An adaptive pedagogical model based on the classifier system and reinforcement learning

Abstract Educational strategies and instructor feedback to learners in pedagogical models in the intelligent tutoring system (ITS) affect students’ academic progress and learning speed. In this paper, we used the learning classifier system (LCS) of genetic algorithms (GAs) and reinforcement learning (RL) to develop instructional policies for ITS. In this research, we introduced three models. In the first model, we used LCS. Second, we used reinforcement learning, and in the third model, we integrated an LCS of GA with reinforcement learning to build a hybrid model called an Adaptive Pedagogical Model Based on a Learning Classifier System and Reinforcement Learning (PMCR). The primary aim of this work is to create and implement an adaptive educational model that utilizes the advantages of GAs and reinforcement learning to construct a learning environment that is personalized and adaptable for learners. The proposed model begins with generating a random population of conditions and actions, where the condition represents the evaluations for the problems presented to the educator, and the action is the teaching policy (tactic) that should be considered for such kind of student. In the end, the system should come up with a new population that can mostly produce the right teaching policy based on the student’s responses to the given problems. The study’s findings indicate that the third model (PMCR), which combines LCS with RL, is the best model compared with the other proposed models. The accuracy of PMCR reached 82% when the model was tested by different problem cases.