ANALYSIS OF COVID-19 DATA USING ARIMA-NEURAL ARTIFICIAL INTELLIGENCE

Abstract

An Auto Regressive Integrating Moving Average (ARIMA) - Neural Artificial Intelligence (Neural AI) are employed the nature and sustainability of the COVID-19 (2019-nCoV/SARS-CoV-2), pandemics, through the four different waves in the country Pakistan (from 26 Feb 2020 to 21 October 2020, first wave of epidemic, 22 October 2020 to 16 March 2021 second wave, 17 March 2021 to 10 July 2021 the third wave and 11 July 2021 to 30 September 2021respectively). the appropriate ARIMA (1, 1, 8) and (1,1,7) models for the 1^st wave (26 Feb 2020 to 21 October 2020) to 4^th wave (11 July 2021   to 30 September 2021) over to Pakistan are obtained on basis of small value of sum of square error (SSE), mean absolute Percentage Error (MAPE) and higher value of AIC/BIC. Afterwards, the Neural AI is also applied, which has one output neuron and many input neuron, inputs neurons for infectious cases reported in different regions of Pakistan (i.e., Punjab, Sindh, Khyber Pakhtunkhwa, Baluchistan, Gilgit Baltistan, Azad Jammu and Kashmir and capital city Islamabad). At this stage, the neural network is performed smoothly when which is delivered by only one predictor. Therefore, it is concluded that Neural AI results are more effective (with lower values of the sum of square error for both trained and testing data, along with related errors) compared to fitted ARIMA models. ARIMA models require a higher number of predictors and exhibit slightly higher error values. This conclusion is drawn to acquire univariate predictions for daily infected COVID-19 data series.