Research on the Risks of Financial Informatization Construction in Colleges and Universities, Its Prevention and Control and Path Optimization

With the development of social economy and the progress of information technology, financial management of colleges and universities are facing new challenges and opportunities. Financial informatization construction has become an inevitable trend of financial management in colleges and universities, providing new ideas and means for colleges and universities to improve the level of financial management, enhance management efficiency, reduce management costs, and provide support for management decisions [1-4]. Financial information construction in universities is not only a simple application of information technology, but also an all-round reform and optimization of financial management concepts, systems and processes in universities. However, in the process of financial informatization construction, the same also faces some problems and challenges, including system imperfection, security risks, lack of quality of personnel and so on [5-8].

The solution to the above problems is of great significance to the financial management of colleges and universities, and the following measures can be taken: Improvement of system functions: the financial informatization system of colleges and universities needs to be continuously improved and optimized to meet the needs of financial management. It is necessary to continuously improve the function and performance of the system through demand research and user feedback to improve the usability and adaptability of the system [9-12]. Data integration and sharing: Universities need to strengthen the information exchange and sharing between various departments to achieve data integration and sharing. A unified data platform and standards can be established to solve the problem of data silos through data integration, sharing and exchange, and to improve the value and utilization efficiency of information [13-16]. Improve information security: The financial informationization system of colleges and universities needs to strengthen information security, take effective technical and management measures to ensure the safety and reliability of financial information. It can establish a perfect information security management system, encrypt data and authority control, establish a complete audit and monitoring mechanism to improve the security and stability of the system [17-20]. Strengthen the construction of talent team: universities need to increase the training and introduction of financial informatization personnel to enhance the quality of talent building and team building efforts. Through the establishment of disciplines and specialties, the establishment of scholarships, job rotation and other ways to cultivate and attract more financial information technology talents, the establishment of a stable talent team [21-24].

Zhao, L. et al [25] designed a financial management (FM) informatization system for universities. By proposing and verifying the BF algorithm based on multiple eigenvalue hash partition and its effectiveness. It also reveals that the FM information management system effectively improves the effectiveness of financial information management in colleges and universities. Yang, L. et al [26] analyzes the problems and solutions encountered by a college and university in the process of financial informatization construction, aiming to provide references for other colleges and universities. The advantages and disadvantages of the university in the process of financial informatization construction are pointed out, and feasible suggestions are put forward. Zhang, T.[27] puts forward a financial risk early warning model for colleges and universities, which is able to identify potential financial risks in colleges and universities in a timely manner, so as to help administrators take measures in order to ensure the stable development of finances. Jun, Y.[28] emphasizes the fusion of information technology and financial management on the financial management of colleges and universities and explains that there are many problems in the current financial management of colleges and universities, which seriously restrict the efficiency of financial management of colleges and universities. Zhao, Y.[29] elaborates that based on the basic situation of colleges and universities, the establishment of a stable and secure network financial information management system under the premise of considering the security factors is an important prerequisite for the healthy and stable development of colleges and universities. Bao, X.[30] utilizes the big data technology and data mining two technologies to design and optimize the enterprise financial management model. Based on the theory of enterprise financial management under the Internet information, it compares the factors affecting the financial management model and emphasizes the importance of designing the enterprise financial management model.

The above study points out the challenges faced by the current financial informationization construction, and puts forward relevant coping strategies, but also can be seen that the “financial informationization construction risk optimization path” in the academic community has not been widely concerned about, and the financial informationization construction risk optimization path for the sustainable development of colleges and universities, enterprises and organizations have an important role to play. Important role, therefore, in the study should not only examine the risk of financial information construction and preventive and control measures, but also focus on the optimization of its path to carry out a systematic analysis.

In this paper, 12 relevant indicators were firstly selected from four aspects to construct the financial risk evaluation system of colleges and universities, and the empirical research was carried out with the financial data of a college or university. The principal components of the financial data were extracted by principal component analysis, and eight principal factor components were obtained as the input data of the model. Then the particle swarm algorithm (PSO) and BP neural network combination of methods to study the financial risk early warning model of colleges and universities, and finally the classical BP prediction model, GA-BP prediction model and the PSO-BP prediction model in this paper on the comparative analysis, through the example to prove that particle swarm algorithm optimization of the BP neural network financial risk prediction model has a better prediction accuracy. Finally, the optimization path of financial informatization construction in colleges and universities is proposed in five aspects: changing the concept of financial informatization, eliminating information silos, optimizing system functions, strengthening information security prevention, and improving the business ability of practitioners.

2

Building a financial risk evaluation system for colleges and universities

2.1

Selection of financial risk indicators for higher education institutions

2.1.1

Financing risk indicators

Financing risk in the choice of indicators is mainly placed on the repayment of debt, many colleges and universities are in the period of debt settlement, so the settlement of past debts still accounts for a large proportion of the financial risk of colleges and universities: 1)

college gearing ratio = total liabilities / total assets amount

Colleges and universities gearing ratio reflects the total debt of colleges and universities accounted for the ratio of total assets, the smaller the ratio shows that the ability of colleges and universities to resolve debt is stronger, the normal operation of colleges and universities as well as related financial activities can be guaranteed, on the contrary, if the ratio of the college or university the larger, or even more than 50% of the amount of debt is too large for colleges and universities is likely to affect the day-to-day operation of colleges and universities, the financial risk of the college or university has been very great, colleges and universities Financial security is low.

2)

Ratio of annual borrowing to total annual income = year-end borrowing balance / annual income

This indicator is the ratio of the year-end debt and annual income of the university, the indicator reflects the ability of the university to pay off debts, the larger the indicator directly indicates that the ability of the university to pay off debts is lower, the lower the degree of protection of debt repayment of colleges and universities. If the indicator is smaller, it means that the university can easily or in a short period of time to pay off the bank borrowing, the ability to pay off the debt is stronger, the university is facing less pressure to repay the debt, and the bank to recover the arrears of money is more guaranteed.

2.1.2

Investment risk indicators

Investment risk-related indicators mainly focus on the university’s investment in school-run colleges and universities and other areas 1)

Return on investment = (income from university-run colleges and universities + income from other inputs) / (investment in university-run colleges and universities + other investments)

The lower the rate of return on investment, the poorer the ability of universities to obtain returns through school-run universities and other investments, and the higher the risk of non-recovery of investments. On the contrary, the higher the rate of return on investment, the stronger the ability of school-run colleges and universities to earn income, the stronger the ability of colleges and universities to self-finance, the higher the rate of return on investment in school-run colleges and universities.

2)

financial risk of infrastructure investment = (total infrastructure borrowings at the end of the year + infrastructure accounts payable at the end of the year, a surplus of infrastructure at the end of the year) / (end of the year, a business fund, an investment fund + end of the year, a special fund, a receivable and a temporary payment) This indicator reflects the risk of the university’s investment in infrastructure, the net amount of investment in infrastructure and the net amount of the fund at the end of the year, the larger the ratio shows that the university of the higher the investment in infrastructure. If the indicator is smaller, the amount of investment in the university borrowing is smaller, and the investment risk is also smaller.

3)

Gearing ratio of school-run colleges and universities = liabilities of school-run colleges and universities / assets of school-run colleges and universities

This indicator reflects the ratio of liabilities and assets of school-run universities, the higher the ratio, the worse the solvency of school-run universities, and the risk that universities may not be able to recover their investment. If the indicator is smaller, it means that the debt of the university is smaller, and the possibility of the university receiving joint and several liabilities is also smaller, and the possibility of the investment not being recovered is also lower.

2.1.3

Overall operational capacity indicators

1)

Self-financing revenue capacity = (business income + income from affiliated units + donations + other income) / income of colleges and universities

The main source of funding for colleges and universities is the state’s financial allocations, but the formation of diversified funding trends, colleges and universities can no longer rely solely on the state’s financial allocations, but also to strengthen the ability to self-funding, so that the increase in college and university income to cope with the financial risks of the ability to be stronger, the degree of flexibility of the university’s economic activities will be stronger.

2)

Funding self-sufficiency rate = (career income + operating income + income from affiliated units + other income) (/ career expenditure + operating expenditure)

This indicator reflects whether the self-financing of the university can meet the daily expenses of the university, if the self-financing of the university can meet the higher ability of the university to raise funds, the overall operation and financial management of the higher level, can be completely self-sufficient, if the indicator is less than 1 or even very low, it shows that the university’s self-financing ability is poorer, for the state appropriation of the degree of dependence on the state appropriations are high, the day-to-day management of the budget and so on, relatively poor, there may be excessive expenditure or insufficient self-financing ability.

3)

Cash disbursement ratio = (cash + bank deposit) / (total annual expenditure / 12)

This indicator is to examine the daily operation ability of universities, the number of months that the bank deposits and cash of universities can be used to pay for university expenditures, the larger the number indicates that the daily operation of universities is more secure, and the ability to cope with daily risks is stronger.

2.1.4

Indicators of growth capacity of higher education institutions

1)

Net assets growth rate = (net assets at the end of the period - the beginning of the net assets) / the beginning of the net assets

This indicator is the net growth in net assets for the year and the ratio of net assets at the end of the previous year, reflecting the proportion of growth in college assets, the larger the ratio indicates that the faster the growth of college assets, the better the growth of colleges and universities, the stronger the competitiveness, and the ability to cope with risk is also stronger.

2)

Fixed asset growth rate = = (total fixed assets at the end of the year - total fixed assets at the end of the previous year) / total fixed assets at the end of the previous year the indicator is the ratio of the net increase in fixed assets of colleges and universities in the current year to the total fixed assets of the previous year, the indicator if too large indicates that the assets of colleges and universities are increasing too quickly, but it is possible that the risk of day-to-day operations increased, the college or university cash holdings are insufficient, but the ratio is too low indicates that the ability of colleges and universities to grow is poor, the development is relatively slow, so the ratio is also a need for an appropriate ratio.

3)

Growth rate of intangible assets = (total intangible assets at the end of the year - total intangible assets at the end of the previous year) / total intangible assets at the end of the previous year

Net increase in intangible assets and intangible assets at the end of the previous year the ratio of the total intangible assets can be reacted to the degree of growth of intangible assets, intangible assets, including the number of patents in colleges and universities should focus on a part of college and university assets, colleges and universities to strengthen the management of intangible assets is to prevent the loss of state-owned assets is also an important part of the loss.

4)

Endowment income growth rate = (last year’s endowment income - this year’s endowment income) / last year’s total income

Access to endowment income is an important part of the diversified financing of a college, the ratio of the increasing value also indicates that the growth of colleges and universities, self-financing ability to improve the development of diversified financing structure of colleges and universities more perfect, due to the endowment income of colleges and universities accounted for a smaller proportion of college and university revenues, so the attention to this indicator is less with this index, which is a new indicator. The analysis of financial risk factors of colleges and universities is shown in Table 1.

Table 1.

Analysis of financial risk factors in colleges and universities

Financial risk indicators for universities(X)	Risk indicators for college financing (debt service) (X1)	College asset ratio(X11)
	Risk indicators for college financing (debt service) (X1)	Annual borrowing accounts for the total annual income(X12)
	University investment risk index(X2)	Investment yield(X21)
		Financial risks of infrastructure investment(X22)
		The credit rate of the school office(X23)
	Overall operational ability indicator(X3)	self-financing(X31)
		Fund feed rate(X32)
		Cash payment rate(X33)
	Index of growth ability of universities(X4)	Net equity growth rate(X41)
		Fixed asset growth rate(X42)
		The growth rate of intangible assets(X43)
		Income growth(X44)

2.2

Principal component analysis

2.2.1

Mathematical model for principal component analysis

Based on the above basic idea of principal component analysis and its geometric significance, it can be known that principal component analysis utilizes the idea of dimensionality reduction to concentrate the information contained in a set of variables onto certain composite variables. These composite variables obtained are uncorrelated with each other [31]. From a geometric point of view, principal component analysis is to rotate the original axes to obtain mutually orthogonal axes, under which all data points can be dispersed most widely, and new axes are obtained by arranging them according to the numerical magnitude of the corresponding eigenvalues.

Analyzing principal component analysis in algebraic terms: for a given set of data sample points $X = {x_{1}, x_{2}, \dots, x_{n}}$ , there are n sample points in the data set, each containing p indicators, i.e. x_i ∈ R^p, i = 1, 2, ⋯, n, then: (1) $X = | \begin{matrix} x_{11} & x_{12} & \dots & x_{1 p} \\ x_{21} & x_{22} & \dots & x_{2 p} \\ ⋮ & ⋮ & ⋮ & ⋮ \\ x_{n 1} & x_{n 2} & \dots & x_{n p} \end{matrix} | = [\begin{array}{l} x_{1} & x_{2} & \dots & x_{p} \end{array}]$

Among them: (2) $x_{i} = {(x_{1 i}, x_{2 i}, \dots, x_{n i})}^{T}, i = 1, 2, \dots, p$

Principal component analysis is the process of linearly combining the original p metrics to obtain a new composite p composite metric, i.e: (3) ${\begin{matrix} z_{1} = w_{11} x_{1} + w_{21} x_{2} + \dots + w_{p 1} x_{p} \\ z_{2} = w_{12} x_{1} + w_{22} x_{2} + \dots + w_{p 2} x_{p} \\ ⋮ \\ z_{p} = w_{1 p} x_{1} + w_{2 p} x_{2} + \dots + w_{p p} x_{p} \end{matrix}$

Chemicalized as: (4) $z_{i} = w_{1 i} x_{1} + w_{2 i} x_{2} + \dots + w_{p i} x_{p}, i = 1, 2, \dots, p$

where x_i is n-dimensional, then z_i is also n-dimensional. Where the coefficient w_ij needs to be satisfied: 1)

z_i, z_j are uncorrelated where $(i \neq j, i, j = 1, 2, \dots, p)$ .

2)

The variance of variable z₁ is greater than or equal to the variance of variable z₂, which is decreasing.

3)

$w_{k 1}^{2} + w_{k 2}^{2} + \dots + w_{k p}^{2} = 1$ , k = 1, 2⋯, p.

When 3 conditions are met, the random variable indicators obtained after the transformation are uncorrelated between two by two. The variance is decreasing sequentially.

From the above, the transformation matrix of p × p can be obtained: (5) $W = | \begin{matrix} w_{11} & w_{12} & \dots & w_{1 p} \\ w_{21} & w_{22} & \dots & w_{2 p} \\ ⋮ & ⋮ & ⋮ & ⋮ \\ w_{p 1} & w_{p 2} & \dots & w_{p p} \end{matrix} |$

So there is: (6) $Z = [z_{1}, z_{2}, \dots, z_{p}] = W^{T} X$

The principal component analysis process is a de-correlation process where the variables obtained are two by two uncorrelated with each other.

2.2.2

Principal component contribution

The contribution rate and cumulative contribution rate of principal components reflect the degree to which the transformed indicators z_i portray the amount of information of the original data X respectively. 1)

Contribution rate: the proportion of the ird largest eigenvalue of the covariance matrix to the sum of all the eigenvalues of the covariance matrix, the larger the value, the more the ith variable indicator presents the information of the original data, the stronger the ability to synthesize the information of the original variable indicators [32]. Where the covariance matrix ith largest eigenvalue λ, then the formula is expressed as: (7) $ω_{i} = \frac{λ_{i}}{\sum_{i = 1}^{p} λ_{i}}$

2)

Cumulative contribution ratio: the ratio of the sum of the first k eigenvalue of the covariance matrix to the sum of all the eigenvalues of the covariance matrix, the larger this value is, the more the first k principal components can fully represent the information of the original data, calculated as follows: (8) $Z_{k} = \frac{\sum_{i = 1}^{k} λ_{i}}{\sum_{i = 1}^{p} λ_{i}}$

In practical applications, usually select the first k large variable indicators, so that its cumulative contribution rate to meet a certain threshold using the first k large composite indicators to replace the original p variable indicators for subsequent data analysis and processing, to achieve the purpose of data compression and dimensionality reduction, but also to extract the main features of the data.

Steps of principal component analysis algorithm: 1)

Calculate the mean $μ = \frac{1}{n} \sum_{i = 1}^{n} x_{i}$ of the samples of the sample data set $X = {x_{1}, x_{2}, \dots, x_{n}}$ , centering the samples, with the centered samples denoted ${\bar{x}}_{i}$ , ${\bar{x}}_{i} = x_{i} - μ$ .

2)

Compute the covariance matrix V for the original data set X, with the matrix denoted $V = \frac{1}{n} \bar{X} {\bar{X}}^{T}$ .

3)

Eigen-decomposition of the covariance matrix V to obtain the eigenvalues of the covariance matrix and the eigenvectors w_i corresponding to the eigenvalues, and at the same time arrange the eigenvalues in descending order.

4)

Take the first k eigenvalues based on the cumulative contribution that has been selected. λ₁, λ₂, ⋯, λ_k are the k eigenvectors corresponding to the k extracted eigenvalues, and the first k eigenvector combination matrix is denoted as $W_{k} = [w_{1}, w_{2}, \dots, w_{k}]$ then the k extracted principal components are $W_{k}^{T} \bar{X}$ .

3

Application of principal component analysis to the treatment of financial data

3.1

Analysis of correlation coefficients of financial data of higher education institutions

In this section, the training sample set (financial data of a university for the year 2023) is analyzed using SPSS software to extract the principal components. Before extracting the principal components for dimensionality compression, the correlation between the 12 sets of financial data is initially analyzed. The matrix of correlation coefficients of financial data is shown in Table 2. From the table we see that the correlation coefficients between the 12 groups of financial data are still relatively large, indicating that there is a large redundancy between them, which can be simplified. The financial data with stronger correlation are the return on investment (X21), the ratio of annual borrowing to total annual income (X12) is 0.95, the return on investment (X21) and the financial risk of infrastructure investment (X22) is 0.94, and the ratio of annual borrowing to total annual income (X12) and the financial risk of infrastructure investment (X22) is 0.89. These three financial indicators respond to the university’s debt repayment capacity and there is room for simplification.

Table 2.

Financial data correlation coefficient matrix

	X11	X12	X21	X22	X23	X31	X32	X33	X41	X42	X43	X44
X11	1	0.68	0.15	0.78	0.92	0.59	0.25	0.08	0.84	0.16	0.07	0.69
X12		1	0.95	0.89	0.27	0.78	0.51	0.94	0.16	0.81	0.4	0.44
X21			1	0.94	0.34	0.46	0.6	0.76	0.59	0.72	0.2	0.58
X22				1	0.68	0.42	0.11	0.77	0.12	0.53	0.29	0.07
X23					1	0.83	0.28	0.79	0.18	0.88	0.64	0.18
X31						1	0.38	0.35	0.63	0.85	0.51	0.58
X32							1	0.8	0.66	0.57	0.31	0.45
X33								1	0.15	0.7	0.72	0.76
X41									1	0.06	0.66	0.54
X42										1	0.37	0.28
X43											1	0.08
X44												1

3.2

Selection of principal components and their interpretation

The eigenvalues and variance contributions of the principal components are shown in Table 3. According to the eigenvalue criterion, only the principal components with eigenvalues more than 1 are selected here. Finally, 8 principal component factors are obtained, which contain 95.084% of the original amount of information.

Table 3.

The eigenvalue and variance contribution rate of the main component

Principal component	Eigenvalue	Individual variance contribution%	Total variance contribution%
1	2.548	21.233	21.233
2	2.32	19.333	40.566
3	2.222	18.517	59.083
4	2.183	18.192	77.275
5	1.037	8.642	85.917
6	0.593	4.942	90.859
7	0.313	2.608	93.467
8	0.194	1.617	95.084

The factor loading matrix is shown in Table 4. It can be seen that the first, second principal components explain each variable adequately. While the last few principal components and the original variables are relatively weak. Among them, the principal component Y1 explains X33, X41, X43, X12 more adequately, which represents the ability of the university in the two categories of profitability and growth. Principal component Y2 explains X11, X12, X21, X22 and it represents the solvency of universities.

Table 4.

Factor load matrix

Original sample size	Main cause
Original sample size	Y1	Y2	Y3	Y4	Y5	Y6	Y7	Y8
X11	0.133	0.888	0.429	0.718	0.267	0.693	0.495	0.651
X12	0.617	0.889	0.328	0.807	0.586	0.151	0.42	0.691
X21	0.59	0.879	0.87	0.262	0.8	0.249	0.202	0.678
X22	0.518	0.824	0.117	0.88	0.846	0.885	0.827	0.546
X23	0.5	0.777	0.853	0.254	0.552	0.489	0.256	0.544
X31	0.347	0.611	0.364	0.626	0.564	0.606	0.408	0.571
X32	0.187	0.771	0.514	0.254	0.232	0.705	0.324	0.258
X33	0.838	0.551	0.752	0.184	0.102	0.291	0.514	0.608
X41	0.815	0.546	0.807	0.744	0.388	0.475	0.162	0.332
X42	0.583	0.225	0.8	0.106	0.183	0.428	0.806	0.72
X43	0.884	0.189	0.625	0.282	0.11	0.847	0.602	0.671
X44	0.496	0.164	0.434	0.204	0.707	0.28	0.715	0.566

4

Neural Network Early Warning Modeling Study

4.1

Neural Network Theory

4.1.1

Neural network theory

The operation mechanism of BP neural network is forward propagation of information and backward propagation of error, and the model can be trained to achieve the predetermined error target by transferring the error in the backward direction. The specific process of BP neural network is as follows: 1)

Initialize the network: design the number of nodes in the input layer n, the number of nodes in the output layer l and the number of nodes in the hidden layer m according to the number of input variables X and output variables Y. Initialize the connection weights ω_ij and ω_jk between neurons in the input layer, hidden layer and output layer, and the thresholds a and b for the hidden layer and output layer.

2)

Calculate the output of the hidden layer: based on the input variable X, the weights between the input layer and the hidden layer ω_ij and the threshold of the hidden layer a, the output of the hidden layer H is calculated, and the signal is passed to the next layer using the activation function Sigmoid of the hidden layer. (9) $H_{j} = f (\sum_{i = 1}^{n} ω_{i j} x_{i} - a_{j})$

Where m is the number of nodes in the implicit layer, j = 1, 2, 3…, m, x_i are the input variables, H_j is the implicit layer output value, and f(x) is the implicit layer activation function with the expression: (10) $f (x) = \frac{1}{1 + e^{- x}}$

3)

Output layer output calculation. Based on the implied layer output H, the connection weights ω_jk, and the output layer threshold b, the predicted output is calculated G. The formula is as follows: (11) $G_{k} = \sum_{j = 1}^{m} H_{j} ω_{j k} - b_{k}$

where G_k is the output value of the output layer and l is the number of nodes in the output layer. k = 1, 2, …, l 4)

Error calculation. Based on the predicted output G_k and output variable Y, the prediction error of the neural network is calculated. The formula is as follows: (12) $e_{k} = Y_{k} - G_{k}$

e_k is the difference between the model’s predicted and actual values. k = 1, 2…l. 5)

Weights update. Recalculate the weights ω_ij and ω_jk based on the error e_k, where η is the learning efficiency: (13) $ω_{i j} = ω_{i j} + η H_{j} (1 - H_{j}) x (i) \sum_{k = 1}^{l} ω_{j k} e_{k}$ (14) $ω_{j k} = ω_{j k} + η H_{j} e_{k}$

Where k = 1, 2…l. j = 1, 2, 3…, m. i = 1, 2, …n. 6)

Threshold update. Substitute the corrected weights ω_ij and ω_jk into the following equation to adjust the threshold size for each node: (15) $a_{j} = a_{j} + η H_{j} (1 - H_{j}) \sum_{k = 1}^{l} ω_{j k} e_{k}$ (16) $b_{k} = b_{k} + e_{k}$

Where k = 1, 2⋯l. j = 1, 2, 3⋯m. 7)

Determine if the algorithm is finished.

4.1.2

Particle Swarm Algorithm Theory

Particle swarm algorithms can be used to explore the best solution to a problem.

Particle position can be expressed as $X_{i} = {(x_{i 1}, x_{i 2} \dots x_{i D})}^{T}$ .

Particle velocity can be expressed as $V_{i} = {(v_{i 1}, v_{i 2} \dots v_{i D})}^{T}$ .

The relationship between particle position and velocity can be expressed as: (17) $x_{i d}^{k + 1} = x_{i d}^{k} + v_{i d}^{k + 1}$ (18) $v_{i d}^{k + 1} = v_{i d}^{k} + c_{1} r_{1}^{k} (p b e s t_{i d}^{k} - x_{i d}^{k}) + c_{2} r_{2}^{k} (g b e s t_{d}^{k} - x_{i d}^{k})$

Where: $x_{i d}^{k}$ - the position of particle i in the dth dimension at the krd evolution.

$v_{i d}^{k}$ - the velocity of particle i in the dth dimension at the kth evolution.

c₁c₂ - is the learning factor i.e. acceleration constant, usually taken as 2.

r - A random number between 0 and 1.

$p b e s t_{i d}^{k}$ - The position of the individual best point of particle i in the dth dimension at the kth evolution.

$g b e s t_{d}^{k}$ - The position of the global optimum of particle i in the dth dimension at the kth evolution.

The velocity and position of the particle changes continuously as it moves, and by constantly comparing the size of the population optimum $(g_{b e s t})$ and the individual optimum $(p_{b e s t})$ , the particle is constantly updated with respect to the velocity and position, and the fitness value is recalculated.

4.1.3

Particle Swarm Algorithm for Optimizing BP Neural Networks

Although BP neural network is widely used and applicable in various industries, it has deficiencies in operation, and the methods of improvement usually include: 1)

Improve the convergence rate. In the BP neural network algorithm, the learning rate is usually set to 0.25, and the change of the learning rate is related to the weights and stability of the neural network, and the higher the weights are, the worse the stability of the model is, so the convergence speed can be improved by adjusting the learning rate.

2)

Optimization of weights. By optimizing the weights and thresholds, the model can overcome the problem of falling into the local optimum easily, so as to improve the prediction accuracy of the model.

This paper adopts the particle swarm algorithm combined with BP neural network to carry out financial risk early warning research on manufacturing colleges and universities. The main optimization steps are: 1)

Initialize the particle population and set the size of the particle population according to the actual problem.

2)

Calculate the fitness value of each particle and update the individual and group optimal solutions. Here the mean square error function is generally used to calculate the error of BP neural network: (19) $f i t n e s s = \frac{1}{N} \sum_{i = i}^{n} {(y_{r} - y_{i})}^{2}$

Where y_i is the actual output of the ind sample, y_r is the desired output of the ith sample, n is the number of training samples, after a number of iterations, the position corresponding to the lowest fitness particle is the optimal solution to the problem. 3)

Compare the adaptation value of each particle with the individual optimal position P_best experienced by that particle, if the current position has a better adaptation value, the point will be taken as the current best position, otherwise it will not be updated.

4)

Compare the fitness value of each particle and the global optimal position experienced by this particle g_best, if the current position fitness value is better, the point will be used as the current best position, otherwise no update.

5)

Iteratively update the velocity and position of the particle. The particle updates its position and velocity through the updated individual optimal solution and group optimal solution with the following equations: K is the current number of iterations, ω is the inertia weight, c₁ is the individual learning factor (non-negative constant), and c₂ is the group learning factor (non-negative constant). (20) $V_{i d}^{k + 1} = ω V_{i d}^{k} + c_{1} r_{1} (P_{i d p b e s t}^{k} - X_{i d}^{k}) + c_{2} r_{2} (P_{d g b e s t}^{k} - X_{i d}^{k})$

Where, r₁, r₂ are random numbers in the interval $[0, 1]$ . 6)

Calculate the new particle fitness value and determine whether to end the iteration. Calculate the new particle fitness value and update it if it is less than the current value.

7)

The optimal values obtained are the initial weights and thresholds given to the BP neural network.

4.2

PSO-BP neural network early warning model construction

4.2.1

Flow of PSO-BP neural network model operation

Constructing PSO-BP neural network model is specifically divided into three steps: 1)

Construct the BP neural network model. Determine the number of nodes in the input, hidden and output layers, and set the relevant initial parameters.

2)

Particle swarm algorithm optimizes the network. Gradually update the position and speed of particles by constantly comparing the adaptation values of particles at different locations, and find the optimal values to be assigned to the BP neural network as the new weights and thresholds.

3)

Input the data of the training sample group into the optimized model for simulation training, and test whether the errors of all the samples meet the predetermined conditions [33]. At this point, the model training has reached a certain accuracy, and the sample data of the prediction group can be input into the model for simulation and prediction.

The specific flow of the model operation is shown in Figure 1.

4.2.2

Steps of PSO-BP neural network model construction

Specifically, it consists of the following steps: 1)

Input layer design: the number of nodes in the input layer of the PSO-BP neural network model depends on the dimension of the input variables. The number of nodes in the input layer of the PSO-BP neural network model is set to 8.

2)

Output layer design: the number of nodes in the output layer depends on the set number of classifications.

3)

Hidden layer design: the role of the hidden layer is to extract the features of the data and process the data information.

4)

Selection of training function: In the BP neural network model, the most commonly used training functions are trainlm, trainbfg, trainrp and traingd. In this paper, trainlm function is chosen to train the model.

5)

Select the transfer function: the most commonly used transfer functions include logsig, tansig and purelin. In this paper, the purelin function is chosen according to the actual situation, which can limit the output value to a specific range.

6)

Setting initial parameters: the particle swarm algorithm is used to optimize the weights and thresholds, and to precisely adjust the levels of the neural network, the number of neurons, the activation function, and the parameters of the particle swarm [34].

5

Simulation studies of financial risks

The financial data used in this experiment were obtained from the Thematic Database for Analyzing Financial Indicators of Listed Universities in China.

5.1

Training process of PSO-BP model

The PSO-BP model is trained using 150 sets of samples from the training set, and the remaining 50 sets of samples from the test set are used to predict the model that meets the accuracy requirements and to validate the effectiveness of the model. Among them, the end of the PSO algorithm calculation is limited to the calculation to reach the maximum number of iterations or to meet the desired error, while the PSO fitness function is adopted as the mean square error function used in the calculation of BP neural network error, and the error analysis is performed on the expected output of the model and the actual prediction output data. The learning and training method of BP is “LM algorithm”. The optimal individual fitness of PSO-BP algorithm is shown in Figure 2. It can be seen from the figure that the PSO algorithm in the iterative process, the fitness value of the particles change significantly, the error is decreasing, and in 160 iterations, the error sum of squares of the test samples is reduced to 0.1405.

The change of mean square error during the training process of PSO-BP network is shown in Figure 3. From the figure, it can be seen that after the principal component analysis method and PSO algorithm optimized BP model for training, the model after 160 times of training, the overall error is stable at 5.2%, the model reaches the convergence state, and effectively avoids the defects that exist in the classical BP model.

5.2

Analysis and comparison of model effects

After the PSO-BP network finished training, the empirical study was conducted using test sample data. In order to verify the effectiveness of the proposed model, this paper uses the same test samples to test and compare the BP network and GA-BP network, and the comparison of the prediction results and the expected outputs of the BP model, the GA-BP model, and the PSO-BP model are shown in Fig. 4, Fig. 5, and Fig. 6. The comparison of the accuracy of the prediction results of the three models is shown in Table 5. As can be seen from the table, the PSO-BP prediction model achieves 91.7% recognition accuracy for the test samples, which is a significant improvement compared to the 68% recognition correctness of the traditional BP model and 83.3% of the GA-BP model. Therefore, the PSO-BP prediction model has better financial risk prediction ability. In practice, the improved BP model established in this paper based on principal component analysis and PSO algorithm can provide a certain reference basis, and colleges and universities and investors can cast different degrees of attention to colleges and universities according to the specific risk judgment value of the model.

Table 5.

The BP model is compared with the PSO-BP and GA-BP model

	Test sample number	Miscalculation number	Accuracy/%
BP prediction model	60	19	68
GA-BP prediction model	60	10	83.3
PSO-BP prediction model	60	5	91.7

Aiming at the problem of low accuracy of financial risk prediction, this paper establishes a model of PSO-optimized BP network through the analysis of particle swarm algorithm and BP model. Using the faster convergence speed of the PSO algorithm to find the global optimal point, the BP network is given the optimal initial weights and thresholds, which is applied to the financial risk prediction of listed colleges and universities, and the prediction effect is compared with the BP model and the GA-BP model. The simulation experiments show that: 1)

The BP neural network model does not have too much restriction on the research sample data, has the ability of self-learning and self-adaptation, has a wide range of applicability, and has a certain degree of superiority in the prediction of financial risk, but it still exists in itself with the problems of easy to fall into the local optimum, long arithmetic time, and low prediction accuracy.

2)

Aiming at the shortcomings of BP network, principal component analysis and particle swarm optimization algorithm are introduced into the neural network, and optimization is carried out in terms of the input and the initial parameters of the model, respectively, and the improved BP model significantly improves the accuracy of financial risk prediction.

3)

The BP model based on principal component and particle swarm optimization not only improves the nonlinear mapping ability of the BP network, but also enhances the convergence speed and overall performance of the network, which provides a better nonlinear fitting ability and higher prediction accuracy for the financial status of colleges and universities compared with the original BP prediction model and GA-BP prediction model.

6

Optimization strategy of financial informationization construction path in colleges and universities

6.1

Transform the concept of financial informatization

Financial informatization construction in colleges and universities should not only increase the investment in hardware and software equipment, but also according to the current stage of financial management of the business process and work mode, the construction of financial management system to match the current development of colleges and universities, in order to improve the quality of financial informatization construction. Part of the traditional financial management mode of colleges and universities can not meet the needs of the new period of financial informatization construction, colleges and universities should be through the work of research, seminars and other ways to guide the relevant personnel to establish the concept of modern financial informatization.

6.2

Addressing information silos

First, it is necessary to increase the investment in hardware and software required for the construction of financial informatization. Colleges and universities can ask software developers to focus on the development of accounting systems, accounting systems, remote reporting systems, payroll systems, tuition management systems, and the integration of systems, the use of a unified database, the formation of a consistent user interface, to open up the data interfaces between the systems, and the development of standards for the entry and sharing of information, to solve the problem of information silos. Second, optimize the business process. Colleges and universities should combine the construction of financial management systems, optimize business processes based on data sharing, and adjust business and financial job settings. Third, realize the sharing of financial data. Colleges and universities should unify data sharing standards, establish corresponding data sharing systems, require all functional departments to maintain communication with the financial department, take the initiative to cooperate with the financial management work, provide standard data in accordance with the requirements of the financial department, standardize the data sharing behavior, and improve the utilization rate of data.

6.3

Optimize the functionality of the financial management system

Colleges and universities should optimize the functions of their financial management systems in an orderly manner in accordance with the principle of overall planning and step-by-step implementation, taking into account the construction of financial informationization. First, to add a convenient financial services, information query function, in the existing system interface through eye-catching way to guide the user to use the function, to ensure that the user can through the system in a timely manner query to the financial business contacts, bidding information and so on. Secondly, the financial service function should be optimized and upgraded, and the software developer should be required to improve the financial reimbursement function on the existing financial system, and add the functions of funding authorization management and online declaration of labor expenses. Thirdly, the travel expense reporting function should be optimized, and the software developer is required to add budget declaration function and advance borrowing ticket function.

6.4

Strengthening information security precautions

The financial management system of colleges and universities in the era of informationization is always in an open state, and there are certain security risks during sharing and use. Therefore, colleges and universities should strengthen the information security precautionary work in the construction of financial informationization.

First of all, it is necessary to establish and improve the financial management network security system, to clarify the financial data security standards in the system, and to implement specific responsibilities to the departments and people, regular and irregular organization of network security checks to find out whether there is any information leakage or security risks, and requires real-time reporting of the results of the investigation to the higher authorities to formulate the next phase of the work plan for information security, to improve the security of the information through the normalization of the investigation. Preventive effect. Secondly, antivirus software and firewalls should be installed. Colleges and universities should organize information technology personnel to install relevant antivirus software and firewalls for existing computer equipment, explain the use of antivirus software and firewalls to the management and use of computer equipment, and ask them to do a good job of virus checking and killing and security precautions independently at a later stage to ensure the stable operation of the system.

Finally, encryption passwords and gateways should be designed. Colleges and universities should assign professional information technology personnel to participate in the construction of financial information technology, strengthen the protection of important financial data according to the type of existing financial data, design encryption passwords and gateways and regularly replace the encryption passwords in order to safeguard the security of financial data and prevent information leakage.

6.5

Enhance the business capacity of information technology construction practitioners

On the one hand, colleges and universities can introduce professional financial personnel through social recruitment. In order to ensure that the business ability of the introduced financial personnel meets the requirements of financial information construction, universities should organize the head of the financial department and information technology personnel to participate in the recruitment process during the recruitment period, and examine whether the business ability of the candidates meets the requirements through on-site question and answer, case analysis and other methods during the interview, and select high-quality financial personnel. On the other hand, colleges and universities should strengthen internal education and training to improve the business ability of information technology construction practitioners. At the same time, colleges and universities can also allow the financial department to organize and summarize the learning videos related to the construction of financial information technology, and recommend them to the practitioners, requiring the practitioners to learn independently and continuously improve their business level.

7

Conclusion

With a series of developments such as the formation of multiple financing methods in universities, the transformation of investment methods, and the continuous improvement of the internal control system in universities, the study of the risk of financial informationization and its control in universities is a process of continuous transformation.

In the analysis of the correlation coefficient of financial data of colleges and universities, the financial data with stronger correlation are the rate of return on investment (X21) and the ratio of annual borrowing to total annual income (X12) of 0.95, the rate of return on investment (X21) and the financial risk of infrastructure investment (X22) of 0.94, and the ratio of annual borrowing to total annual income (X12) and the financial risk of infrastructure investment (X22) of 0.89. These three financial indicators respond to the ability of universities to service their debt has room for simplification.

In the experiments of comparing the prediction results and expected output of BP model, GA-BP model and PSO-BP model, the PSO-BP prediction model achieves 91.7% of identification accuracy for the test samples, which is a significant improvement compared to the identification correctness of traditional BP model and GA-BP model. The PSO-BP neural network model has a better prediction rate and a higher degree of fitting. It is effective to apply PSO-BP neural network model to predict financial risk in colleges and universities, and it is practical and operable in practical application.

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

Research on the Risks of Financial Informatization Construction in Colleges and Universities, Its Prevention and Control and Path Optimization

Xin Lv

Pubblicato online: 26 set 2025

Ricevuto: 09 gen 2025

Accettato: 29 apr 2025

DOI: https://doi.org/10.2478/amns-2025-1039

Parole chiaveParticle swarm algorithm, Principal component analysis, PSO-BP neural network, Risk prediction, College finance

© 2025 Xin Lv, published by Sciendo.

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

Parole chiave
Particle swarm algorithm, Principal component analysis, PSO-BP neural network, Risk prediction, College finance