Application of Neural Network Models Based on Multi-Objective Optimization in the Sustainable Evaluation and Intervention of Educational Equity
Year: 2025
Type: conference-paper
Abstract: Educational disparities remain a critical global challenge, hindering the achievement of equitable and inclusive education for all, as emphasized by the United Nations Sustainable Development Goal 4 (SDG 4) on Quality Education. This study explores the application of neural network models based on multi-objective optimization in the evaluation and intervention of educational equity. The proposed integrated framework combines Deep Convolutional Generative Adversarial Networks (DCGAN), Long Short-Term Memory (LSTM), Gated Recurrent Units (GRU), Attention Mechanism, and Multi-Objective Optimization (NSGA-II) to improve the accuracy and fairness of educational resource distribution. Experimental results demonstrate that the methodology outperforms traditional models, including rule-based approaches, historical data regression, and machine learning models such as Decision Trees, Random Forests, and DNN. Furthermore, fairness-aware baselines, such as adversarial debiasing and fair representation learning, were included for comparison. The integrated approach significantly improves both prediction accuracy and fairness, offering a more effective and equitable solution for policy interventions. The combination of multi-objective optimization with neural networks presents a promising solution for achieving sustainable educational outcomes aligned with SDG 4.
Authors: Yuying Jiang, Shan Luan, Mario Di Nardo
Language: English
Subfield: Education
Field: Social Sciences
Domain: Social Sciences
Open Access status: hybrid
OpenAlex ID: W7161309331