Artificial Neural Network dalam Meramalkan Kasus Covid-19 di Indonesia Menggunakan Metode Backpropagation

Abstract

COVID-19 cases have become a threat to the whole world, including Indonesia. The graph of cases of the spread of covid-19 in Indonesia has been getting out of control since the beginning of the first case of covid-19 in Indonesia. Prediction of daily confirmed cases of covid-19 is needed to break the chain of spread of covid-19 in the future. In this study, through an artificial neural network approach using the backpropagation method, forecasting of daily confirmed cases of COVID-19 in Indonesia was carried out using previous daily case data starting from March 2, 2020 to February 18, 2023. The best architecture and parameters in this forecasting process are obtained there are 14 neurons in the input layer, 128 neurons in the hidden layer and 1 neuron in the output layer. While the best parameter values during model learning obtained a learning rate value of 0.001 and the activation function used was ReLu in 1000 training epochs. In this forecasting, the forecast results will be obtained in the next 14 days with an accuracy level assessed based on the smallest MAPE value of 1.16% and the coefficient of determination R2=98%.