Infection rates forecasting of Novel Covid-19 using ensembled machine learning algorithm for better resource and response planning in Oman

Abstract

The COVID-19 pandemic, declared a global health tragedy by the WHO, has profoundly impacted the world economy, leading to recession, unemployment, and unprecedented debt. This manuscript introduces an innovative approach to outbreak analytics for Novel COVID-19, aiming to move beyond the conventional assessment of confirmed cases and fatalities. The primary focus is forecasting the number of individuals likely to be infected by the novel coronavirus, utilizing an ensemble machine-learning algorithm to predict infection and death rates. The study's specific goal is to assist government entities in anticipating the impact of Novel COVID-19 across the Sultanate states. The proposed machine learning-based model will forecast future cases and guide private and government institutions' needed resource and response planning. The contributions of this work encompass the development of a machine learning-based method to predict the spread of COVID-19 among the governorates of the Sultanate. It optimizes the current detection method by introducing new features for swift and accurate disease detection. Additionally, a deep learning method is proposed to determine COVID-19 status, issues, and issue alerts rapidly. The introduced approach is designed to function seamlessly in both offline and online systems, easily integrating with any framework. The decision Tree model has a high training score of 99.41% but a low testing score of 53.39%, indicating overfitting. The Random Forest model has a lower training score of 94.66% but a higher testing score of 61.32%, suggesting better generalization performance. The Gradient Boost model has a training score of 92.96% and a testing score of 59.44%, placing it between Decision Tree and Random Forest models in terms of accuracy. The heat map results analysis reveals a robust positive correlation between the variable representing Daily New Cases and the variable denoting Active Cases. This suggests that instances of elevated new cases align with an augmentation in the number of active cases. Conversely, negative correlations between the variables representing Total Coronavirus Cases and Total Coronavirus Deaths might infer that the mortality rate diminishes as the aggregate number of cases increases. This negative correlation could be attributed to efficient healthcare services and vaccination efforts. Nonetheless, it is imperative to conduct further statistical analyses to validate this hypothesis rigorously.