A CNN Model with Data Imbalance Handling for Course-Level Student Prediction Based on Forum Texts

Abstract

Nowadays teaching and learning activities in a course are greatly supported by information technologies. Forums are among information technologies utilized in a course to encourage students to communicate with lecturers more outside a traditional class. Free-styled textual posts in those communications express the problems that the students are facing as well as the interest and activeness of the students with respect to each topic of a course. Exploiting such textual data in a course forum for course-level student prediction is considered in our work. Due to hierarchical structures in course forum texts, we propose a method in this paper which combines a deep convolutional neural network (CNN) and an adopted and adapted loss function for more correct recognitions of instances of the minority class which includes students with failure. A CNN model with data imbalance handling is a novel method appropriate for the course-level student prediction task. Indeed, through an empirical evaluation, our method has been confirmed to be an effective solution. Compared to other methods such as C4.5, Support Vector Machines, and Long-Short Term Memory networks, the proposed method can provide higher Accuracy, Precision, Recall, and F-measure on average for early predictions of the students with either success or failure in two different real courses. Such better predictions can help both students and lecturers beware of students’ study and support them in time for ultimate success in a course.