Student Adaptability Level Prediction in Online Education During COVID-19 Using Machine Learning Techniques

Abstract

Distance learning (DL) is a method of instruction that makes use of technological advancements to allow for indirect connections between students and teachers who are separated by physical distance. Since COVID-19 swept the globe, the term "online education" has gained popularity. Most schools have moved their operations online so that instruction may continue even while they expand. It took a long time for a country like Bangladesh to guarantee online education at all educational levels. Our real objective is to contribute to this discussion by researching important aspects of online education. In this research, we conducted physical and online questionnaires to gather data from students at all three academic levels (school, college, and university) and also from Kaggle. The sociodemographic characteristics of a person are included in the survey form. A total of 14 variables were used: student gender, student type, age range, educational level of an institution, type of educational institution, IT student, student location, load shading level, family financial situation, category of internet, used device type, network connection type, and adaptability level of the learner. Our dataset was used to predict the level of student adaptability to online education using several machine learning algorithms, including Random Forest Classifier (RF), Decision Tree Classifier (DT), K-Nearest Neighbor (KNN), Logistic Regression (LR), Support Vector Classifier (SVC), and XGBoost Algorithm (XGB). The decision tree classifier surpassed other algorithms and had the highest accuracy (93%), compared to those that were used.