A data analytics approach for COVID-19 spread and end prediction (with a case study in Iran)

Arman Behnam, Roohollah Jahanmahin

Abstract

The world is currently navigating the challenges of the COVID-19 pandemic, with Iran being one of the significantly affected nations. From the initial outbreak through April 30, 2020, Iran reported over 90,000 confirmed cases. Given the profound socio-economic implications, there is a critical need to accurately predict outbreak trends. This study compiles a comprehensive dataset of confirmed, death, and recovered cases. By integrating time-varying variables such as public policy and behavioral shifts with machine learning algorithms—specifically logistic functions and inflection point analysis—we developed novel metrics including the "Life Rate" and refined mortality and recovery rates. Our findings indicate that Gaussian functions offer superior performance in forecasting, providing governments with vital insights into the factors influencing future viral trends.

Introduction

Early in the COVID-19 pandemic, global governments struggled to transition from reactive crisis management to proactive planning. Traditional exponential models often failed to capture the "leveling off" effect resulting from social distancing and healthcare interventions. This research addresses a critical gap: the need for more nuanced metrics beyond raw case counts to evaluate the "health" of a national response. By analyzing the relationship between recovery and mortality rates over time, we can identify the "inflection point"—the moment viral spread begins to decelerate. This study specifically focuses on predicting the peak and conclusion of the outbreak in Iran to assist decision-makers in mitigating socio-economic fallout.

Methodology

The researchers implemented a three-tiered analytics framework designed to move from historical observation to future prediction.

The multi-stage data analytics framework used to study and predict COVID-19 trends in Iran.
  1. Descriptive Analytics: Historical data visualization to identify patterns, with a specific focus on demographic factors such as age-related recovery and mortality.
  2. Prescriptive Analytics: The definition of dynamic metrics to evaluate policy effectiveness, including:
  3. Weekly Recovered/Death Rates: Smoothed ratios that provide a clearer view of long-term trends.
  4. Life Rate: A composite index calculated as the difference between the weekly recovery rate and the weekly death rate, serving as a pulse for the pandemic's status.
  5. Predictive Analytics: A comparison of various statistical and machine learning tools:
  6. Logistic Functions: Employed to identify the inflection point where growth shifts from acceleration to deceleration.
  7. Gaussian Functions: Used for distribution fitting to predict the long-term "tail" and eventual end of the outbreak.
  8. Time-Series Forecasting: Utilization of Support Vector Machines (SVM), Linear Regression, and Facebook’s Prophet.

Datasets

The study utilized a compiled dataset of confirmed, recovered, and death cases in Iran, sourced from the World Health Organization (WHO), Worldometer, and the Mendeley Database. The primary analysis window spanned from February 19th to June 25th, 2020.

Key Insights

  • The Significance of the Inflection Point: Infectious diseases typically follow a logistic curve rather than a purely exponential one. Identifying the inflection point is essential for accurately predicting the maximum case ceiling.
  • The "Life Rate" Metric: This metric provides a single indicator of national healthcare success. A positive, rising Life Rate suggests that the healthcare system is effectively managing the patient load.
  • Demographic Drivers: Data analysis confirmed a stark age divide; recoveries were most frequent in the 30–50 age bracket, while mortality rates were significantly higher for individuals aged 55–85.
  • Model Performance: While various models are suitable for short-term "nowcasting," Gaussian functions demonstrated the most stability for long-term predictions regarding the end of the outbreak.

Results and Discussion

The analysis provided several key projections for the trajectory of the virus in Iran.

Trends of confirmed, active, recovered, and death cases in Iran from March to August 2020.
  • Peak and Duration: The model predicted the outbreak's peak to occur approximately 150 days after onset, with the conclusion estimated between 290 and 320 days (projecting an end in November 2020).
  • Case Forecasts: Based on early data, the study forecasted a maximum of 200,000 confirmed cases and 10,000 deaths under a worst-case scenario, contingent on the maintenance of isolation policies.
  • Growth Stabilization: By April 6th, 2020, the growth rate in Iran had stabilized near zero, suggesting that early mitigation efforts successfully transitioned the country out of its initial exponential growth phase.

Limitations and Future Work

While the framework provides valuable insights, several limitations remain. The time-series nature of the data restricts the application of certain machine learning algorithms. Furthermore, the current models do not account for "soft" variables such as specific policy shifts, social behavior changes, or climatic parameters (e.g., humidity and solar radiation). Future research should aim for higher granularity, incorporating state-level data and sociological factors to enhance local predictive accuracy.

Summary updated

Abstract

World is now experiencing the new pandemic caused by COVID-19 virus and all countries are affected by this disease specially Iran. From the beginning of the outbreak until April 30, 2020, over 90,000 confirmed cases of COVID-19 have been reported in Iran. Due to socio-economic problems of this disease, it is required to predict the trend of the outbreak and propose a beneficial method to find out the correct trend. In this paper, we compiled a dataset including the number of confirmed cases, the daily number of death cases and the number of recovered cases. Furthermore, by combining case number variables like behavior and policies that are changing over time and machine-learning (ML) algorithms such as logistic function using inflection point, we created new rates such as weekly death rate, life rate and new approaches to mortality rate and recovery rate. Gaussian functions show superior performance which is helpful for government to improve its awareness about important factors that have significant impacts on future trends of this virus.
BibTeX
@inproceedings{behnam2021data,
  author = {Arman Behnam and Roohollah Jahanmahin},
  title = {{A data analytics approach for COVID-19 spread and end prediction (with a case study in Iran)}},
  booktitle = {Modeling},
  year = {2021},
  doi = {10.1007/s40808-021-01086-8},
}
Figure