Prediction and Optimization of Student Grades Based on Genetic Algorithm and Graph Convolutional Neural Networks

• Published in: International journal of computational intelligence systems Vol. 18; no. 1; pp. 1 - 21
• Main Author: Li, Ting
• Format: Journal Article
• Language: English
• Published: Dordrecht Springer Netherlands 17.03.2025; Springer
• Subjects: Artificial Intelligence; Computational Intelligence; Control; Dependency measure and academic performance; Engineering; Genetic algorithm; Graph convolution neural networks; Graph structure; Mathematical Logic and Foundations; Mechatronics; Research Article; Robotics; Student grade; Graph structure; Genetic algorithm; Student grade; Graph convolution neural networks; Dependency measure and academic performance
• ISSN: 1875-6883; 1875-6883
• DOI: 10.1007/s44196-025-00775-x

Due to limited support in registered courses, students frequently struggle to complete their courses in higher education institutions. To combat this, educational systems are incorporating intelligent prediction tools to help students improve their academic performance by predicting their grades. Students' demographic information, past performance in the subject, and course characteristics are some of the factors used by the grade prediction system to foretell how they will do in future. Complexity and non-linearity in the analysis of inter-variable connections pose problems for traditional prediction models. Our solution to these problems is a GGCNN, or Genetic Algorithm with Graph Convolutional Neural Networks. In order to improve the accuracy of predictions, GGCNN examines educational data and finds intricate linkages. Using a graph structure, the graph convolution model emphasizes the interdependencies among academic metrics, course features, and student performance. Relationships and correlations can be better predicted with the use of this dependency metric. Use of the genetic algorithm improves the grade prediction system by optimizing the network and making better use of features. Administrators and teachers alike can find ways to boost their kids' grades through the optimization process. To test how well the system performs on different measures, we utilize the Student Performance Kaggle dataset. This continues until the convergence requirements are satisfied. With Python as its implementation, the system was able to get an accuracy of 0.98% after 100 epochs and 0.97% after 1000 epochs.