A Model Proposal With Deep Learning: On Student Success

This study aims to investigate predicting student academic performance using interaction data from online learning environments and course and demographic information from student information systems using deep learning methods. Deep learning enables high-precision prediction in large and complex datasets, enabling more accurate and reliable student prediction. In this context, modelling student performance based on data from online learning environments and student information systems not only identifies students at risk of nonattendance, but also allows prediction of potential failures by the end of the semester, enabling timely intervention at an early stage in the process. This model can be used to grade student performance and provide feedback on an individual or group basis. In addition, it will make a significant contribution to tutors managing online courses by enabling them to monitor students’ overall performance levels and organise learning processes more effectively. The dataset used in the study includes interaction data from 4,470 students enrolled in various online courses at Ya¸sar University during the autumn semester of 2019-2020, as well as course and demographic information from the student information system. The deep learning model developed in the study achieved 99.81% classification accuracy on training data and 98.36% on test data, predicting students’ final grades.