The research aims at adding to the knowledge in using big data technology in athlete selection and training in the sports domain. The approach is established based on a data-driven selection and training athlete model. This study is an inquiry of using in-game performance, physiological indicators, and training data to maximize selection and training. Random forest, gradient-boosted trees, and other statistical models and machine learning algorithms were used to execute characteristic selection and performance prediction. The case analysis results show that the application of these models has really optimized the level of athlete performance. The research does not just prove the actual effectiveness of a data-driven model in sports training but proves that the improvement of athlete performance via personal training planning indeed has practical meaning in athlete training. The practical value of the big data and intelligent analytic tool validation in athlete training is hereby achieved.