Student academic success prediction in multimedia-supported virtual learning system using ensemble learning approach

Multimedia systems and metaverse has been gaining increasing interest for education in virtual environments. With wide adoption of these technologies, the data is expected to grow exponentially. The increasing rise of educational data indicates that standard processing methods may be limited. As a result, constructing data mining-based academic performance assessment methods is becoming increasingly important. Students’ academic performance prediction can raise an alert for students at risk of failure and can help academia plan accordingly. This research suggests utilizing features derived from a convolutional neural network (CNN) in conjunction with machine learning models to enhance the overall performance of the model. By doing so, the necessity for manual feature extraction is eliminated, leading to superior outcomes compared to using machine learning and deep learning models separately. At the outset, nine machine learning models are employed to assess both the original features and the convoluted features. The top two individual models are later used to build an ensemble model using soft voting criteria. The ensemble model is built using random forest (RF) and support vector machine (SVM) for predicting student academic performance. The performance of the proposed approach is compared to existing models for validation indicating the superior performance of the former. With an accuracy of 98.99