agencies (federal, state, and city) were found to be less active on Twitter and mostly discussed safety measures, project management, and so forth. By contrast, the transit agencies (local bus and light, heavy, and commuter rail) were more active on Twitter and shared information about crashes, schedule information, passenger services, and so forth. Moreover, transportation agencies shared minimal pandemic safety information than transit agencies. Dynamic network analysis reveals interaction patterns among different transportation actors that are poorly connected and coordinated among themselves and with different health agencies (e.g., Centers for Disease Control and Prevention [CDC] and the Federal Emergency Management Agency [FEMA]). The outcome of this study provides understanding to improve existing communication plans, critical information dissemination efficacy, and the coordination of different transportation actors in general and during unprecedented health crises.« less
  5. ; ; ; ; ; ; ; ; ; ; et al ( , PLOS ONE)
    Adrish, Muhammad (Ed.)
    Mexico has experienced one of the highest COVID-19 mortality rates in the world. A delayed implementation of social distancing interventions in late March 2020 and a phased reopening of the country in June 2020 has facilitated sustained disease transmission in the region. In this study we systematically generate and compare 30-day ahead forecasts using previously validated growth models based on mortality trends from the Institute for Health Metrics and Evaluation for Mexico and Mexico City in near real-time. Moreover, we estimate reproduction numbers for SARS-CoV-2 based on the methods that rely on genomic data as well as case incidence data.more »Subsequently, functional data analysis techniques are utilized to analyze the shapes of COVID-19 growth rate curves at the state level to characterize the spatiotemporal transmission patterns of SARS-CoV-2. The early estimates of the reproduction number for Mexico were estimated between R t ~1.1–1.3 from the genomic and case incidence data. Moreover, the mean estimate of R t has fluctuated around ~1.0 from late July till end of September 2020. The spatial analysis characterizes the state-level dynamics of COVID-19 into four groups with distinct epidemic trajectories based on epidemic growth rates. Our results show that the sequential mortality forecasts from the GLM and Richards model predict a downward trend in the number of deaths for all thirteen forecast periods for Mexico and Mexico City. However, the sub-epidemic and IHME models perform better predicting a more realistic stable trajectory of COVID-19 mortality trends for the last three forecast periods (09/21-10/21, 09/28-10/27, 09/28-10/27) for Mexico and Mexico City. Our findings indicate that phenomenological models are useful tools for short-term epidemic forecasting albeit forecasts need to be interpreted with caution given the dynamic implementation and lifting of social distancing measures.« less
  • Citation Formats
  • MLA
  • APA
  • Chicago
  • BibTeX
  • Save / Share this Record
  • Send to Email