The role of information technology-based accounting internal control strategies in corporate risk prevention

A data warehouse is a database system consisting of one or more applications and a database for analyses and reporting that derives data from other data sources. There is usually a loading of data into the database followed by periodic snapshots or incremental updates of the data to provide the information needed for more important business decisions.

The data warehouse snowflake model is shown in Figure 1. In addition to having the functionality of a dimension table in the star model, the dimension table is connected to a detailed category table that provides a detailed description of the fact table. The detailed category table achieves the purpose of narrowing down the fact table and improving query efficiency by providing a detailed description of the fact table in the dimension in question.

Figure 1.

Structure diagram of snowflake model

The function of a data warehouse is to screen, format, convert, clean and analyse data from other information systems or data collection tools and to bring together the processed data into a stored whole for value discovery by data mining systems.

Data mining is the act of using computer algorithms to discover valuable information from enterprise databases, without including the initial data collection and processing, etc. [26]. In the actual processing, the object of data mining may be either structured data, semi-structured data, or even other structured data. Figure 2 illustrates the specific process of data mining.

Figure 2.

Data mining process diagram

The complete data mining architecture is the process of combining the data warehouse with other information systems. Data mining should first determine the goal of the problem to be solved, and then the data mining professionals use the relevant algorithmic tools to explore the data warehouse, find the laws and generate models; finally, the model is nested in the target data and the results are obtained, thus generating reports and other information that can be used.

ETL technology helps companies to solve the problem of data integration and consistency, which is, i.e., the process of data extraction, transformation and loading, which is an important part of constituting a data warehouse [27]. Firstly, the data needs to be extracted from different data sources; secondly, the data is converted into the specified format by tools; and finally, the data is inserted into the target database, i.e., the data warehouse.

Multidimensional analysis in OLAP refers to analysing data in a multidimensional dataset by slicing, dicing, and rotating the data so that the user can look at the data in the data warehouse from multiple perspectives and sides [28]. In OLAP analysis, there are also drill-down operations such as “scroll up”, “drill down” and “drill through”. The essence of these operations is to transform the granularity of the analysis by changing the level of dimensions. “Drilling up” is to generalise from low-level detail data to high-level summary data in a certain dimension, or to reduce the number of dimensions, while “drilling down” is the reverse of “drilling up”, which drills down from the summary data to the detail data for observation. Aggregated data can be used to drill down to the details to observe or add new dimensions.

By entering one or more variables into a linear combination to the expected value of the target is called multiple linear regression [29-30]. The general form of the multiple linear regression method can be expressed as: (1)y=β0+β1x1+β2x2+⋯+βkxk+ε$$y = {\beta _0} + {\beta _1}{x_1} + {\beta _2}{x_2} + \cdots + {\beta _k}{x_k} + \varepsilon$$ 1)

Function Fitting

When the data has multiple eigenvalues, set to x₁, x₂ etc., and one target value set to h, based on the existing eigenvalues and target value the equation can be assumed to be h = α₀ + α₁x₁ + α₂x₂, and then assuming that the value of Definition x₀ is equal to 1, then the above mathematical equation can be expressed as h=∑i=0nαixi=αTX$$h = \sum\limits_i = {0^n}{\alpha _i}{x_i} = {\alpha ^T}X$$:

Since the error follows a Gaussian distribution, taking the error into a normal distribution function, where μ = 0, can be obtained: (3)p(y(i)|x(i);α)=1σ2πe−(y(i)−αTx(i))22σ2$$p({y^{(i)}}|{x^{(i)}};\alpha ) = \frac{1}{{\sigma \sqrt {2\pi } }}{e^{ - \frac{{{{({y^{(i)}} - {\alpha ^T}{x^{(i)}})}^2}}}{{2{\sigma ^2}}}}}$$ 3)

Likelihood function

Likelihood function is mainly used to describe the likelihood in the parameters of the model, when the results obtained from certain observations are known, the general likelihood is usually used to estimate the parameters of the characteristics of the things associated with them.

If the output is x, the probability that the likelihood function L(θ|x) is equal to the variable X given the parameter θ is: (4)L(θ|x)=p(X=x|θ)$$L(\theta |x) = p(X = x|\theta )$$

Maximum Likelihood Estimation: maximum likelihood estimation means to maximise this function over all values of θ. And the best parameter of the solution required in this paper is the α that needs to satisfy the maximum likelihood estimation , so to bring in formula (4) to get: (5)L(α)=∏i=1mp(y(i)|x(i);α)=∏i=1m1σ2πe−(y(i)−αTx(i))22σ2$$L(\alpha ) = \prod\limits_{i = 1}^m p ({y^{(i)}}|{x^{(i)}};\alpha ) = \prod\limits_{i = 1}^m {\frac{1}{{\sigma \sqrt {2\pi } }}} {e^{ - \frac{{{{({y^{(i)}} - {\alpha ^T}{x^{(i)}})}^2}}}{{2{\sigma ^2}}}}}$$

The log-likelihood function is often used in maximum likelihood estimation and related fields with the following formula: (6)logL(α)=log∏i=1mp(y(i)|x(i);α)=log∏i=1m1σ2πe−(y(i)−αTx(i))22σ2=∑i=1mlog1σ2πe−(y(i)−αTx(i))22σ2=mlog1σ2π−1σ2×12∑i=1m(y(i)−αTx(i))2$$\begin{array}{rl} \log L(\alpha ) = \log \prod\limits_{i=1}^{m}{p}({{y}^{(i)}}|{{x}^{(i)}};\alpha )=\log \prod\limits_{i=1}^{m}{\frac{1}{\sigma \sqrt{2\pi }}}{{e}^{-\frac{{{({{y}^{(i)}}-{{\alpha }^{T}}{{x}^{(i)}})}^{2}}}{2{{\sigma }^{2}}}}} \\ {} = \sum\limits_{i=1}^{m}{\log }\frac{1}{\sigma \sqrt{2\pi }}{{e}^{-\frac{{{({{y}^{(i)}}-{{\alpha }^{T}}{{x}^{(i)}})}^{2}}}{2{{\sigma }^{2}}}}} \\ {} = m\log \frac{1}{\sigma \sqrt{2\pi }}-\frac{1}{{{\sigma }^{2}}}\times \frac{1}{2}\sum\limits_{i=1}^{m}{{{({{y}^{(i)}}-{{\alpha }^{T}}{{x}^{(i)}})}^{2}}} \\ \end{array}$$

In order to maximise the log-likelihood function, mlog1σ2π$$m\log \frac{1}{{\sigma \sqrt {2\pi } }}$$ is a constant, and it is then necessary to make the values subtracted after it smaller to obtain the objective function: (7)J(α)=12∑i=1m(y(i)−αTx(i))2=12∑i=1m(Y(x(i))−y(i))2=12(Xα−y)T(Xα−y)$$\begin{array}{l} J(\alpha ) = \frac{1}{2}\sum\limits_{i = 1}^m {{{({y^{(i)}} - {\alpha ^T}{x^{(i)}})}^2}} = \frac{1}{2}\sum\limits_{i = 1}^m {{{(Y({x^{(i)}}) - {y^{(i)}})}^2}} \\\quad\quad{} = \frac{1}{2}{(X\alpha - y)^T}(X\alpha - y) \\ \end{array}$$

Apply the partial derivative to α: (8)12(2XTXα−XTy−(yTX)T=XTXα−XTy)$$\frac{1}{2}(2{X^T}X\alpha - {X^T}y - {({y^T}X)^T} = {X^T}X\alpha - {X^T}y)$$

Making the partial derivative equal to 0 gives: (9)α=(XTX)−1XT(Least square)$$\alpha = {({X^T}X)^{ - 1}}{X^T}{\text{(Least square)}}$$ 4)

Least Squares

In the study of the relationship between two variables, usually, you can get a series of pairs of data, such as the variable (x, y) , then can get the data for the ((x₁, y₁)……(x_m, y_m); these data will be mapped in the right-angled coordinate system, you can get the corresponding points distributed around a straight line, set the equation of this straight line for: (10)yi=a0+a1x(a0 and a1 Is any real number)$${y_i} = {a_0} + {a_1}x({a_0}{\text{ and }}{a_1}{\text{ Is any real number}})$$

To determine the values of a₀ and a₁, the sum of the squares of the differences between the actual y_i and the calculated value of y_j(y_i = a₀ + a₁x) is required: (11)φ=∑i=1n(yi−yj)2$$\varphi = \sum\limits_{i = 1}^n {{{({y_i} - {y_j})}^2}}$$

Bringing Eq. (3-10) into Eq. (3-111) gives: (12)φ=∑i=1n(yi−a0−a1xi)2$$\varphi = \sum\limits_{i = 1}^n {{{({y_i} - {a_0} - {a_1}{x_i})}^2}}$$

In order to minimise the value of φ, let the function φ take a partial derivative with respect to a₀ and a₁ so that its partial derivative is equal to zero, and solve for the minimum value to obtain the system of equations: (13){∑i=1n2(a0+a1xi−yi)=0∑i=1n2xi(a0+a1xi−yi)=0$$\left\{ {\begin{array}{*{20}{l}} {\sum\limits_{i = 1}^n 2 ({a_0} + {a_1}{x_i} - {y_i}) = 0} \\ {\sum\limits_{i = 1}^n 2 {x_i}({a_0} + {a_1}{x_i} - {y_i}) = 0} \end{array}} \right.$$

Solving the system of equations gives: (14)a0=1n∑i=1n(yi−a1xi)$${a_0} = \frac{1}{n}\sum\limits_{i = 1}^n {({y_i} - {a_1}{x_i})}$$ (15)a1=n∑i=1nxiyi−∑i=1nxi∑i=1nyin∑i=1nxi2∑i=1nxi∑i=1nxi$${a_1} = \frac{{n\sum\limits_{i = 1}^n {{x_i}} {y_i} - \sum\limits_{i = 1}^n {{x_i}} \sum\limits_{i = 1}^n {{y_i}} }}{{n{{\sum\limits_{i = 1}^n {{x_i}} }^2}\sum\limits_{i = 1}^n {{x_i}} \sum\limits_{i = 1}^n {{x_i}} }}$$

Finally, the results from the solution are brought into the above equation to obtain the desired linear equation.

Internal control of an accounting information system is the process of ensuring the normal operation of the accounting information system through some management tools and system technology, as well as ensuring that the system truly reflects and records information and securely protects and stores information. Accounting information system internal control system functions include:

Preventive control measures are functional measures taken to avoid risks beforehand. Detective control functions refer to treatment and compensatory measures taken to prevent ongoing incidents from causing greater losses. Corrective control functions are designed to remedy incidents that have already occurred.

In this paper, the COBIT 5.0 model is used to design an information-based accounting internal control system. The main processes of internal control of the system include the following five aspects: 1)

Guidance and Monitoring

Guidance and monitoring refers to resource optimisation and risk control in the process of setting up the internal control framework of the accounting information system, taking into account the interests of all stakeholders, and making the financial and management information of the enterprise transparent to those who have the appropriate information authority

The architecture of the accounting big data analysis platform construction is shown in Figure 3. It mainly contains the following layers:

Figure 3.

Overall architecture diagram of the platform

Cloud service platform layer: network equipment, storage equipment, operating system, etc., for system management construction. Since the platform is based on the cloud computing service model, the basic IT environment is provided by the cloud service provider.

Data Acquisition Layer: Accounting data obtained from various internal departmental systems and external networks of the enterprise, including accounting business data, accounting and financial management data, industry development, and publicly disclosed accounting information of competing enterprises.

Data Processing and Storage Layer: Unified processing and integration of the acquired accounting big data and categorisation of the processed data to be stored in different databases.

Data Output Display Layer: Process accounting data with a variety of mining and analysis tools and output the information obtained after processing from different functional modules. 1)

Basic IT environment deployment

The accounting big data analytics platform constructed in this paper selects the platform-as-a-service (PaaS) model. It is located in the middle layer of the cloud computing architecture, and enterprises can build an accounting big data analysis platform based on this service model without defining accounting data storage scalability or managing and controlling cloud infrastructure, networks, servers, operating systems or storage.

1)

Comprehensive financial analysis

Introducing the Harvard analysis framework for financial analysis of the enterprise’s accounting big data. In strategic analysis, establish an analysis system of industry development, competitive strategy, and enterprise business strategy situation. In accounting analysis, the accounting policy analysis of key accounting items of the enterprise can be combined with the horizontal analysis of competitors or the vertical analysis based on previous years. In financial analysis, an in-depth analysis of the enterprise’s financial statements is carried out by combining the trend analysis method, comparison analysis method, and ratio analysis method. In prospect analysis, data mining technology is used to achieve prospect prediction analysis after loading the knowledge base by combining factors such as internal innovation, external policy, customer demand, and internal business development trends.