A novel metaheuristic prediction approach for COVID-19 cases using XGBoost algorithm

Abstract

COVID-19 prediction is of great importance to build stronger government prevention and control of the global pandemic. This pandemic has a devastating impact on people and their lives. Many industries are suffering and are struggling to overcome unexpected pandemic challenges. Therefore, it is extremely important to develop up to date and practicable models for time series prediction of COVID-19 that would give promising results. In this paper, we build a model for predicting daily confirmed n w cases for the time series data of Europe by applying the Extreme Gradient Boosting (XGBoost) algorithm and using two performance parameters such as RMSE and MAPE to evaluate the effect of model fitting. Since machine learning methods (ML) have shown favorable results recently, among the tested models, XGBoost has shown to give the best results among the other models tested. Another advantage of the model is that we can use the XGBoost model to determine the robustness of the predictive model by adjusting parameter features. Our results are in line with the expectations of the performance of the model based on the available data and the variability of the data. By analyzing important features and updating the prediction for the new cases of the pandemic, the governments all around the globe could manage the situation much more effortlessly and with greater impact.