Covid-19 and its impact on school closures: a predictive analysis using machine learning algorithms

Abstract

This research presents an extensive point of reference for investigating the operation of several machine learning (ML) algorithms in postulating the multiclass classification problem regarding the forthcoming effects of Covid-19 on school closures. With the prompt closure of schools across the world in response to this pandemic, school-going children and teenagers are ruptured both mentally and physically. Hence, ML has come across to be a reliable component to forecast the scenario effectively. A dataset from UNESCO is trained and tested by ten supervised ML algorithms. A comprehensive analysis among the predictive ML models was executed which bought satisfactory results with regard to accuracy, precision, sensitivity, F1 score, ROC-AUC by hyper parameter optimization. In this regard, grid search cross validation (GridSearchCV) was utilized in order to obtain the optimal parameters. However, the performance of Artificial Neural Network (ANN) was also investigated and compared with the supervised ML models where ANN displayed maximum accuracy of 80.37%. After rigorous comparative analysis, Decision Tree (DT) portrayed the highest accuracy of 90.75%. Hence, it is evident that machine learning algorithm holds strong promise in forecasting the upcoming scenario of school closures due to Covid-19 and can contribute significantly in decision making for the welfare of the education system.