Learning management systems are typically used by teacher professional development to support teachers in managing and administrating training courses. However, teacher learning management systems are rarely adaptive. With the increasing amount of training courses, course selection becomes a difficult and time-consuming problem. In this study, we focused on the design and development of an adaptive course recommender system, which was integrated into Shanghai Teacher Learning Management System. Model-based collaborative filtering recommendation techniques with machine learning and data mining methods were used. The online methods, offline methods and user studies were designed to evaluate the recommender system. Results and implications are discussed.