Learning Path Construction Using Reinforcement Learning and Bloom's Taxonomy

Massive Open Online Courses (MOOC) often face low course retention rates due to lack of adaptability. We consider the personalized recommendation of learning content units to improve the learning experience, thus increasing retention rates. We propose a deep learning-based learning path construction model for personalized learning, based on knowledge tracing and reinforcement learning. We first trace a student's knowledge using a deep learning-based knowledge tracing model to estimate its current knowledge state. Then, we adopt a deep reinforcement learning approach and use a student simulator to train a policy for exercise recommendation. During the recommendation process, we incorporate Bloom's taxonomy's cognitive level to enhance the recommendation quality. We evaluate our model through a user study and verify its usefulness as a learning tool that supports effective learning.