How to dig out informations from courses and conduct cluster analysis through effective data mining for online education is one of the problems to be solved. The topic distribution and classification of 1 472 courses from an online education platform were analyzed experimentally based on the text description. The text informations of 1 472 courses from the platform were collected, a customized dictionary and stop word list were constructed to do the word segmentation, and then the TF-IDF was employed to calculate the word frequency weighting. The topic distribution was recognized by using LDA and 230 topics were discovered.Both the document-topic distribution and topic-word distribution for each course text were obtained under the 230 topics. The hierarchical clustering for courses was completed based on the distribution similarity function and it is found that the courses were interrelated based on different levels of abstract topics. In the end, informations of 16 topics were visualized. This discovery of topics hidden in the semantics reflects the topic feature and the aggregate distribution of massive courses.