Analyzing Learning Behavior of Student Persona toward Non-Negative Matrix Factorization

Abstract

Abstract-Estimation of motivation and learning strategy of students is crucial for a teacher to engage them in programming. Let us consider a persona, which is a virtual student representing a student group similar in motivation and learning strategy to learn programming. Personas enable the teacher to predict student behavior during the programming education course. The paper proposes a method to figure out the weight each student belongs to a specific persona. It uses non negative matrix factorization (NMF) to decompose a matrix of portfolio, which is extracted from their real learning behavior, into the product of 2 matrices. A matrix represents the weight of each student belonging to certain personas. The other represents persona features. For the NMF, determining persona feature matrix is essential to achieve the good factorization. From the learning behavior of 66 students, we found that the trends of motivation features along the course, such as learning time, test score, submissions before deadline is good indicator for the feature matrix.