Intelligent financial management of company based on neural network and fuzzy volatility evaluation

Abstract

By integrating business processes, financial intelligent management system can provide effective data information with high quality and low cost for strategic decision-making. In this paper, the authors analyze the intelligent financial management of company based on neural network and fuzzy volatility evaluation. With the help of the development and support of information technology, the construction of financial intelligent management center realizes the efficient and standardized business process by building the information system center platform. It is the basis for the smooth operation of process reengineering and can guarantee the successful and safe operation of financial intelligent management center. The correlation dimension information reflecting the financial data abnormal feature is extracted to construct discrimination statistic and test criterion. The abnormal feature of financial data is mined according to the significant difference of discrimination statistic to realize anomaly analysis of financial data. The results show that the accuracy of financial data anomaly mining with this method is better. It has good application value in the financial audit and economic investigation and other fields.

Keywords

Neural network volatility assessment corporate finance intelligent management system

1 Introduction

In recent years, with the acceleration of economic development, information technology is also constantly updated, more and more enterprises began to seek ways of scale expansion [1]. In order to maintain their advantages in the fierce competition, some large-scale enterprise groups set up their branches to expand in a large scale across regions [2, 3]. The branches of the enterprise groups are all over the country, and their business types are widely involved. But at the same time of opening up the market, the increase of branches also has a negative impact on the management efficiency of enterprise groups [4]. Each increased business unit needs to be equipped with corresponding financial department and relevant personnel, which directly leads to the rise of human cost. At the same time, due to the scattered organization, it is easy to cause the information communication is not timely and the information transmission is not accurate. At the same time, there are certain differences in the accounting standards between branches and subsidiaries [5, 6]. The differences in standards will lead to the difficulty of enterprise management, at the same time, it will cause certain risks to the financial management and production and operation of the enterprise group, and it will produce a sustainable and stable development Set obstacles. Therefore, if the enterprise group wants to expand the scale and reduce the management cost, and strengthen the financial supervision of its branches, it needs to change the traditional management concept and seek a new operation management mode [7]. The exploration of the operation management mode often starts from the internal core departments. As the hub of the whole operation process of the enterprise group, the financial management of the company is the key point for the enterprise to realize the transformation of the management operation mode [8, 9]. The company makes a breakthrough from the direction of the financial management mode, so as to transform to the financial intelligent management mode of the company. It unifies the financial work repeatedly set up by different business units in the enterprise, so as to solve the problem of resource waste in the finance of large-scale groups, and change the existing disadvantages through the rational allocation of resources. At the same time, it will produce obvious results for business efficiency and process management, so as to solve the problems brought by multi regional management [10].

With the development of large data information processing technology, it provides strong data support for the statistical departments and financial audit departments to use large data information processing methods for financial data processing and analysis. The financial data includes bank journal data, financial expenditure data, and financial income data etc. The data mining and information feature extraction technology are used to analyze the anomaly of financial data [11, 12]. The attribute features reflecting abnormal information of financial data are mined to realize scientific analysis of financial data. The study of abnormal analysis method of financial data has important application value in economic crime investigation and audit investigation.

The financial intelligent management system integrates the business process through standardized and unified standards, provides high-quality and low-cost effective data information for strategic decision-making, and releases more basic financial personnel to invest in business activities to create value for enterprises, which becomes an important way of financial transformation [13]. According to the research, the advantages of financial intelligent management system are mainly reflected in improving the standardization process of enterprise standards, cost control and promoting the development of enterprise core business. By centralizing the financial departments and information systems of each branch, the enterprise financial intelligent management system no longer sets up corresponding supporting departments separately. After realizing the basic work of centralized integration, standardized operation is formed in the aspects of financial workflow, management method and data caliber by establishing unified operation mode and standard. Under this operation mode, the financial situation of each branch is under centralized control, information communication and transmission will be more accurate and timely, and the management efficiency and development speed of the enterprise will be improved.

The establishment of the financial intelligent management system is to centralize the repeated businesses of different branches of the enterprise, which will release a large number of financial personnel in basic positions. Only a part of the staff need to be retained to handle the business. Because of the standardization of the management operation and the flow of the business scope, the working time of the financial personnel will be greatly reduced, and the efficiency of the management will be improved, so as to achieve Cost reduction caused by reduction of personnel on the premise that business volume increases or remains unchanged. At the same time, the centralized management of funds is realized through the centralized business process, so as to form the scale effect. By using the characteristics of scale effect of financial intelligent management, the bargaining power of enterprises is improved and the financing cost of enterprises is reduced, so as to lay a foundation for improving the efficiency of enterprises.

2 Related work

There are many factors affecting the abnormalities of financial data. The distribution features of abnormal attributes have big randomness, with the features of time-varying and autocorrelation coupling [14, 15]. It is difficult to analyze the anomaly of financial data. The construction of abnormal features mining model for financial data is a kind of abnormal features mining and data prediction problem for a group of economic sample sequences. In the traditional method, the anomaly analysis method for financial data mainly includes supporting vector machine method, decision tree analysis method, and statistical feature analysis and autocorrelation feature mining method [16]. The abnormal attribute mining of financial data is achieved with the above method decomposing the economic analytic model of financial data as the statistics including multi-linear components, so as to realize the feature reconstitution and multivariate linear equation of financial data by combining the statistics information processing of financial data and empirical data analysis method, thus to test your financial data anomaly analysis, and achieve some research result.

In this paper, a statistical feature analysis and abnormal behavior mining method of financial data based on CART decision tree data mining model is presented in literature [17, 18]. Semantic similarity statistical analysis method is combined to realize entity identification and attribute correlation analysis of financial data to extract information feature mutually for data flow and anomaly mining of financial data, but this method cannot effectively realize the decoupling of financial data in the analysis [19]. The singularity in data mining process is bigger and the cost of parallel computing is larger. Literature proposes a method for analyzing the abnormal financial data based on supporting vector machine and analytical method of PCA main component [20]. Firstly, redundant information of financial data is filtered and carried out with reduplication processing to reduce computational overhead of financial data analysis, extract spectral features of financial data flow, combining high order spectral analysis for financial data abnormal feature extraction, to achieve nonlinear time series analysis of financial data and significant difference analysis of abnormal data, but the immunity of this method is not good in performing the anomaly analysis of financial data, and accuracy of anomaly feature mining is not very high.

In order to solve the above problems, this paper proposes a method of financial data anomaly analysis based on correlation data feature mining technology. Firstly, the phase space reconstitution method is used to carry out high dimensional feature expansion of financial data. The principal component analysis method is used for auto-correlation feature matching of financial data, combining with the alternative data method for randomization processing of financial data, and then feature compression method is used for non-correlation redundancy processing [21]. High-dimensional multivariate statistical analysis is combined for correlation mapping of financial data, to extract the correlation data reflecting the abnormal features of financial data, construct the discriminant statistics and test criteria. Financial data anomaly feature mining is carried out according to the significant difference of discriminant statistics to achieve abnormal financial data analysis. Finally, the empirical data analysis and simulation experiments are carried out [22]. The method in this paper is used to realize the abnormal analysis of financial data, which is shown, to draw the conclusion of validity.

By centralizing the financial departments and information systems of each branch, the enterprise financial intelligent management system no longer sets up corresponding supporting departments separately. After realizing the basic work of centralized integration, standardized operation is formed in the aspects of financial workflow, management method and data caliber by establishing unified operation mode and standard. Under this operation mode, the financial situation of each branch is under centralized control, information communication and transmission will be more accurate and timely, and the management efficiency and development speed of the enterprise will be improved.

With the help of the development and support of information technology, the construction of financial intelligent management center realizes the efficient and standardized business process by building the information system center platform. It is the basis for the smooth operation of process reengineering and can guarantee the successful and safe operation of financial intelligent management center. The construction of the information system platform is the core of the construction of the financial intelligent management service center, which strengthens the coordination between the systems. The company has formulated the construction goal of “one platform (Financial intelligent management platform) and five systems". It mainly includes comprehensive budget management system, fund management system, new tax management system, later management accounting report system and performance appraisal system for comprehensive evaluation. Mainly on the basis of budget, build a comprehensive budget management system, at the same time carry out a comprehensive supervision of funds, and add the tax control module to the management scope, improve the whole chain of financial management.

The company’s financial intelligent management information system platform will implement a management mode with cost target budget as the guide, project contract management and control as the core and image system and reimbursement system as the support. It will provide financial information basis for management through the accounting business of the sharing center, from business bidding management to contract signing, and then subcontract according to the project cost budget, through contracting and subcontracting The business forms creditor’s rights and debt management respectively, implements the collection and payment plan according to the corresponding needs, and finally enters the financial Intelligent Management Center for business processing. It realizes the mutual connection of Financial Sharing Center, business accounting system, cost system, capital system and tax system, and integrates the core business of each business department into the financial intelligent management center.

At present, nine subsystems have been built on the basis of the five system construction goals. The data generated by each information management system is transferred to the financial Intelligent Management Center for processing, forming a unified and complete system, and realizing the mutual integration of financial business. In order to supervise the operation efficiency of the information management platform, the company formulated relevant systems to regularly evaluate and supervise the information construction process, and accelerated the docking of ERP system with financial intelligent management platform and OA system.

3 Information flow reconstitution and preprocessing of financial data

3.1 Financial data flow reconstitution

In order to realize the anomaly analysis of financial data, information flow model of financial data needs to be constructed to extract the abnormal characteristics of financial data, analyzing the change and increasing situation of financial data through the data mining technology, to provide the reference index for financial statistics and auditing department. The paper analyzes the statistical characteristics of financial data through constructing a mining model for financial data and journal trend effectively and synthesizing financial data feature distribution, adopting certain mathematical model construction method to construct information flow model of sampling financial data, combining economic sequence analysis method to make financial data flow reconstitution and feature reorganization.

The financial data is a set of nonlinear economic sequences. The nonlinear economic sequence analysis method can be used to mine the abnormal characteristics of financial data. Through statistics and sampling of apriori information of financial department, inversion economic sequence of univariate financial data can be constructed as {x_n}. High-dimensional phase space reconstitution technology is used for financial data correlation feature reorganization. The abnormal characteristics of financial data is analyzed in the phase space. The information flow model of financial data high-dimensional characteristics is ${x_{i}}_{i = 1}^{N}$ . Optimal correlation decomposition of the data is done. The phase space vector model of financial data obtained in the reconstituted dimension space is described as: $L = [\begin{matrix} x_{1}^{T} \\ x_{2}^{T} \\ ⋮ \\ x_{N}^{T} \end{matrix}] = [\begin{matrix} x_{1} & x_{1 + τ} & \dots & x_{1 + (m - 1) τ} \\ x_{2} & x_{2 + τ} & \dots & x_{2 + (m - 1) τ} \\ ⋮ & ⋮ & ⋱ & ⋮ \\ x_{N - 1} & x_{N - 1 + τ} & \dots & x_{N - 1 + (m - 1) τ} \end{matrix}]$ (1)

In the formula, the embedding dimension of financial data in the phase space is m. The sampling time delay of financial data is τ. In the m dimension space of reconstituted data map, the attribute characteristics reflecting financial data category is extracted. The test and statistic model of financial data L is constructed to generate a Gaussian time series y(n) and a discriminant statistical model of financial data is constructed to generate the corresponding alternative sequence: $x_{n} = φ_{0} + \sum_{i = 1}^{p} φ_{i} x_{n - i} + \sum_{j = 0}^{q} θ_{j} η_{n - j}$ (2)

In the equation, {η_i} is financial data redundancy interference item with mean value of 0, and variance of σ², φ₀, φ₁, φ₂, ⋯ φ_p is known as randomization coefficient, θ₁, θ₂, ⋯ θ_q is known as time window. The singular value decomposition method is used to obtain the inversion irreversible value of financial data anomaly feature mining. The singular value decomposition process is L = U * S * C , U and C is a phase randomized matrix, and $C = (c_{1}, c_{2}, \dots, c_{n})$ (3)

Fourier conversion is carried out for the statistical series of original financial data. Sigma test criterion is used to obtain the third order autocorrelation value of financial data to get the generalized inverse positive solution of financial data distributed in the phase space. Here, S is the singular value of L . The main component vector of financial data redundancy information is: $S = diag (σ_{1}, σ_{2}, \dots, σ_{n}), σ_{1} \geq σ_{2} \geq \dots \geq σ_{n} \geq 0$ (4)

For any orthogonal matrix, in the linear subspace, phase space reconstitution track matrix L and self-organizing map functional is used to get the phase space control matrix of N × m dimensions through balance point control: $X = [\begin{matrix} x_{1} \\ x_{2} \\ ⋮ \\ x_{3} \end{matrix}] = [\begin{matrix} a_{1}^{T} c_{1} & a_{1}^{T} c_{2} & \dots & a_{1}^{T} c_{m} \\ a_{2}^{T} c_{1} & a_{2}^{T} c_{2} & \dots & a_{2}^{T} c_{m} \\ ⋮ & ⋮ & ⋱ & ⋮ \\ a_{N}^{T} c_{1} & a_{N}^{T} c_{2} & \dots & a_{N}^{T} c_{m} \end{matrix}]$ (5)

In this paper, $J_{x}^{(N)}$ is obtained through singular value decomposition of the ith eigenvector $J_{x_{i}}^{(N)}$ , combining the non-linear anomaly feature mining method for the maximum Lyapunove index functional of financial data. Set up a separation coefficient $J_{x}^{(1)}$ for the separation processing of redundant information of financial data according to the number of training samples of financial data and abundance of redundant data. Blind source separation technology is adopted to get the blind source separation covariance matrix C of financial data: $C = \frac{1}{N} [X - \bar{X_{l}}] {[X - \bar{X_{l}}]}^{T}$ (6)

In the equation: $l = {[1, 1, \dots 1]}_{1 \times N}$ (7) $\bar{x_{i}} = \frac{1}{N} \sum_{k = 1}^{N} x_{ik}$ (8) $X = [X_{1}, X_{2}, \dots, X_{m}]$ (9)

Remove the dimension of original financial data to get the slack variable of financial data in the phase space: $δ x_{i + 1} \approx J_{x_{i}}^{(1)} δ x_{i}$ (10)

In the equation δ x _i = x _j- x _i, x _j is the vector quantization information of the original data sample sequence in phase space, and the corresponding m principal component feature subset is x_j ∈ R ^m. So let x_n+1 = μx_n (1-x_n) be the state information parameter of abnormal distribution of financial data. When the detection statistic meets δ x _i+1 = x _j+1- x _i+1, autoregression ARMA estimation is carried out through financial economic sample sequence of the original statistics [8] to get δ x _i and δ x _i+1. And when the number of principal components satisfies the constraint condition U = { u (t) |u (t) ∈ X , ∥ u ∥ ≤ d, t ∈ I } , cumulative variance of financial data is calculated in the reconstituted phase space, and the Euler distance of financial data flow distribution track in high-dimensional phase space meets: $∥ x_{j} - x_{i} ∥ \leq ɛ$ (11)

In the formula, (ɛ is a small positive number), thus the relevance mapping vector x _j of financial data is determined, thus forming a ontology model writing matrix form of financial data parallel mining: $B_{x_{i}} = {(δ x_{i} (j_{1}), δ x_{i} (j_{2}), \dots, δ x_{i} (j_{N_{b}}))}^{T}$ (12)

Through the above phase processing, phase space reconstitution method is used to carry out high dimensional feature expansion of financial data, providing the original data input basis for the abnormal analysis of financial data.

3.2 Autocorrelation feature matching of financial data

Based on high-dimensional feature expansion of financial data by using phase space reconstitution method, the principal component analysis method is used for auto-correlation feature matching of financial data, to improve the accuracy for mining of financial data anomaly. Time-frequency analysis method is used to obtain the correlation dimension of financial data as: $x (t) = Re {a_{n} (t) e^{- j 2 π f_{c} τ_{n} (t)} s_{l} (t - τ_{n} (t)) e^{- j 2 π f_{c} t}}$ (13)

In the formula, on the time scale of information flow distribution of financial data, multiple wavelets are decomposed to get the transient disturbance of financial data as: $c (τ, t) = \sum_{n} a_{n} (t) e^{- j 2 π f_{c} τ_{n} (t)} δ (t - τ_{n} (t))$ (14)

In the equation above, a_n(t) is thee inversion integral function of abnormal feature performed on the nth phase space distribution trajectory. τ_n (t) is the time delay of financial empirical sampling. f_c is the sampling frequency of main feature quantity of financial data, and s_l(t) is the transitional information of simple component.

The grid segmentation method is used for transient state disturbance optimization of financial data, and the optimal correlation distribution function of financial data is obtained as follows: $h (t) = \sum_{i = 1}^{P} a_{i} p (t - τ_{i})$ (15)

In the equation, a_i and τ_i are the similarity coefficient and disturbance amplitude respectively. The relevance mapping of financial data is carried out in the reconstituted high-dimensional feature space of financial data, and the attribute characteristics reflecting financial data category are extracted, to get the autocorrelation feature distribution function of financial data as: ${\begin{matrix} y (t) = x (t - t_{0}) \Rightarrow W_{y} (t, v) = W_{x} (t - t_{0}, v) \\ y (t) = x (t) e^{j 2 π v_{0} t} \Rightarrow W_{y} (t, v) = W_{x} (t, v - v_{0}) \end{matrix}$ (16)

Through the time-frequency scaling, spectral analysis technology is used for the abnormal feature clustering process of financial data. The data clustering center for the construction of financial data abnormal feature mining is described as: $y (t) = \sqrt{k} x (kt), k > 0$ (17) $W_{y} (t, v) = W_{x} (kt, v / k)$ (18)

Here, k refers to the disturbance amplitude of data clustering, v is expressed as the sampling spectrum value of financial data, W_x is the time window function. Set the amplitude of financial data anomaly feature mining information flow as A. Main feature quantity N_k(x) of optimal correlation data of financial data is obtained. The alternative data analysis method is used to get the information classification error of financial data: $W_{x} (t, v) = \int_{- \infty}^{+ \infty} x (t + τ / 2) x^{*} (t - τ / 2) e^{- j 2 π v τ} d τ$ (19)

In the finite domain, the frequency domain distribution of abnormal characteristics of financial data is expressed as: $W_{x} (t, v) = \int_{- \infty}^{+ \infty} X (v + ξ / p - 2) X^{*} (v - ξ / 2) e^{j 2 π ξ t} d ξ$ (20)

In the above equation, ξ is the attenuation coefficient in the clustering process of financial data abnormal features, and X is the bifurcation distance of financial data classification vector. X^* refers to looking for the conjugate of the financial data. Hyperbolic frequency modulation is carried out for financial data anomaly detection system. According to autocorrelation r feature matching technique, the matching result of abnormal feature is: $p (Q_{s}) = a_{k} \frac{1}{\sqrt{2 π} σ_{s}} exp [- \frac{{(Q_{s} - 〈 Q_{s} 〉)}^{2}}{2 {σ_{s}}^{2}}]$ (21)

Here, a_k is recognition coefficient of financial data attribute. If a_k = 0, it means the kth abnormal data clustering center tends to zero. If a_k = 1, it means there is a disturbance in the clustering space, it indicates that there is abnormal data.

3.3 Purchasing and accounts payable business

Before the process reengineering of the financial intelligent management center, first analyze the business processes, connect with the business processes of each department, define and confirm the scope of business processes that need to be adjusted due to the construction of the financial intelligent management center, and then redesign and transform it. To fully analyze the existing constraints between departments and the interrelationship between business processes, it is necessary to cooperate with information system in the process of business process design and reengineering to optimize business processes to the greatest extent.

The company’s new financial process reengineering framework has been transformed into transferring the financial data of each branch’s business department directly to the financial Intelligent Management Center for processing, which has changed the three-level management mode under the traditional mode. This process has realized the standardization of standards, reduced the processing level and the total amount of work, and improved the work efficiency. Through the business types transferred to the financial intelligent management center to confirm the scope of business process, and through standardized processing to reengineer each process, the important link of the construction of financial intelligent management is realized. The company follows the principle of business process reengineering, optimizes the business process, and improves the efficiency of financial processing by formulating corresponding standards. It mainly reengineers the corresponding purchase and payment business, cost and expense management, sales and collection business, asset management, general ledger and report business.

Procurement and accounts payable business is the earliest and most basic content of financial intelligent management application. Under the traditional financial management mode, each secondary organization has the financial autonomy, and each unit is responsible for the procurement and accounts payable business. The financial department of the headquarters does not interfere, which is not conducive to the understanding and management of the headquarters on the business situation. However, in the financial intelligent management mode, it is necessary to redesign the process of procurement and accounts payable business, and no longer use the previous three-level management system, and concentrate the procurement and accounts payable business to the Financial Sharing Center for processing. The headquarters of the company can monitor the operation status of each branch in real time, and connect the procurement and accounts payable business with the budget system, so as to fully develop them We should play the role of budget management, strengthen cost control and reduce financial risks. The company’s procurement business mainly refers to the cost control of the construction unit through budget management control and the procurement of large-scale production equipment, engineering production materials and other bulk businesses according to the needs of the project schedule during the construction process. There are a few sporadic purchases with high frequency and less funds involved, which can be classified into expense management for business processing according to the situation.

In the process of procurement and accounts payable business process design, it is necessary to combine risk prevention and control, and divide the business process design into three important links. First, budget generation and review is to make corresponding budget according to the construction plan, which is reviewed by relevant personnel. The basic basis of business data information is to complete the approved data, and to calculate the cost and budget data Comparison can strengthen cost control and risk prevention, and strictly control budget to prevent over expenditure. Then the business is preliminarily reviewed. According to the budget entry contract, the application is submitted through the scanning data of the image system and is waiting for review. The main purpose is to preliminarily review the authenticity of the business. After passing the preliminary review, it will enter the business review of the sharing center, control the business compliance and whether it is within the budget through the financial intelligent management center, and feed back to the performance department and relevant business departments, so as to control the problems in time, and pay for the approved online banking, and the financial intelligent management center will feed back in time in case of failure for special reasons.

4 Realization of financial data abnormal mining

4.1 Randomization and de-redundancy processing of financial data

Based on the information flow reconstitution and autocorrelation feature matching of financial data performed above for abnormal feature mining, this paper presents a method to analyze the abnormal financial data based on the correlation data feature mining technology. Alternative data method is used for the randomization of financial data processing. Alternative data method is from the modern statistics Bootstrap theory, and the realization process is described as:

Phase-randomize the original financial data to generate a set of Gaussian economic sequences y(n) to obtain a linearly related Gaussian process;

Take the rank of the original financial data x(n) as the association rule mapping vector set, for the financial data association rule mapping

r_{n} = y (x_{n}) = y_{rank (x_{n})}, n = 1, 2, \dots, N

(22)

In the association rule mapping system, the quaternary group of financial data attribute classification is constructed;

${r (n)}_{n = 1}^{N}$ is made Fourier transformation, and optimal matching of financial data is carried out to get a new sequence y′ (n), and the association rules and autoregressive eigenvectors are the same;

According to the rank of y′ (n), financial data is rearranged to achieve the randomization processing of financial data and generate alternative sequences.

The attribute characteristics reflecting abnormal information of financial data are extracted, and elimination of redundancy is made for the substitute data x(k). The redundant vector of the financial data is: $W_{y} (t, v) = \int_{- \infty}^{+ \infty} W_{h} (t - s, v) W_{x} (s, v) ds$ (23)

In the equation, t is data sampling time, s is correlation coupling coefficient of financial data. In the data clustering center, the self-adaption weighting W_e = (ω_j^(e), 0) of financial data classification is carried out non-correlation redundancy removal process by feature compression method. K-L transformation is used to obtain characteristic compression results:

$\begin{matrix} {| \int_{- \infty}^{+ \infty} x (t) y^{*} (t) dt |}^{2} = \int_{- \infty}^{+ \infty} \int_{- \infty}^{+ \infty} \\ W_{x} (t, v) W_{y}^{*} (t, v) dtdv \end{matrix}$ (24)

Set the error of abnormal information mining of financial data as $\bar{p} = \sum p_{i} \in N_{k} (x) p_{i} θ_{p_{i}} (x)$ . The nonlinear abnormal feature mining is used to get the non-correlation redundancy blind source separation results of the financial data as: $f_{x} (t) = \frac{\int_{- \infty}^{+ \infty} v W_{x_{α}} (t, v) dv}{\int_{- \infty}^{+ \infty} W_{x_{α}} (t, v) dv}$ (25)

In the equation, x_α is the number of main components reflecting abnormal information in the financial data information flow.

4.2 Analysis of abnormal characteristics of financial data

In the nonlinear economic sequence of financial data, high-dimensional multivariate statistical analysis method is used to map the financial data. Assume that the financial data is generated by the linearly related nonlinear economic sequence, and the following ARMA model is used: $x_{n} = a_{0} + \sum_{i = 1}^{M_{AR}} a_{i} x_{n - i} + \sum_{j = 0}^{M_{MA}} b_{j} η_{n - j}$ (26)

In the equation, a₀ is the sampling amplitude of initial financial data, x_n-i is the financial data scalar economic sequence with the same mean, variance, b_j is the oscillation amplitude of financial data. Third-order autocorrelation statistic is used as the inspection statistical data of financial data abnormal analysis: $C_{or 3} = \frac{〈 (x_{n} - \bar{x}) (x_{n - d} - \bar{x}) (x_{n - D} - \bar{x}) 〉}{〈 {(x_{n} - \bar{x})}^{3} 〉}$ (27)

In the equation, x_n refers to non-linear economic sequence of financial data, d refers to the time interval for sampling financial data, D = 2d, $\bar{x}$ represents the mean, 〈x (n)〉 represents taking the mean of the financial data sample x(n): $〈 x (n) 〉 = 1 / N \sum_{n = 1}^{N} x (n)$ (28)

The trajectory of non-linear economic sequence and vector feature economic sequence of financial data in the high-dimensional phase space is {x (t₀ + iΔt)}, i = 0, 1, ⋯ , N-1. Decision tree model is used for the anomaly of financial data. The basic idea is to measure the time-varying characteristics of economic data and randomness characteristics x and x_n_+τ. Through the linear correlation processing, the average mutual information C(τ) of financial data is defined as: $C (τ) = lim_{T \to \infty} \frac{1}{T} \int_{- \frac{T}{2}}^{\frac{T}{2}} x (t) x (t + τ) d τ$ (29)

In the equation, τ is the time delay of financial data in the reconstituted phase space, representing the correlation degree of financial data change at the time of t and t+τ. The abnormal behavior trend and abnormal feature of financial data are mined based on the correlation degree to get the correlation dimension information in vector space of financial data economic sequence as: $I (τ) = - \sum_{ij} p_{ij} (τ) ln \frac{p_{ij} (τ)}{p_{i} p_{j}}$ (30)

A point in the economic sequence phase space of reconstituted financial data is expressed as X _n, and the nearest neighbor point in the set of stochastic phases of financial data is expressed as X _η(n), thus the economic sequence exported by financial data anomaly feature mining is: $X (n) = {x (n), x (n + τ), \dots, x (n + (m - 1) τ)}$ (31)

Finally, Sigma test method is used to construct the test criterion of abnormal feature mining of financial data. According to the significant difference of discriminant statistic, the accuracy of abnormal feature mining of financial data is tested. The test criteria is: $p (Q_{s}) = \frac{1}{\sqrt{2 π} σ_{s}} exp [- \frac{{(Q_{s} - 〈 Q_{s} 〉)}^{2}}{2 {σ_{s}}^{2}}]$ (32) $\int_{- \infty}^{\infty} p (Q_{s}) d Q_{s} = 1$ (33)

In the equation, p (Q_s) ∼ (Q_s) curve is shown as Fig 1.

Fig. 1

p(Q_s)∼(Q_s) curve distribution of test criterion.

According to Fig. 1, financial data abnormal feature mining distribution meets the standard normal distribution, and if the difference between Q₀ and 〈Q_s〉 exceeds a certain threshold Q_c, making:

$p (| Q_{0} - 〈 Q_{s} 〉 | > Q_{c}) \leq 0.05$ (34)

At this point the confidence of financial data mining is 95%, because the normal distribution is symmetrical on both sides of 〈Q_s〉, there should be: $0.025 = \int_{- \infty}^{z_{2}} p (Q_{s}) d Q_{s} = 1 - \int_{- \infty}^{z_{1}} p (Q_{s}) d Q_{s}$ (35)

In the equation, z₂ =-z₁, when S ≥ 2.00, the abnormal distribution of financial data is not established with 95% probability, accepting the original hypothesis, the financial data anomaly mining results meet the convergence conditions.

5 Simulation experiment and result analysis

In order to test the performance of this method in the realization of financial data anomaly analysis, simulation experiments and empirical data analysis are carried out. The hardware environment of simulation experiment is PC. The configuration parameters are CPU 3.0 G, Core (TM) CPU T6600, 12G internal storage. The empirical data analysis software includes Excel 2007 and SPSS19.0. The related parameters of financial data test statistics are: Q = 200, c₁ = 30, c₂ = 10, c_r = 2, μ₁ = μ₂ = 0.01, ρ₁ = ρ₂ = 0.01, δ = 0.8. The financial data is from a large group, and the statistical time is from January to April 2017. The similarity correlation coefficient of financial data is μ=12. The abnormal feature sampling rate of financial data is f_s = 10 * f₀Hz = 10KHz, and the frequency band of data distribution is 4∼25 KHz. According to the F test and correlation analysis of financial data carried out with Hausman test, the correlation coefficient is shown in Table 1.

Table 1
Analysis Result of Financial Data

Correlation coefficient x ₁ x ₂ x ₃ x ₄ x ₅ x ₆

x ₁ 1 0.543 0.454 0.454 0.235 0.543

x ₂ 0.345 1 0.454 0.456 0.365 0.432

x ₃ 0.454 0.435 1 0.654 0.432 0.212

x ₄ 0.543 0.654 0.465 1 0.638 0.345

x ₅ 0.875 0.432 0.543 0.854 1 0.512

x ₆ 0.754 0.345 0.323 0.843 0.654 1

Correlation coefficient	x ₁	x ₂	x ₃	x ₄	x ₅	x ₆
x ₁	1	0.543	0.454	0.454	0.235	0.543
x ₂	0.345	1	0.454	0.456	0.365	0.432
x ₃	0.454	0.435	1	0.654	0.432	0.212
x ₄	0.543	0.654	0.465	1	0.638	0.345
x ₅	0.875	0.432	0.543	0.854	1	0.512
x ₆	0.754	0.345	0.323	0.843	0.654	1

The moral hazard and adverse selection risk of managers and employees also directly or indirectly affect the business performance of the company. In addition, governments and opportunities are also major variables that affect stakeholders. As a party to stakeholders, their financial policies and fiscal policies directly affect the interests of other stakeholders. Technological innovation and technological breakthroughs not only provide opportunities for enterprises and stakeholders, but also bring risks and challenges. Therefore, enterprises are supported by cooperation with stakeholders, and they also bear risks in the game with stakeholders. The construction of branch functions as show in Table 1.

It is reflected by the business risk early warning index, investment risk early warning index and financing risk early warning index. The financial stakeholders risk index early warning subsystem is its lagged early warning index system, including the supply value chain risk warning and the interest relationship value risk warning, which are reflected through the supply value chain risk warning index and the interest relationship value risk early warning index.

In terms of financing risk, the resource support required by the company’s operations ultimately comes down to the support of capital, and the effective operation of capital ultimately comes down to the company’s cash flow. Fundraising brings the cost of capital to the company. The blindness and failure of capital operation will inevitably affect the liquidity and solvency of the company. As a result, the financial relationship will be extremely deteriorated and financial constraints will be encountered. Therefore, enterprises are constantly acquiring resources in the operations, financing, and investment operations, and are also resolving risks in the allocation of resources for operations, financing, and investment.

According to the priori statistical results of financial data samples, the time-domain waveforms of the samples of 4 groups of financial data are shown in Fig. 2.

Fig. 2

Financial data samples.

The abnormal feature mining is carried out for the sampled financial data. The correlation dimension feature of financial data is mined to get the abnormal feature mining results of financial data, as shown in Fig. 3.

Fig. 3

Abnormal feature mining of financial data.

After analyzing the results of Fig. 3, it shows that the method in this paper for the associated information mining of financial data anomaly characteristics has obvious beam directivity, indicating that the anti-interference ability is obvious. In order to compare performance, the method in this paper and traditional method are used to analyze the accuracy of financial data anomaly feature mining. The results are shown in Fig. 4. The analysis shows that the method in this paper for financial data anomaly mining is more accurate. It improves the right mining ability for the abnormal data mining.

Fig. 4

Comparison of abnormal feature mining accuracy of financial data.

6 Company’s intelligent financial optimization strategy

6.1 Improve the effectiveness of business process

Based on the construction of the information system platform, the company realizes the main link of the construction of the financial intelligent management service center through the design and adjustment of the business process. The smooth connection between the business process and the financial intelligent management platform determines the effectiveness of the operation. When building the financial intelligent management center, the design and improvement of the business process should be carried out at any time according to the strategic objectives and development needs Appropriate adjustment, try to make each business process to achieve effective use and high-speed operation. For the management of business process, we can take the way of analysis and discussion to find solutions to the problems in the use process. At the same time, the optimization of business processes should be investigated in-depth by the business departments, the business processes necessary for the implementation of the operation of the financial intelligent management center should be preserved, and the business processes with certain restrictions and low use frequency should be eliminated and refined to ensure the effectiveness of all business processes and achieve the goal of making full use of resources. The construction of branch functions as show in Table 2.

Table 2
The construction of branch functions

X Y R

E1>E2 E2–E1 <

E1> =E2 E2–E1 < =

E1<E2 E1–E2 <

E1< =E2 E1–E2 < =

E1>E2 E2–E1 <

E1> =E2 E2–E1 < =

E1<E2 E1–E2 <

E1< =E2 E1–E2 < =

E1==E2 abs(E1–E2) ==

E1> =E2 E2–E1 <

E1<E2 E1–E2 ==

E1< =E2 E1–E2 <

E1==E2 abs(E1–E2) <

E1> =E2 E2–E1 < =

E1<E2 E1–E2 <

E1< =E2 E1–E2 < =

E1==E2 abs(E1–E2) ==

E1!=E2 –abs(E1–E2) <

X	Y	R
E1>E2	E2–E1	<
E1> =E2	E2–E1	< =
E1<E2	E1–E2	<
E1< =E2	E1–E2	< =
E1>E2	E2–E1	<
E1> =E2	E2–E1	< =
E1<E2	E1–E2	<
E1< =E2	E1–E2	< =
E1==E2	abs(E1–E2)	==
E1> =E2	E2–E1	<
E1<E2	E1–E2	==
E1< =E2	E1–E2	<
E1==E2	abs(E1–E2)	<
E1> =E2	E2–E1	< =
E1<E2	E1–E2	<
E1< =E2	E1–E2	< =
E1==E2	abs(E1–E2)	==
E1!=E2	–abs(E1–E2)	<

6.2 Continuous improvement of business process

To adjust and improve the business process, it is necessary to run through the whole business process. It needs to be carried out continuously for a long time. According to the development of the enterprise and the internal and external environment, the company needs to make corresponding adjustments in time. To improve the business process, the company needs a professional process management team to be responsible for the confirmation and management of the business process scope, and must have professional technology Only in this way can we find out the problems in the business process and optimize the plan in time. There should be strict requirements and standards for the selection of professional business process management personnel, as well as the ability of good communication and coordination, and be able to achieve information exchange between the financial personnel and the staff of each business department, excavate the information related to the business process from the exchange and conduct in-depth analysis to find solutions to the problems. At the same time, the participants of business process should be encouraged to put forward more use feelings and suggestions related to business process, cultivate the participation consciousness of all employees, publicize the importance of business process management, advocate the staff team to actively make suggestions and suggestions, and the opinions and suggestions from the practice of grass-roots staff team are more practical. Therefore, some incentive measures should be given to the employees who put forward valuable suggestions To mobilize all employees to contribute to the goal of the financial intelligent management center.

6.3 Adjust risk control measures in time

In the process of transformation of financial management mode, resources will be re integrated with big data and other information technologies. The corresponding organizational structure and business process will be redesigned and adjusted. The concentration of financial accounting also brings the concentration of risk. The change of management mode will have a certain impact on the original internal control management. Therefore, when building the financial intelligent management center, the It is necessary to strengthen the management of internal control, formulate internal control standards more suitable for the new management mode of financial intelligent management, and provide guarantee for the company to achieve the new goal of overall development. Individuals in the top 10 as show in Table 3, and risk evaluation shows in Table 4.

Table 3
Individuals in the top 10

Number Chromosome coding Parameter value Adaptability

A B C

1 110101011100011100 53 28 28 67

2 100110101001100111 38 41 39 58

3 011011011010010110 27 26 22 55

4 000111001100001000 7 12 8 55

5 100110101101101100 38 45 44 54

6 011011100011100010 27 35 34 53

7 111011111010101011 59 58 43 51

8 110110111000111011 54 56 59 49

9 100001011110011100 33 30 28 49

10 110011111000111010 51 56 58 48

Number	Chromosome coding	Parameter value	Adaptability
1	110101011100011100	53	28	28	67
2	100110101001100111	38	41	39	58
3	011011011010010110	27	26	22	55
4	000111001100001000	7	12	8	55
5	100110101101101100	38	45	44	54
6	011011100011100010	27	35	34	53
7	111011111010101011	59	58	43	51
8	110110111000111011	54	56	59	49
9	100001011110011100	33	30	28	49
10	110011111000111010	51	56	58	48

Table 4

Risk evaluation

FIR	FAR	FRR
N	0.000001	0.01
N	0.01	0.01
N	0.001	3.3
N	0.1	0.4
M	0.0001	2.6
M	0.0002	0.05
M	0.0001	0.0001
M	0.0001	0.003
M	0.0001	0.03
P	0.001	3.3
P	0.002	0.05
P	0.0001	3.3
P	0.0002	0.05
Q	0.0001	0.0001
Q	0.0001	0.005
Q	0.0001	0.5

6.4 Improve the internal control system

Strengthening the control activities of the financial intelligent management center and paying attention to the risk management of receivables can promote the expansion of income scale. The company’s financial intelligent management center has the characteristics of large-scale project and long construction time due to the particularity of its industry, which leads to the long recovery time of accounts receivable. For a / R and a / P management, the financial intelligent management center should set up a collection department and a credit rating approval management department to audit and supervise the business, do a good job in risk assessment, select a reputable enterprise cooperation to reduce the possibility of bad debts and dead accounts, and set up a feedback agency to control and monitor the situation of receivables. By tracking the payment, we try to avoid the improper behavior of intentionally defaulting payment, and strengthen the cooperation between departments to strengthen the management of collection and payment. At the same time, the financial department should provide the corresponding basis for the internal control management according to the real-time situation of accounts receivable, and adjust the risk control measures in time according to the dynamic situation.

7 Conclusion

In this paper, financial data anomaly analysis method is studied. A method of financial data anomaly analysis based on correlation data feature mining technology is proposed. The phase space reconstitution method is used to carry out high dimensional feature. The principal component analysis method is used for correlated features matching, combined with alternative data method for randomization of financial data. Feature compression method is used for non-correlation redundancy removal processing. High-dimensional multivariate statistical analysis method is used for correlation mapping of financial data. The correlation dimension information reflecting the financial data abnormal feature is extracted to construct discrimination statistic and test criterion. The abnormal feature of financial data is mined according to the significant difference of discrimination statistic to realize anomaly analysis of financial data. The results show that the accuracy of financial data anomaly mining is better than that of financial data, and the accuracy of financial data is analyzed. The results show that the accuracy of financial data anomaly mining with this method is better. It has good application value in the financial audit and economic investigation and other fields.

References

Quan

, Srinivasan

and Khosravi

, Particle swarm optimization for construction of neural network-based prediction intervals, Neurocomputing127 (2014), 172–180.

Zhang

, Gong

, Hu

et al., Feature selection algorithm based on bare bones particle swarm optimization, Neurocomputing148(1) (2015), 150–157.

Das

, Pattnaik

P.K.

and Padhy

S.K.

, Artificial Neural Network trained by Particle Swarm Optimization for non-linear channel equalization, Expert Systems with Applications41(7) (2014), 3491–3496.

Prediction of seismic slope stability through combination of particle swarm optimization and neural network, Engineering with Computers32(1) (2016), 85–97.

Pulido

, Melin

and Castillo

, Particle swarm optimization of ensemble neural networks with fuzzy aggregation for time series prediction of the Mexican Stock Exchange, Information Sciences280 (2014), 188–204.

Chatterjee

, Sarkar

, Hore

et al., Particle Swarm Optimization Trained Neural Network for Structural Failure Prediction of Multi-storied RC Buildings, Neural Computing and Applications28(8) (2016), 2005–2016.

Jadoun

V.K.

, Gupta

, Niazi

K.R.

et al., Modulated particle swarm optimization for economic emission dispatch, International Journal of Electrical Power & Energy Systems73 (2015), 80–88.

Rashid

T.A.

and Ahmad

H.A.

, Lecturer performance system using neural network with Particle Swarm Optimization, Computer Applications in Engineering Education (2016).

Hajihassani

, Armaghani

D.J.

, Monjezi

et al., Blast-induced air and ground vibration prediction: a particle swarm optimization-based artificial neural network approach, Environmental Earth Sciences74(4) (2015), 2799–2817.

10.

Sheng-Feng

, Xiao-Hua

and Lu

, Application of wavelet neural network with improved particle swarm optimization algorithm in power transformer fault diagnosis, Power System Protection & Control42(19) (2014), 37–42.

11.

Kuo

R.J.

, Hung

S.Y.

and Cheng

W.C.

, Application of an optimization artificial immune network and particle swarm optimization-based fuzzy neural network to an RFID-based positioning system, Information Sciences262(3) (2014), 78–98.

12.

, Qian

and Li

et al., Resource allocation based on quantum particle swarm optimization and RBF neural network for overlay cognitive OFDM System, Neurocomputing, 2015:S092523121501303X.

13.

, Hu

and Bai

, Generalized radial basis function neural network based on an improved dynamic particle swarm optimization and AdaBoost algorithm, Neurocomputing152(C) (2015), 305–315.

14.

Zhuowei

H.U.

, Yiming

, Zhiheng

et al., Research on Spatial Characteristics of Regional Poverty Based on BP Neural Network: A Case Study of Wuling Mountain Area, Journal of Geo-Information Sci-ence17(1) (2015), 69–77.

15.

Huckans

, Fuller

B.E.

, Chalker

A.L.N.

et al., Plasma Inflammatory Factors Are Associated with Anxiety, Depression, and Cognitive Problems in Adults with and without Methamphetamine Dependence: An Ex-ploratory Protein Array Study, Frontiers in Psychiatry6(4) (2015), 178.

16.

Chand

, Alam

M.A.

and Tejaswini

Y.R.S.N.

, Performance comparison of artificial neural networks learning algorithms and activation functions in predicting severity of autism, Network Modeling Analysis in Health Informatics and Bioinformatics4(1) (2015), 2.

17.

Kedia

, Mussweiler

and Linden

D.E.

, Brain mechanisms of social comparison and their influence on the reward system, Neuroreport25(16) (2014), 1255–1265.

18.

Factor

, Innovation, intellectual property, and economic growth, Science-Technology and Management6(5) (2014), 347–348.

19.

Sehrawat

, Giri

A.K.

and Wagner

N.F.

, Financial development and economic growth: empirical evidence from India, South Asia Economic Journal11(2) (2015), 287–308.

20.

Zhang

K.H.

, Does Foreign Direct Investment Promote Economic Growth?, Evidence From East Asia And Latin America, Contemporary Economic Policy19(1) (2014), 175–85.

21.

Menyah

, Nazlioglu

and Wolde-Rufael

, Financial development, trade openness and economic growth in African countries: New insights from a panel causality approach, Economic Modelling37(574) (2014), 386–394.

22.

Cuaresma

J.C.

, Doppelhofer

and Feldkircher

, The Determinants of Economic Growth in European Regions, Regional Studies48(1) (2014), 44–67.