Physical education performance in primary and secondary school classroom education seems mediocre, and many students treat it as a minor subject. With parents and schools paying attention to physical education, physical education performance is very important now, whether in primary and secondary schools or universities. Nowadays, college students have less and less physical exercise, and sports achievements are one of the index achievements for evaluating scholarships. Many teachers also arrange appropriate physical training reasonably in order to improve students’ sports achievements. Forecasting sports achievements is the key to making scientific sports training plans. According to the study of college students’ group sports achievements, we can predict students’ follow-up learning achievements, collect, sort out, and study students’ sports achievements information regularly, so as to better guarantee the quality of college sports teaching. This paper compares and analyzes the sports achievements of various schools from the aspects of strength, endurance, and sensitivity. At the same time, it compares and analyzes the sports achievements of major universities. Finally, from the perspective of physical education, this paper analyzes and studies the situation of individual schools and then finds out the pedagogical factors that make them achieve excellent physical education results and puts forward the teaching strategies to improve physical education results, as well as the positive influence of physical education teachers on teaching, school physical education results, and students’ physical quality. When teaching courses, use big data mining technology to find qualified students to study the actual teaching needs, recommend an efficient learning curriculum system to students with reference to research conclusions, and give rich material resources. Teachers can teach students in accordance with their aptitude with reference to their usual learning situation, instead of “one size fits all.”

1. Introduction

In today’s new era, physical education is particularly important. Under the expectation of the country, schools, and parents, students’ physical education achievements in schools always affect students’ general subject achievements. For the prediction of students’ sports achievements, teachers arrange a series of different sports trainings to teach students in accordance with their aptitude. Literature [1] expounds the influence of extracurricular sports on academic achievements. References [2, 3], respectively, use particle swarm optimization and PSO-SVM learning methods to predict sports performance. Literature [4] proposes a sports performance prediction algorithm combining gray prediction features with CNNs and optimizes the sports competition performance prediction model of extreme learning machine based on the Drosophila algorithm in literature [5], so that the performance prediction can achieve higher prediction accuracy and computational efficiency. Reference [6] uses factor analysis to study and analyze the performance of the prediction model. Literature [7] puts forward a prediction model with higher accuracy, which provides a new research idea for the prediction of sports achievements. Using the information technology in reference [8], this paper predicts and analyzes the sports training plan and quantitative load arrangement. Based on the machine learning in literature [9] and literature [10], this paper predicts and analyzes the college sports achievements. For the calculation of students’ endurance scores, the management system of students’ exercise prescription based on BP neural network in literature [11] is used. Through the hybrid genetic neural network in literature [12], the special performance of athletes is predicted. Literature [13] expounds the neural network prediction model to predict the performance of speed skating. Literature [14] expounds the application and prospect analysis of artificial neural network in competitive sports. Literature [15] studies the analysis of influencing factors of sports performance and the ways to improve sports performance mentioned in literature [16]. Literatures [17, 18] analyze various factors affecting sports performance. This paper analyzes the influence of school factors on students’ sports achievements [19]. Literature [20] uses intelligent technology to improve the reference value of academic performance and models based on the LSTM model in literature [21] and puts forward the main factors affecting academic performance and judgment methods. The trend of performance change is proposed to analyze the factors affecting performance in the literature [22]. Literature [23] expounds the application of multiple linear regression in college achievement prediction. Literature [24] predicts and empirically analyzes CET-4 scores, while literature [25] judges the difficulty of learning and studies and analyzes the accuracy of scores.

2. Influencing Factors and Improvement Methods of Sports Achievements

2.1. Influence of Physical Training on Sports Performance

Physical fitness plays a significant role in promoting human health, developing physical fitness, and improving immunity. The diversity of training also determines the enthusiasm of students to carry out physical training in physical education. Due to the relative lag of physical training, there are few studies on the influence of physical training methods on sports achievements. Under the background of the new curriculum standard of physical education and health, students cannot do without physical training if they want to have sports skills and get strong physique. Offering physical education courses in schools can effectively improve students’ strength, speed, sensitivity, and other sports skills and promote students’ healthy development. Training in this area has a positive role in promoting. It should be pointed out that in the teaching process, teachers should, according to the characteristics of sports events, focus on developing students’ physical fitness. According to different training purposes, physical training to improve students’ fitness ability can usually be subdivided into general physical training and special physical training. Among them, general physical training is the basis of special physical training. The main purpose is to improve the individual’s physical function level and promote the all-round development of physical quality, while special physical training is highly related to special physical training, which mainly develops the specific sports quality needed to complete specific skills and tactics in special projects, and is the basis for improving and enhancing special sports ability. Reasonable training of physical fitness can certainly promote the best sports performance. The framework diagram of the physical fitness system is shown in Figure 1.

2.2. Effective Means to Improve Sports Performance

In recent years, with the implementation and promotion of physical education entrance examination in all parts of the country, the scores of physical education entrance examination have been improved in some areas, which make the importance of preparing for physical education entrance examination increasingly prominent. This paper analyzes the present situation of endurance quality of junior high school students. It is found that the third grade students’ knowledge and understanding of endurance events are not complete, their scores in endurance events are relatively poor in physical examination, and their enthusiasm for physical training is not high at ordinary times. A large number of students feel that endurance events have great room for improvement. Junior three students are not interested in endurance quality training, and their grades improve slowly. The reasons can be summarized as three aspects: monotonous practice methods, insufficient venues and facilities, and insufficient attention from schools and parents. For endurance training, change the traditional training methods, increase the interest of training content, and stimulate students’ training interest and participation enthusiasm. Increase the publicity of endurance quality, improve students’ understanding of endurance quality, and strengthen the communication and contact between schools and parents through the combination of home and school. Increase capital investment, improve equipment and site construction, improve teachers’ teaching ability and professional quality, and create good conditions for students’ endurance quality training. The framework diagram of endurance training is shown in Figure 2.

2.3. The Influence of Physical Condition on Performance

Students’ training performance is determined by their comprehensive physical condition. In the research, it is difficult to judge that a certain physical health index has a direct impact on sports performance. This time we studied the influence of a single index on 1000 m long-distance running. The morning pulse is analyzed, as shown in Figure 3.

In Figure 3, we can see that the morning pulse is unstable because the physical strength and physical condition are exhausted during training, so the performance drops obviously in the subsequent training.

3. Deep Learning Sports Performance Prediction

Deep learning framework diagram is shown in Figure 4.

3.1. Neural Network

BP neural network is the fastest and most flexible algorithm among all neural networks, which can accurately describe the curve change characteristics of the system. Let the input of the system be x (i), i = 1, 2, …, n, and the expressions of the input vector and output vector of the hidden layer of the neural network are as follows:

The input vectors and output vectors of the output layer are as follows:

3.2. Optimize Neural Network Algorithm

According to the luminous form of fireflies to attract other fireflies [26], different positions of fireflies represent different solutions of the problem and the luminous brightness corresponds to the fitness function. The greater the fitness value, the stronger the luminous brightness. The firefly with weak brightness is getting closer and closer to the firefly with strong brightness and keeps approaching the firefly with the strongest brightness. Because the position of the firefly with the strongest brightness indicates the optimal solution of the problem, the following assumptions are made first:

Brightness is determined by the fitness function of the problem: on the basis of satisfying the above assumptions, the brightness and attractiveness of fireflies are defined to represent the brightness of a firefly itself, so the received brightness of fireflies is as follows:where is the light intensity absorption factor and r is the distance.

The attractivity calculation formula is as follows:

If r = 0, the attractivity is .

If the two fireflies are i and j, respectively, and the condition  >  is satisfied, then i is attracted by j and changes its position; that is, there arewhere is a random number.

The attractivity coefficient is improved, and the change rule is as follows:

At the initial stage, the value of is relatively large and the search range is large, as the number of moves increases. With the increase in iteration times, the attractiveness becomes smaller, and the optimal position can be found quickly.

3.3. Sports Performance Prediction Model

The data value of sports achievements can be regarded as a group of irregular time series, and the fitting state model of sports achievements described by a multivariate statistical characteristic equation is as follows:

Through discrete analytic processing of data, the information entropy of distribution characteristics of sports achievements is as follows:

The auxiliary spatial test cumulant of time series X for sports achievements is as follows:

Take the average value

For the statistical sequence x (n) of sports achievements with multivariate variables, the statistical characteristic quantities of sports achievements expressed by scores and generalized integro-differential equations are as follows:

Constraints are as follows:

The constraint variables of environmental factors are

In order to maintain the initial data characteristics of sports achievements, the statistical model is reconstructed:

The information characteristic state equation of statistical sports achievements obtained by the mathematical method of least square estimation is as follows:

The control objective function of sports performance prediction is as follows:

denotes prediction factor and denotes prediction plus carrier.

The probability density functional coefficient of sports performance prediction is as follows:

The confidence degree of the accuracy of sports performance prediction is as follows:

The eigendecomposition values of the self-similar regression model meet the following requirements:

4. Experiments

4.1. Contents of the Experiment

A 1000 m long-distance running of 30 college students in a university is selected as an experimental object. The neural network adopted has three input nodes and one output node. The population of the particle swarm optimization method is m = 20, and the initial value of inertia weight is 1. With the gradual decrease in iteration times, the inertia weight is reduced to 0.5. When the iteration times reach the maximum, the iteration stops. Comparing the prediction data of this method with other common neural network optimization methods, the results are shown in Table 1.

In the data of Table 1, the prediction error and convergence time of the BP neural network algorithm are obviously better than those of the other two sports performance prediction methods. Judging from the prediction error of college students’ sports achievements, the prediction error of this method is always lower than that of the other two methods, which shows that BP neural network method has the best prediction effect.

Ten college students were randomly selected to predict the performance of 100-meter sprint, which were compared by the BP neural network method, GDX method, and LM method under deep learning. Channel theory formula is a method of buying and selling stocks by technical means and empirical judgment. This formula smoothes and corrects the trend line, which reflects the running law of stock price more accurately. When the stock price rises to the pressure line, investors sell the stock, and when the stock price falls to the support line, investors make up accordingly. The difference between the predicted results and the actual demerits is tested as shown in Figures 57.

It can be seen from Figures 57 that the difference between the actual value and the predicted value of sports achievements predicted by the BP neural network algorithm is not big, and the other two algorithms have large errors, which shows that the prediction efficiency and accuracy of sports achievements predicted by the BP neural algorithm are higher.

4.2. Simulation Experiment

The samples are selected from the sports achievements of college students, including 3 km long-distance running, 100 m sprint, skipping rope, archery, and yoga. Various sports can also test the universality of the prediction model under deep learning. All the data are analyzed according to three methods, and the waveform description diagram of sports performance prediction is obtained as shown in Figure 8.

The prediction effect of sports performance is universal in various events, as shown in Table 2.

The results of five sports events were compared with the prediction model such as genetic algorithm [27] and particle swarm optimization algorithm [28] as shown in Table 3.

The fitting accuracy is the deviation between the predicted value and the actual value in the past five years after we make a time series model and determine the model parameters.

Taking the sports performance statistics collected above as the test sample set, the simulation analysis of the sports performance prediction model is carried out, and the comparison results of prediction errors under different algorithms are obtained as shown in Figure 9.

Test the performance of the model to evaluate whether the model for predicting sports performance has practical applicability. The test results are as shown in Table 4.

4.3. Model Comparison

The prediction accuracy of the model is compared, as shown in Figure 10.

4.4. Comparison of Optimization Algorithms

Comparison diagram combined with the firefly algorithm and neural network BP algorithm for optimization and improvement is as shown in Figure 11.

5. Conclusion

Reasonable physical training in schools can improve students’ physical quality on the one hand and get higher physical performance on the other hand. In order to improve students’ overall academic performance, teachers make different physical training plans according to different physical indicators of each student, so as to make students’ predicted physical performance reach the best. Combined with the deep learning method, this paper constructs an optimization model that can make the predicted value of sports achievements achieve high precision and low error, as follows:(1)The BP algorithm, LM algorithm, and GDX algorithm are compared with the difference of prediction error and convergence time. It is obvious that the error value of the BP neural network algorithm is obviously lower than that of the other two algorithms, and the convergence time is also higher than the other two algorithms.(2)According to the three algorithms, the difference between the actual value and the predicted value of sports achievements is compared, and it is obvious that the actual value and the predicted value of sports achievements predicted by the other two algorithms are quite different.(3)According to the simulation experiment, different sports items are selected to test the universality of the model. The fitting accuracy and prediction accuracy of the firefly optimized BP algorithm are obviously more than 5% higher than those of the three models.(4)By comparing the performance of three algorithms under deep learning, it can be concluded that the BP neural network model is more practical.(5)Select five different samples, mathematical model, neural network algorithm, optimization algorithm, and other four different models to test and compare the prediction accuracy. The error under the mathematical model is the largest, and the firefly algorithm based on the neural network algorithm is the best.

Data Availability

The experimental data used to support the findings of this study are available from the corresponding author upon request.

Conflicts of Interest

The authors declare that they have no conflicts of interest.