ESTIMATING AND ANALYZING COVID-19 TRANSMISSION IN THAILAND THROUGH DEEP NEURAL NETWORK

Abstract

The COVID-19 pandemic has caused significant global suffering and mortality, and effective control measures have been elusive. This study aims to develop an accurate and reliable prediction model using deep neural networks (DNN) to estimate the epidemic size and trends of COVID-19 cases, as well as the effective reproduction number, R(t). The efficacy of various control measures for COVID-19 has been questioned, and an efficient prediction model can aid in decision-making and planning. Overfitting is a common issue in neural networks, which can limit their accuracy and reliability. A modified dropout regularization technique and particle swarm optimization (PSO) are employed to enhance the accuracy of the DNN. The proposed model outperforms conventional neural networks and previous studies in terms of accuracy and reliability. The estimated R(t) values are consistent with measured values, which demonstrates the usefulness of this model in analyzing the situation and informing effective intervention strategies. The developed dropout-DNN-PSO model is an accurate and reliable predictor of COVID-19 trends and R(t) values. It can aid decision-makers in planning and implementing effective control measures. The proposed model can be extended to other countries to analyze and predict the trends of COVID-19 cases.