Abstract
The focus of this article is to explore the application of artificial intelligence in university sports information services based on the development trend of artificial intelligence technology. Research and analyze the characteristics and functions of new intelligent information service tools, and explore the effects of artificial intelligence in optimizing university sports information services from the three aspects of intelligent evolution information service, intelligent push information, and intelligent retrieval information, and the connotation of intelligent environment Analyze characteristics and technical support to promote the optimization and upgrading of university sports information services, research on the transformation of evaluation methods from manual evaluation to intelligent evaluation, and from standardized evaluation to differential evaluation, and specifically analyze the connotation of intelligent evaluation and differential evaluation, Features and key technologies, analyze the general process of intelligent evaluation, and summarize the implementation suggestions for intelligent evaluation. And discuss the application of artificial intelligence in university sports information services from scientific decision-making and automated management.
Introduction
With the development of intelligence, machines will increasingly replace human physical labor. Humans who use labor to promote social progress will rely more on innovation and creation to promote social progress, and humans will begin to enter the intelligent society and the intelligent era. While artificial intelligence has a huge impact on people’s lives and work, it is also accelerating its integration into education [1]. A new era of disruptive smart technology transformation education is approaching. A new round of technological transformation education will be led by smart education for education [2]. Teaching injects new ideas, provides new methods and tools, drives the fundamental transformation of education and teaching models, and promotes the quality of teaching. At present, artificial intelligence has been able to improve teaching and learning efficiency in all aspects such as learning guidance, teaching evaluation and teaching space optimization, enhance the learning experience, and make personalized learning a reality [3]. Artificial intelligence is leading the innovation of education and teaching, and has become an important factor in the development of education informatization.
The applications of artificial intelligence in education mainly include: educational robots, intelligent tutor systems, intelligent answering systems, intelligent agents, learning analysis, etc. The research and application of artificial intelligence in education abroad started early, and the research scope is wide, including the technical research of educational artificial intelligence, such as robots, artificial neural networks, etc., as well as practical research on applying these technologies to teaching. Foreign education Artificial intelligence research shows diversified characteristics [4]. When discussing the role of Information Communications Technology(ICT) in education reform, Chan-Olmsted et al. proposed an analytical framework that can help understand the integration of information technology and teaching [5]. Professor Chen pointed out that “change” refers to the entire education Structural changes, rather than “minor repairs” to the education system [6]. Chin discussed how to improve the possibility of using artificial intelligence in the education process [7]. In an overview of the cognitive research progress of feedforward neural networks and teaching support systems, experimental methods are used to compare the effects of traditional teaching and personalized teaching [8]. Gutierrez-Estevez studied the artificial intelligence technology used by the adaptive education system in the online learning platform to develop an effective online learning environment, and discussed the use of these technologies to achieve better importance of an intelligent, adaptive online learning environment [9].
Through the analysis of relevant policy documents and academic literature on teaching reform, it is found that with the development of society, experts and scholars have made more in-depth research on teaching, starting from the generality to in-depth research on specific aspects [10]. Huang focused on the theory and practice of robot-assisted education based on analyzing the knowledge of educational robots, to stimulate the development of future education [11]. He emphasized that robot education can develop students’ potential and expertise, cultivate children’s creativity, and is of great significance to promote the development of future education. Hussain analyzed the reasons for the change in teaching methods, researched the direction of teaching method changes, and proposed that the reform of teaching methods is not a complete denial of teaching traditions, but is to optimize teaching on the basis of inheriting the advantages of traditional teaching or integrating traditional teaching wisdom [12]. It also analyzes the direction of the reform of teaching methods from four aspects: the optimization of teaching resources, the change of teaching organization, the change of learning activities, and the innovation of learning evaluation methods. Starting from the key characteristics of the big data era, Hwang analyzed the development trend of big data reform teaching, as well as the changes in the resource environment view, teaching view and teacher development view brought by the teaching reform, and analyzed specific teaching cases [13]. Kaartemo studied the teaching reform induced by big data from the perspective of big data, pointed out that big data strongly impacted teaching activities, and specifically discussed the reform of teaching thinking, teaching structure and teaching methods induced by big data, and analyzed and used big data [14]. The supporting conditions for inducing in-depth teaching reforms put forward teaching reform strategies from teaching content, teaching methods, and teaching organization [15–18].
University sports information services should be updated. This era requires us to use innovative thinking and methods to resolve various contradictions in university sports information services, so as to improve the quality of university sports information services. In the context of the coexistence of challenges and opportunities, optimizing university physical education information services is particularly important for improving the level of education information services and solving the current difficulties faced by university sports information services. The research is based on artificial intelligence technology, comprehensively optimizes and upgrades the level of university sports information services, considers the reform and innovation of university sports information services, aims to comprehensively improve the physical fitness of college students, and also build a more reasonable, scientific and optimized university sports information The service model ultimately achieves the goal of improving the physical fitness of students and optimizing and upgrading the quality of university information services.
Research on artificial intelligence models
Model basis
In the field of university sports applications, the premise of providing learners with personalized learning support services is to collect learners’ voice, emotions and other physical signs data, and through mining and analysis of these data, provide basic data models for subsequent personalized learning stand by. The application of pattern recognition in college physical education mainly includes: in practical classrooms, the recognized student action patterns can be compared with standard action patterns to guide students in operation; intelligently identify learners’ learning status, and provide learning assistance and encouragement in a timely manner; Learners use voice to search for learning resources.
The combination of big data and artificial intelligence will bring new opportunities to education and teaching [19]. Massive data is the cornerstone of machine intelligence. During the teaching process, the teacher interacts with the students online and offline through the random roll call function. During the period, teachers-initiated voting on the teaching content or released in-class time-limited exercises. The dynamic data generated during the teaching process included student check-in data, barrage data, random roll-in submission data, in-class voting data, and in-class answer data [20]. After class, the teacher pushes learning materials such as after-school exercises or extracurricular resources for students to complete the assessment or review after class. The ice and snow talent training class will present the students’ completion status to understand the students’ learning effect. The students in the after school will evaluate the course content or suggestions or individual teachers, therefore, dynamically generated data, including after-school students in after-school test data and feedback data. Big data has effectively promoted the advancement of machine learning and other technologies, releasing unlimited potential in the application of intelligent services. Big data has greatly boosted the development of artificial intelligence. The combination of big data and artificial intelligence will give full play to the advantages of big data. For example, there are a lot of teaching design and teaching data in the education and teaching process. The artificial intelligence model trained based on these data can assist teachers to find the deficiencies in teaching and make improvements [21]. Figure 1 is a working model of learning analysis in university sports information services. Learning analysis is a new concept derived from the rise of big data and data mining. It collects learner data related to learning activities, uses a variety of methods and tools to interpret the data comprehensively, and explores the learning environment and learning trajectory. Discover learning laws, predict learning results, provide learners with corresponding intervention measures, and promote effective learning. It can be seen that big data is the basis for learning analysis, and learning analysis can realize the value of big data.

AI model of information service work.
Intelligent resources and environment
Sports information resources are the general term for various elements in sports information activities, which include not only sports information itself, but also sports-related personnel, equipment, technology, and funds. Through in-depth research on artificial intelligence technology, and analysis and comparison of the advantages and disadvantages of several commonly used intelligent algorithms such as content-based intelligence, association rule-based intelligence, knowledge-based intelligence, and collaborative filtering intelligence [22]. The choice of personalized intelligent algorithms depends on the items to be intelligent zed and the characteristics of users. The collection of data generated after class mainly includes the completion of students’ after-class assessments and the results of after-class assessments, including students’ scores on test papers, class rankings, distribution of class answers, and the total duration of their after-class tests.
In the past, teaching only relied on homework and classroom attendance to evaluate teaching. It was unobjective and rigorous, leading students to ignore the usual course study and “complete the assignment” to “submit the assignment.” Forced by the pressure of the school on credit requirements. Surprise learning at the end of the course is less effective. The preview test questions in the evaluation of artificial intelligence, the creation of classroom tasks, and the creation of completed work can all be used as the evaluation content and evaluation method.
Therefore, the sports information resource guarantee system will be used as the research object to analyze and explore the characteristics of sports information resources and users to realize the use of algorithms for intelligent technology. Choose to provide a reference basis. The theory of human-environment matching has extended a complex multi-dimensional structure, covering five dimensions of matching between people and occupations, people and groups, people and jobs, people and organizations, and people and people. Each has its own analysis model [23].
The development of the teaching environment is the basis for promoting teaching reform. The new generation of learners put forward higher requirements for the construction of the teaching environment, such as intelligently perceiving the needs of learners and providing personalized learning services [24]. To satisfy learners’ demands for the teaching environment, the intelligent teaching environment has become an inevitable trend in the development of contemporary education environment.
The environment is smart, and the school is smart. It is necessary to cultivate smart people with the help of an intelligent teaching environment.
The single interpretation of R and E variables is completely inferior to the unique meaning of special behaviors or behavior results under special R-E interactions. Second, there are two basic structures in the analysis of the combination of human and environment: consistency and complementarity. Consistency refers to the assessment of the degree of fit between R and E through the same dimensions. Taking skills as an example, some teachers have the skills needed for a job, some teachers do not, or some jobs require teachers to have skills that other jobs do not need [25]. Complementarity refers to the behavior and reaction of R and E in mutual exchange. After all, teachers and the environment are not static and immutable entities, on the contrary, under their respective influences, they can and will inevitably change. The R-E matching model can well present the dynamic process of the interaction between the teacher and the environment (Fig. 2).

R-E matching model.
Based on the research of artificial intelligence technology and the characteristics of sports information resources and their users, this paper determines the selection of collaborative filtering algorithm as the basic algorithm for the construction of sports information resource intelligent models, and then further studies the implementation process and general model of collaborative filtering algorithm to determine. The construction of the user interest model and the sports information resource characteristic model is the basis for the construction of the sports information resource artificial intelligence model. A suitable similarity calculation algorithm and a suitable collaborative filtering model are selected to finally complete the construction of the sports information service artificial intelligence model to realize information services Intelligent. Therefore, when visually mapping and expressing the data in mixed teaching, we must combine the data type, attributes, and data values to map them to the corresponding elements of the visualization, obtain the visual structure of the data. We selected the appropriate visualization means to design and express its structure, convert the visual structure into a specific view for display, and intuitively display the content contained in the data so that users can feel the information more intuitively. The construction of the feature model of sports information resources needs to extract metadata that can represent the typical characteristics of various sports information resources from the complex data, and perform preprocessing operations on various sports information resources based on the extracted metadata, and re-describe the target. Imported and stored in the database of the system platform.
Therefore, the key technology of intelligent technology is how to obtain the set of items with the highest recommendation R for user c among the massive items, as shown in Equation (1).
Because users have a strong subjectivity in scoring items, and the criteria for scoring items are not completely consistent, only considering the user’s rating without considering the scoring criteria and personal subjectivity, the results obtained have a high error. The purpose of the modified cosine similarity algorithm is to solve the problem that the cosine similarity algorithm only considers the similarity in the dimensional direction of the vector and does not take into account the differences in the dimensions of each dimension. When calculating the similarity, the modified cosine similarity algorithm does each dimension subtraction. The correction operation of the mean value reduces the error caused by different user scoring standards by using the average scoring method. The modified cosine similarity calculation formula is as follows.
Determine the users nearest neighbor set based on the level of similarity, and then determine the neighbor user’s preference based on the nearest neighbor set and intelligentize the list of items that the current user has not tried.
Based on the characteristics of large amount of data of sports information resources, various types, multiple unstructured or semi-structured data, the algorithm of the intelligent technology of sports information resources chooses collaborative filtering algorithms, and collaborative filtering algorithms are usually divided into user-based collaborative filtering and There are two types of project-based collaborative filtering. However, in the actual application of intelligent technology, whether it is user-based or item-based collaborative filtering algorithms, there are certain defects, such as data sparsity problems, cold start problems, etc. The combined collaborative filtering algorithm that combines these two collaborative filtering algorithms can better solve these problems. The intelligentization of sports information resources also chooses to combine collaborative filtering algorithms, which can not only solve the cold start and data sparse problems caused by the relatively stable users of sports colleges, but also provide users with more accurate service information.
Resource optimization
The virtuous circle of high-quality resources, the intelligent identification and elimination of inferior resources, and the intelligent convergence and selection of resources with the same theme are still major research topics facing the evolution of teaching resources. Resource evolution requires stronger evolutionary power, a more complete evolutionary guarantee mechanism and more suitable evolutionary technical support. The goal of intelligent evolution of teaching resources is to achieve continuous self-renewal, continuous maturity development, and continuous adaptation to learners’ learning needs. Therefore, this research attempts to conduct a preliminary analysis of the evolution of learning resources from the perspective of autonomous and intelligent evolution of resources. The general processing flow based on artificial intelligence, the semantic modeling technology of integrated resources, the dynamic semantic association and the control technology of aggregated orderly evolution, etc. are constructed. The intelligent evolution process of teaching resources is shown in Fig. 3.

Intelligent learning resources.
With the support of big data, learning analysis, data mining and other technologies, providing teachers and students with a personalized teaching environment is an important direction for the development of the teaching environment. The intelligent teaching environment perceives physical location and environmental information, records the cognitive style, knowledge background and individual preferences formed in the teaching and learning process of teachers and students, thereby providing them with personalized teaching resources, tools and services. Intelligent learning is a learner-centered learning activity carried out in an intelligent learning environment. Learners can not only obtain the required resources and evaluation feedback in time, but also enjoy personalized learning support services, making learning easier, more efficient and interesting.
The intelligent information service platform uses artificial intelligence to understand each student’s knowledge points and skill operation level, reasonably group them, and complete specific tasks. Encourage students to study cooperatively. In an artificial intelligence society, a lot of work cannot be done alone. It requires a team to complete it. In the team, everyone uses their own advantages and sincerely cooperates. Supervise and guide each other through group learning, preview the learning materials pushed by teachers together before class, discover and solve problems together, and effectively cultivate students’ exploration ability; in class, teachers can discuss and express opinions freely on the problems raised by teachers. Through this process, you can also understand the students’ learning mentality and thinking; after class, you can complete group assignments together to cultivate students’ communication and cooperation skills. Through the exercise of cooperative ability, it is helpful to cultivate students’ advantages in competition with intelligent machines.
Information service quality refers to the service process of an information service organization or department and the level of the quality of the information products it ultimately provides. It is specifically reflected in the benefits obtained by the information service, the extent to which the purpose is achieved, and the result of problem solving, which is finally reflected in the user. The degree of satisfaction or satisfaction of both parties and the service organization. Information service quality evaluation refers to the evaluation of the pros and cons of information service work and its information service products and value estimation; the evaluation of pros and cons requires qualitative analysis, and value estimation requires quantitative analysis. The two are unified and complementary to form a complete quality evaluation system. Since information service itself has the characteristics of vagueness, perishability, lagging and indirectness, certain principles and procedures must be followed in the evaluation. A scientific, complete, and operable evaluation index system is the key to the smooth completion of the information service quality evaluation and the improvement of the reliability and validity of the evaluation results [3]. The evaluation index of information service quality refers to the identification that reflects the characteristics of a certain aspect of information service quality, and is the determination and description of the characteristics of the results of each stage of information service. The evaluation index system is a collection of these indicators, the standard and basis for information service quality evaluation, and a composite system with multiple levels and multiple indicators. The construction of the information service quality evaluation index system should follow the principles of objectivity, integrity, hierarchy and flexibility.
In information service quality evaluation, the evaluation techniques that are often used or for reference are mainly qualitative evaluation, quantitative evaluation, and evaluation methods that combine qualitative and quantitative. The semi-quantitative evaluation methods that combine qualitative and quantitative mainly include Delphi. There are two types: statistical analysis method, user questionnaire evaluation method + weighting method. However, no matter which evaluation method and technology are used, whether it is from the perspective of evaluating service level or service benefit, a set of appropriate evaluation index system must be constructed. The structure of the indicator system should pay attention to the following relationships: the hierarchical structure and weight of the indicator system, service costs and user perception, social and economic benefits, quantitative issues and measurement levels, short-term and long-term utility, direct utility and indirect utility.
Results and analysis
Intelligent model analysis
All sports resources of the university are not open to the outside world, which affects sports activities of residents in surrounding communities to a certain extent; resource venues for public welfare open services account for only 9.4%, and paid services and some paid services account for 28.7% and 14.3% respectively. The value of information-based instructional design lies in learners as the center, encouraging learners to actively participate and achieve the optimization of the teaching process. Although the latter has a certain positive role in promoting education informatization and education innovation, neither of them has a revolutionary impact on school education and teaching.
Most of the venues for external services are outdoor venues, such as table tennis, badminton courts, and venues for the three major events. The management costs of these venues are relatively low, and the opening objects are mainly for teachers and students of the school and surrounding residents; most of the paid openings are indoors. Venue resources, such as swimming pools, table tennis and badminton gymnasiums, gyms, tennis courts, etc., due to the special requirements of site materials, management personnel costs, and maintenance costs, so corresponding fees are charged to reduce the burden on the school. The sports resource service situation is shown in Fig. 4.

University sports resource services.
With the development of national fitness, some university facilities and resources have been opened to the outside world, witnessing surrounding residents using university sports resources for exercise. At the same time, more and more residents and organizations like to go to universities to carry out sports activities, alleviating the shortage of sports venues. However, the lack of communication with the university in the early stage and the lack of understanding of the specific management system of the university’s external services not only affects the residents’ sports development, but also brings great difficulties to normal teaching tasks, such as the outdoor venues of Shaanxi Normal University, residents’ exercise and University physical education is mixed together, if not handled properly, it will cause conflicts, which not only hinder residents’ physical fitness, but also interfere with the teaching order. In addition, the development of community sports activities requires relevant university planning and organization personnel and referees; secondly, the development of community sports activities is not limited by time, but due to work and curriculum reasons, teachers and students cannot leave their posts at will, and students cannot casually Skip class. Therefore, there is a certain conflict between university sports resources and community service time. From the error change curve of the training process, it can be seen that the LM algorithm converges very quickly, and the expected network error index is reached after only 4 epochs. The network mean square error MSE = 1.1246, and the regression analysis from the calculation results (as shown in the Fig. 5) It can be seen that the training accuracy of the network model is also quite high, which proves that the number of hidden nodes selected is reasonable.

Network training results.
The verification result of the university sports information service evaluation model is shown in Fig. 6. It can be seen from the figure that the average no difference between the evaluation scores output by the network and the scores of actual experts is less than 1%, reaching a very high accuracy, indicating that the evaluation model based on artificial intelligence technology is technically feasible, and the results are relatively reliable.

Model verification results.
With the rapid development of information technology, evaluation methods are becoming more automated and intelligent. Using technology to assist the evaluation of information services not only saves the cost of manual evaluation, but also greatly improves the timeliness and accuracy of evaluation feedback, thereby increasing the utilization rate of university sports information services. This experiment uses the CNN model for intelligent evaluation of university sports information services. Since learning efficiency and batch affect the classification accuracy and operating cost of the model, this experiment mainly optimizes the model for learning efficiency and batch, and uses the grid search method to determine the best To optimize the learning efficiency and batch, the ROC curve of the classification result is used as the evaluation index, and the best combination of learning efficiency and batch is finally obtained. The running result is shown in Fig. 7.

Intelligent evaluation model training classification results.
Among them, the accuracy rate of the category I evaluation result is 1, the accuracy rate of the category II evaluation result is 1, the accuracy rate of the category III evaluation result is 1, the accuracy rate of the category IV evaluation structure is 0.93, and the accuracy rate of the category V evaluation result is 0.5. As the learning rate increases, the learning ability of the CNN intelligent evaluation model gets better and better, but when the learning rate increases to a certain level, the prediction accuracy of the model does not improve; on the contrary, the classification accuracy will decrease. The reason is that as the learning efficiency increases, the learning ability of the model becomes stronger and stronger, and the prediction accuracy of the model is improved within a certain range. If the learning efficiency is too high, the network reconstruction error will increase rapidly, and the connection. The weight will also increase the most. As the Batch Size increases, the learning ability of the CNN intelligent evaluation model is getting better and better, but when the Batch Size increases to a certain level, the prediction accuracy of the model does not improve; on the contrary, the classification accuracy will decrease. The reason is that appropriately increasing the batch can effectively improve the utilization of data samples and increase the running time of the model, thereby improving the classification accuracy of the CNN model. When the batch is too large, the data sample information will be partially omitted, and the information cannot be guaranteed. Make full use of, thereby affecting the classification results of the model. In summary, when the learning rate is 0.1 and the batch is 256, the prediction effect of the CNN intelligent classification model is the best and can effectively provide classification results.
Aiming at the evaluation index prediction model, it can effectively fit the variation trend of evaluation index G1, G2, G3 and G4. Although the variation trend of G1, G2, G3 and G4 in the original data is very large, the LSTM prediction model can be simulated with smaller errors. The change trend of the evaluation index system can be combined to better reflect the superiority of the LSTM evaluation index system prediction model. In addition, the LSTM evaluation index system prediction model has a smaller floating range of fit error variation of 0.1, which has a higher accuracy, as shown in Fig. 8.

Fitting effect diagram of G1, G2, G3, G4 evaluation index system.
The university’s sports venue resources have some time for external services, but the university does not know what the community needs, and the residents do not know what sports resources the university has for external services. Therefore, the information between the two parties is in a state of “generation gap"; secondly, the residents do not know. The university’s opening management system does not know the specific management department; on the other hand, the university does not have a relevant management platform, and the usual relevant management system and opening hours are only posted in the venues, which does not give full play to the role of “Internet+”, which is important for fitness knowledge. Publicity is even trivial. The interruption of information between the university and the community is an important factor hindering residents from going to the university to participate in exercise. University sports resources serving community sports lies in the attitude of the management. The shortage of community sports resources has the problem of lack of communication and interaction between the managers of both sides. The community has not been brought in, and the university has not gone out. It only develops relevant management systems in various fields., Failing to maximize the value of existing resources; in terms of human resources, college teachers and students have insufficient service awareness and rarely participate in community sports practice activities, which gradually widens the distance between the two, thereby blocking university sports resources from serving the community Sports development. In order to improve the current situation of resource imbalance, the two parties should establish corresponding institutions to coordinate the development of the university and community sports planning and construction, coordinate the development between the two, communicate with each other, solve current problems, and make the university’s material and human resources move towards the community, towards the market, and serve. Society (Fig. 9).

Intelligent application results.
After a lot of searches for related documents on “University Sports Information Service”, “University Sports”, and “University Sports Index System”. After the questionnaire was designed, the validity of the questionnaire was tested by expert judgment, which was carried out from the overall design, content design, and structure design of the questionnaire. The results of the expert’s test of the questionnaire validity are shown in Fig. 10.

Evaluation results of expert questionnaire.
Open the way for the university to serve the community. The concept is the forerunner of innovation. University leaders change new ideas and thoughts, change the way of development, and plan new strategies for university sports resources and community development from the direction of national fitness needs, and give full play to the university’s superior resources, In order to facilitate the physical activities of the fitness group and stimulate more people to participate in physical exercise. The implementation of the plan is inseparable from the guidance of the system, so the new system is running, which can not only promote the development of university sports, but also activate the whole people’s interest in fitness, thereby broadening the university development market, and enabling people of all ages to participate in sports. While improving the physical and mental health of the whole people, it is also changing the public’s awareness of sports and providing protection for lifelong sports.
After artificial intelligence entered the field of information services, changes in technical support resources and the environment prompted a series of changes in information service methods. In terms of information services, the development process of university sports and intelligent information services are analyzed, and an intelligent information service model is constructed. In terms of learning, the development process of learning and the connotation of intelligent learning are analyzed, and the general process of intelligent learning is explored from the aspects of adaptive previewing new knowledge, intelligent interactive learning, intelligent companion practice, and intelligent guided deep learning. The university sports information service management supported by artificial intelligence is becoming more and more automated, intelligent and scientific. First of all, in terms of evaluation, the research pointed out that the evaluation method in the artificial intelligence era is changing from manual evaluation to intelligent evaluation, and from standardized evaluation to differentiated evaluation. In the aspect of university sports information service management, the application of artificial intelligence in university sports information service management is discussed from two aspects of scientific decision-making and automatic management. Future research should strengthen the cooperation between artificial intelligence professionals and experts in the field of information services, pinpoint the points of convergence between artificial intelligence and information services, strengthen the practical application of artificial intelligence in information services, and promote faster and better development of information services.
