As people use non-face-to-face learning frequently, interest in adaptive learning for efficient learning is increasing. In order for students to easily learn, it is necessary to set up a learning path that considers students' learning outcomes. However, as everyone recognizes, learning pathways are not created based on learning logs, but rather rely on the curriculum that experienced teachers have used. However, the answer to the question “Can today's students learn efficiently with the existing curriculum system?” is perhaps not certain. In addition to studies whose subjects were creating a learning path, this study uses a structural equation model, Hidden Markov Model (HMM) and inductive reasoning to create a hierarchical learning path between latent factors at a higher level. Using the mathematics learning data of company D in Korea, a learning path of concepts rather than a learning path of subdivided concepts was created. It was confirmed that the subdivided concepts showed reliable observation values for the upper concepts and that the subdivided concepts conveyed the contents whose concepts were conveyed with validity.