Whether the selection of sports injury repair program is reasonable directly affects the rehabilitation cycle, the quality of functional recovery and the effect of regression training. Aiming at the problems that it is difficult to compare the rehabilitation process under different repair paths and the cycle judgment depends on experience, this paper constructs an analysis model that integrates rehabilitation data preprocessing, multi-source feature representation, cycle prediction and strategy optimization. Based on 192 sports injury samples and 5696 stage observation records, the clinical indicators, functional tests, training load and dynamic monitoring information were uniformly coded, and the rehabilitation cycle prediction framework of different repair schemes was established. The results show that the mean absolute error (MAE) of the proposed model on the test set is 5.0 d, the root mean square error (RMSE) is 6.4 d, and the determination coefficient reaches 0.95. After optimization, the average rehabilitation period was reduced from 78.6 days to 68.9 days, and the function standard rate was increased from 81.4% to 88.7%. The results show that this method can effectively reveal the recovery differences of different sports injury repair schemes, and provide data support for rehabilitation path optimization.
