All educational organizations strive to improve the overall quality of education by raising students' academic performance. In this regard, Educational Data Mining (EDM) is a rapidly growing research field that employs the the core ideas of data mining (DM) to assist academic institutions in determining useful details about how satisfied students are with the online learning experience (SSL) (OL) during the COVID-19 lock-down. To provide the optimum educational environments, several approaches have been explored using EDM to anticipate students' behavior. As a result, Feature Selection (FS) is often used to discover the most relevant subset of characteristics with the lowest cardinality. Because the FS process has a substantial impact on the predicted accuracy of a satisfaction model, the usefulness of the SSL model in conjunction with FS approaches is thoroughly investigated in this research. In this regard, a dataset of student evaluations of OL courses was initially gathered online through a questionnaire. The performance of wrapper FS approaches in DM and classification algorithms was evaluated in terms of fitness values using this datasets. Finally, the goodness of subsets with various cardinalities is assessed in terms of prediction accuracy and the number of chosen features through evaluating the performance of 11 wrapper-based FS algorithms in addition to Support Vector Machine (SVM) and k-Nearest Neighbor (k-NN) as baseline classifiers. WThe studies indicated the optimum dimensionality of the feature subset as well as the best technique. The current study's results clearly corroborate the well-known association between the presence of a small number of characteristics and an improvement in prediction accuracy. The relevance of FS for high-accuracy SSL prediction is outstanding, as the necessary collection of traits may effectively aid in the development of constructive instructional initiatives. On the used real-time dataset, our work offers a feature size reduction of up to 80% as well as up to 100% classification accuracy.
Article Details
Unique Paper ID: 162514
Publication Volume & Issue: Volume 10, Issue 10
Page(s): 226 - 232
Article Preview & Download
Share This Article
Join our RMS
Conference Alert
NCSEM 2024
National Conference on Sustainable Engineering and Management - 2024