Prediction of engineering students’ virtual lab understanding and implementation rates using SVM classification

Abstract

In 2020 many universities were forced to switch to a distant form of education because of the COVID-19 lockdown. This was especially challenging for the engineering specialties, where laboratory and practical exercises are a fundamental part of the educational process. This study presents results from the electrical engineering education in two Bulgarian universities, where the Engine for Virtual Electrical Engineering Equipment was used as a tool for providing virtual labs. At the end of the semester the students were asked to fill in a survey, accounting for their learning program, years of studying, experience with virtual and real labs and the instructions delivery methods used. Data mining algorithms were utilized with the aim to predict students’ rate of understanding and rate of implementation when dealing with virtual labs. Initially, a regression analysis model was created which achieved R-squared above 95%. However, the verification of the model showed an unsatisfactory prediction success rate of 37%. Next, SVM classification was utilized. The verification showed its success rates for predicting the rate of understanding and rate of implementation were 83% and 86%, respectively. This approach could be used to optimize the educational experience of students, using virtual labs, as well as for identification of students that might need additional support and instructions. © Italian e-Learning Association.