Predicting the Growth and Trend of COVID-19 Pandemic

This study applies an improved mathematical model to analyse and predict the growth of the epidemic. An ML-based improved model has been applied to predict the potential threat of COVID-19 in countries worldwide. We show that using iterative weighting for fitting Generalized Inverse Weibull distribution, a better fit can be obtained to develop a prediction framework. This has been deployed on a cloud computing platform for more accurate and real-time prediction of the growth behavior of the epidemic. Interactive prediction graphs can be seen at the following links:

Static model: https://collaboration.coraltele.com/covid/.
Dynamic LSTM model: https://collaboration.coraltele.com/covid2/.
Multi-peak dynamic model*: https://shreshthtuli.github.io/projects/covid/.

* This model is also being used by the National Health Service (NHS) of the UK.

Quick installation of real-time prediction webapp

To install and run the dynamic real-time prediction webapp on your server run the following commands:

$ git clone https://github.com/shreshthtuli/covid-19-prediction.git
$ mv covid-19-prediction covid
$ cd covid
$ chmod +x run.sh
$ ./run.sh

To access your server go to $HOSTNAME/covid/ from your browser. The webapp is hosted on https://shreshthtuli.github.io/projects/covid/ where graphs get updated daily based on new data.

Dataset

We use the Our World in Data dataset for predicting number of new cases and deaths in various countries.

Model contributions

Weibull Distribution

The model uses weibull distribution with the following function:

Robust Curve Fitting

The model uses robust curve fitting as described in [1]. This is to give low weightage to outliers for curve fitting. The iterative loop of robust curve fitting is shown below.

Dynamic Parameter Updates

The model uses LSTM model to calculate the coefficients of the weibull distribution as described in [2]. This is to adapt to the data and give higher weightage to recent data.

Mixed Weibull distribution

The model uses mixed weibull model to handle multiple peaks where each peak is modelled using a separate weibull distribution. This is summation of upto four weibull functions as described before but with same and values to share the trend of the virus in a country.

Without mixed distribution (for UK):

With mixed distribution (for UK):

Developer

Shreshth Tuli (shreshthtuli@gmail.com)

Cite this work

If you use our static model, please cite:

@article{tuli2020predicting,
title = "Predicting the Growth and Trend of COVID-19 Pandemic using Machine Learning and Cloud Computing",
journal = "Internet of Things",
pages = "100--222",
year = "2020",
issn = "2542-6605",
doi = "https://doi.org/10.1016/j.iot.2020.100222",
url = "http://www.sciencedirect.com/science/article/pii/S254266052030055X",
author = "Shreshth Tuli and Shikhar Tuli and Rakesh Tuli and Sukhpal Singh Gill",
}

If you use our dynamic model, please cite:

@article{tuli2020modelling,
  title={Modelling for prediction of the spread and severity of COVID-19 and its association with socioeconomic factors and virus types},
  author={Tuli, Shreshth and Tuli, Shikhar and Verma, Ruchi and Tuli, Rakesh},
  journal={Biomedical Research and Clinical Reviews},
  year={2020},
  volume={1},
  issue={3},
  doi={10.31579/2692-9406/014}
  publisher={Auctores}
}

References

Shreshth Tuli, Shikhar Tuli, Rakesh Tuli and Sukhpal Singh Gill, Predicting the Growth and Trend of COVID-19 Pandemic using Machine Learning and Cloud Computing. Internet of Things, ISSN: 2542-6605, Elsevier Press, Amsterdam, The Netherlands, May 2020. (Open access link)
Shreshth Tuli, Shikhar Tuli, Ruchi Verma and Rakesh Tuli, Modelling for prediction of the spread and severity of COVID-19 and its association with socioeconomic factors and virus types. Biomedical Research and Clinical Reviews. 1(3); DOI: 10.31579/2692-9406/014. (Open access link)

shreshthtuli / covid-19-prediction