Non-negative Matrix Factorization to Identify Motivation and Learning Strategies from Portfolio

摘要

The paper provides an effective method to give appropriate supervision. Students are classified into personas, which are student groups similar in motivation and learning strategies. The paper uses the non-negative matrix factorization to classify students into personas. The characteristics of the each student group are identified from past student learning behavior, which is represented with their portfolio. A matrix indicating the portfolio of past students is decomposed into 2 matrixes: a wait matrix and persona matrixes. The former represents the degree of each student belonging to each persona, while the latter represents features each persona has. Assuming the persona matrix remains unchanged over years, the portfolio matrix of the current year are decomposed into the current weight matrix and the persona matrix. Based on the current weight matrix, supervisors can grasp motivation and learning strategies of them.