This paper focuses on the application of the five metaheuristic algorithms (Jaya, Teaching Learning-Based Optimization, Rao1, Rao2, and Rao3) to optimize the factors that influence students’ academic performance and their success in studies. It identifies key elements of student engagement—such as self- confidence, active learning perception, goal setting, effort, self-assessment, and feedback response—through a six-question survey and academic data (grades from exams, homework, projects, and tests). The collected data are preprocessed using ANOVA, principal component analysis (PCA), and gray relational analysis (GRA), and then the artificial neural network (ANN) is applied to model the relationship between the engagement factors and the overall academic success. The above five algorithms are employed to optimize academic performance based on the neural model, with the primary focus on comparing their efficiency in predicting student success. This comparison uses metrics like convergence speed, solution quality (model accuracy), and computational time, along with an analysis of the population size effects. Findings indicate that the Rao2 and Jaya algorithms deliver the highest solution quality and stable results, while Rao3 achieves the fastest convergence.