A Modified Artificial Neural Network (ANN)-Based Time Series Prediction of COVID-19 Cases from Multi-Country Data

J. Inst. Eng. India Ser. B https://doi.org/10.1007/s40031-022-00849-w ORIGINAL CONTRIBUTION A Modified Artificial Neural Network (ANN)‑Based Time Series Prediction of COVID‑19 Cases from Multi‑Country Data Babita Majhi1 Received: 26 March 2022 / Accepted: 21 December 2022 © The Institution of Engineers (India) 2023 Abstract The deadly Corona virus that first appeared in a seafood market in the Wuhan city of China in December 2019 has been causing global distress by claiming lives and collapsing economies. Given its serious nature, there is an urgent need to understand the virus’s future trajectory. The current study predicts the next day confirmed, death and recovery cases of COVID-19 pandemic for India, Italy, Spain, and the USA by using a modified multilayer neural network (MMLNN) model. The spread of the COVID-19 data is collected from the Kaggle website for the period of 22nd January 2020 to 20th April 2020 (i.e., for 90 days). The predicted figures of the spread of the disease have been estimated and compared with the actual values. Higher precision of the estimates has been observed from the MMLNN model compared to the conventional multilayer neural network (MLANN) model. Specifically, the MMLNN model does faster and more efficient training of the data resulting in less error. The paper forecasts the next day figures (i.e., for 21st April) for all the three cases and does the comparison of the results with the actual values reported. A deviation of 6% is obtained for India, and for the other three countries the deviation is below 3.5%. Given the high accuracy predictive power, the authors recommend that the MMLNN model can be integrated into the health policy of the countries that are struggling with the spread of the virus. Specifically, a decision on health policies such as restrictions on movement can be based on the short-range predictions of the spread of the virus infection. * Babita Majhi 1 Department of Computer Science and Information Technology, Guru Ghasidas Vishwavidyalaya, Central University, Bilaspur 495009, India Keywords Time series analysis · COVID-19 data · Multilayer artificial neural network (MLANN) · Modified multilayer neural network (MMLNN) Introduction About 367,499 persons have been infected with the fatal corona virus known as "COVID-19," and about 254,199 have lost their lives as of the publication of this report. The Director-General of the World Health Organization (WHO) declared COVID-19 a Public Health Emergency of International Concern [1] on January 30, 2020, because to the increasing risk it poses to human life and economic stability. The novel infection is transmitted from one human to the other by respiratory beads, resulting in symptoms such as fever, headache, and breathing problem within a period of 2 to 14 days after the contamination, as per the Centers for Disease Control and Prevention (CDC) [1–3]. Infecting over two-hundred-fifteen nations, this virus poses several problems for humankind and science. It is also affecting the world economy due to restrictions on movement in the affected countries to control the disease as there is no alternative other than the social distancing. Policymakers worldwide are struggling to strike a balance between keeping the illness under control by enforcing the lockdown and preventing the disruption of economic operations, which would protect the employment and incomes of many people [4]. Strict lockout policies have been shown to disrupt supply chains significantly, leading to an increase in joblessness [5, 6]. In the immediate aftermath of lockdown in India, the sight of a large number of migrant laborers trying to travel to their native rural places by foot and other means has made the government extremely worried. Similarly, a record number 13 Vol.:(0123456789) J. Inst. Eng. India Ser. B of unemployment benefits applications have been filed in the USA after the lockdown has rendered several thousand citizens unemployed. This is, therefore, a key policy issue as to when and how much restriction is necessary to contain the COVID-19 pandemic. Governments may benefit from accurate predictions of infection and mortality rates in developing or adjusting policies, creating mitigation measures, and making efficient use of healthcare facilities. The spread of the COVID-19 pandemic has been predicted using numerous statistical methods, including regression models [7], the Susceptible-InfectiousQuarantined-Recovered (SIQR)model [8], autoregressive integrated moving average (ARIMA) and suttee ARIMA models [9], and the Composite Monte-Carlo (CMC) simulation model [10]. Although each of these models has its distinctive advantages, none of them is adaptive meaning that the prediction error is not used in an iterative process to improve the model. The prediction task relating to COVID-19 is a burning issue and many researchers are working on this problem. Out of various models, the one which offers better performance and demonstrate faster training is preferable. Hence, an attempt has been made in this paper to predict the number of confirmed, death and recovery cases for the next day for the USA, Spain, Italy, and India using the data available between 22nd January 2020 and 20th April 2020 employing a modified neural network model [11]. The model recently developed by Panda and Panda [12] has a few computational and predictive advantages over the other models used in the past. Specifically, the proposed ANN architecture in which activation and summation operations are interchanged has provided superior training and testing performance in a few applications. This observation has inspired us to choose this ANN model to use for the COVID-19 prediction task. The same authors in another paper suggested an improved BP algorithm for a multilayer neural network model with an adaptive slope of the activation function [12]. A large number of articles forecasting COVID-19 cases by employing ANN models have already been published. However, there are certain advantages and limitations for each of them. The novelties of the proposed model in this paper are the following: 1. The proposed ANN model uses the activation function of the neuron before summation tasks which have not been used by other models so far for COVID 19 prediction. 2. The resultant model offers faster training and improved prediction performance. 3. The simulation study has demonstrated that the used ANN model offers robust and consistent performance. 13 For a given dataset the prediction performance of the supervised ANN model mostly depends on extracted features from the raw data, the architecture of the model, and the learning algorithm employed. In the present investigation, the focus has been given on the selection of the last two aspects. In this paper’s paradigm, the nonlinear activation function of a synthetic neuron is initialized at the top of each layer. Such architecture had not been used by any of the recently publis (...truncated)