Data Analysis and Optimization of Youth Physical Fitness Training Based on Deep Learning

Abstract

Adolescents are the future of national development, but according to effective surveys, it can be found that the health of the youth system in my country is in a state of decline. At present, the reasons for the decline of the youth system in our country are caused by many factors, such as poor sports awareness and too much academic stress. The main reason is lack of exercise. Based on the deep learning method, this paper analyzes the importance of physical fitness training for adolescents, and proposes to improve the service system of physical fitness training for adolescents, and promote the formation of a guarantee mechanism for physical fitness training for adolescents. The research results of the article are as follows: (1) Before receiving the training, the test results of various indicators of the experimental group and the control group were basically the same, and there was no major difference. The T test results showed that the $P$ values ​of the two groups were both above 0.05. Explain that before training, the initial situation of the two groups can be regarded as the same. After receiving the special training, the general condition of the members of the conventional training group was slightly improved compared with the test before the training. Compared with the experimental group, they performed pull-ups, throwing a 2Kg medicine ball on the spot, running 30 meters, reaching a height on approach, moving half a meter, and repeating. The $P$ values of the cross-test scores are all less than 0.05, indicating a large gap between the two. Among them, the $P$ value of the 30-meter run is lower than 0.000, which has a very significant difference, while the $P$ value of the fast clean and jerk 20 kg and the 60s double shake is greater than 0.05. It can be seen that there is no significant difference between the control group and the experimental group after these two assessments. The experimental results also show that the trainees who received the mode training method have been improved in various indicators of physical fitness, and the experimental results and the traditional mode training have been greatly optimized. (2) In the simulation test analysis experiment, the statistical average of exercise time is 5.784, which is the highest statistical average among the five variables, and the statistical average of physical fitness is 2.436, which is the lowest in the statistical results. There is no significant difference between the statistical average of the quality and the daily exercise situation. In the sensitivity test, the evaluation accuracy of the deep learning methods is the highest among all models. When the number of iterations reaches 50, the evaluation accuracy can reach 1. (3) After running on the test set, the article proposes that the accuracy rate of the physical training model based on the deep learning algorithm is 89.12%, and the improved accuracy rate can reach 92.46%, which is the one with the highest index value among the four models in the experiment. The AUC curve values ​of the article and the improved system are very stable. The AUC value before the improvement remains around 0.90, and the AUC value after the model improvement also remains at 0.97. After running on the mixed test set, the performance of the four methods has declined to a certain extent, but the performance of the model proposed in this article is still the highest among the four models, and the AUC curve values ​of the improved system are very stable. Yes, the AUC value has been maintained at 0.95, and the AUC value before the improvement is stable within the range of 0.90-095. The research data also show that the recognition accuracy of the physical training method of the deep learning algorithm is the highest.