The Influence and Analysis of Artificial Neural Network on Higher Education Management System

It is a correct but challenging experiment to establish a system by BP neural network, and the challenge lies in the correctness and rationality of information evaluation. Therefore, it is also an important part to communicate with the Ministry of Education of every university. Secondly, using the correct model building method can save a lot of time, and it can also be compared with the real situation to obtain more correct data [1]. Through the understanding of software development and BP neural network, in order to be more authentic, connecting students and teachers can not only shorten the distance between students and teachers, but also make the data more accurate and provide the most intuitive data for observers. [2]. Combined with BP neural network, it is convenient and efficient, and it can also provide more accurate data for college appraisal, making college appraisal more fair [3].The system has strong scientificity, combined with the characteristics and characteristics of neural network on the basis of BP neural algorithm, and is convenient, accurate and reliable, with high identification accuracy and good operability [4]. In order to reflect learners' learning situation and diversify learners' evaluation methods, a distance education learner evaluation model is constructed on BP neural network, and then experiments are carried out, and an accurate experimental report is obtained [5]. The algorithm and basic idea of BP neural network are described, and the model is established based on this [6]. Through the evaluation data of experts, the influencing factors of the evaluation system are explained, the defects and deficiencies of the system are analyzed, and the practicability is enhanced [7]. According to the different characteristics of college education, a humanized evaluation model of college teaching quality is established, which makes the operation convenient and fast. Compared with the traditional evaluation method, it is more authentic and can better reflect the teaching quality of colleges and universities [8]. Students are the future of the country and the foundation of a country's development. Therefore, it is the most important thing to find out the loopholes in teaching and improve the quality of teaching. It is what every university should do to keep up with the national development policy and implement the national development requirements [9].The successful establishment of the model has enhanced the teaching capabilities of university educators and improved the overall quality of education in higher learning institutions, encouraging more self-directed learning among students and enabling teachers to impart knowledge beyond textbooks, thus broadening students' horizons [10]. While there are many methods to improve teaching quality, most are not effectively guided. However, the construction of a Backpropagation (BP) neural network can identify issues from various perspectives, leading to more scientific solutions that enhance teaching quality [11]. Any experiment must be evaluated based on its outcomes and practical applications, including education; therefore, a teaching evaluation model based on BP neural networks is particularly important. The success of this model's establishment is evident from the results [12]. By utilizing BP neural networks, collected samples are input into the model according to relevant indicators for testing. This process analyzed the primary factors influencing tuition fees in higher education and provided a relatively reasonable formula for calculating average costs [13]. Using student grades as inputs and final scores as outputs within a BP neural network framework allows conclusions to be drawn faster, more conveniently, and directly [14]. Considering the characteristics of higher education combined with the features of BP neural networks, a teaching quality assessment model was designed for universities. Even when multiple teachers participate in the evaluation, quick conclusions can be reached [8]. For students of different academic levels, various calculation methods have been established, creating a fairer learning environment and strengthening students' motivation to learn [15]. Keeping up with the trends of the times, building new media concepts, and integrating modern information technology have become a trend. Therefore, combining contemporary information technology with BP neural networks presents a new challenge [16]. Using Matlab for simulation training showcased the work quality of class advisors [17]. Military academies differ from civilian universities in management style, leaning more towards military management. Obtaining teaching information from military academies is very challenging, so aligning closely with potential military academy management models is a significant challenge for collaboration opportunities [18]. Applying the principles and ideas of basic BP neural networks, a model suitable for various universities can be established and further refined according to each institution's characteristics, better aligning with higher education philosophies [19].

2

Artificial neural network theory

2.1

BP neural network

Initially, neural networks were not called artificial neural networks; they had another name: perceptrons. Later on, the single-layer perceptron network emerged, which is the earliest model of a neural network. Its construction was simple yet clearly demonstrated its characteristics. However, as neural networks evolved, scientists identified several limitations, particularly in addressing nonlinear problems. They struggled to find convenient and direct solutions for these issues and could not achieve certain fundamental functions, limiting the application scope of single-layer perceptron networks.To better address these challenges, researchers delved deeper into the study of single-layer perceptrons and eventually developed a novel algorithm: the "Backpropagation (BP) algorithm." This method not only resolved issues within the hidden layers of neural networks but also enabled the network to make predictions. Combining the input layer and the hidden layer to form a BP neural network, and an output layer. Its propagation mechanism involves both forward and backward passes. Like biological neurons, it possesses strong mapping capabilities and a flexible network structure. Each layer contains numerous nodes, hence the term neural network.

The introduction of the BP algorithm marked a significant advancement in neural network technology, enabling the processing of complex, non-linear data and expanding the practical applications of neural networks far beyond what was possible with single-layer perceptrons.

Like the distribution of nerves, it has a very strong mapping ability and flexible neural network structure. Each layer of it has many nodes, so it can be designed differently for each layer, and with the slight difference in the structure, its performance will also be different. Neural networks have the following disadvantages:

The learning speed is slow. Even a small problem requires hundreds of times of study;

Prone to local minima;

The number of layers and the number of neurons cannot be determined;

Limited network promotion ability.

2.2

Neural network model

The BP neural network selects the function φ (y) according to the characteristics of the neural network, and activates the function expression: (1) $φ (y) = \frac{1}{1 + e^{- y}}$

The value input in the input layer needs to be naturalized, so that the value range of the result is [0, 1]. Finally, as required, the result is standardized to make it look more standardized, and to achieve this purpose, the function adopted is: (2) $x_{i} = \frac{(x_{i} - x_{\min})}{(x_{\max} - x_{\min})} \times μ + v, 2 μ + v = 1$

Forward propagation is a propagation mode of BP neural network, that is, Transport layer to hidden layer process, and then to the output layer, so the input and output expressions between the k-th layer and the k+1-th layer are: (3) $y_{n}^{k} + φ_{n}^{k} (\sum w_{i n}^{k} y_{i}^{k} - θ_{n}^{k + 1})$

Wherein n is any constant; n_k is also any constant, which is the number of neurons in the k-th layer; M Total number of layers.

Another propagation mode of BP neural network is back propagation, that is, when the vector is input in the input layer, a new vector Xp is input, and the vector Y_p,k+1 will be directly output in the output layer after being calculated by the calculation mode of the neural network. However, the results obtained in this way will be different from the theoretical results, and the results on both sides will be different. Therefore, the total square error of the output vector of the output layer is: (4) $E(n) = \frac{1}{2} \sum e_{j}^{2} (n)$

When the number of samples in the entire calculation process is N, the corresponding average square error is: (5) $E_{A} = \frac{1}{N} \sum_{n = 1}^{N} E (n)$

In the process of correlation test, a condition must be met, that is, the correlation between the experimental prediction result and the actual result is not less than 0.6; Therefore, when finally checking the two points, you can calculate according to the following formula: (6) $C= \frac{s_{2}}{s_{1}} = \frac{\sqrt{\frac{\sum {(x_{i} - x)}^{2}}{n - 1}}}{\sqrt{\frac{\sum {(Δ_{i} - Δ)}^{2}}{n - 1}}}$ (7) $P=P {| Δ_{i} - Δ | < 0.6745 S_{1}}$

If the result is greater than 0.95 and C is less than 0.35, then the prediction success rate is high; If the result is greater than 0.8 and C is less than 0.5, then the prediction success rate is qualified; If the result is less than 0.7 and C is greater than 0.65, then the prediction success rate is a failure.

To determine the number of hidden layers of neural network, the method is mainly to find out [5, 6, 7] through experiments. By comparing and analyzing with previous data, we can get the desired results, and some formulas can be used, such as: (8) $L= \sqrt{M + N} + A$

M represents the initial number, N represents the last number, L represents the number of middle parts, and the value range of A is (1, 10).

The algorithm model of BP neural network is usually described by the following first-order differential equation: (9) ${\begin{matrix} y_{i} (t) \end{matrix} \begin{array}{l} τ \frac{d μ_{i}}{d t} \\ = f [u_{i} (t)] \end{array} = - μ_{i} (t) + \sum w_{i j} x_{j} (t) - θ_{i}$

The output function based on this is the following expression: (10) $f (μ_{i}) = \frac{1}{1 + \exp {(- μ_{i} / c)}^{2}}$

The function is shown in Figure 2:

In order to eliminate the influence of the attributes of the data itself on the prediction accuracy, the first step is to normalize the input and output data, and take any number between [0, 1]. Its formula is: (11) $Y = \frac{X - X_{\min}}{X_{\max} - X_{\min}}$

The inverse change is: (12) $X = X_{\min} + Y (X_{\max} - X_{\min})$

According to empirical formula: (13) $N = \sqrt{n + 1} + a$

N, n, and 1 represent the number of starting, intermediate nodes, and resulting nodes, respectively; The value range of a is any constant between 1 and 10;

From this, it can be deduced that the correction value of the weight from neuron i to neuron j is: (14) $Δ ω_{i j} (n) = η \times δ_{j} (n) y_{i} (n)$ (15) $δ_{j} (n) = (d_{j} (n) - y_{j} (n)) φ_{j}^{'} (\sum_{i = 0}^{m} ω_{i j} (n) y_{i} (n))$

In the hidden layer, the number of neurons is expressed as: (16) $l = \sqrt{n + m} + a$ n indicates the starting number; m indicates that it is the last number; The value range of a is any constant between [1, 10].

Suppose there are m nodes at the beginning, n nodes at the end, U is the number of intermediate nodes. Its output function formula: (17) $b_{y} = f (W^{T} X - θ)$

The output function of the node in the output layer is: (18) $c_{j} = f (V^{T} B - φ)$

The definition of mean square error MSE is: (19) $MSE = \frac{\sum_{i = 1}^{n} {(t_{i} - a_{i})}^{2}}{n}$ Where t_i is the expected output, it is the actual output of the network.

Optimizing the BP neural network algorithm is to find the initial weights W of a set of BP neural networks so that it can minimize the corresponding prediction error value E. Its function is: (20) $E = \frac{1}{2 N} \sum_{i}^{N} {(y_{i} - y_{i})}^{2}$

3

Impact and analysis of artificial neural network on higher education management system

3.1

Number of studies

Through the collection and analysis of previous related documents, the collected documents are made into a table, and then converted into the following line chart. According to the analysis of the publication year of each paper, the paper on this subject takes a long time to research. Before that, someone published a related paper, which described the evaluation of the combination of BP neural network and teaching quality evaluation system. It can be seen from his literature that he found an educational evaluation method that is highly like neural network after research. However, 44 related documents appeared in 2010, which shows that this research project has reached its peak, but in 2024, there are currently 146 related documents, which shows that the participation and awareness of this research project are not enough. All these signs indicate that artificial neural networks can reach a higher position in the application and research in the field of education, but basically few people study it. Therefore, even if we can reach a higher peak in this field, we still need more researchers to participate and make contributions to the country's education.

From 2016 to 2022, in the research of education system structure, the trend in literature data is relatively obvious. It reached 166 articles by 2022, a decline in 2023. It shows that the research content in the past two years has decreased compared with before, and the main reason is that the lack of funds leads to the decline of research hotspots.

3.2

Correlation Analysis and Evaluation of Teaching Quality

3.2.1

Index

The evaluation standards of each university are different. Therefore, if you want to be more practical, you need to carry out in-depth cooperation with each university. After knowing what indicators will be included in the assessment requirements, teachers can conduct assessments according to the prescribed indicators, which is simple and convenient. But because of this, the establishment and selection of each indicator are very important. Therefore, to be fair and not to hit students' self-confidence because it is difficult to complete, the determination of the evaluation index is formulated by all teachers in the school together, and the setting of the evaluation index is completed in combination with the opinions of all teachers in the school. According to the contents of the official documents in the table, after analysis, investigation and summary, a system of factors and indicators on how to evaluate the classroom teaching quality is compiled. See Table 1 for details: Table 1.

Teaching quality evaluation index table

Level 1 indicators	Serial number	Secondary indicators	Scoring Range
X1 Teaching attitude	X11	Put an end to being late, leaving early and missing classes, don't skip class	0-100
	X12	Caring for students, being able to communicate friendly with students, and caring for their physical and mental health	0-100
	X13	Do things rigorously, prepare lessons adequately, be responsible for homework, and be able to answer questions for students	0-100
X2 Teaching content	X21	Clear teaching purpose, highlighting key points, Give different teaching and education to different students	0-100
	X22	The language is accurate and focused, and the ability to skillfully use teaching tools and auxiliary teaching methods	0-100
	X23	Be able to combine theory with display to encourage students	0-100
X3 Teaching objective	X31	High theoretical knowledge, hands-on ability to reflect the changes of modern society	0-100
	X32	Instead of completely breaking away from the outline, connecting with practice	0-100
	X33	The atmosphere of class discussion is strong, and students can actively participate	0-100
X4 Teaching effect	X41	Through teaching, students can master basic theoretical knowledge and improve their practical ability	0-100
	X42	Through teaching, students can develop and learn more extracurricular knowledge by themselves	0-100
	X43	Through teaching, students can find teaching or homework problems and solve them	0-100

For different teaching effects, the highest score of 100 is used for evaluation. If the score of a single index is comprehensively evaluated by multiple weights, other indicators are comprehensively evaluated.

(21)

S_{i} = \sum_{j = 1}^{n} w_{i j} X_{i j}

3.2.2

Teaching quality evaluation model

To realize the construction of the model, we can use high-performance numerical calculation visualization software-commercial mathematics software. The commercial mathematics software comes with its own neural network toolbox, in which many functions can be directly called through the analysis and design of BP neural network system, which can also reduce some burden for programmers and effectively improve the working efficiency of the software. The program block diagram of network implementation of teaching quality evaluation system is shown in Figure 4

Take 10 sets of data as the initial values of the system, and learn how to apply the system in the neural network toolbox. After passing the practice, you can accomplish the goal. After the network training, the last five sets of test data are tested, After checking the actual results of BP neural network, the error value can be obtained. The error is shown in the table below.

Table 2.

Statistical table of experimental results

Experiment No.	Expert evaluation value	Final evaluation results	Evaluation value of the system	Systematic evaluation results
1	0.66	Failure	0.6599	Failure
2	0.67	Pass	0.6533	Pass
3	0.64	Failure	0.6537	Failure
4	0.69	Medium	0.6732	Medium
5	0.74	Good	0.7222	Good

By comparing the actual evaluation results with the expert evaluation results, the results obtained from the teaching quality evaluation model based on BP neural network are in good agreement with the initial prediction. Figure 5 and Figure 6 can show the whole process and results. Therefore, this BP neural network evaluation software can determine whether to achieve the desired effect through various evaluation indicators.

3.3

Index system of higher education management system

To judge whether a simple, practical, fair and just teaching evaluation index model conforms to the classroom teaching quality evaluation system of military academy. The indicators of high-quality higher education management system must be combined with the information of the Ministry of Education on teaching content, and then analyzed the factors that will affect them, so as to get a reasonable system. The evaluation criteria and indicators are shown in Table 3.

Table 3.

Index System of Higher Education Management System

Level 1 indicators	Serial number	Secondary indicators	Scoring Range
Y1 Teaching content	Y11	Meet the teaching requirements of the course and can change the teaching content according to the characteristics of students	0-100
Y1 Teaching content	Y12	Pay attention to the teaching of cultural courses, be able to keep up with the trend of the times, stand at the forefront of scientific development, and the teaching of each arm is closer to actual combat	0-100
Y2 instructional design	Y21	Clearly recognize the importance of talent cultivation	0-100
Y2 instructional design	Y22	Take students as the main body, pay attention to teaching objectives, content and methods	0-100
Y3 Teaching Concept	Y31	Pay attention to each student's moral character, ability and development needs, and make them think independently and positively	0-100
Y3 Teaching Concept	Y32	Pay attention to the cultivation of modern information technology and keep pace with the times	0-100
Y4 Teaching Methods	Y41	Teaching by combining the characteristics of curriculum and students	0-100
Y4 Teaching Methods	Y42	Be able to guide students to learn independently, explore and think problems	0-100
Y5 teaching methods	Y51	According to the curriculum requirements and the realistic teaching environment, the teaching methods should be used reasonably and combined with arms teaching	0-100
Y5 teaching methods	Y52	Pay attention to the combination of curriculum and modern science and technology to improve students' professional knowledge	0-100

Input layer, hidden layer and output layer are important components of BP neural network, and each layer is connected, so the relationship between each layer is also very close. However, depending on the difference between the military academy and other schools, when redesigning, The hidden layer is divided into one or more layers. According to the content of the teaching quality evaluation model, we set the hidden layers as one layer. However, there is more than one corresponding evaluation indicator. From Table 4, we can see that there are 10 secondary indicators in total, so we can use these 10 indicators as input values of the modelThe number of neurons in the output layer n = 1. Through the analysis of experimental results, the expression of the method for calculating the number m of neurons in the hidden layer is as follows: (22) $m = \sqrt{l + n} + \frac{q}{2}$

Through the analysis of the formula, l = 10, n = 1, q = 10, the result m = 8.3 is obtained. FIG. 7 is a configuration diagram of an evaluation model.

The evaluation and analysis of the first-level indicators in colleges and universities are carried out to evaluate the rationality and application effect of the higher education system structure from the teaching content, teaching design, teaching philosophy, teaching methods and teaching means, as shown in Figure 8-Figure 12.

4

Conclusion

The establishment of BP neural network is like a bridge, which strengthens the connection between developers and users, and the construction of neural network is inseparable from the active cooperation of users, so the three of them are interdependent. Artificial neural network is also applicable to the evaluation of music teaching quality. It needs the cooperation between builders and colleges to complete its construction. The teaching quality evaluation system has good significance. For teachers, it can make it easier for teachers to understand students' situation, communicate in a targeted manner, and it will be easier for teachers to teach. For students, it can enable them to find their own direction, clearly know their strengths and weaknesses, and regulate their behavior. In response to the call of the state, we have been at the forefront of the times, it plays an important role in the construction and development of the national education system and the training of talents.

Lingua:: Inglese

Frequenza di pubblicazione:: 1 volte all'anno
Argomenti della rivista:: Scienze biologiche, Scienze della vita, altro, Matematica, Matematica applicata, Matematica generale, Fisica, Fisica, altro

Feed RSS della rivista

The Influence and Analysis of Artificial Neural Network on Higher Education Management System

Zhichao Shen

Pubblicato online: 27 feb 2025

Ricevuto: 26 set 2024

Accettato: 09 gen 2025

DOI: https://doi.org/10.2478/amns-2025-0127

Parole chiaveBP neural network, high quality, practice, education

© 2025 Zhichao Shen, published by Sciendo

This work is licensed under the Creative Commons Attribution 4.0 International License.

Parole chiave
BP neural network, high quality, practice, education