Nonlinear Autoregressive Neural Network for Forecasting COVID-19 Confirmed Cases in Malaysia

Abstract

A nonlinear autoregressive neural network (NARNN) model is a feedforward neural network for handling complex nonlinear time series problems. In this study, the tangent sigmoid (tansig) activation function with the different numbers of past values and different numbers of hidden neurons for the NARNN modelis determined. The COVID-19 daily confirmed cases in Malaysia are collected with different amounts of samples used, which are 100, 500 and 900. Therefore, data from 100, 500 and 900 days before 21 September 2022 are extracted for the NARNN model training, validation and testing procedure. The lowest average mean squared error (MSE) becomes the best combination. The result shows that the past value is 1:10 and the number of neurons of 10 when the sample size is 100. At sample size 500, past values of 1:10 and neurons of 8 enable the model to perform at its best. Whereas for sample size 900, the network setting of 1:5 past value and five hidden neurons gives the least MSE. Multi-step ahead time series forecasting is conducted to forecast the number of confirmed COVID-19 cases in 7 days from 22 to 28 September 2022. The result shown for 7-days-ahead confirmed cases indicating Malaysia datasets, the best forecasting outcome occurs when 900 samples are inputted.