Similar Problems Recommendation Model to Support Programming Education

Abstract

This paper proposes a recommendation model for similar programming problems to support programming education. In the proposed model, problem similarity is determined according to the similarity of source codes, in terms of the term frequency-inverse document frequency and the effort required to solve the given problem, as calculated according to Halstead metrics. The proposed model can be used to improve student understanding of a programming concept by solving many similar problems simultaneously. In addition, teachers can diversify similar programming problems during practical exercises, assignments, quizzes, and exams. The first experiment carried out in the Aizu Online Judge showed that the user’s accuracy when solving a problem was correlated to the user’s accuracy for a similar problem and, the second experiment showed a matching rate of 70% between the result of our recommendation model and the observations of a teaching assistant involved in programming classes.