The automatic evaluation model of physical education teaching based on two screening algorithms

Abstract

In a certain network environment, the use of teaching evaluation assistant decision-making system can further promote the rationality and fairness of teaching evaluation. Two screening algorithms are proposed, which combine with the influence factors in the automatic evaluation model of physical education teaching, delete the relevant factors and leave them behind. After two deep screening, the accuracy of the results is improved. By introducing the artificial neural network technology into the evaluation of physical education teachers’ teaching quality, the evaluation factors of neurons are calculated to establish the evaluation model of BP neural network. Secondly, the factors affecting the evaluation results in the evaluation model of BP neural network are decomposed and screened by using the second screening method, and a certain amount of training and learning is carried out for the teaching quality data. The experimental results show that the second screening algorithm is effective and can improve the accuracy of the results of automatic evaluation of physical education teaching. By establishing the automatic evaluation model of physical education teaching, it can provide reference for the evaluation and assistant decision-making of physical education teaching quality in Vocational colleges.

Keywords

Twice screening algorithm physical education teaching automatic evaluation

1. Introduction

With the progress of Internet and computer technology, more and more applications are applied in all walks of life. As a kind of intelligent algorithm, the twice-screening algorithm has been applied in many aspects. This algorithm has been introduced into various evaluation and assessment systems. Through data screening, it can further improve the scientificity and effectiveness of the evaluation and assessment system [1]. For the education industry, most colleges and universities are still using the traditional teaching evaluation model, but in the actual teaching process, the traditional teaching evaluation model can no longer meet the teaching requirements. The traditional evaluation data is usually expressed by specific scores. When the students are assessed, some specific data are used to express the students’ assessment results. The assessment data are too general to get more useful information from this data [2]. Therefore, two screening algorithms are introduced into the automatic evaluation of teaching system in order to further improve the scientificity and accuracy of the evaluation system.

The specific screening method is to arrange N natural numbers in order and screen the solutions satisfying the conditions according to certain conditions. The solutions that do not satisfy the conditions are deleted, and the solutions satisfying the conditions are saved, cycling repeatedly until the solution of the objective function is screened out, so as to form the sum of the solutions satisfying the set conditions finally. The two-stage filtering algorithm is to filter the objective function solutions satisfying the set conditions twice in order to improve the objective accuracy of the results [3]. In order to improve the reasonableness of the automatic evaluation of physical education teaching, the two-screening algorithm can be used to screen the factors affecting the results of the automatic evaluation model of physical education teaching twice, and then use the screened data to evaluate physical education teaching automatically. Therefore, the automatic evaluation model of physical education teaching is analyzed by using two screening algorithms.

The research is mainly divided into four parts: the first part is about the current development of teaching quality assessment in China and the background of establishing teaching evaluation system in various countries, which focuses on the significance and advantages of building teaching evaluation system; the second part is based on the theory and technology of systematic development, mainly introducing the methods used in traditional teaching evaluation and the establishment of teaching. The related technologies are used in the learning evaluation system; in the third part, two screening algorithms are put forward, analyzing the automatic evaluation model of physical education teaching; in the fourth part, the experimental test is carried on to the algorithm, according to the test cases used in the design of each plate, carrying on the preliminary test, continuously adjusted in the test process, so as to make the individual plate run smoothly. The purpose of cohesion of the whole system is unambiguous.

2. Related work

In recent years, the two screening algorithms have been widely used to improve the accuracy of evaluation results, which has attracted the attention of many scholars, and a lot of related research has been carried out. Zhou S et al. deduced and validated a new screening algorithm, combining N-terminal B-type natriuretic peptide with PFT to reduce the dependence on TTE. The research proves that this algorithm has certain limitations [4]. Wei X designed a screening algorithm for type 2 diabetes mellitus to optimize the sensitivity and specificity of identifying undiagnosed patients with DM, as well as the affordability of health systems and individuals. Baseline demographic and clinical data were used in mathematical models and analyzed by target screening algorithm. It is found that the combination of risk assessment and fasting blood glucose measurement algorithm is superior to most resource-constrained settings (sensitivity 68%, sensitivity 99% and cost of $2.94 per diabetic patient) [5]. Yang X et al. proposed a simple screening algorithm that combines fasting blood glucose (FPG) measurements and multivariate risk estimation models to screen fasting blood glucose values in 25 diabetic patients to screen individuals with normal FPG levels and decide whether to direct further treatment [6]. Li M et al. proposed an indirect hidden Markov model (IHMM), which estimates the transmission probability by feature similarity and the conversion probability by a distorted model based on heuristic distance. The reliability of the two filtering algorithms was analyzed [7]. Rebar A L et al. designed an interval screening method based on known primes to find unknown primes, improving the speed of the algorithm by optimizing the screening process [8]. Osorio G et al. proposed an automatic screening system based on transfer learning algorithm to solve the problem of insufficient data in physical education teaching evaluation. By constructing a fusion model of CNN+ SVM, the correct data classification is realized, and the high accuracy classification and recognition [9]. Svennberg E et al. proposed a weighted polynomial screening algorithm based on least squares polynomial curve simulation data sequence. Relevant experiments are carried out using Chuhe River reservoir water level data. The results show that the screening algorithm can effectively screen aging data, improve the accuracy and efficiency of data analysis results, and thus assist in improving decision-making ability [10]. Koppers L et al. proposed an effective network feature selection algorithm for APT samples based on k-means++clustering. The original feature set extracted was divided into APT traffic feature set and background traffic feature set based on clustering idea. Then, the degree of clustering performance change after removing one dimension feature vector was calculated. Finally, the result was used to evaluate the clustering performance. The experimental results show that the algorithm is feasible and has some advantages over other filtering algorithms [11]. Gu X et al. applied intelligent algorithm to measure the longitudinal change of childrens’ sports and sports motivation, making an effective evaluation of childrens’ sports teaching. The results show that the combination of neural network and analytic hierarchy process can make a good evaluation of sports teaching [12]. Bozkurt et al. analyzed the students’ understanding and views on value education, classifies the students’ evaluation results using clustering algorithm, the types of students summarized based on different motivations, providing reference for the teaching of physical education teachers [13]. Mayorgavega D. et al. conducted a randomized controlled trial to analyze the influencing factors of health-related physical fitness of senior high school students, which provides a reference for health education evaluation of students [14]. Kwon E H et al. put forward a network learning scheme for implementing adaptive sports in P.E. teacher education project. Two screening algorithms were used to process the data. The results show that the efficiency of two screening algorithms for health data processing is higher than that of traditional algorithms [15].

According to the research on the two-time screening algorithm by scholars of China and other countries, most of the current research focuses on the hospital data evaluation and sample data screening. Therefore, the research on the two-time screening algorithms’ sports automatic evaluation model has important theoretical and practical significance [16].

3. The algorithm of automatic evaluation model in physical education teaching

3.1 Artificial network technology under twice screening method

Screening method is also called screening method. The concrete method is to arrange N natural numbers in order first. 1 is not a prime number, nor is it a composite number [17, 18]. 1 should be crossed out. The second number 2 is the prime number left behind, and all the numbers that can be divided by two after two are delimited. The first number left behind is 3. Leave 3 behind. Then cross out all the numbers that can be divided by 3 after 3. The first number that is not crossed out after 3 is 5 [19]. Leave 5, and then cross out all the numbers that can be divided by 5 after 5. If keep doing this, all the summations will be screened out that do not exceed N, leaving behind all the prime numbers that do not exceed N. Because the Greeks wrote numbers on the wax-coated board, and every time they crossed out a number, they wrote small dots on it. After the work of seeking prime numbers was finished, many small dots were like a sieve, so the method of Eratosthenia was called “Eratosthenia sieve”, or “sieve method” for short. Another explanation is that the numbers were written on papyrus at that time. Every time a number was crossed out, the number was dug out. After the work of seeking prime numbers was finished, many small holes were like a sieve. Based on this, artificial neural network (ANN) is a new subject after the popularization of computer. ANN is a complex calculation method to simulate the connection structure of human brain neurons and neurons. The ANN technology mainly takes the workflow of human nerve as the design model and calculates by the way of processing related content of human nerve. ANN describes the brain function with mathematical method. The nature and ability of information processing has become a hot topic of common concern in many disciplines, such as biology, cognitive science, mathematical science, computer science, automatic control, robotics, microelectronics and so on. It has become a typical interdisciplinary science involving a wide range of fields and has been applied more and more widely [20]. Figure 1 is a basic neuron model. The main activation functions in this model include threshold function, piecewise linear function, sigmoid function, and a threshold value of theta (or bias-theta). This is one of the three basic components of the basic neuron model. The other two are a group of connections and a sum unit. In the connection, the sum of the data is a regular representation of activation, which can be expressed by activation function, and vice versa, it can be suppressed.

Fig. 1

Basic neuron model.

The basic structure of a BP neural network is shown in Fig. 2 below.

Fig. 2

Three layer BP neural network structure diagram.

3.2 Neural network algorithms based on the automatic evaluation model of physical education teaching

The mathematical formulas involved in the training process of BP neural network are as follows: $x_{i} = \frac{x_{max} - x_{i}}{x_{max} - x_{min}}$ (1)

In formula (1), x_i is the input component, and is the first neuron cell quantity before pretreatment. x_min and x_max are the minimum and maximum values, respectively, and are the minimum and maximum values of all input components of the first neuron cell.

In the first step, we need to prepare the data that meet the requirements of the indicators, and take the qualified data as the input training samples of the network, then correctly construct the BP neural network and train the network. In the second step, the training network established needs to be evaluated in the first step. The third step is to get the training effect of the network, and then get a real and effective evaluation model of the neural network. In BP neural network, set the input layer as X₁ = (x₁, x₂, … x_n), where the n value ranges from 19 to 22, and can also write the input of X1, X3, X4, Xn accordingly; the output layer is 1, which can be written as 0; the hidden layer is: $h_{j} = f (\sum_{i = 1}^{22} v_{ij} x_{j})$ . The volume neural network model of the teacher’s teaching evaluation system based on neural network is shown in Fig. 4. From the above, the value of J in the graph is in the range of 5 to 14. Among them: X1 to X n are n input values. In the system, it refers to the variables of teaching evaluation indicators and corresponding quantified values. N equals 20, V1 to V2*n*j refers to the connection error value; Y1 to Y n refers to the number of hidden layer nodes and corresponding values. Among them, n is controlled to the range of 10-14, W 1 to Wj are the weight correction weights of each layer, and finally an output layer of evaluation results.

Fig. 3

BP neural network algorithm flow chart.

Fig. 4

Teaching quality evaluation model based on BP neural network.

3.3 Twice screening algorithms

When quadratic screening (QS) is used to decompose the modulus N of RSA, it is necessary to solve the linear correlation series of large sparse matrices, and the computational complexity of various algorithms is directly related to the size of the matrices. At present, the most commonly used linear correlation column algorithm in quadratic screening method is block Lanczos algorithm. For sparse matrices with size m₁ × n₁ (m₁ × n₁), the computational complexity is about $O ({dn}_{1}^{2} / W) + (n_{1}^{2})$ , where D is the average number of non-zero elements in the matrix, and W is the machine word length. Two sparse matrices E and V are involved in QS. E stores the decomposition coefficients with smooth relation obtained by screening. V is the matrix on which E is transformed into GF (2). The scale of E is m × n (m > n) and N is about $g (\frac{1}{2} + α (1)) \sqrt{log M log log V}$ . It can be seen that with the increase of N, the scale of E and V increases very rapidly. Therefore, effectively reducing the size of matrix can not only save storage space, but also reduce a lot of computing time. When using QS to decompose N of decimal 70 bits, we find that if we do not delete all 0 rows and columns in matrix V, because the rank of matrix is too small, the block Lanczos algorithm cannot find the linear correlation columns, which will eventually lead to the failure of decomposition N. Therefore, when the number of N increases to a certain extent, it is necessary for QS to reduce the size of the matrix. In order to store sparse matrices in computers, how to save storage space should be considered, but also consider the storage structure to facilitate various operations on matrices. According to the principle of quadratic screening, there is at least one non-zero element in each row of the sparse matrix. Based on the above factors, a special linked list is used to store the sparse matrix (Fig. 5).

Fig. 5

Sparse matrix storage structure.

In this storage structure, each non-zero element of a matrix is represented by a node. It consists of three parts: the value of the matrix element, the number of columns in the matrix and the pointer to the next node in the same row. All the nodes in each row of the matrix form a list. The linked list corresponding to different rows of a matrix is associated with a one-dimensional pointer array H of length M. Each element of this one-dimensional pointer array corresponds to a row of sparse matrix and points to the first node of the linked list corresponding to that row. The row numbers of each element of a matrix are not stored in row numbers’ corresponding nodes, but can be calculated by row numbers’ subscripts to the elements that should be in H. In order to describe the algorithm conveniently, the symbols in QS are as follows: H(i): indicating the pointer to the first node in line I of E; let Po be the pointer to a node >po.e., po.col and po.next respectively to indicate the element value, column number and address of the subsequent node as indicated.

4. Experimental design and analysis

4.1 Experimental environment and conditions

The test is divided into two parts, first is the test tool part, and secondly is the test hardware. The test requirements mainly include the hard disk, CPU, memory and other requirements. The hardware also includes the hardware requirements of the client and the system pressure test tools. The following are the main requirements for testing host hardware: hard disk: 500GB; CPU clock frequency: >2.8GHZ; CPU secondary buffer capacity: >8MB; memory: higher than DDR generation 2, and capacity at least 2GB. There are not only hardware requirements, but also stress testing of the system. Stress testing is required in a variety of browser environments. Browsers, including commonly used browsers, need a smooth network to test the systems’ pressure resistance. After the creation of network and test environment, the test tool used is Load Runnerya, which is mainly used for the systems’ pressure test. At the same time, the tool can accurately test the data flow of the system, count the feedback time of the system, and also conduct statistical tests on various access information of various media objects. The tool not only needs to test the function and performance individually, but also needs to do multi-point distributed testing. At the same time, the tool has a common short message interface.

The test plan is designed according to the system requirements and research results, and the test case is used as a small unit in the test plan. Testing is divided into performance testing and functional testing, after the completion of the test, further improvements will be made, and more perfect products will be released eventually. When testing, test cases should be made according to the needs of system application. Based on test plan, test cases should be written. At the same time, the requirements and design of the system should be verified. The internal program testing of the system needs to be enhanced. The system program code testing should be completed by unit testing to ensure the accuracy of data in the operation of micro-units. When testing, all units need to be integrated, and the operation and results of the integrated units need to be tested. Usually there are no problems found in unit testing. Once the integration unit is integrated, there will be some problems, that is, there are no problems in part. After integration is considered as a whole, there will be problems in testing and running. Before the formal operation of the system, all functional modules need to be integrated. The main part includes hardware and software. It can be seen that the integration test in the system test belongs to a high stage. The integration mode must be adopted in the test to ensure that all functional modules can be unified and coordinated in operation and calculation. The overall test scheme of the system is shown in Fig. 6 below.

Fig. 6

Flow chart of overall system test plan.

When testing, test cases should be formulated according to the actual needs of the system. Based on the test plan, test cases should be compiled, and then the system test simulation environment should be built. Then, according to the test plan of the test case test system, the bugs in the software should be modified according to the BUG detected by the test. After the modification, the product will be delivered after further testing.

4.2 Teaching evaluation model experiments based on neural network

Sample training is an indispensable operation step in neural networks. As mentioned in the above chapters, in this topic, the original teaching-related data should be normalized to make the pre-processed data meet the requirements of training samples of neural networks. The formula of normalization treatment is as follows: $X = \frac{T - T_{min}}{T_{max} - T_{min}}$ (2)

Here X is the input value of the normalized feed-in neural network, T is the input value of the untreated task, Tmax and Tmin are the maximum and minimum values of the input of the neural network, respectively.

The simulation results are shown in Table 1, Figs 7 and 8.

Table 1

Comparison between experimental results and manual evaluation results

Entry name serial number	1	2	3	4	5	6	7
Experimental result	0.7754	0.7748	0.8969	0.8052	0.7424	0.6762	0.8262
Experimental evaluation level	in	in	good	good	in	qualified	good
Expert and peer test results	0.7939	0.7753	0.887	0.8058	0.7488	0.6758	0.8257
Manual rating	in	in	good	good	in	qualified	good
Error	0.0015	0.0005	0.0001	0.0004	0.0006	0.0004	0.0005
Entry name serial number	8	9	10	11	12	13	–
Experimental result	0.7710	0.7883	0.7979	0.7965	0.7877	0.7643	–
Experimental evaluation level	in	in	in	in	in	in	–
Expert and peer test results	0.7703	0.7741	0.7975	0.7984	0.7879	0.7648	–
Manual rating	in	in	in	in	in	in	–
Error	0.0007	0.0008	0.0006	0.0004	0.0008	0.0005	–

Fig. 7

Comparison between experimental results and manual evaluation results.

Fig. 8

Error curves of experimental results and manual evaluation results.

Table 2

Test cases for teacher evaluation system

Project software	Teachers’ teaching quality evaluation system		Program version	1.0
Functional module	frmXSXXLR		Compactor
Functional characteristics	User evaluation of teachers’ teaching quality
Test purpose	Can the test form evaluate the teaching quality of teachers?
Operation steps	Pedagogical operation	data	Expected result	Actual results	Test state
1	Input numbers in Text1	10	Display number 10	Same as expected result	good
2	Input numbers in Text2	10	Display number 10	Same as expected result	good
3	Input numbers in Text3	10	Display number 10	Same as expected result	good
4	Input numbers in Text4	10	Display number 10	Same as expected result	good
5	Input numbers in Text5	10	Display number 10	Same as expected result	good
6	Input numbers in Text6	10	Display number 10	Same as expected result	good
7	Click calculation total score	nothing	Total score 60	Same as expected result	good

According to the final results, it can be found that in the model of teaching evaluation, when it involves the evaluation and judgment of teachers’ teaching in relevant chemistry and biology disciplines, the training model needs to be judged by the teachers’ teaching evaluation system of neural network. Similarly, different majors and disciplines must choose the neural network adapted to it for sample training, so as to more accurately evaluate the teaching quality of each specialty or discipline.

5. Conclusion

A teacher evaluation model based on the theory of neural network has been built, and a new system has been developed. The model includes five aspects: login module, management module, teacher teaching scoring module, teacher teaching evaluation information query and neural network evaluation module. Login module is used to verify user name and password; system management module is used for user management and system data management; teacher teaching evaluation module is composed of evaluation backstage service component and front-end operation interface. Teacher evaluation result module mainly realizes the teaching ranking management, open display and comprehensive query of evaluation result content. Neural network modules use neural network model. Type B evaluates teachers’ teaching. Based on the future development of information management platform and network technology, the following tasks need to be carried out in the future, to further improve the degree of component and enhance the maintenance and scalability of the system. Fusion of Wechat platform, transfer part of the query service to Wechat platform, the improved convenience of system use, follow-up can also consider the use of H5-based APP system to expand core business to mobile platform. The system has great limitations in the future development, so the system also needs many means and methods to enhance the systems’ develop ability. The BP model neural network adopted in this system can also try to replace Hopfield, Boltzmann machine or adaptive resonance network model, or find better network structure.

References

J.R.

Botkin ,

Rothwell and

R.A.

Anderson , Prenatal Education of Parents About Newborn Screening and Residual Dried Blood Spots: A Randomized Clinical Trial, Jama Pediatr 170(6) (2016), 1.

H.A.

Kalantari and

Esmaeilzadeh , Association between academic achievement and physical status including physical activity, aerobic and muscular fitness tests in adolescent boys, Environmental Health & Preventive Medicine 21(1) (2016), 27–33.

A.M.

Baldini ,

Bao , Baracchini E. Muon polarization in the MEG experiment: Predictions and measurements, European Physical Journal C 76(4) (2016), 223.

Zhou ,

Yang and

L.I.

Liming , Research on the Basis of Database Applications Open Experimental Platform and Automatic Evaluation System, Computer Education 89(8) (2016), 747–754.

Wei , The design and application of physical education teacher evaluation automation system, Electronic Design Engineering 54(8) (2016), 597–600.

Yang ,

L.I.

Bin and

X.U.

Sheng-You , Study on Evaluation of Teaching Quality for the Principle of Automatic Control Based on MISO Model, Journal of Electrical & Electronic Education 27(8) (2017), 644–649.

Li ,

Wang and

Li , Modeling Monolingual. Character Alignment for Automatic Evaluation of Chinese Translation, ACM Transactions on Asian and Low-Resource Language Information Processing 15(3) (2016), 1–18.

A.L.

Rebar ,

Stephanie and

S.J.

Alley , Automatic Evaluation Stimuli – The Most Frequently Used Words to Describe Physical Activity and the Pleasantness of Physical Activity, Frontiers in Psychology 27(5) (2016), 1149–1164.

Osorio ,

Hoenigl and

Quartarolo , Evaluation of opt-out inpatient HIV screening at an urban teaching hospital, Aids Care 29(8)(2017), 1–5.

10.

Svennberg ,

Stridh and

Engdahl , Safe automatic one-lead electrocardiogram analysis in screening for atrial fibrillation, Europace: European pacing, arrhythmias, and cardiac electrophysiology: Journal of the working groups on cardiac pacing, arrhythmias, and cardiac cellular electrophysiology of the European Society of Cardiology 19(9) (2017), 1449.

11.

Koppers ,

Wormer and

Ickstadt , Towards a Systematic Screening Tool for Quality Assurance and Semiautomatic Fraud Detection for Images in the Life Sciences, Science & Engineering Ethics 23(4)(2017), 1113–1128.

12.

Gu and

Zhang , Changes of Children's Motivation in Physical Education and Physical Activity: A Longitudinal Perspective, Research Quarterly for Exercise & Sport 84 (2016), A42–A43.

13.

Eyüp. Bozkurt, Mikail. Tel, Opinions and Perceptions of Physical Education Students about Value Education, Educational Research & Reviews 11(20) (2016), 1918–1924.

14.

Mayorgavega , Montoroesca, ntilde, J.o and

Merinomarban , Effects of a Physical Education-Based Programme on Health-Related Physical Fitness and Its Maintenance in High School Students: A Cluster-Randomized Controlled Trial, European Physical Education Review 22(2) (2016), 243–259.

15.

E.H.

Kwon and

M.E.

Block , Implementing the adapted physical education E-learning program into physical education teacher education program, Res Dev Disabil 69 (2017), 18–29.

16.

N.L.D.

Chatzisarantis ,

M.S.

Hagger , Effects of an intervention based on self-determination theory on self-reported leisure-time physical activity participation, Psychology and Health 24(1) (2009), 29–48.

17.

L.D.

Feisel and

A.J.

Rosa , The role of the laboratory in undergraduate engineering education, Journal of Engineering Education 94(1) (2005), 121–130.

18.

P.A.

Ertmer and

T.J.

Newby , Behaviorism, cognitivism, constructivism: Comparing critical features from an instructional design perspective, Performance improvement quarterly 6(4) (1993), 50–72.

19.

M.D.

Curtner–Smith , The more things change the more they stay the same: Factors influencing teachers’ interpretations and delivery of national curriculum physical education, Sport, Education and Society 4(1) (1999), 75–97.

20.

G.A.

Gunter ,

R.F.

Kenny and

E H.

Vick , Taking educational games seriously: Using the RETAIN model to design endogenous fantasy into standalone educational games, Educational technology research and Development 56(5-6) (2008), 511–537.