A Novel Hyper parameter Tuned Deep Learning Model and Optimal Feature Selection Based Student’s Performance Prediction with Data Balancing

Abstract

Nowadays, predicting students' performance is one of the most specific topics for learning environments, such as universities and schools, since it leads to the development of effective mechanisms that can enhance academic outcomes and avoid destruction. This paper proposes a novel hyperparameter tuned deep learning model and optimal feature selection-based student’s performance prediction with data balancing. The proposed system comprises the following phases: Data collection, Dataset balancing, Preprocessing, Feature selection, and Classification. To begin, the student performance data is collected from the OULAD dataset. Then, the system uses K-Nearest Neighbor (KNN) algorithm to solve the class imbalance problem. After that, the system performs preprocessing to improve the quality of the dataset. Then, the system uses the hummingbird flight based tunicate optimization algorithm (HFTOA) to select the best features from the dataset. After that, the system uses the Hyperparameter tuned with Hard swish activation based Gated Recurrent Unit (H²GRU) algorithm to classify the student’s performance into pass or fail. The findings showed that the proposed system achieves an average result of 98.96% accuracy, 99.04% precision, 98.86% recall, 98.99% f-measure, and 98.91% AUC, which is better than the state-of-the-art methods.