Application of ARIMA and Artificial Neural Networks Models for Daily Cumulative Confirmed Covid-19 Prediction in Nigeria

Abstract

Coronavirus 2019, commonly referred to as COVID-19, is a disease discovered in China towards the end of December 2019. This novel and highly infectious virus spreads rapidly across the globe. In Nigeria, as of June 2020, the cumulative number of COVID-19 cases reported was 25,694: out of this, 9746 cases were treated and 590 cases lost their lives. This research was aimed at comparing the prediction ability of ARIMA and ANN models. The aggregate COVID-19 cases reported in Nigeria was subjected to Box-Jenkins time series and  Back propagation gradient-based Artificial Neural Network Approaches for the prediction purpose. The data obtained from the Nigeria Centre for Disease Control (NCDC) was used. The data were identified to follow ARIMA (1, 2, 1) and were best trained by Bayesian  Regularization Artificial Neural Network algorithm. The prediction performance of the two models were compared using RMSE, MAE and MAPE. The empirical results obtained show that the Artificial Neural Network model gives better predictions and forecasts over the ARIMA
model.