International Journal of Science, Engineering and TechnologyAuthors: Research Scholar Satish Kumar, Associate Professor Dr Surendra Singh Vishwakarma
Abstract: The growing availability of educational data has created new opportunities for predicting student academic performance, although privacy protection and appropriate feature selection remain significant challenges. This paper proposes a School Scholar Grade Prediction by Neural Network (SSGPNN) model for privacy-aware student grade prediction. Initially, distributed educational datasets are cleaned by removing sensitive information and transforming categorical attributes into suitable numerical representations. A Genetic Algorithm (GA) is employed to select an optimized set of relevant features and eliminate less informative attributes. The selected features are subsequently used to train a neural network for grade prediction. The proposed model is evaluated at dataset percentages of 12%, 25%, 50%, and 75% using precision, recall, F-measure, and accuracy. Experimental results show that SSGPNN outperforms K-PPD-ERT and achieves maximum precision, recall, F-measure, and accuracy values.
