Enhancing University Performance Evaluation through Digital Technology: A Deep Learning Approach for Sustainable Development

In the era of sustainable development, the evaluation of higher education institutions' performance has gained paramount importance. This paper presents a novel approach to university performance evaluation by harnessing the power of digital technology, specifically deep learning techniques. Building upon the foundation of sustainable development theory and the experiences of developed countries, we propose a multi-classification model for university performance evaluation, offering both technical methods and theoretical underpinning for educational reform. Our research focuses on the integration of digital technology, including artificial intelligence, deep learning, and data mining, into the assessment of university performance. We employ a multi-layer restricted Boltzmann machines (RBMs) feature extraction approach coupled with the SoftMax classifier to enhance the accuracy of university performance predictions. The paper provides a comprehensive description of the model's architecture, the forward propagation process, and the solution methodology. Comparative experiments are conducted to evaluate various feature extraction methods, highlighting the superior feature expression capabilities of RBMs over traditional approaches. The results demonstrate that our proposed model surpasses the SoftMax classifier and Deep Belief Networks (DBN) in terms of prediction accuracy, average accuracy, and average recall rate, indicating its practical significance in performance evaluation. While our study offers valuable insights and advancements in university performance assessment, we acknowledge the need for further exploration of potential trade-offs, computational complexity, model interpretability, and generalization performance. These aspects warrant continued investigation to refine and optimize our approach for the benefit of the knowledge economy, innovation, entrepreneurship, and society at large.