Machine Learning Model Construction and Practice for Personalized Training Programs in Athletics Training

Big data science is a complexity science produced in the new era, and machine learning models belong to its main branch, which has characteristic methodological features and provides new ideas to scientifically solve the personalized formulation of training programs in track and field training. In this paper, firstly, the athletes’ sports data are collected by installing sensors in the key sports parts of the athletes, then the real-time state estimation of the athletes’ sports data is given by Kalman filtering, and the estimation is optimized by microelectromechanical technology. The obtained solution results are inputted into the important movement joint model of the human body so as to realize the motion capture of track and field athletes. Based on this, a personalized training model for track and field has been constructed using an ant colony algorithm. The generation of a personalized training plan is varied into an optimization problem with constraints, containing discrete and continuous variables. Then, the method of adaptation evaluation with constraints and the method of updating related solutions were proposed, thus completing the construction of the machine learning model. The experimental group improved much more in track and field events than the control group, and the experimental group improved 24.96% more in girls’ shot put. It shows that the training program developed through the personalized training model based on machine learning is more in line with the different students’ own needs, and the training program generated based on the machine learning method can provide track and field athletes with more efficient and personalized guidance, which verifies the effectiveness of the model constructed in this paper through the practice of the method and the design of the experiment.