The Application of Artificial Intelligence and Deep Learning Technologies in Sustainable Education Management Decision Systems

Against the backdrop of digital transformation in educational management, student performance prediction, as a core function of educational decision support systems, directly affects the development of personalized teaching strategies and the optimization of educational resource allocation.Traditional prediction methods, such as linear regression and machine learning models, often have low accuracy due to neglecting the correlation between disciplinary features and the non-linear distribution of data.Although deep learning models such as LSTM and Transformer have advantages in feature extraction and sequence modeling, their hyperparameter sensitivity and local optimal traps still constrain the model's generalization ability.Therefore, how to further improve the accuracy of the model through intelligent optimization algorithms has become a key issue in enhancing the reliability of educational prediction.Therefore, in response to the problem of student performance prediction in educational management decision-making, this study proposes an improved algorithm (PSO-MPA) based on a mixture of particle swarm optimization (PSO) and marine predator algorithm (MPA) to optimize the predictive performance of deep learning models.The experimental results showed that PSO-MPA achieved accuracy rates of 81.81%, 86.85%, and 83.39% in predicting English, mathematics, and psychology, respectively, with an average accuracy rate of 84.02%, which is 0.53% and 0.84% higher than the traditional MPA algorithm (83.49%) and the basic PSO algorithm (83.18%), respectively.Especially in mathematics subject prediction, PSO-MPA exhibits significant advantages (86.85%), verifying the synergistic effect of hybrid strategies in feature optimization and global search capabilities, providing a more accurate intelligent prediction tool for educational decision-making.