This paper collects physiological index data such as heart rate through real-time monitoring of wearable devices. Mining and extracting the relevant features of the physiological index data, we constructed a dynamic correlation model between the physiological indexes and the exercise load, and predicted the future physiological state of the athletes. The non-dominated sorting genetic (NSGA-II) algorithm is introduced to realize the multiobjective optimization and regulation of heart rate in the prediction of training load to enhance the training effect. The practical value of this paper’s method of combining real-time monitoring and genetic algorithm modeling is verified through multiple sets of experiments. The results show that the physiological data of athletes can be monitored and collected in real time at a frequency of 1 time per second by using a wearable device, and the data have research value. During the 4 stages of incremental load exercise, the muscle oxygen saturation of different muscle parts showed a decreasing trend. Combined with the method of this paper, real-time regulation was performed to maintain the decreasing muscle oxygen saturation at the 4th stage. In the physical fitness training experiment, the real-time heart rate prediction error of the athletes was optimized by the algorithm and adjusted to be consistent with the actual monitoring value, so as to realize the real-time accurate regulation of the exercise load in the training process.