SuccessNet: An Automated Approach to Predict Student Academic Performance Using PCA Extracted CNN Novel Features and RF-SVM Ensemble Model

Abstract

With the rapid growth of data in the education sector, traditional techniques have failed to predict student academic success effectively. This research work uses features extracted from Convolutional Neural Networks (CNN) with a Random Forest (RF) and Support Vector Machine (SVM) ensemble model to predict the academic performance of students. We called this novel framework SuccessNet. It obviates manual feature extraction and surpasses independent Deep Learning (DL) and Machine Learning (ML) models in performance. The experiments are carried out in two sets. First, the original features are used to apply nine ML algorithms. The second set of experiments contains features extracted by CNN. The SuccessNet is formed with a soft voting mechanism that combines the top models generated during the above two sets of experiments based on academic performance prediction for students using an ensemble of RF and SVM. A comparison of performance with existing models shows auspicious results. SuccessNet gives an accuracy of 99.35% with a precision, recall, and F-Score of 99%.