Hypergraph Neural Network Approach with Imbalanced Sampling to Predict Graduation Outcomes for Diverse Student Populations

  • Unique Paper ID: 168652
  • Volume: 11
  • Issue: 5
  • PageNo: 1646-1649
  • Abstract:
  • Predicting educational graduation outcomes is highly critical, even though current models are unable to handle the tail nature of the data, and imbalanced data, leading to biased or wrong predictions for the majority of the population. We introduce the novel idea of hypergraph construction along with optimization of neural networks for enhanced accurate representation of diverse students using imbalanced sampling techniques. The methodology begins with comprehensive data on the performance and social behavior of students, capturing complex relations not explored with other model traditional methodologies. To alleviate the drawbacks of imbalanced data, we devise a self-updating hypergraph neural network that optimizes the hyperedge representation and overcomes long-tail distribution problems. In addition, contrastive learning is further introduced to address the minority outcome classes. The study explored the impacts of extracurricular activities, such as sporting and community service, on student academic success and career readiness as well. This improved our perspective on what elements affect the development of students. The multifaceted source of data thus allowed our approach to give educational institutions much more accurate and actionable information for identifying at-risk students and designing targeted interventions. The outcomes of this study contribute to this field of educational data mining and construct practical tools to enhance graduation rates for different kinds of diverse student populations.

Related Articles