SIR Model-based Prediction of Infected Population of Coronavirus in Hubei Province

Jun Zhang, Lihong Wang, Ji WangarXiv 2020 · DOI · arXiv

Abstract

After the sudden outbreak of Coronavirus in Wuhan, continuous and rich data of the epidemic has been made public as the vital fact for decision support in control measures and aggressive implementation of containment strategies and plans. With the further growth and spreading of the virus, future resource allocation and planning under updated strategies and measures rely on careful study of the epidemic data and characteristics for accurate prediction and estimation. By using the SIR model and reported data, key parameters are obtained from least sum of squared errors for an accurate prediction of epidemic trend in the last four weeks.

Introduction

In the early weeks of the COVID-19 pandemic, uncertainty was the primary challenge for public health officials. Governments and healthcare providers needed to know when the "peak" of infections would arrive to effectively allocate medical resources, plan lockdowns, and manage public expectations. This paper provides a retrospective and predictive analysis of the COVID-19 outbreak in Hubei Province using the classic SIR (Susceptible-Infected-Recovered) epidemiological model. By fitting the model to early reported data through non-linear optimization, the study successfully identified the "inflection point" of the epidemic, demonstrating how simple mathematical frameworks can provide critical decision support during a public health crisis.

Methodology

The researchers utilized the SIR Model, which divides a population into three compartments:

  1. Susceptible (S): Individuals at risk of contracting the virus.
  2. Infected (I): Individuals currently carrying and spreading the virus.
  3. Recovered/Removed (R): Individuals who have recovered or died, and are no longer spreading the disease.

Technical Approach

Because the full SIR equations are non-linear and coupled, they are difficult to solve analytically. The authors first used an exponential growth approximation ($I(t) = e^{(\beta-\gamma)t}$) for the early outbreak phase to establish baseline values for the transmission rate ($\beta$) and the recovery/removal rate ($\gamma$).

They then employed the Runge-Kutta method to solve the differential equations of the full SIR model numerically. To optimize these parameters, the Least Sum of Squared Errors (SSE) principle was applied, searching for $\beta$ and $\gamma$ values that minimized the difference between the model's predicted "Infected" count and the actual reported "confirmed cases" from Hubei. The model was re-evaluated using different time windows (7, 14, 21, and 28 days) to track how parameters evolved as the epidemic progressed and interventions were implemented.

Datasets

The study relied on official daily epidemic reports for Hubei Province provided by the Hubei Provincial Health Commission and the National Health Commission of China. The dataset began on January 24, 2020, and extended through late February 2020, focusing on the daily counts of "currently confirmed cases" (active infections).

Key Results

A primary finding of the study was the model's accuracy in identifying the epidemic's inflection point. As the actual data approached the peak in mid-February, the model’s prediction for the peak date stabilized. By February 14, the model consistently pointed to a peak on February 17, which was remarkably close to the actual peak observed on February 18.

Figure 4: SIR model prediction vs. actual confirmed cases for Hubei Province using a 28-day data window, showing the full trajectory and inflection point.

For the 28-day data window, the optimized parameters were found to be $\beta = 5.3130$ and $\gamma = 5.0839$. The visual correlation between the model's generated curves and the actual reported cases is high, validating the robustness of the fit.

Figure 5: Detailed comparison table of predicted versus actual confirmed cases in Hubei Province, highlighting the model's precision near the peak.

The method proved highly effective at short-term forecasting, particularly for the three-day window following the data sample. Furthermore, the authors noted that once the inflection point is passed, subsequent data does not significantly change the identified location of the peak, suggesting the model is highly reliable for identifying the "turn" in an epidemic.

Discussion and Conclusion

The values of $\beta$ and $\gamma$ were not static; they served as social indicators, shifting over time to reflect the "composite effect" of human intervention, such as the lockdown of Wuhan and improved medical response. While more complex models like SEIR (which includes an "Exposed" period) exist, the authors argue that a well-tuned SIR model is robust enough for rapid, real-world estimation when data is limited.

However, the model has limitations. It is "blind" to specific policy changes, capturing their effects only implicitly through shifting parameters. Its accuracy is also entirely dependent on the quality of reported data; early changes in diagnostic criteria caused data spikes that required smoothing. Despite these limitations, this work provides a vital foundation for more granular epidemiological modeling and highlights the power of mathematical simplicity in crisis management.

Summary updated

Abstract

After the sudden outbreak of Coronavirus in Wuhan, continuous and rich data of the epidemic has been made public as the vital fact for decision support in control measures and aggressive implementation of containment strategies and plans. With the further growth and spreading of the virus, future resource allocation and planning under updated strategies and measures rely on careful study of the epidemic data and characteristics for accurate prediction and estimation. By using the SIR model and reported data, key parameters are obtained from least sum of squared errors for an accurate prediction of epidemic trend in the last four weeks.
BibTeX
@misc{zhang2020sir,
  author = {Jun Zhang and Lihong Wang and Ji Wang},
  title = {{SIR Model-based Prediction of Infected Population of Coronavirus in Hubei Province}},
  year = {2020},
  eprint = {2003.06419},
  archivePrefix = {arXiv},
  url = {https://arxiv.org/abs/2003.06419},
}
Figure