A Hybrid Neural Network-based Approach for Predicting Course Grades

Abstract

For the academic performance prediction problem, a hybrid neural network model (HNNM) is constructed using the advantages of CNN and GRU in processing temporal and spatial data, respectively. Through an online teaching platform, rich data types are collected and the HNNM model is used to study students’ life behavior and classroom learning behavior. Through experiments in several courses, the results show that the HNNM model has high prediction accuracy and can provide more reasonable guidance and suggestions for course teaching improvement and student academic warning.