An Ensemble Machine Learning Model for Automatic Prediction of Perceived Personal Well-Being of Indian University Students During COVID-19 Lockdown

Abstract

COVID-19 has impacted personal well-being globally in a disruptive manner. Frequent lockdowns have slowed down dramatically the economy of every nation. There is a fear of future insecurity cropping up in the minds of the people. The paper aims to restructure the popular Personal Well-being Index (PWI) according to the relevant indicators that impacted students’ life in India during the second wave of COVID-19. The students at Delhi state university participated in the research. The researchers use various machine learning algorithms such as Lasso Regressor (LR), Support Vector Regressor (SVR), and Decision Tree Regressor (DTR) to predict the perceived PWI. The R-squared value for LR, SVR and DTR are 0.9103, 0.9159 and 0.5339. Mean squared errors are 0.0034, 0.0035 and 0.0105 respectively. The five most influential determinants of perceived PWI were extracted. An ensemble model of the three mentioned base learners was designed to remove the overfitting and underfitting problems. The algorithm has demonstrated impressive performance, with an R-squared value of 0.9839 and MSE of 0.0014. A GUI-based prediction model was implemented in Python that triggered the ensemble model at the back end to predict PWI based on five questions only, along with recommendations for the respondents.