Student Single Subject Grade Prediction Model Based on Feedforward Neural Networks

Abstract

Predicting student grades can assist teachers and educational administrators in timely under-standing of students' learning situations, thereby intervening and tutoring students who may fail in academic tests in advance, or providing personalized academic advice and support. This optimizes and enhances teaching quality, promoting the transformation of education towards digitization and intelligence. This paper analyzes and preprocesses academic performance data collected during online and offline teaching, considering various factors affecting college stu-dents' grades, to build a dataset with 10 features and 3102 real data entries of college students' academic performance. Based on this, a multilayer neural network model using feedforward neural network technology was constructed to predict grades in a college computer basics course, assess if students can pass the final exam, and the model was trained and tested with real data. Experimental results show that our proposed model achieved 92.15% and 90.74% accuracy on training and testing datasets, respectively, supporting further development of educational support applications.