Alternative Approaches for Modelling COVID-19: High-Accuracy Low-Data Predictions

  1. Dewang Kumar Agarwal
  2. Soham De
  3. Ojas Shukla
  4. Archit Checker
  5. Apoorvi Mittal
  6. Ankita Borah
  7. Debayan Gupta

Reviewed by

Read the full article See related articles

Discuss this preprint

Start a discussion What are Sciety discussions?

Listed in

Log in to save this article

Abstract

Background

Numerous models have tried to predict the spread of COVID-19. Many involve myriad assumptions and parameters which cannot be reliably calculated under current conditions. We describe machine-learning and curve-fitting based models using fewer assumptions and readily available data.

Methods

Instead of relying on highly parameterized models, we design and train multiple neural networks with data on a national and state level, from 9 COVID-19 affected countries, including Indian and US states and territories. Further, we use an array of curve-fitting techniques on government-reported numbers of COVID-19 infections and deaths, separately projecting and collating curves from multiple regions across the globe, at multiple levels of granularity, combining heavily-localized extrapolations to create accurate national predictions.

Findings

We achieve an R 2 of 0·999 on average through the use of curve-fits and fine-tuned statistical learning methods on historical, global data. Using neural network implementations, we consistently predict the number of reported cases in 9 geographically- and demographically-varied countries and states with an accuracy of 99·53% for 14 days of forecast and 99·1% for 24 days of forecast.

Interpretation

We have shown that curve-fitting and machine-learning methods applied on reported COVID-19 data almost perfectly reproduce the results of far more complex and data-intensive epidemiological models. Using our methods, several other parameters may be established, such as the average detection rate of COVID-19. As an example, we find that the detection rate of cases in India (even with our most lenient estimates) is 2.38% - almost a fourth of the world average of 9% [1].

Article activity feed

  1. ScreenIT

    SciScore for 10.1101/2020.07.22.20159731: (What is this?)

    Please note, not all rigor criteria are appropriate for all manuscripts.

    Table 1: Rigor

    NIH rigor criteria are not applicable to paper type.

    Table 2: Resources

    No key resources detected.

    Results from OddPub: Thank you for sharing your code and data.

    Results from LimitationRecognizer: An explicit section about the limitations of the techniques employed in this study was not found. We encourage authors to address study limitations.

    Results from TrialIdentifier: No clinical trial numbers were referenced.

    Results from Barzooka: We did not find any issues relating to the usage of bar graphs.

    Results from JetFighter: We did not find any issues relating to colormaps.

    Results from rtransparent:
    • Thank you for including a conflict of interest statement. Authors are encouraged to include this statement when submitting to a journal.
    • Thank you for including a funding statement. Authors are encouraged to include this statement when submitting to a journal.
    • No protocol registration statement was detected.

    About SciScore

    SciScore is an automated tool that is designed to assist expert reviewers by finding and presenting formulaic information scattered throughout a paper in a standard, easy to digest format. SciScore checks for the presence and correctness of RRIDs (research resource identifiers), and for rigor criteria such as sex and investigator blinding. For details on the theoretical underpinning of rigor criteria and the tools shown here, including references cited, please follow this link.

  2. Version published to 10.1101/2020.07.22.20159731 on medRxiv