This study aims to create learning path navigation for target learners by discovering the correlation among micro-learning units. In this study, the learning path is defined as a sequence of learning units used to realize a learning goal, and a period used for realizing the learning goal is regarded as a learning cycle. Furthermore, the learning unit datasets are extracted according to the learning cycle. In order to discover the correlations of learning units, we proposed an algorithm named Bayesian Network Association Rule (BNAR), which is used to establish a dynamic learning path according to the learning history of reference learners group who achieved learning goals. Based on the successful learning history, the dynamic learning path navigation will help target learners to improve learning efficiency.