Epidemic Forecasting Framework Combining Agent-Based Models and Smart Beam Particle Filtering

Abstract

Over the past decades, numerous techniques have been developed to forecast the temporal evolution of epidemic outbreaks. This paper proposes an approach that combines high resolution agent-based models using realistic social contact networks for simulating epidemic evolution with a particle filter based method for assimilation based forecasting. Agent-based modeling using realistic social contact networks provides two key advantages: (i) they capture the causal processes underlying the epidemic and hence are useful to understand the role of interventions on the course of the epidemics -typically time series models cannot capture this and as a result often do not perform well in such situations; (ii) they provide detailed forecast information -this allows us to produce forecast at high levels of temporal, spatial and social granularity. We also propose a new variation of particle filter technique called beam search particle filtering. The modification allows us to more efficiently search the parameter space which is necessitated by the fact that agentbased techniques are computationally expensive.We illustrate our methodology on the synthetic dataset of Ebola provided as a part of the NSF/NIH Ebola forecasting challenge. Our results show the efficacy of the proposed approach and suggest that agent-based causal models can be combined with filtering techniques to yield a new class of assimilation models for infectious disease forecasting.

Towards Real-Time Crowd Simulation Under Uncertainty Using an Agent-Based Model and an Unscented Kalman Filter

“…This paper has developed an Unscented Kalman Filter (UKF) that can be used to perform data assimilation on an agent-based model. Although previous efforts have used particle filters for this task [18,13,8,9,6] and variants of the Kalman filter for simpler models [19], this is the first time that an unscented Kalman filter has been used to optimise an agent-based model that includes heterogeneous, interacting agents. Importantly, the UKF is able to predict the locations of the agents with relatively little information about the crowd.…”

“…Examples of these approaches are often developed under the banner of 'Data-Driven Agent-Based Modelling' (DDABM), which itself emerged from a broader work in data-driven application systems [2]. A number of recent attempts have been made to allow agent-based models to react to new data [12,18,20,11,7,10,19,13,8,6]. However, whilst promising these applications all exhibit a number of limitations that this work will begin to address.…”

Mathematical Models for COVID-19 Pandemic: A Comparative Analysis

“…This includes: (i) social media data, (ii) weather data, (iii) incidence curves and (iv) demographic data. Causal models can be used for epidemic forecasting in a natural manner 30,3,37,32,38,39 . These models calibrate the internal model parameters using the disease incidence data seen until a given day and then execute the model forward in time to produce the future time series.…”

“…The choice of the models depends on the specific question at hand and the computational and data resource constraints. One of the key ideas in forecasting is to develop ensemble models-models that combine forecasts from multiple models 40,6,38,39 . The idea which originated in the domain of weather forecasting has found methodological advances in the machine learning literature.…”

Applying particle filtering in both aggregated and age-structured population compartmental models of pre-vaccination measles

“…57 This paper seeks to support more accurate estimation and prediction of measles 58 dynamics by applying a computational statistics technique that combines the best 59 features of insights from ongoing (although noisy) empirical data and dynamic models 60 (although fraught by systematic errors, omissions, and stochastic divergence over time) 61 while mitigating important weaknesses of each. The use of sequential Monte Carlo 62 methods in the form of particle filtering [16][17][18][19][20][21][22][23][24][25][26][27][28] has provided an effective and versatile 63 approach to solving this problem in other infectious diseases, such as the influenza. 64 Specifically, this paper investigates the combination of particle filtering methods with a 65 compartmental model (SEIR model) of measles to recurrently estimate the latent state 66 of the population with respect to the natural history of infection with measles, to 67 anticipate measles evolution and outbreak transitions in pre-vaccination era.…”