Author:
Rumiko Azuma
Affiliation:
College of Commerce, Nihon University, Setagaya-ku, Tokyo, Japan
Keyword(s):
Learning Analytics, Machine Learning, Text Mining, Educational Support, Reflection Sheet.
Abstract:
In recent years, schools and universities have become more focused on how to allow learners to learn successfully, and it has become an expectation to design instruction in a way that takes into account the individual differences of learners. Accordingly, the purpose of this study is to predict, at an earlier stage in a course, which students are likely to fail, so that adequate support can be provided for them. We proposed a new approach to identify such students using free-response self-reflection sheets. This method uses the unrestricted comments from the students to create a comment vector that can be used to predict who are likely to fail the course. Subsequently, we conducted experiments to verify the effectiveness of this prediction. In comparison to methods used in existing research which predict potential failures using quiz scores and the students’ subjective level of understanding, our proposed method was able to improve the prediction performance. In addition, when cumula
tive data after several sessions were used to predict which students were likely to fail, the predictions made by the support vector machine (SVM) algorithm showed a consistent prediction performance, and the prediction accuracy was higher than that of other algorithms.
(More)