Grades Prediction Through Use of The BP Nerual Network Model

This article is centered on BP neural network technology, wherein sample data is standardized and different data sets are organically integrated. The method leverages information entropy to determine the number of hidden layers, and dynamic learning rate is employed to circumvent the limitations of traditional static learning rates, such as low learning rates, slow convergence speed, large learning rates, and a lack of decrease in error function. Utilizing historical academic performance and student behavior data, the study aims to predict the future graduation grades of students. Through experimental comparisons, it is determined that using both past academic performance and behavior as factors affecting students' academic performance yields higher prediction accuracy than solely relying on academic performance history. The outcomes are more favorable, compensating for the deficiencies of low prediction accuracy when solely depending on past academic performance. The article delves into the correlation between student behavior and academic performance. The model, being scientific, incorporates not only past grades but also considers student behavior.