<p>With the increasing awareness of health, the demand for personalized exercise training is growing. Traditional training methods cannot meet individualized needs, so modern technology is needed to achieve personalized recommendations. The aim of this study is to design a personalized exercise training recommendation system based on motion sensors and data mining. By analyzing exercise data, it provides users with scientific and effective exercise recommendations. A personalized sports training recommendation system was designed, and the overall architecture of the system was designed to ensure its scalability and maintainability. Evaluate the recommendation effectiveness of the system, collect real-time user speed and acceleration data, extract effective motion features, and store and manage them. By combining hybrid recommendation methods, we aim to improve the diversity and accuracy of recommendation results, as well as address the issue of cold start. By introducing basic user information and initial testing data, we can quickly generate initial recommendations. The experimental results show that the designed system can effectively collect and process sports data, and accurately recommend personalized sports training plans through data mining algorithms, with high practicality and accuracy.</p>