CN116227939A - Enterprise credit rating method and device based on graph convolution neural network and EM algorithm - Google Patents
Enterprise credit rating method and device based on graph convolution neural network and EM algorithm Download PDFInfo
- Publication number
- CN116227939A CN116227939A CN202310490149.1A CN202310490149A CN116227939A CN 116227939 A CN116227939 A CN 116227939A CN 202310490149 A CN202310490149 A CN 202310490149A CN 116227939 A CN116227939 A CN 116227939A
- Authority
- CN
- China
- Prior art keywords
- credit
- credit rating
- enterprise
- node
- rating
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
- 238000000034 method Methods 0.000 title claims abstract description 52
- 238000004422 calculation algorithm Methods 0.000 title claims abstract description 42
- 238000013528 artificial neural network Methods 0.000 title claims abstract description 24
- 239000011159 matrix material Substances 0.000 claims abstract description 77
- 239000013598 vector Substances 0.000 claims abstract description 59
- 238000010586 diagram Methods 0.000 claims abstract description 41
- 238000003062 neural network model Methods 0.000 claims abstract description 26
- 238000000605 extraction Methods 0.000 claims abstract description 12
- 238000004364 calculation method Methods 0.000 claims description 41
- 230000006870 function Effects 0.000 claims description 26
- 230000008569 process Effects 0.000 claims description 11
- 238000005457 optimization Methods 0.000 claims description 6
- 230000004913 activation Effects 0.000 claims description 5
- 238000010276 construction Methods 0.000 claims description 3
- 206010063385 Intellectualisation Diseases 0.000 abstract description 3
- 238000013527 convolutional neural network Methods 0.000 description 3
- 230000009286 beneficial effect Effects 0.000 description 2
- 238000012986 modification Methods 0.000 description 2
- 230000004048 modification Effects 0.000 description 2
- 238000005096 rolling process Methods 0.000 description 2
- 238000007619 statistical method Methods 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q10/00—Administration; Management
- G06Q10/06—Resources, workflows, human or project management; Enterprise or organisation planning; Enterprise or organisation modelling
- G06Q10/063—Operations research, analysis or management
- G06Q10/0635—Risk analysis of enterprise or organisation activities
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02P—CLIMATE CHANGE MITIGATION TECHNOLOGIES IN THE PRODUCTION OR PROCESSING OF GOODS
- Y02P90/00—Enabling technologies with a potential contribution to greenhouse gas [GHG] emissions mitigation
- Y02P90/30—Computing systems specially adapted for manufacturing
Landscapes
- Engineering & Computer Science (AREA)
- Business, Economics & Management (AREA)
- Human Resources & Organizations (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Strategic Management (AREA)
- Economics (AREA)
- Entrepreneurship & Innovation (AREA)
- General Physics & Mathematics (AREA)
- Life Sciences & Earth Sciences (AREA)
- Biomedical Technology (AREA)
- Quality & Reliability (AREA)
- Tourism & Hospitality (AREA)
- Marketing (AREA)
- General Business, Economics & Management (AREA)
- Educational Administration (AREA)
- Development Economics (AREA)
- Health & Medical Sciences (AREA)
- Game Theory and Decision Science (AREA)
- Artificial Intelligence (AREA)
- Operations Research (AREA)
- Biophysics (AREA)
- Computational Linguistics (AREA)
- Data Mining & Analysis (AREA)
- Evolutionary Computation (AREA)
- General Health & Medical Sciences (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- General Engineering & Computer Science (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Financial Or Insurance-Related Operations Such As Payment And Settlement (AREA)
Abstract
The invention discloses an enterprise credit rating method and device based on a graph convolution neural network and an EM algorithm, wherein the method comprises the following steps: constructing a credit relation network diagram between enterprises; feature extraction is carried out on the credit relation network graph based on the graph convolution neural network model, and an enterprise credit feature matrix is obtained; the enterprise credit feature matrix comprises credit feature vectors of all nodes in the credit relation network diagram; acquiring credit rating of an enterprise, and classifying the credit rating into a plurality of categories; calculating the matching degree of the credit feature vector of each node and each credit rating class by adopting an EM algorithm; a credit rating corresponding to each node is determined based on the degree of matching. The enterprise credit rating method based on the graph convolution neural network and the EM algorithm provided by the invention can accurately credit rating enterprises by utilizing the credit relation among the enterprises; the method realizes automation and intellectualization of enterprise credit rating, and improves rating accuracy and efficiency.
Description
Technical Field
The application relates to the technical field of enterprise wind control, in particular to an enterprise credit rating method and device based on a graph convolution neural network and an EM algorithm.
Background
With the accelerated development of globalization and marketization, transactions and cooperations between enterprises are more and more frequent, and credit risks are brought. Enterprise credit rating is an important means of assessing enterprise credit status, providing important information about enterprise credit risk to enterprise stakeholders such as financial institutions, suppliers, investors, customers, and the like. The credit rating of the enterprise can also better help the enterprise to know the credit condition of the enterprise, so that the credit risk is effectively reduced, and the stable development of the enterprise is ensured.
Currently, most existing enterprise credit rating methods rely on manual intervention, are low in efficiency, and are difficult to deal with large-scale data processing. Traditional rating methods based on statistical analysis ignore complex credit relationships among enterprises and result in inaccurate rating results.
Disclosure of Invention
Based on the above, it is necessary to provide an enterprise credit rating method and device based on a graph convolution neural network and an EM algorithm, aiming at the problems that the existing rating method ignores complex credit relation among enterprises and the rating result is not accurate enough.
In a first aspect, the present application provides a method for enterprise credit rating based on a graph roll-up neural network and an EM algorithm, the method comprising:
s1: constructing a credit relation network diagram between enterprises;
s2: performing feature extraction on the credit relation network graph based on a graph convolution neural network model to obtain an enterprise credit feature matrix; the enterprise credit feature matrix comprises credit feature vectors of all nodes in a credit relation network diagram;
s3: acquiring credit rating of an enterprise, and classifying the credit rating into a plurality of categories;
s4: calculating the matching degree of the credit characteristic vector of each node and each credit rating class by adopting an expected maximization algorithm; and determining the credit rating corresponding to each node based on the matching degree.
Preferably, in S1, constructing a credit relation network graph between enterprises includes:
acquiring enterprise attribute data, transaction relations among enterprises and transaction amounts among enterprises; the business attribute data includes registered capital, business income, net profit, and balance ratio of the business; normalizing the enterprise attribute data and then splicing to obtain an attribute vector of an enterprise; normalizing the transaction amount between enterprises to a range of [0,1 ];
and constructing a credit relation network diagram between enterprises by taking the enterprises as nodes, attribute vectors of the enterprises as attributes of the nodes, transaction relations between the enterprises as directed edges and normalized transaction amounts between the enterprises as weights of the directed edges.
Preferably, in S2, feature extraction is performed on the credit relation network graph based on a graph convolution neural network model to obtain an enterprise credit feature matrix; the enterprise credit feature matrix comprises credit feature vectors of all nodes in a credit relation network diagram;
determining an adjacency matrix and a feature matrix of the credit relation network diagram according to the credit relation network diagram; and taking the adjacency matrix and the feature matrix of the credit relation network diagram as inputs of a diagram convolution neural network model, and outputting the enterprise credit feature matrix; the calculation formula is as follows:
wherein ,represent the firstlEnterprise credit feature matrix of +1 layer; />Representing an activation function; />Representation->A degree matrix of (2); />A sum of the adjacency matrix and the identity matrix representing the credit relation network diagram; />Represent the firstlAn enterprise credit feature matrix of the layer; />Represent the firstlA weight matrix of layers.
Preferably, in S2, the method further includes optimizing the graph roll-up neural network model, and the process includes:
obtaining a true enterprise credit feature label; and calculating a loss function based on the real enterprise credit feature labels and the credit feature vectors, wherein the calculation formula is as follows:
wherein ,L(θ) Representing a loss function;Nrepresenting the total number of nodes of the credit relation network diagram;h i represent the firstiCredit feature vectors for individual nodes;y i represent the firstiMarking the real enterprise credit characteristics of the individual nodes;θparameters representing a graph convolution neural network model;
optimizing parameters of a graph convolution neural network model by adopting a gradient descent back propagation algorithm based on a loss function; in the iterative optimization process, the gradient of the loss function with respect to the parameter is calculated based on the loss function, and the calculation formula is as follows:
wherein ,indicating lossGradient of the function with respect to the parameter; />Representing bias leads;
updating the parameter based on the calculated gradient of the loss function with respect to the parameter; the calculation formula is as follows:
wherein ,representing the updated parameters; />Parameters representing the last iteration;αrepresenting a learning rate;
and (5) ending the optimization until the graph convolution neural network model converges or reaches the maximum iteration number.
Preferably, in S3, initializing a mean vector, a covariance matrix and a mixing coefficient of each credit rating class; the initialized mean vector, covariance matrix and mixing coefficient of each credit rating class are used for matching degree of the credit characteristic vector of each node with each credit rating class.
Preferably, in S4, the process of calculating the matching degree between the credit feature vector of each node and each credit rating class by using the expectation maximization algorithm is as follows:
the expectation maximization algorithm comprises an expectation step and a maximization step;
establishing a matching degree calculation model; calculating posterior probability of each node under each credit rating class in the matching degree calculation model through the expected step; the calculation formula is as follows:
wherein ,γ ik represent the firstiThe individual node is at the firstkPosterior probability under the individual credit rating category;π k represent initialized firstkMixing coefficients for the individual credit rating categories;h i represent the firstiCredit feature vectors for individual nodes;μ k represent initialized firstkA mean vector of the individual credit rating categories; sigma (sigma) k Represent initialized firstkCovariance matrix of individual credit rating class; n%h i |μ k ,∑ k ) Expressed in terms ofμ k Is the mean value of Sigma k Is a multi-element normal distribution of covariance matrixh i Probability density at;Krepresenting a total number of credit rating categories;π j represent initialized firstjMixing coefficients for the individual credit rating categories;h i represent the firstiCredit feature vectors for individual nodes;μ j represent initialized firstjA mean vector of the individual credit rating categories; sigma (sigma) j Represent initialized firstjCovariance matrix of individual credit rating class; n%h i |μ j ,∑ j ) Expressed in terms ofμ j Is the mean value of Sigma j Is a multi-element normal distribution of covariance matrixh i Probability density at;
updating parameters of a matching degree calculation model by using the posterior probability calculated at present in the maximizing step, wherein the parameters of the matching degree calculation model comprise a mean value vector, a covariance matrix and a mixing coefficient; the calculation formula comprises:
wherein ,representing the updated mean vector; />Representing the updated covariance matrix; />Representing the updated mixing coefficients;N k indicating that all nodes belong to the first nodeKProbability of individual credit rating categories;Nrepresenting the total number of nodes;Trepresenting a transpose;
repeating the expected step and the maximizing step until the parameters of the matching degree calculation model converge or reach the maximum iteration times; the final posterior probability is taken as the firstiThe individual nodekMatching degree of the individual credit rating categories.
Preferably, in S4, determining the credit rating corresponding to each node based on the matching degree includes:
for each node, selecting the credit rating class with the highest matching degree as the credit rating of the node; the calculation formula is as follows:
wherein ,a credit rating representing a node; />Represent the firstiThe individual nodekMatching degree of the individual credit rating categories.
Preferably, in S3, the credit rating is divided into a plurality of categories from high to low.
Preferably, the method further comprises risk early warning according to the matching degree of the node and the credit rating class, and the process is as follows:
setting an early warning threshold and a second threshold; a credit rating class below a second threshold is taken as a lower credit rating class, otherwise a high credit rating class;
if the matching degree of the node and the category with the lower credit rating exceeds the early warning threshold, marking the node as a potential inauguration enterprise, otherwise, marking the node as a inauguration enterprise; the calculation formula is as follows:
wherein, risk @ isv) Representing nodesvThe corresponding enterprise is a potential risk enterprise;ue Low Credit represents a Credit rating category belonging to a lower Credit rating categoryu;Represent the firstvThe individual nodeuMatching degree of the individual credit rating categories;τis an early warning threshold.
In a second aspect, the present application provides an enterprise credit rating apparatus based on a graph roll-up neural network and an EM algorithm, the apparatus comprising:
the construction module is used for constructing a credit relation network diagram among enterprises;
the extraction module is used for carrying out feature extraction on the credit relation network graph based on the graph convolution neural network model to obtain an enterprise credit feature matrix; the enterprise credit feature matrix comprises credit feature vectors of all nodes in a credit relation network diagram;
the acquisition and classification module is used for acquiring the credit rating of the enterprise and classifying the credit rating into a plurality of categories;
the calculation module is used for calculating the matching degree of the credit characteristic vector of each node and each credit rating class by adopting an expected maximization algorithm;
and the credit rating determining module is used for determining the credit rating corresponding to each node based on the matching degree.
The beneficial effects are that: the enterprise credit rating method based on the graph convolution neural network and the EM algorithm provided by the invention can accurately credit rating enterprises by utilizing the credit relation among the enterprises; the method realizes automation and intellectualization of enterprise credit rating, and improves rating accuracy and efficiency.
Drawings
In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the drawings that are required in the embodiments or the description of the prior art will be briefly described below, it being obvious that the drawings in the following description are only some embodiments of the present application, and that other drawings may be obtained according to these drawings without inventive effort for a person skilled in the art.
FIG. 1 is a flow chart of an enterprise credit rating method based on graph convolutional neural network and EM algorithm in accordance with an embodiment of the present application;
fig. 2 is a schematic structural diagram of an enterprise credit rating apparatus 200 based on a graph convolutional neural network and an EM algorithm according to an embodiment of the present application.
Detailed Description
In order to make the above objects, features and advantages of the present application more comprehensible, embodiments accompanied with figures are described in detail below. In the following description, numerous specific details are set forth in order to provide a thorough understanding of the present application. This application is, however, susceptible of embodiment in many other forms than those described herein and similar modifications can be made by those skilled in the art without departing from the spirit of the application, and therefore the application is not to be limited to the specific embodiments disclosed below.
Furthermore, the terms "first," "second," and the like, are used for descriptive purposes only and are not to be construed as indicating or implying a relative importance or implicitly indicating the number of technical features indicated. Thus, a feature defining "a first" or "a second" may explicitly or implicitly include at least one such feature. In the description of the present application, the meaning of "plurality" is at least two, such as two, three, etc., unless explicitly defined otherwise.
Referring to fig. 1, which is a flowchart illustrating an enterprise credit rating method based on a graph roll-up neural network and an EM algorithm according to some embodiments of the present application, as shown in fig. 1, the enterprise credit rating method based on the graph roll-up neural network and the EM algorithm may include the following steps:
s1: and constructing a credit relation network diagram among enterprises.
Specifically, constructing a credit relation network graph between enterprises includes:
acquiring enterprise attribute data, transaction relations among enterprises and transaction amounts among enterprises; the business attribute data includes registered capital, business income, net profit, and balance ratio of the business; normalizing the enterprise attribute data and then splicing to obtain an attribute vector of an enterprise; normalizing the transaction amount between enterprises to a range of [0,1 ];
and constructing a credit relation network diagram between enterprises by taking the enterprises as nodes, attribute vectors of the enterprises as attributes of the nodes, transaction relations between the enterprises as directed edges and normalized transaction amounts between the enterprises as weights of the directed edges.
S2: performing feature extraction on the credit relation network graph based on a graph convolution neural network model to obtain an enterprise credit feature matrix; the enterprise credit feature matrix comprises credit feature vectors of all nodes in the credit relation network diagram.
Specifically, feature extraction is carried out on the credit relation network graph based on a graph convolution neural network model, so that an enterprise credit feature matrix is obtained; the enterprise credit feature matrix comprises credit feature vectors of all nodes in a credit relation network diagram;
determining an adjacency matrix and a feature matrix of the credit relation network diagram according to the credit relation network diagram; and taking the adjacency matrix and the feature matrix of the credit relation network diagram as inputs of a diagram convolution neural network model, and outputting the enterprise credit feature matrix; the calculation formula is as follows:
wherein ,represent the firstlEnterprise credit feature matrix of +1 layer; />Representing an activation function; />Representation->A degree matrix of (2); />A sum of the adjacency matrix and the identity matrix representing the credit relation network diagram; />Represent the firstlAn enterprise credit feature matrix of the layer; />Represent the firstlA weight matrix of layers.
In this embodiment, to improve the accuracy of the features extracted from the graph rolling neural network model, the method further includes optimizing the graph rolling neural network model, and the process includes:
obtaining a true enterprise credit feature label; and calculating a loss function based on the real enterprise credit feature labels and the credit feature vectors, wherein the calculation formula is as follows:
wherein ,L(θ) Representing a loss function;Nrepresenting the total number of nodes of the credit relation network diagram;h i represent the firstiCredit feature vectors for individual nodes;y i represent the firstiMarking the real enterprise credit characteristics of the individual nodes;θparameters representing a graph convolution neural network model;
optimizing parameters of a graph convolution neural network model by adopting a gradient descent back propagation algorithm based on a loss function; in the iterative optimization process, the gradient of the loss function with respect to the parameter is calculated based on the loss function, and the calculation formula is as follows:
wherein ,representing the gradient of the loss function with respect to the parameter; />Representing bias leads;
updating the parameter based on the calculated gradient of the loss function with respect to the parameter; the calculation formula is as follows:
wherein ,representing the updated parameters; />Parameters representing the last iteration;αrepresenting a learning rate;
and (5) ending the optimization until the graph convolution neural network model converges or reaches the maximum iteration number.
In this embodiment, the graph convolutional neural network model includes two graph convolutional layers, the dimension of the first graph convolutional layer is 256, the length of the feature vector extracted by the second graph convolutional layer is 64, and the activation function adopted by the second graph convolutional layer is a ReLU activation function.
S3: the credit rating of the business is obtained and the credit rating is divided into a plurality of categories from high to low.
Specifically, the method further comprises initializing a mean vector, a covariance matrix and a mixing coefficient of each credit rating class; the initialized mean vector, covariance matrix and mixing coefficient of each credit rating class are used for matching degree of the credit characteristic vector of each node with each credit rating class.
S4: calculating the matching degree of the credit characteristic vector of each node and each credit rating class by adopting an expected maximization algorithm; and determining the credit rating corresponding to each node based on the matching degree.
Specifically, the process of calculating the matching degree of the credit feature vector of each node and each credit rating class by adopting an expected maximization algorithm is as follows:
the expectation maximization algorithm comprises an expectation step and a maximization step;
establishing a matching degree calculation model; calculating posterior probability of each node under each credit rating class in the matching degree calculation model through the expected step; the calculation formula is as follows:
wherein ,γ ik represent the firstiThe individual node is at the firstkPosterior probability under the individual credit rating category;π k represent initialized firstkMixing coefficients for the individual credit rating categories;h i represent the firstiCredit feature vectors for individual nodes;μ k represent initialized firstkA mean vector of the individual credit rating categories; sigma (sigma) k Represent initialized firstkCovariance matrix of individual credit rating class; n%h i |μ k ,∑ k ) Expressed in terms ofμ k Is the mean value of Sigma k Is a multi-element normal distribution of covariance matrixh i Probability density at;Krepresenting a total number of credit rating categories;π j represent initialized firstjMixing coefficients for the individual credit rating categories;h i represent the firstiCredit feature vectors for individual nodes;μ j represent initialized firstjA mean vector of the individual credit rating categories; sigma (sigma) j Represent initialized firstjPersonal credit rating classIs a covariance matrix of (a); n%h i |μ j ,∑ j ) Expressed in terms ofμ j Is the mean value of Sigma j Is a multi-element normal distribution of covariance matrixh i Probability density at;
updating parameters of a matching degree calculation model by using the posterior probability calculated at present in the maximizing step, wherein the parameters of the matching degree calculation model comprise a mean value vector, a covariance matrix and a mixing coefficient; the calculation formula comprises:
wherein ,representing the updated mean vector; />Representing the updated covariance matrix; />Representing the updated mixing coefficients;N k indicating that all nodes belong to the first nodeKProbability of individual credit rating categories;Nrepresenting the total number of nodes;Trepresenting a transpose;
repeating the expected step and the maximizing step until the parameters of the matching degree calculation model converge or reach the maximum iteration times; the final posterior probability is taken as the firstiThe individual nodekMatching degree of the individual credit rating categories.
Determining a credit rating for each node based on the degree of matching includes:
for each node, selecting the credit rating class with the highest matching degree as the credit rating of the node; the calculation formula is as follows:
wherein ,a credit rating representing a node; />Represent the firstiThe individual nodekMatching degree of the individual credit rating categories.
The enterprise credit rating method provided by the embodiment further comprises risk early warning according to the matching degree of the nodes and the credit rating class, and the process is as follows:
setting an early warning threshold and a second threshold; a credit rating class below a second threshold is taken as a lower credit rating class, otherwise a high credit rating class;
if the matching degree of the node and the category with the lower credit rating exceeds the early warning threshold, marking the node as a potential inauguration enterprise, otherwise, marking the node as a inauguration enterprise; the calculation formula is as follows:
wherein, risk @ isv) Representing nodesvThe corresponding enterprise is a potential risk enterprise;ue Low Credit represents a Credit rating category belonging to a lower Credit rating categoryu;Represent the firstvThe individual nodeuMatching degree of the individual credit rating categories;τis an early warning threshold.
The enterprise credit rating method based on the graph convolution neural network and the EM algorithm provided by the embodiment of the application can accurately credit rating the enterprise by utilizing the credit relation among enterprises; the method realizes automation and intellectualization of enterprise credit rating, and improves rating accuracy and efficiency.
In the above embodiment, an enterprise credit rating method based on a graph roll-up neural network and an EM algorithm is provided, and correspondingly, the application also provides an enterprise credit rating device based on the graph roll-up neural network and the EM algorithm. The enterprise credit rating device based on the graph roll-up neural network and the EM algorithm provided by the embodiment of the present application may implement the enterprise credit rating method based on the graph roll-up neural network and the EM algorithm, and the enterprise credit rating device based on the graph roll-up neural network and the EM algorithm may be implemented by software, hardware or a combination of software and hardware. For example, the graph roll-up neural network and EM algorithm based enterprise credit rating apparatus may include integrated or separate functional modules or units to perform the corresponding steps in the methods described above.
Referring to fig. 2, a schematic diagram of an enterprise credit rating apparatus based on a graph convolution neural network and an EM algorithm according to some embodiments of the present application is shown. Since the apparatus embodiments are substantially similar to the method embodiments, the description is relatively simple, and reference is made to the description of the method embodiments for relevant points. The device embodiments described below are merely illustrative.
As shown in fig. 2, an enterprise credit rating apparatus 200 based on a graph roll-up neural network and an EM algorithm may include:
a construction module 201, configured to construct a credit relation network diagram between enterprises;
the extracting module 202 is configured to perform feature extraction on the credit relation network graph based on a graph convolution neural network model to obtain an enterprise credit feature matrix; the enterprise credit feature matrix comprises credit feature vectors of all nodes in a credit relation network diagram;
the acquiring and classifying module 203 is configured to acquire a credit rating of an enterprise, and classify the credit rating into a plurality of categories;
a calculating module 204, configured to calculate, using an expectation maximization algorithm, a matching degree between the credit feature vector of each node and each credit rating class;
the credit rating determining module 205 is configured to determine a credit rating corresponding to each node based on the matching degree.
In some implementations of the embodiments of the present application, the enterprise credit rating device 200 based on the graph roll-up neural network and the EM algorithm provided by the embodiments of the present application has the same beneficial effects as the enterprise credit rating device method based on the graph roll-up neural network and the EM algorithm provided by the foregoing embodiments of the present application due to the same inventive concept.
The technical features of the above-described embodiments may be arbitrarily combined, and all possible combinations of the technical features in the above-described embodiments are not described for brevity of description, however, as long as there is no contradiction between the combinations of the technical features, they should be considered as the scope of the description.
The above examples only represent a few embodiments of the present application, which are described in more detail and are not to be construed as limiting the scope of the claims. It should be noted that it would be apparent to those skilled in the art that various modifications and improvements could be made without departing from the spirit of the present application, which would be within the scope of the present application. Accordingly, the scope of protection of the present application is to be determined by the claims appended hereto.
Claims (10)
1. An enterprise credit rating method based on a graph convolution neural network and an EM algorithm, the method comprising:
s1: constructing a credit relation network diagram between enterprises;
s2: performing feature extraction on the credit relation network graph based on a graph convolution neural network model to obtain an enterprise credit feature matrix; the enterprise credit feature matrix comprises credit feature vectors of all nodes in a credit relation network diagram;
s3: acquiring credit rating of an enterprise, and classifying the credit rating into a plurality of categories;
s4: calculating the matching degree of the credit feature vector of each node and each credit rating class by adopting an EM algorithm; and determining the credit rating corresponding to each node based on the matching degree.
2. The method for rating credit of enterprises according to claim 1, wherein in S1, constructing a credit relation network diagram between enterprises comprises:
acquiring enterprise attribute data, transaction relations among enterprises and transaction amounts among enterprises; the business attribute data includes registered capital, business income, net profit, and balance ratio of the business; normalizing the enterprise attribute data and then splicing to obtain an attribute vector of an enterprise; normalizing the transaction amount between enterprises to a range of [0,1 ];
and constructing a credit relation network diagram between enterprises by taking the enterprises as nodes, attribute vectors of the enterprises as attributes of the nodes, transaction relations between the enterprises as directed edges and normalized transaction amounts between the enterprises as weights of the directed edges.
3. The enterprise credit rating method according to claim 1, wherein in S2, feature extraction is performed on the credit relation network graph based on a graph convolution neural network model to obtain an enterprise credit feature matrix; the enterprise credit feature matrix comprises credit feature vectors of all nodes in a credit relation network diagram;
determining an adjacency matrix and a feature matrix of the credit relation network diagram according to the credit relation network diagram; and taking the adjacency matrix and the feature matrix of the credit relation network diagram as inputs of a diagram convolution neural network model, and outputting the enterprise credit feature matrix; the calculation formula is as follows:
wherein ,represent the firstlEnterprise credit feature matrix of +1 layer; />Representing an activation function; />Representation->A degree matrix of (2); />A sum of the adjacency matrix and the identity matrix representing the credit relation network diagram; />Represent the firstlAn enterprise credit feature matrix of the layer; />Represent the firstlA weight matrix of layers.
4. The enterprise credit rating method of claim 3, further comprising optimizing the graph roll-up neural network model in S2, the process comprising:
obtaining a true enterprise credit feature label; and calculating a loss function based on the real enterprise credit feature labels and the credit feature vectors, wherein the calculation formula is as follows:
wherein ,L(θ) Representing a loss function;Nrepresenting the total number of nodes of the credit relation network diagram;h i represent the firstiCredit feature vectors for individual nodes;y i represent the firstiMarking the real enterprise credit characteristics of the individual nodes;θparameters representing a graph convolution neural network model;
optimizing parameters of a graph convolution neural network model by adopting a gradient descent back propagation algorithm based on a loss function; in the iterative optimization process, the gradient of the loss function with respect to the parameter is calculated based on the loss function, and the calculation formula is as follows:
wherein ,representing the gradient of the loss function with respect to the parameter; />Representing bias leads; />
Updating the parameter based on the calculated gradient of the loss function with respect to the parameter; the calculation formula is as follows:
wherein ,representing the updated parameters; />Parameters representing the last iteration;αrepresenting a learning rate;
and (5) ending the optimization until the graph convolution neural network model converges or reaches the maximum iteration number.
5. The enterprise credit rating method of claim 1, wherein S3 further comprises initializing a mean vector, a covariance matrix, and a mixing coefficient for each credit rating class; the initialized mean vector, covariance matrix and mixing coefficient of each credit rating class are used for matching degree of the credit characteristic vector of each node with each credit rating class.
6. The enterprise credit rating method according to claim 5, wherein in S4, the process of calculating the matching degree of the credit feature vector of each node and each credit rating class by using the EM algorithm is as follows:
the EM algorithm includes a desired step and a maximizing step;
establishing a matching degree calculation model; calculating posterior probability of each node under each credit rating class in the matching degree calculation model through the expected step; the calculation formula is as follows:
wherein ,γ ik represent the firstiThe individual node is at the firstkPosterior probability under the individual credit rating category;π k represent initialized firstkMixing coefficients for the individual credit rating categories;h i represent the firstiCredit feature vectors for individual nodes;μ k represent initialized firstkA mean vector of the individual credit rating categories; sigma (sigma) k Represent initialized firstkCovariance matrix of individual credit rating class; n%h i |μ k ,∑ k ) Expressed in terms ofμ k Is the mean value of Sigma k Is a multi-element normal distribution of covariance matrixh i Probability density at;Krepresenting a total number of credit rating categories;π j represent initialized firstjMixing coefficients for the individual credit rating categories;μ j represent initialized firstjA mean vector of the individual credit rating categories; sigma (sigma) j Represent initialized firstjCovariance matrix of individual credit rating class; n%h i |μ j ,∑ j ) Expressed in terms ofμ j Is the mean value of Sigma j Is a multi-element normal distribution of covariance matrixh i Probability density at;
updating parameters of a matching degree calculation model by using the posterior probability calculated at present in the maximizing step, wherein the parameters of the matching degree calculation model comprise a mean value vector, a covariance matrix and a mixing coefficient; the calculation formula comprises:
wherein ,representing the updated mean vector; />Representing the updated covariance matrix; />Representing the updated mixing coefficients;N k indicating that all nodes belong to the first nodeKProbability of individual credit rating categories;Nrepresenting the total number of nodes;Trepresenting a transpose;
repeating the expected step and the maximizing step until the parameters of the matching degree calculation model converge or reach the maximum iteration times; the final posterior probability is taken as the firstiThe individual nodekMatching degree of the individual credit rating categories.
7. The enterprise credit rating method of claim 6, wherein determining a credit rating for each node based on the degree of matching in S4 comprises:
for each node, selecting the credit rating class with the highest matching degree as the credit rating of the node; the calculation formula is as follows:
8. The enterprise credit rating method of claim 7, wherein in S3, the credit ratings are divided into categories from high to low.
9. The method of claim 8, further comprising performing risk early warning according to a degree of matching between the node and the credit rating class, wherein the steps of:
setting an early warning threshold and a second threshold; a credit rating class below a second threshold is taken as a lower credit rating class, otherwise a high credit rating class;
if the matching degree of the node and the category with the lower credit rating exceeds the early warning threshold, marking the node as a potential inauguration enterprise, otherwise, marking the node as a inauguration enterprise; the calculation formula is as follows:
wherein, risk @ isv) Representing nodesvThe corresponding enterprise is a potential risk enterprise;ue Low Credit represents a letter belonging to a lower Credit rating categoryBy rating classu;Represent the firstvThe individual nodeuMatching degree of the individual credit rating categories;τis an early warning threshold.
10. An enterprise credit rating apparatus based on a graph roll-up neural network and an EM algorithm, the apparatus comprising:
the construction module is used for constructing a credit relation network diagram among enterprises;
the extraction module is used for carrying out feature extraction on the credit relation network graph based on the graph convolution neural network model to obtain an enterprise credit feature matrix; the enterprise credit feature matrix comprises credit feature vectors of all nodes in a credit relation network diagram;
the acquisition and classification module is used for acquiring the credit rating of the enterprise and classifying the credit rating into a plurality of categories;
the calculation module is used for calculating the matching degree of the credit characteristic vector of each node and each credit rating class by adopting an EM algorithm;
and the credit rating determining module is used for determining the credit rating corresponding to each node based on the matching degree.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202310490149.1A CN116227939A (en) | 2023-05-04 | 2023-05-04 | Enterprise credit rating method and device based on graph convolution neural network and EM algorithm |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202310490149.1A CN116227939A (en) | 2023-05-04 | 2023-05-04 | Enterprise credit rating method and device based on graph convolution neural network and EM algorithm |
Publications (1)
Publication Number | Publication Date |
---|---|
CN116227939A true CN116227939A (en) | 2023-06-06 |
Family
ID=86579052
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN202310490149.1A Pending CN116227939A (en) | 2023-05-04 | 2023-05-04 | Enterprise credit rating method and device based on graph convolution neural network and EM algorithm |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN116227939A (en) |
Cited By (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN116452007A (en) * | 2023-06-15 | 2023-07-18 | 深圳市迪博企业风险管理技术有限公司 | Enterprise tax compliance risk assessment method based on capsule network |
Citations (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN105160598A (en) * | 2015-08-28 | 2015-12-16 | 国网智能电网研究院 | Power grid service classification method based on improved EM algorithm |
CN111046664A (en) * | 2019-11-26 | 2020-04-21 | 哈尔滨工业大学(深圳) | False news detection method and system based on multi-granularity graph convolution neural network |
CN111859166A (en) * | 2020-07-28 | 2020-10-30 | 重庆邮电大学 | Article scoring prediction method based on improved graph convolution neural network |
CN112686498A (en) * | 2020-12-11 | 2021-04-20 | 天津中科智能识别产业技术研究院有限公司 | Enterprise credit rating method based on deep convolutional network |
CN113362158A (en) * | 2021-05-31 | 2021-09-07 | 中国银联股份有限公司 | Credit evaluation method, device and computer readable storage medium |
CN113849580A (en) * | 2021-09-28 | 2021-12-28 | 天弘基金管理有限公司 | Subject rating prediction method and device, electronic equipment and storage medium |
-
2023
- 2023-05-04 CN CN202310490149.1A patent/CN116227939A/en active Pending
Patent Citations (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN105160598A (en) * | 2015-08-28 | 2015-12-16 | 国网智能电网研究院 | Power grid service classification method based on improved EM algorithm |
CN111046664A (en) * | 2019-11-26 | 2020-04-21 | 哈尔滨工业大学(深圳) | False news detection method and system based on multi-granularity graph convolution neural network |
CN111859166A (en) * | 2020-07-28 | 2020-10-30 | 重庆邮电大学 | Article scoring prediction method based on improved graph convolution neural network |
CN112686498A (en) * | 2020-12-11 | 2021-04-20 | 天津中科智能识别产业技术研究院有限公司 | Enterprise credit rating method based on deep convolutional network |
CN113362158A (en) * | 2021-05-31 | 2021-09-07 | 中国银联股份有限公司 | Credit evaluation method, device and computer readable storage medium |
CN113849580A (en) * | 2021-09-28 | 2021-12-28 | 天弘基金管理有限公司 | Subject rating prediction method and device, electronic equipment and storage medium |
Non-Patent Citations (2)
Title |
---|
SHENGMING ZHANG ET AL: ""Scalable Heterogeneous Graph Neural Networks for Predicting High-potential Early-stage Startups"", 《KDD \'21:PROCEEDINGS OF THE 27TH ACM SIGKDD CONFERENCE ON KNOWLEDGE DISCOVER & DATA MINING》, pages 2202 - 2211 * |
周文安: "《人工智能视域下机器学习在教育研究中的应用》", 北京邮电大学出版社, pages: 112 - 113 * |
Cited By (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN116452007A (en) * | 2023-06-15 | 2023-07-18 | 深圳市迪博企业风险管理技术有限公司 | Enterprise tax compliance risk assessment method based on capsule network |
CN116452007B (en) * | 2023-06-15 | 2023-09-19 | 深圳市迪博企业风险管理技术有限公司 | Enterprise tax compliance risk assessment method based on capsule network |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN109492026B (en) | Telecommunication fraud classification detection method based on improved active learning technology | |
CN112102073A (en) | Credit risk control method and system, electronic device and readable storage medium | |
CN109740642A (en) | Invoice category recognition methods, device, electronic equipment and readable storage medium storing program for executing | |
US20230342606A1 (en) | Training method and apparatus for graph neural network | |
CN114663002A (en) | Method and equipment for automatically matching performance assessment indexes | |
CN110866832A (en) | Risk control method, system, storage medium and computing device | |
CN116227939A (en) | Enterprise credit rating method and device based on graph convolution neural network and EM algorithm | |
CN113674087A (en) | Enterprise credit rating method, apparatus, electronic device and medium | |
CN112836750A (en) | System resource allocation method, device and equipment | |
CN110348516B (en) | Data processing method, data processing device, storage medium and electronic equipment | |
CN116227940B (en) | Enterprise fund flow anomaly detection method based on fund flow diagram | |
CN116542763A (en) | Internet financial credit default prediction method based on big data | |
CN107402984B (en) | A kind of classification method and device based on theme | |
CN116611911A (en) | Credit risk prediction method and device based on support vector machine | |
CN113177733B (en) | Middle and small micro enterprise data modeling method and system based on convolutional neural network | |
CN116010875A (en) | Method and device for classifying ammeter faults, electronic equipment and computer storage medium | |
CN115439079A (en) | Item classification method and device | |
CN115146904A (en) | Artificial intelligence-based security asset risk monitoring system | |
CN110570301B (en) | Risk identification method, device, equipment and medium | |
IMBALANCE | Ensemble Adaboost in classification and regression trees to overcome class imbalance in credit status of bank customers | |
CN113191771A (en) | Buyer account period risk prediction method | |
Nurhidayat et al. | Analysis and Classification of Customer Churn Using Machine Learning Models | |
Zimal et al. | Customer churn prediction using machine learning | |
CN117291740B (en) | Receivables data authenticity intelligent identification auditing system based on big data | |
CN110119453A (en) | A kind of local chronicle document method for visualizing and system based on WebGIS |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
PB01 | Publication | ||
PB01 | Publication | ||
SE01 | Entry into force of request for substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
RJ01 | Rejection of invention patent application after publication | ||
RJ01 | Rejection of invention patent application after publication |
Application publication date: 20230606 |