Predicting Student Performance Using Teacher Observation Reports
Menna Fateen, Tsunenori Mine
Jul 01, 2021 14:50 UTC+2
—
Session D3
—
Zoom link
Keywords: text mining, student grade prediction, teacher observation reports, machine learning
Abstract:
Studying for entrance examinations to qualify for higher educational institutes can be a distressing period for numerous students. Consequently, many students decide to attend cram schools to assist them in preparing for these exams. For such schools and for all educational institutes, it is necessary to obtain the best tools to provide the highest quality of learning and guidance. Performance prediction is one tool that can serve as a resource for insights that are valuable to teachers, students, and administrators alike.With accurate predictions of their grades, students can be further guided and fostered in order to achieve their optimal learning goals. In this regard, we target middle school students to be able to guide them on their educational journey as early as possible. In this paper, we propose a method to predict the students' performance in entrance examinations using the comments that cram school teachers made throughout the lessons. Teachers in cram schools observe their student's behavior closely and give reports on the efforts taken in their homework and subject material. We show that the teachers' comments are qualified to construct a tool that is capable of predicting students' grades efficiently. This is a new method because previous studies focus on predicting final grades mainly using student data such as their reflection comments or earlier scores.To create our model, we use and compare different Natural Language Processing techniques to represent the teachers' reports effectively. Furthermore, we compare the performance of our proposed model with a model that only uses student data. Finally, our results show that using readily available feedback from teachers can remarkably contribute to the accuracy of student performance prediction.