Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Pandemic model with data-driven phase detection, a study using COVID-19 data

Yuansan Liu, Saransh Srivastava, Zuo Huang, Felisa J. Vázquez-Abad

TL;DR

The paper introduces a data-driven, piecewise SEIDR framework that couples a two-group SIR-type ODE with statistical change detection to automatically identify phase transitions in COVID-19 dynamics. It defines a nine-parameter model, fits it to death data, and uses a vertical adjustment to account for under-reporting, validating the approach on UK and US data. Phase detection via CUSUM reveals regime changes aligned with major social and policy events, while a two-way detection and online, sliding-window extension enable robust, real-time phase tracking. The work provides actionable insights into which parameters drive shifts in spread, offers a mechanism to compare data quality across regions, and demonstrates a practical pathway for data-driven epidemiological reasoning during a pandemic.

Abstract

The recent COVID-19 pandemic has promoted vigorous scientific activity in an effort to understand, advice and control the pandemic. Data is now freely available at a staggering rate worldwide. Unfortunately, this unprecedented level of information contains a variety of data sources and formats, and the models do not always conform to the description of the data. Health officials have recognized the need for more accurate models that can adjust to sudden changes, such as produced by changes in behavior or social restrictions. In this work we formulate a model that fits a ``SIR''-type model concurrently with a statistical change detection test on the data. The result is a piece wise autonomous ordinary differential equation, whose parameters change at various points in time (automatically learned from the data). The main contributions of our model are: (a) providing interpretation of the parameters, (b) determining which parameters of the model are more important to produce changes in the spread of the disease, and (c) using data-driven discovery of sudden changes in the evolution of the pandemic. Together, these characteristics provide a new model that better describes the situation and thus, provides better quality of information for decision making.

Pandemic model with data-driven phase detection, a study using COVID-19 data

TL;DR

The paper introduces a data-driven, piecewise SEIDR framework that couples a two-group SIR-type ODE with statistical change detection to automatically identify phase transitions in COVID-19 dynamics. It defines a nine-parameter model, fits it to death data, and uses a vertical adjustment to account for under-reporting, validating the approach on UK and US data. Phase detection via CUSUM reveals regime changes aligned with major social and policy events, while a two-way detection and online, sliding-window extension enable robust, real-time phase tracking. The work provides actionable insights into which parameters drive shifts in spread, offers a mechanism to compare data quality across regions, and demonstrates a practical pathway for data-driven epidemiological reasoning during a pandemic.

Abstract

The recent COVID-19 pandemic has promoted vigorous scientific activity in an effort to understand, advice and control the pandemic. Data is now freely available at a staggering rate worldwide. Unfortunately, this unprecedented level of information contains a variety of data sources and formats, and the models do not always conform to the description of the data. Health officials have recognized the need for more accurate models that can adjust to sudden changes, such as produced by changes in behavior or social restrictions. In this work we formulate a model that fits a ``SIR''-type model concurrently with a statistical change detection test on the data. The result is a piece wise autonomous ordinary differential equation, whose parameters change at various points in time (automatically learned from the data). The main contributions of our model are: (a) providing interpretation of the parameters, (b) determining which parameters of the model are more important to produce changes in the spread of the disease, and (c) using data-driven discovery of sudden changes in the evolution of the pandemic. Together, these characteristics provide a new model that better describes the situation and thus, provides better quality of information for decision making.

Paper Structure

This paper contains 15 sections, 9 equations, 12 figures, 4 tables, 1 algorithm.

Table of Contents

  1. Introduction
  2. Literature Review
  3. SEIDR model
  4. Statistical Fitting
  5. Selection criteria for datasets
  6. Model fitting
  7. Validation of the SEIDR Model for COVID-19 Data
  8. Vertical Adjustment
  9. Discussion
  10. Phase detection
  11. Motivation
  12. Phase change detection on the COVID-19 data
  13. Two-way detection
  14. Data-driven Phase Detection
  15. Concluding Remarks

Figures (12)

  • Figure 1: The flowchart of model
  • Figure 2: The left plot is the predicted number of deaths of UK model, red dots are reported number of deaths used to fit the model, from 2020-04-01 to 2020-06-09; middle plot is validation of the prediction of UK model in 30 days (from 2020-06-10 to 2020-07-09); right shows the residuals of the projection.
  • Figure 3: The left plot is the predicted number of deaths of US model from 2020-04-11 to 2020-05-30; middle is validation in 30 days (from 2020-5-31 to 2020-6-29); right shows the residuals of the prediction.
  • Figure 4: Number of infections reported against data. Left is UK data from 2020-04-01; Right is USA data from 2020-5-31.
  • Figure 5: U.K., U.S. infection prediction with vertical adjustment
  • ...and 7 more figures