Skip to main page content
U.S. flag

An official website of the United States government

Dot gov

The .gov means it’s official.
Federal government websites often end in .gov or .mil. Before sharing sensitive information, make sure you’re on a federal government site.

Https

The site is secure.
The https:// ensures that you are connecting to the official website and that any information you provide is encrypted and transmitted securely.

Access keys NCBI Homepage MyNCBI Homepage Main Content Main Navigation
. 2021 Jan 21;5(1):79-91.
doi: 10.1109/TETCI.2020.3046012. eCollection 2021 Feb.

Optimisation of Non-Pharmaceutical Measures in COVID-19 Growth via Neural Networks

Annalisa Riccardi  1 , Jessica Gemignani  2   3 , Francisco Fernandez-Navarro  4 , Anna Heffernan  5   6
Affiliations

Optimisation of Non-Pharmaceutical Measures in COVID-19 Growth via Neural Networks

Annalisa Riccardi et al. IEEE Trans Emerg Top Comput Intell. .

Abstract

On [Formula: see text] March, the World Health Organisation declared a pandemic. Through this global spread, many nations have witnessed exponential growth of confirmed cases brought under control by severe mass quarantine or lockdown measures. However, some have, through a different timeline of actions, prevented this exponential growth. Currently as some continue to tackle growth, others attempt to safely lift restrictions whilst avoiding a resurgence. This study seeks to quantify the impact of government actions in mitigating viral transmission of SARS-CoV-2 by a novel soft computing approach that makes concurrent use of a neural network model, to predict the daily slope increase of cumulative infected, and an optimiser, with a parametrisation of the government restriction time series, to understand the best set of mitigating actions. Data for two territories, Italy and Taiwan, have been gathered to model government restrictions in travelling, testing and enforcement of social distance measures as well as people connectivity and adherence to government actions. It is found that a larger and earlier testing campaign with tighter entry restrictions benefit both regions, resulting in significantly less confirmed cases. Interestingly, this scenario couples with an earlier but milder implementation of nationwide restrictions for Italy, thus supporting Taiwan's lack of nationwide lockdown, i.e. earlier government actions could have contained the growth to a degree that a widespread lockdown would have been avoided, or at least delayed. The results, found with a purely data-driven approach, are in line with the main findings of mathematical epidemiological models, proving that the proposed approach has value and that the data alone contains valuable knowledge to inform decision makers.

Keywords: BiLSTM; CNN; Covid-19; ELM; SARS-CoV-2; data gathering; epidemiology; optimisation.

PubMed Disclaimer

Figures

Fig. 1.
Fig. 1.
The top panel portrays the distributions of the growth curves in Italy (left) and Taiwan (right), along with the indication of some key moments characterising the chronological development of the epidemics. The areas shaded in grey represent the timeframe employed by the neural networks to predict the growth, while the areas shaded in red indicate the timeframe (April 7th-24th) employed to compute the optimisation of the combination of control measures via genetic algorithm. The bottom panel illustrates, for each territory, the set of variables collected and utilised to describe and predict the evolution of the epidemics.
Fig. 2.
Fig. 2.
Graphical representation of the BiLSTM network model used for Italy and Taiwan. Parameter Number of Hidden Units refers to the size of the LSTM layers, parameter First Layer Size refers to the size of the fully connected layer.
Fig. 3.
Fig. 3.
Slope prediction (red) as computed by the best BiLSTMNN Neural Network model. Grey shaded area indicates model time window for which predictions cannot be made, histogram represents real slope values.
Fig. 4.
Fig. 4.
Optimal solution for Italy. On the right axis, black solid line represents the original variable, red solid line is the optimised solution. On the left axis the histograms of slope values for original solution (grey) and optimal one (red) are reported. The last plot represents the cumulative value of infected people in the country for original solution.
Fig. 5.
Fig. 5.
Optimal solution for Taiwan. On the right axis, black solid line represents the original variable, red solid line is the optimised solution. On the left axis the histograms of slope values for original solution (grey) and optimal one (red) are reported. The last plot represents the cumulative value of infected people in the region for original solution.
See this image and copyright information in PMC

Similar articles

See all similar articles

Cited by

References

    1. W. H. Organization, “Novel coronavirus (2019-ncov): Situation report, 1,” , 2020-01-21.
    1. Lu R. et al., “Genomic characterisation and epidemiology of 2019 novel coronavirus: Implications for virus origins and receptor binding,” The Lancet, vol. 395, no. 10224, pp. 565–574, 2020. - PMC - PubMed
    1. Wu F. et al., “A new coronavirus associated with human respiratory disease in china,” Nature, vol. 579, no. 7798, pp. 265–269, 2020. - PMC - PubMed
    1. Zhou P. et al., “A pneumonia outbreak associated with a new coronavirus of probable bat origin,” Nature., vol. 579, no. 7798, pp. 270–273, 2020. - PMC - PubMed
    1. Tian H. et al., “An investigation of transmission control measures during the first 50 days of the COVID-19 epidemic in china,” Science, vol. 368, no. 6491, pp. 638–642, 2020. - PMC - PubMed

Grants and funding

Research at Perimeter Institute is supported in part by the Government of Canada through the Department of Innovation, Science and Economic Development Canada and in part by the Province of Ontario through the Ministry of Colleges and Universities. The work of A. Riccardi was supported by SFC Scottish Funding Council. The work of J. Gemignani was supported by the ERC Consolidator under Grant 773202 ERC-2017-COG BabyRhythm. The work of F.-F. Navarro was supported by the Spanish Ministry of Science under Project ENE2017-88889-C2-1-R. The work of A. Heffernan was supported by the Natural Sciences and Engineering Research Council of Canada.

LinkOut - more resources