<p>During the post-pandemic period, educational video-based systems were prominent and influential in e-learning contexts. In our prior study, we developed a rule-based personalized video environment to enhance the learners’ learning experience in programming instruction. The environment delivered instructional videos based on learners’ learning styles and preferences by leveraging various learner attributes. One significant limitation of this technique is the static nature of the video prediction rules, which constrains the adaptability of learners’ evolving requirements and cognitive progress. To address this issue, we are proposing a learning classifier system (LCS) for e-learning environments that rely on a rule-based machine learning paradigm. It dynamically generates video prediction rules from the learning environment, particularly incorporating cognitive feedback from students. An improved version of LCS called educational LCS (eLCS) is adopted, which autonomously adjusts to changes in learner preferences and updates the rules through the LCS architecture. The eLCS is facilitated by a genetic algorithm for the dynamic adaptability of video prediction rules. Results revealed that the proposed eLCS model was more accurate than traditional LCS models. Therefore, the proposed model is strongly recommended for cognitive feedback-enabled personalized instructional video environments, particularly for programming instruction.</p>