Surrogate assisted by genetic algorithm for simulation of a hospital-resource allocation: lockdown policy analysis for COVID-19 pandemic case

Abstract

The government and healthcare sector plays a major role in managing the transmission upsurge and unpleasant effects of COVID-19. The dynamics of the pandemic is an emerging question for these parties, as this will guide their planning and strategy implementation of disease prevention and treatment policy. The emphasis of this study is to develop a simulation-based system that can estimate the future trans- mission dynamics of COVID-19. The simulation will predict the current and future trends of COVID-19 in South Africa, Italy, India, Mexico, and Fiji (SIIMCF). The pre- diction will assist hospital management with capacity planning and assist government in policy formulation around lockdown measures. Agent-Based Modeling and Simulation (ABMS) is an excellent computational tool for this simulation process. However, there are several constraints to using Agent-Based Modeling and Simulation (ABMS) to simulate a real-world event. Such constraints include the duration of the time required to perform simulation, the need for High performance computational resources, and the complexity of parameter tuning. Our study consider surrogate model as a substitute for Agent-Based Modeling and Simulation (ABMS). Our findings show the surrogate can overcome the constraints identified by ABMS for simulating the COVID-19 pandemic. Further, we optimize the surrogate with the Genetic Algorithm (GA) by running a local search and extracting the best hyper-parameter values for the surrogate. The optimization of the surrogate will enhance the quality of its results when compared using real-world COVID-19 datasets. To evaluate the performance of the ADRIANA system, we used three evaluation met- rics, which are, Mean Absolute Error (MAE), Root Mean Squared Error (RMSE), and R-squared (R2).