A Recommendation Module based on Reinforcement Learning to an Intelligent Tutoring System for Software Maintenance

Abstract

The demand for qualified professionals to work with Software Maintenance (SM) brings challenges to computer education. These challenges are related to SM’s inherent complexity and the teacher’s significant work in providing adequate support in practical SM activities. In this context, Artificial Intelligence (AI) based techniques, such as recommendations, can play a central role in developing Intelligent Tutoring Systems (ITS) to focus the teaching-learning process. The literature points out a lack of ITS to SM and that most of them do not use AI-based techniques to recommend content to the students. In this work, we present an Expert Knowledge Module (EKM) for an ITS specially designed for SM. To model the EKM content, we did a deep analysis of the ACM curricula regarding SM topics and the syllabus related to SM from all Brazilian public universities. The content recommendation engine uses the Q-Learning algorithm, a well-known Reinforcement Learning (RL) AI-based technique. Using simulation-based experiments, we could verify the efficiency of the Q-Learning-based recommendation mechanism to propose contents using the ITS’s EKM properly. This work highlights how AI-based techniques can enhance and improve SM’s teaching-learning process using ITS and advance this research area.