1. A multi-dimensional credit evaluation model construction method based on performance ability and behaviors is characterized by comprising the following steps:

the method comprises the following steps:

a, acquiring credit level of the commercial and negotiable enterprises to construct a target layer, constructing a criterion layer and an index layer through the performance behavior and performance capability of the credit level of the commercial and negotiable enterprises,

b, establishing a principal component of a multi-dimensional credit evaluation model, and establishing an index layer and a criterion layer for the principal component;

c, evaluating the index layer by using a principal component analysis method;

evaluating the criterion layer by using an entropy method;

and E, weighting the index layer comprehensive evaluation value data after the standardization processing to obtain the final evaluation value of the mode.

2. The method of claim 1, wherein the multidimensional credit evaluation model based on performance and behavior comprises: the performance behaviors comprise management capacity and transaction capacity, the performance capacity comprises financial conditions and social behaviors, the management capacity, the transaction capacity, the financial conditions and the social behaviors are constructed into a criterion layer, the management capacity comprises human resource management, business management and intangible asset management, the transaction capacity comprises main products, sales income and sales objects, the financial conditions comprise financial quality, financial statements and debt repayment, and the social behaviors comprise public credit information and market credit information.

3. The method of claim 1, wherein the multidimensional credit evaluation model based on performance and behavior comprises: the method for evaluating the index layer by the principal component analysis method comprises the following steps:

setting m evaluation objects, using n evaluation indexes x₁,x₂,L x_nEvaluation was carried out. The index value can form an m × n order matrix x ═ x_ij)_m×nLet x_k＝(x_1k,x_2k,L x_nk)^TThe kth column vector of x. Obtaining expected mu-mu of n indexes₁,μ₂,L μ_n]^TLet v be_ij＝cov(x_i,x_j)(i,j∈[1,n]Wherein cov (x)_i,x_j) Denotes x_iAnd x_jThe covariance between. Thereby obtaining an nxn order covariance matrix V ═ V_ij]。

Establishing a mathematical model:

in the formula: a ═ α₁,α,L α₁]^TA＝[α₁,α₂,L α_n]^T；α₁,α₂,Lα_nCoefficients for n indices; d (y) is the variance of y; y is a linear function of the configuration of A and x.

The solution by using the Lagrange multiplier method comprises the following steps:

D(y)＝A^TVA＝λA^TA＝λ (2)

wherein λ ═ λ₁,λ₂,L λ_n]The characteristic value of V is from large to small.

Let λ_iCorresponding feature vector is gamma_i＝(γ_i1,γ_i2,L γ_in)^TThen the ith principal component is

The contribution rate of the ith principal component is

β_iThe contribution rate used for measuring the ith principal component is larger, the larger the contribution rate is, the larger the cumulative contribution rate of the first q principal components is

If the cumulative contribution rate exceeds 85%, the q principal components may be used. In order to fully utilize the original information without discarding any component, the contribution rate of each principal component is used as its weight. Then, using the obtained principal component and its weight, the overall score of each mode can be obtained as

In the formula, P is an m-dimensional column vector.

Standardizing data, calculating comprehensive evaluation values of different criteria according to divided levels, firstly obtaining an evaluation matrix formed by indexes associated with a criterion for one criterion, further obtaining a covariance matrix V of the evaluation matrix, and further obtaining corresponding eigenvalues lambda of each principal component_iAnd a feature vector gamma_iThe principal component y is obtained by using the equations (3) and (4), respectively_iAnd its contribution ratio beta_iFinally, the overall evaluation value P of each mode under this criterion is obtained by using equation (6).

4. The method of claim 3, wherein the model is constructed based on performance and behavior, and comprises: the method for standardizing the data comprises the following steps:

in the formula:the maximum value of the j index of the m evaluation objects is shown;the minimum value of the j index of the m evaluation objects is shown; b representsA benefit type index set; c represents a cost index set.

5. The method of claim 1, wherein the multidimensional credit evaluation model based on performance and behavior comprises: in the method for evaluating the criterion layer by using the entropy method, m evaluation objects are assumed in the application of the criterion layer, q evaluation criteria are assumed, and an evaluation matrix e is formed as (e)_ij)_m×qFor one evaluation criterion, if the difference between the m values is larger, the function of the index in evaluation is larger.

Entropy of defining the ith evaluation criterion is

In the formula:k is 1/ln m; and, assume f_ijWhen equal to 0, f_ijln f_ij＝0。

An entropy weight of the ith evaluation criterion may be defined as

In the formula, 0 is not more than omega_iLess than or equal to 1 and

and obtaining the comprehensive evaluation value of each criterion, and if the cost index and the benefit index are distinguished during standardization, the subsequent standardization is considered as the benefit index, and the index layer comprehensive evaluation value after standardization is obtained. Thus, the entropy value of each criterion can be obtained by using the formula (7), the weight vector omega can be obtained by using the formula (8), and the final evaluation values of different evaluation objects can be obtained by simply weighting the overall evaluation value data of the index layer after standardization processing, so that the score ranking of the evaluation objects can be obtained. When the evaluation values obtained by other methods are compared in a ranking mode, sensitivity analysis of the evaluation values can be carried out, and a sensitivity analysis formula is as follows:

in the formula (I), the compound is shown in the specification,represents the maximum value of the evaluation value;the second largest value of the evaluation value is indicated. The meaning of sensitivity is: the higher the sensitivity is, the better the evaluation value discrimination obtained by the used evaluation algorithm is, and the better the evaluation effect is.

Background

Quality credits are the willingness and ability of an enterprise to fulfill quality commitments, and quality credits are essentially the manifestation of contractual relationships between the enterprise and consumers regarding product quality. The fact proves that the product has good image of high quality and high credit, and has the function of improving the consumption confidence of people and pulling economic growth which is difficult to replace. Due to imperfect economic order of the market, part of enterprises lack quality credit consciousness, and quality loss behaviors such as fake-making and selling, poor-quality buildings, toxic food, service default and the like still occur in various fields.

As a main body of quality credit, enterprises may go to risk in advance of huge profits brought by low cost and sacrifice the quality credit of the enterprises. Therefore, the construction of a quality credit system is strengthened, the relation between enterprise quality credit and consumers needs to be objectively analyzed, and effective progress can be achieved only by adopting targeted measures and guiding, so that a multi-dimensional credit evaluation model construction method based on performance capability and behavior is provided.

Disclosure of Invention

The invention aims to provide a multi-dimensional credit evaluation model construction method based on performance capability and behaviors, which aims to solve the problems in the prior art.

In order to achieve the purpose, the invention adopts the following technical scheme:

the invention comprises the following steps:

a, acquiring credit level of the commercial and trade circulation enterprises to construct a target layer, constructing a criterion layer and an index layer through the performance behavior and performance capability of the credit level of the commercial and trade circulation enterprises,

b, establishing a principal component of a multi-dimensional credit evaluation model, and establishing an index layer and a criterion layer for the principal component;

c, evaluating the index layer by using a principal component analysis method;

evaluating the criterion layer by using an entropy method;

and E, weighting the index layer comprehensive evaluation value data after the standardization processing to obtain the final evaluation value of the mode.

Further, the performance includes operational capability and transaction capability, the performance includes financial status and social behavior, the operational capability, the transaction capability, the financial status and the social behavior are constructed into a criterion layer, the operational capability includes human resource management, business management and intangible asset management, the transaction capability includes main products, sales income and sales objects, the financial status includes financial quality, financial statement and debt repayment, and the social behavior includes public credit information and market credit information.

Further, the method for evaluating the index layer by the principal component analysis method comprises the following steps:

setting m evaluation objects, using n evaluation indexes x₁,x₂,L x_nEvaluation was carried out. The index value can form an m × n order matrix x ═ x_ij)_m×nLet x_k＝(x_1k,x_2k,L x_nk)^TThe kth column vector of x. Obtaining expected mu-mu of n indexes₁,μ₂,Lμ_n]^TLet v be_ij＝cov(x_i,x_j)(i,j∈[1,n]Wherein cov (x)_i,x_j) Denotes x_iAnd x_jThe covariance between. Thereby obtaining an nxn order covariance matrix V ═ V_ij]。

Establishing a mathematical model:

in the formula: a ═ α₁,α,Lα₁]^T A＝[α₁,α₂,Lα_n]^T；α₁,α₂,Lα_nIs a coefficient of n indices; d (y) is the variance of y; y is a linear function of the configuration of A and x.

The solution by using the Lagrange multiplier method comprises the following steps:

D(y)＝A^TVA＝λA^TA＝λ (2)

wherein λ ═ λ₁,λ₂,Lλ_n]The characteristic value of V is from large to small.

Let λ_iCorresponding feature vector is gamma_i＝(γ_i1,γ_i2,Lγ_in)^TThen the ith principal component is

The contribution rate of the ith principal component is

β_iThe contribution rate used to measure the ith principal component is greater, indicating a greater contribution. The cumulative contribution rate of the first q principal components is

In general, if the cumulative contribution rate exceeds 85%, the q principal components may be used. In order to fully utilize the original information without discarding any component, the contribution rate of each principal component is used as its weight. Then, using the obtained principal component and its weight, the overall score of each mode can be obtained as

In the formula, P is an m-dimensional column vector.

Standardizing data, calculating the comprehensive evaluation values of different criteria of different evaluation objects according to the divided levels, firstly obtaining an evaluation matrix formed by the criteria associated with one of the criteria, further obtaining a covariance matrix V of the evaluation matrix, and further obtaining the corresponding eigenvalue lambda of each principal component_iAnd a feature vector gamma_iThe principal component y is obtained by using the equations (3) and (4), respectively_iAnd its contribution ratio beta_iFinally, the overall evaluation value P of each mode under this criterion is obtained by using the formula (6).

Further, the method for normalizing data includes:

in the formula:the maximum value of the j index of the m evaluation objects is shown;the minimum value of the j index of the m evaluation objects is shown; b represents a benefit type index set; c represents a cost index set.

Further, the method for evaluating the criterion layer by using the entropy method comprises the step of forming an evaluation matrix e-e (e) by assuming that m evaluation objects exist and q evaluation criteria are applied to the criterion layer_ij)_m×qFor one evaluation criterion, if the difference between the m values is larger, the function of the index in evaluation is larger.

Entropy of defining the ith evaluation criterion is

In the formula:k is 1/ln m; and, assume f_ijWhen equal to 0, f_ijlnf_ij＝0。

An entropy weight of the ith evaluation criterion may be defined as

In the formula, 0 is not more than omega_iLess than or equal to 1 and

and (4) obtaining the comprehensive evaluation value of each criterion, and if the cost index and the benefit index are distinguished during standardization processing, the subsequent standardization processing is regarded as the benefit index, and the standardized comprehensive evaluation value of the index layer is obtained by the formula (9).

Therefore, the entropy value of each criterion can be obtained by using the formula (7), the weight vector omega of each criterion can be obtained by using the formula (8), and the final evaluation values of different evaluation objects can be obtained by simply weighting the overall evaluation value data of the index layers after standardization processing, so that the score ranking of the evaluation objects can be obtained. When the evaluation values obtained by other methods are subjected to ranking comparison, the sensitivity analysis of the evaluation values can be carried out, and the sensitivity analysis formula is as follows:

in the formula (I), the compound is shown in the specification,represents the maximum value of the evaluation value;the second largest value of the evaluation value is indicated. Sensitivity contains: the higher the sensitivity is, the better the evaluation value discrimination obtained by the used evaluation algorithm is, and the better the evaluation effect is.

Compared with the prior art, the invention has the beneficial effects that:

according to the method, the evaluation is carried out by utilizing principal component analysis and an entropy method through a multi-dimensional enterprise credit evaluation model based on the performance capability and the performance behavior, so that the difference of the credit level of the enterprise can be obtained well, and the main influence elements of the credit evaluation can be identified.

Drawings

FIG. 1 is a schematic diagram of a multi-dimensional credit evaluation model construction method based on performance and behavior according to the present invention;

FIG. 2 is a schematic diagram of credit evaluation indexes of commercial and trade negotiable enterprises;

Detailed Description

The technical solutions in the embodiments of the present invention will be described clearly and completely with reference to the accompanying drawings in the embodiments of the present invention, and it is obvious that the described embodiments are only some embodiments of the present invention, not all embodiments.

As shown in fig. 1, in the present embodiment,

a, acquiring credit level of a commercial and trade circulation enterprise to construct a construction target layer, and constructing a criterion layer and an index layer through a performance behavior and performance capability of the credit level of the commercial and trade circulation enterprise;

10 typical enterprises in the trade and trade circulation industry are selected as research samples. Because the enterprises in the same industry are adopted, the interference of different industry factors is avoided, and the credit evaluation of the enterprises is more scientific and reasonable. The data obtained by calculating the scores using the relevant data indexes disclosed in the fourth quarter of 2020 each enterprise is shown in table 1.

TABLE 1 Enterprise index score data sheet

B, establishing a principal component of a multi-dimensional credit evaluation model, and establishing an index layer and a criterion layer for the principal component;

referring to fig. 2, the credit evaluation index of the business circulation enterprise is established according to the requirements of the credit evaluation index of the business circulation enterprise in the national standard GB/T39450-:

c, evaluating the index layer by using a principal component analysis method;

the method for evaluating the index layer by the principal component analysis method comprises the following steps:

setting m evaluation objects, using n evaluation indexes x₁,x₂,L x_nEvaluation was carried out. The index value can form an m × n order matrix x ═ x_ij)_m×nLet x_k＝(x_1k,x_2k,L x_nk)^TThe kth column vector of x. Obtaining expected mu-mu of n indexes₁,μ₂,Lμ_n]^TLet v be_ij＝cov(x_i,x_j)(i,j∈[1,n]Wherein cov (x)_i,x_j) Denotes x_iAnd x_jThe covariance between. Thereby obtaining an nxn order covariance matrix V ═ V_ij]。

Establishing a mathematical model:

in the formula: a ═ α₁,α,Lα₁]^T A＝[α₁,α₂,Lα_n]^T；α₁,α₂,Lα_nIs a coefficient of n indices; d (y) is the variance of y; y is a linear function of the configuration of A and x.

The solution by using the Lagrange multiplier method comprises the following steps:

D(y)＝A^TVA＝λA^TA＝λ (2)

wherein λ ═ λ₁,λ₂,Lλ_n]The characteristic value of V is from large to small.

Let λ_iCorresponding feature vector is gamma_i＝(γ_i1,γ_i2,Lγ_in)^TThen the ith principal component is

The contribution rate of the ith principal component is

β_iThe contribution rate used for measuring the ith principal component is larger, the larger the contribution rate is, the larger the cumulative contribution rate of the first q principal components is

If the cumulative contribution rate exceeds 85%, the q principal components may be used. In order to fully utilize the original information without discarding any component, the contribution rate of each principal component is used as its weight. Then, using the obtained principal component and its weight, the overall score of each mode can be obtained as

In the formula, P is an m-dimensional column vector.

Standardizing data, calculating the comprehensive evaluation values of different criteria of different evaluation objects according to the divided levels, firstly obtaining an evaluation matrix formed by the criteria associated with one of the criteria, further obtaining a covariance matrix V of the evaluation matrix, and further obtaining the corresponding eigenvalue lambda of each principal component_iAnd a feature vector gamma_iThe principal component y is obtained by using the equations (3) and (4), respectively_iAnd its contribution ratio beta_iFinally, the overall evaluation value P of each mode under this criterion is obtained by using the formula (6).

The data are first normalized using equation (9). And then, calculating the comprehensive evaluation values of all the criteria of different evaluation objects according to the levels divided in the graph. For one criterion, an evaluation matrix formed by indexes related to the criterion is firstly obtained, a covariance matrix V of the evaluation matrix is further obtained, and corresponding eigenvalues lambda of each principal component are further obtained_iAnd a feature vector gamma_iThe principal component y is obtained by using the equations (3) and (4), respectively_iAnd its contribution ratio beta_iFinally, the overall evaluation value P of each mode under this criterion is obtained by using equation (6).

TABLE 2 comprehensive evaluation value of index layer

Evaluating the criterion layer by using an entropy method;

6. in the method for evaluating the criterion layer by using the entropy method, m evaluation objects are assumed in the application of the criterion layer, q evaluation criteria are assumed, and an evaluation matrix e is formed as (e)_ij)_m×qFor a certain evaluation criterion, if the difference between m values is larger, the action of the index in evaluation is larger.

Entropy of defining the ith evaluation criterion is

In the formula:k is 1/ln m; and, assume f_ijWhen equal to 0, f_ij ln f_ij＝0。

An entropy weight of the ith evaluation criterion may be defined as

In the formula, 0 is not more than omega_iLess than or equal to 1 and

and obtaining the comprehensive evaluation value of each criterion, and if the cost index and the benefit index are distinguished during standardization, the subsequent standardization is considered as the benefit index, and the index layer comprehensive evaluation value after standardization is obtained. Therefore, the entropy value of each criterion can be obtained by using the formula (7), the weight vector omega of each criterion can be obtained by using the formula (8), and the final evaluation values of different evaluation objects can be obtained by simply weighting the overall evaluation value data of the index layers after standardization processing, so that the score ranking of the evaluation objects can be obtained. When the evaluation values obtained by other methods are compared in a ranking mode, sensitivity analysis of the evaluation values can be carried out, and a sensitivity analysis formula is as follows:

in the formula (I), the compound is shown in the specification,represents the maximum value of the evaluation value;the second largest value of the evaluation value is indicated. Sensitivity contains: the higher the sensitivity is, the better the evaluation value discrimination obtained by the used evaluation algorithm is, and the better the evaluation effect is.

The above-described comprehensive evaluation value of each criterion can be obtained, and the normalized overall evaluation value of the index layer can be obtained from expression (9). The entropy values for each criterion obtained using equation (7) are (0.7321,0.2734,0.3253,0.7683), respectively.

And E, weighting the index layer comprehensive evaluation value data after the standardization processing to obtain the final evaluation value of the mode.

Using equation (8), the weight vector ω is obtained (0.1324,0.6148,0.1486,0.1042),

therefore, the final evaluation values of the modes obtained after simple weighting by using the normalized index layer comprehensive evaluation value data are sequentially (0.6002, 0.5306, 0.3180, 0.3877, 0.7103, 0.5622, 0.4928, 0.3254, 0.3048 and 0.4873), so that the enterprise credit level ranking is obtained, wherein the credit level of enterprise 5 is the highest, enterprise 1 and enterprise 6 are the second, and enterprise 9 is the last.

The above description is only for the preferred embodiment of the present invention, but the scope of the present invention is not limited thereto, and any person skilled in the art should be considered to be within the technical scope of the present invention, and equivalent alternatives or modifications according to the technical solution and the inventive concept of the present invention should be covered by the scope of the present invention.