CN114970926A - Model training method, enterprise operation risk prediction method and device - Google Patents
Model training method, enterprise operation risk prediction method and device Download PDFInfo
- Publication number
- CN114970926A CN114970926A CN202110218834.XA CN202110218834A CN114970926A CN 114970926 A CN114970926 A CN 114970926A CN 202110218834 A CN202110218834 A CN 202110218834A CN 114970926 A CN114970926 A CN 114970926A
- Authority
- CN
- China
- Prior art keywords
- enterprise
- model
- enterprises
- data
- training
- 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
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
- G06Q10/04—Forecasting or optimisation specially adapted for administrative or management purposes, e.g. linear programming or "cutting stock problem"
-
- 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
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/045—Combinations of networks
-
- 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
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/047—Probabilistic or stochastic networks
-
- 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
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/048—Activation functions
-
- 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
-
- 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
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Business, Economics & Management (AREA)
- General Physics & Mathematics (AREA)
- Human Resources & Organizations (AREA)
- Software Systems (AREA)
- Life Sciences & Earth Sciences (AREA)
- Data Mining & Analysis (AREA)
- Evolutionary Computation (AREA)
- General Health & Medical Sciences (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- General Engineering & Computer Science (AREA)
- Biophysics (AREA)
- Mathematical Physics (AREA)
- Biomedical Technology (AREA)
- Artificial Intelligence (AREA)
- Strategic Management (AREA)
- Computational Linguistics (AREA)
- Health & Medical Sciences (AREA)
- Economics (AREA)
- Entrepreneurship & Innovation (AREA)
- Game Theory and Decision Science (AREA)
- Development Economics (AREA)
- Marketing (AREA)
- Operations Research (AREA)
- Quality & Reliability (AREA)
- Tourism & Hospitality (AREA)
- General Business, Economics & Management (AREA)
- Probability & Statistics with Applications (AREA)
- Educational Administration (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The invention relates to the technical field of computer data processing, in particular to a model training method, an enterprise operation risk prediction method and an enterprise operation risk prediction device. The method comprises the following steps: constructing a training set; constructing an initial prediction model; the initial prediction model comprises a graph convolution network layer and a full-connection network layer; the graph convolution network layer comprises a characteristic input end, an adjacent matrix input end and at least one graph convolution network; the graph convolution network comprises a first graph convolution layer and a second graph convolution layer; and iteratively training the initial prediction model according to the training set to obtain a target prediction model for enterprise operation risk prediction. The invention constructs a training set by utilizing enterprise operation data and operation relationship data of enterprises, so that the model can learn the relationship between the enterprise operation risk result and the enterprise operation data and the operation relationship data when the initial prediction model is trained, thereby realizing accurate and reliable enterprise operation risk prediction.
Description
Technical Field
The invention relates to the technical field of computer data processing, in particular to a model training method, an enterprise operation risk prediction method and an enterprise operation risk prediction device.
Background
The finance industry is concerned with the life pulse of national economic development, and the long-term stability of the finance system is always concerned by the nation. The enterprise operation risk prediction is an important research subject in the field of financial wind control, and each large bank analyzes the operation condition of small and micro enterprises to control the risk of loan release, so that the deep research of the enterprise operation risk prediction has important significance for improving the stability of the financial industry.
Therefore, how to accurately and reliably predict the enterprise operation risk is a technical problem which needs to be solved urgently at present.
Disclosure of Invention
The invention aims to provide a model training method, an enterprise operation risk prediction method and an enterprise operation risk prediction device so as to realize accurate and reliable enterprise operation risk prediction.
In order to achieve the above object, the embodiments of the present invention provide the following solutions:
in a first aspect, an embodiment of the present invention provides a model training method for enterprise operation risk prediction, where the method includes:
constructing a training set; the training set comprises enterprise business data of at least two enterprises and business relation data of the at least two enterprises;
constructing an initial prediction model; wherein the initial prediction model comprises a graph convolution network layer and a fully connected network layer; the graph volume network layer comprises a characteristic input end used for inputting the enterprise business data, an adjacent matrix input end used for inputting the business relationship data and at least one graph volume network; the graph convolution network comprises a first graph convolution layer and a second graph convolution layer;
and iteratively training the initial prediction model according to the training set to obtain a target prediction model for enterprise operation risk prediction.
In one possible embodiment, the constructing the training set includes:
removing non-numerical characteristics in the original enterprise operation data of the at least two enterprises to obtain numerical enterprise operation data of the at least two enterprises; the original enterprise operation data comprises one or more of enterprise operation age, enterprise registered capital, enterprise related litigation information and enterprise recruitment information;
unifying the dimension and the characteristic vacancy value of the numerical enterprise operation data to obtain enterprise operation data of the at least two enterprises;
constructing an adjacency matrix according to the enterprise business relationship information between the at least two enterprises, and acquiring business relationship data of the at least two enterprises; wherein the enterprise business relationship information comprises one or more of investment relationship information, project participation relationship information and infringement litigation relationship information between the at least two enterprises.
In one possible embodiment, the constructing the initial prediction model includes:
constructing an activation function of a first graph convolution layer of the graph convolution network by using a linear rectification function ReLU;
constructing an activation function of a second graph convolution layer of the graph convolution network by utilizing a Softmax function;
and constructing a Loss function of the initial prediction model by utilizing a Focal local function.
In a possible embodiment, the iteratively training the initial prediction model according to the training set to obtain a target prediction model for enterprise operation risk prediction includes:
training the initial prediction model by using the training set, obtaining a plurality of prediction models, and constructing a first model set;
obtaining a verification result of each model in the first model set by using a verification set;
screening one or more prediction models according to the verification result of each model to construct a second model set;
testing the models in the second model set by using a test set to obtain the change rate of the loss function value of each model in the second model set;
and taking the model with the second model centralized loss function value change rate smaller than a set threshold value as the target prediction model.
In a second aspect, an embodiment of the present invention provides a model training apparatus for enterprise operation risk prediction, where the apparatus includes:
the first construction module is used for constructing a training set; the training set comprises enterprise business data of at least two enterprises and business relation data of the at least two enterprises;
the second construction model is used for constructing the initial prediction model; wherein the initial prediction model comprises a graph convolution network layer and a fully connected network layer; the graph volume network layer comprises a characteristic input end used for inputting the enterprise business data, an adjacent matrix input end used for inputting the business relationship data and at least one graph volume network; the graph convolution network includes a first graph convolution layer and a second graph convolution layer;
and the first acquisition module is used for iteratively training the initial prediction model according to the training set to acquire a target prediction model for enterprise operation risk prediction.
In a possible embodiment, the first building block comprises:
the first processing module is used for eliminating non-numerical characteristics in the original enterprise operation data of the at least two enterprises and acquiring numerical enterprise operation data of the at least two enterprises; wherein the original enterprise business data comprises one or more of enterprise business age, enterprise registered capital, enterprise related litigation information, and enterprise recruitment information;
the second processing module is used for unifying the dimension and the characteristic vacancy value of the numerical enterprise operation data and acquiring enterprise operation data of at least two enterprises;
the third construction module is used for constructing an adjacency matrix according to the enterprise business relationship information between the at least two enterprises and acquiring the business relationship data of the at least two enterprises; wherein the enterprise business relationship information comprises one or more of investment relationship information, project participation relationship information and infringement litigation relationship information between the at least two enterprises.
In a possible embodiment, the second building block comprises:
a fourth construction module, configured to construct an activation function of a first graph convolution layer of the graph convolution network using a linear rectification function ReLU;
a fifth constructing module, configured to construct, using a Softmax function, an activation function of a second graph convolution layer of the graph convolution network;
and the sixth construction module is used for constructing a Loss function of the initial prediction model by utilizing a Focal local function.
In a possible embodiment, the first obtaining module includes:
a seventh construction module, configured to train the initial prediction model by using the training set, obtain multiple prediction models, and construct a first model set;
the second obtaining module is used for obtaining the verification result of each model in the first model set by using the verification set;
the eighth construction module is used for screening one or more prediction models according to the verification result of each model and constructing a second model set;
a third obtaining module, configured to test the models in the second model set by using a test set, and obtain a change rate of a loss function value of each model in the second model set;
and the fourth acquisition module is used for taking the model with the second model centralized loss function value change rate smaller than a set threshold value as the target prediction model.
In a third aspect, an embodiment of the present invention provides an enterprise operation risk prediction method, including:
acquiring enterprise operation data of at least two enterprises to be forecasted and operation relation data of the at least two enterprises to be forecasted;
and acquiring enterprise operation risk prediction results of the at least two enterprises to be predicted by using the target prediction model obtained by the model training method in the first aspect.
In a fourth aspect, an embodiment of the present invention provides an enterprise operation risk prediction apparatus, including:
the fifth acquisition module is used for acquiring enterprise operation data of at least two enterprises to be predicted and operation relation data of the at least two enterprises to be detected;
a sixth obtaining module, configured to obtain enterprise operation risk prediction results of the at least two to-be-predicted enterprises by using the target prediction model obtained by using any one of the model training methods in the first aspect.
In a fifth aspect, an embodiment of the present invention provides an electronic device, including:
a memory for storing a computer program;
a processor for executing the computer program to implement the steps of the method of any of the first or third aspects.
In a sixth aspect, an embodiment of the present invention provides a computer-readable storage medium, on which a computer program is stored, the computer program being executed by a processor to implement the steps of the method of any one of the first aspect or the third aspect.
Compared with the prior art, the invention has the following advantages and beneficial effects:
the invention constructs a training set by utilizing enterprise operation data and operation relationship data of enterprises, so that when an initial prediction model is trained, the model can learn the relationship between the enterprise operation risk result and the enterprise operation data and the operation relationship data, not only can capture the complex relationship between the enterprises, but also can improve the effect of enterprise operation risk prediction through Bagging thought, thereby realizing accurate and reliable enterprise operation risk prediction.
Drawings
In order to more clearly illustrate the embodiments of the present specification or the technical solutions in the prior art, the drawings needed to be used in the embodiments will be briefly described below, and it is obvious that the drawings in the following description are only some embodiments of the present specification, and it is obvious for those skilled in the art that other drawings can be obtained according to the drawings without creative efforts.
FIG. 1 is a flowchart of a model training method for enterprise business risk prediction according to an embodiment of the present invention;
FIG. 2 is a schematic structural diagram of an initial prediction model according to an embodiment of the present invention;
FIG. 3 is a schematic structural diagram of a model training apparatus for enterprise operation risk prediction according to an embodiment of the present invention;
FIG. 4 is a flowchart of a method for predicting risk of enterprise operation according to an embodiment of the present invention;
fig. 5 is a schematic structural diagram of an enterprise operation risk prediction apparatus according to an embodiment of the present invention.
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 a part of the embodiments of the present invention, rather than all embodiments, and all other embodiments obtained by those skilled in the art based on the embodiments of the present invention belong to the scope of protection of the embodiments of the present invention.
The inventor of the invention discovers, after analysis and research, that the existing enterprise operation risk prediction scheme based on the feedforward neural network can be used for shallow-level representation of the operation data of enterprises, but cannot depict complex associated information among the enterprises, and a large amount of data characteristic information is lost, so that the prediction result is not accurate enough. The existing enterprise operation risk prediction scheme based on the random forest only carries out modeling analysis on operation information of a single enterprise, ignores the incidence relation among different enterprises, and tends to select characteristics with more values when a decision tree of the random forest is split, so that the overall model of the random forest is influenced.
In order to realize accurate and reliable enterprise business risk prediction, the invention hopes to introduce the incidence relation between enterprises into model training, so the following scheme is provided.
Referring to fig. 1, fig. 1 is a flowchart of a model training method for enterprise operation risk prediction according to an embodiment of the present invention, including steps 11 to 13.
And 11, constructing a training set.
The training set comprises enterprise business data of at least two enterprises and business relation data of the at least two enterprises.
Specifically, the enterprise operation data belongs to the model input feature quantity, and may include one or more of enterprise operation age, enterprise registered capital, enterprise related litigation information, and enterprise recruitment information of the corresponding enterprise.
The business relationship data may be an association relationship between enterprises, such as investment relationship information, project participation relationship information, and infringement litigation between enterprises.
Here, the embodiment of the present invention further provides a specific construction scheme of the training set, which includes steps 21 to 23.
And 21, removing non-numerical characteristics in the original enterprise operation data of the at least two enterprises to obtain numerical enterprise operation data of the at least two enterprises.
Wherein the original enterprise business data comprises one or more of enterprise business age, enterprise registered capital, enterprise-related litigation information, and enterprise recruitment information.
Specifically, in this embodiment, the enterprise operation data is numerical data, which facilitates model training. Of course, before this step, the original enterprise operation data may be preprocessed to convert some of the data into a numerical type, for example, by using a certain numerical generation. For specific business personnel
And step 22, unifying the dimension and the characteristic vacancy value of the numerical enterprise operation data to obtain the enterprise operation data of the at least two enterprises.
Specifically, the numerical enterprise operation data may be normalized to unify the dimensions of the data, and the characteristic vacancy values are required to be used to process the numerical enterprise operation data at the vacancy positions, so that the enterprise operation data corresponding to each enterprise has a unified format.
And step 23, constructing an adjacency matrix according to the enterprise business relationship information between the at least two enterprises, and acquiring the business relationship data of the at least two enterprises.
Wherein the enterprise business relationship information comprises one or more of investment relationship information, project participation relationship information and infringement litigation relationship information between the at least two enterprises.
Specifically, the step adopts an adjacency matrix form to display the business operation relationship between two or more enterprises, so that the model training is convenient.
And step 12, constructing an initial prediction model.
Wherein the initial prediction model comprises a graph convolution network layer and a fully connected network layer; the graph volume network layer comprises a characteristic input end used for inputting the enterprise business data, an adjacency matrix input end used for inputting the business relationship data and at least one graph volume network; the graph convolution network includes a first graph convolution layer and a second graph convolution layer.
Specifically, the number of the graph convolution networks in the graph convolution network layer can be two or more, the effect of the graph neural network is enhanced by adopting an ensemble learning method, a plurality of graph neural networks are constructed based on Bagging thought to train data respectively, and finally, a final classification result is selected by a plurality of graph model voting methods.
Specifically, as shown in fig. 2, a schematic structural diagram of an initial prediction model provided by an embodiment of the present invention is provided, where a graph convolution network layer includes 5 graph convolution networks, each graph convolution network includes two graph convolution layers, and graph convolution operations may aggregate features of neighboring nodes to update feature representation of a central node, so as to introduce information of a neighborhood of the central node into the central node.
The 5 graph convolution networks in the graph convolution network layer are arranged in parallel, and the characteristic input end (characteristic X) and the adjacent matrix input end (A) are respectively connected with the corresponding input ends of the 5 graph convolution networks.
The full-connection network layer comprises two full-connection layers (Dense), and the graph convolution network layer is connected with the full-connection network layer in a splicing mode.
Here, the embodiment of the present invention further provides a specific initial prediction model construction scheme, which includes steps 31 to 33.
Step 31, constructing an activation function of a first map convolution layer of the map convolution network using a linear rectification function ReLU.
Specifically, the step enhances the expressive power of the graph network by mapping the activation function nonlinearities.
And step 32, constructing an activation function of a second graph convolution layer of the graph convolution network by utilizing a Softmax function.
Specifically, the step enhances the expressive power of the graph network by mapping the activation function nonlinearities.
And step 33, constructing a Loss function of the initial prediction model by utilizing a Focal local function.
Specifically, the enterprise operation risk prediction data set is generally unbalanced, namely the number of dead enterprises is far smaller than that of living enterprises, the punishment of common cross entropy Loss functions on errors of positive and negative samples is the same, and the method is not suitable for data with greatly different numbers of positive and negative samples. In addition, the penalty degree proportion of errors of positive and negative samples can be adjusted by adjusting the modulation coefficient of Focal local.
And step 13, iteratively training the initial prediction model according to the training set to obtain a target prediction model for enterprise operation risk prediction.
Here, a preferred model training scheme is provided, which specifically includes steps 41 to 45.
And 41, training the initial prediction model by using the training set, obtaining a plurality of prediction models, and constructing a first model set.
Specifically, the training set not only includes the relevant input features of the enterprise, but also corresponds to the final risk result of the enterprise, and the risk result may be simplified to survival or death, may also be a survival probability or a death probability, and may also be other enterprise risks such as a probability of loss occurrence, and the like.
In the step, a training set is input into an initial prediction model, and different punishment degrees are given to the model by a loss function according to the difference between a prediction result and an actual result, so that the parameters of the model are adjusted, and the iterative training of the model is realized.
And 42, acquiring a verification result of each model in the first model set by using the verification set.
Specifically, the data type and the training set in the validation set are the same, but the specific data and the training set in the validation set are different.
The validation result for each model in the first set of models may be an overall prediction accuracy for that model.
And 43, screening one or more prediction models according to the verification result of each model, and constructing a second model set.
Specifically, the second model set is used for screening out a final target prediction model.
And 44, testing the models in the second model set by using the test set to obtain the change rate of the loss function value of each model in the second model set.
And step 45, taking the model with the loss function value change rate in the second model set smaller than a set threshold value as the target prediction model.
Specifically, the closer the change rate of the loss function value approaches 0, the better the result of the iterative training is, and a suitable model can be selected as the target prediction model according to the characteristic.
Compared with a single shallow model based on a feedforward neural network or a model based on a random forest, the method based on the integrated graph neural network can capture the complex relation among enterprises and improve the prediction effect of enterprise operation risk through Bagging thought, in addition, the Loss function of the graph neural network is improved, the penalty strength for few sample errors is increased by adopting Focal Loss, and the prediction effect is further improved.
The embodiment belongs to an analysis processing scheme aiming at computer data, and is particularly applied to prediction of enterprise operation risks. In this embodiment, the integrated graph neural network is trained by using two types of data related to enterprise operation (enterprise operation data and operation relationship data), and enterprise operation risk prediction is performed by using the trained integrated graph neural network, which essentially belongs to processing and analyzing data.
Based on the same inventive concept as the method, an embodiment of the present invention further provides a model training apparatus for enterprise operation risk prediction, as shown in fig. 3, which is a schematic structural diagram of the embodiment of the apparatus, and the apparatus includes:
a first construction module 51 for constructing a training set; the training set comprises enterprise business data of at least two enterprises and business relation data of the at least two enterprises;
a second construction model 52 for constructing an initial prediction model; wherein the initial prediction model comprises a graph convolution network layer and a fully connected network layer; the graph volume network layer comprises a characteristic input end used for inputting the enterprise business data, an adjacent matrix input end used for inputting the business relationship data and at least one graph volume network; the graph convolution network includes a first graph convolution layer and a second graph convolution layer;
and a first obtaining module 53, configured to iteratively train the initial prediction model according to the training set, so as to obtain a target prediction model for enterprise operation risk prediction.
In a possible embodiment, the first building block comprises:
the first processing module is used for eliminating non-numerical characteristics in the original enterprise operation data of the at least two enterprises and acquiring numerical enterprise operation data of the at least two enterprises; wherein the original enterprise business data comprises one or more of enterprise business age, enterprise registered capital, enterprise related litigation information, and enterprise recruitment information;
the second processing module is used for unifying the dimension and the characteristic vacancy value of the numerical enterprise operation data and acquiring enterprise operation data of at least two enterprises;
the third construction module is used for constructing an adjacency matrix according to the enterprise business relationship information between the at least two enterprises and acquiring the business relationship data of the at least two enterprises; wherein the enterprise business relationship information comprises one or more of investment relationship information, project participation relationship information and infringement litigation relationship information between the at least two enterprises.
In a possible embodiment, the second building block comprises:
a fourth construction module, configured to construct an activation function of a first graph convolution layer of the graph convolution network using a linear rectification function ReLU;
a fifth constructing module, configured to construct, using a Softmax function, an activation function of a second graph convolution layer of the graph convolution network;
and the sixth construction module is used for constructing a Loss function of the initial prediction model by utilizing a Focal local function.
In a possible embodiment, the first obtaining module includes:
a seventh construction module, configured to train the initial prediction model using the training set, obtain multiple prediction models, and construct a first model set;
the second obtaining module is used for obtaining the verification result of each model in the first model set by using the verification set;
the eighth construction module is used for screening one or more prediction models according to the verification result of each model to construct a second model set;
a third obtaining module, configured to test the models in the second model set by using a test set, and obtain a change rate of a loss function value of each model in the second model set;
and the fourth acquisition module is used for taking the model with the second model centralized loss function value change rate smaller than a set threshold value as the target prediction model.
Based on the same inventive concept as the method, the embodiment of the present invention further provides an enterprise operation risk prediction method, as shown in fig. 4, which is a flowchart corresponding to the embodiment of the method, and includes steps 61 to 62.
And step 61, acquiring enterprise operation data of at least two enterprises to be forecasted and operation relation data of the at least two enterprises to be forecasted.
And 62, acquiring enterprise operation risk prediction results of the at least two enterprises to be predicted by using the target prediction model obtained by the model training method.
Based on the same inventive concept as the method, the embodiment of the present invention further provides an enterprise operation risk prediction apparatus, and fig. 5 is a schematic structural diagram of the embodiment of the apparatus, including:
a fifth obtaining module 71, configured to obtain enterprise business data of at least two to-be-predicted enterprises and business relationship data of the at least two to-be-detected enterprises;
a sixth obtaining module 72, configured to obtain the business operation risk prediction results of the at least two enterprises to be predicted by using the target prediction model obtained by the model training method as described in any one of the above.
Based on the same inventive concept as in the previous embodiments, an embodiment of the present invention further provides an electronic device, which includes a memory, a processor, and a computer program stored on the memory and executable on the processor, and the processor implements the steps of any one of the methods when executing the program.
Based on the same inventive concept as in the previous embodiments, embodiments of the present invention further provide a computer-readable storage medium, on which a computer program is stored, which, when being executed by a processor, implements the steps of any of the methods described above.
The technical scheme provided by the embodiment of the invention at least has the following technical effects or advantages:
the embodiment of the invention utilizes the enterprise operation data and the operation relationship data of the enterprise to construct the training set, so that when the initial prediction model is trained, the model can learn the relationship between the enterprise operation risk result and the enterprise operation data and the operation relationship data, thereby realizing accurate and reliable enterprise operation risk prediction.
As will be appreciated by one skilled in the art, embodiments of the present invention may be provided as a method, system, or computer program product. Accordingly, the present invention may take the form of an entirely hardware embodiment, an entirely software embodiment or an embodiment combining software and hardware aspects. Furthermore, the present invention may take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, and the like) having computer-usable program code embodied therein.
The present invention is described with reference to flowchart illustrations and/or block diagrams of methods, apparatus (modules, systems), and computer program products according to embodiments of the invention. It will be understood that each flow and/or block of the flow diagrams and/or block diagrams, and combinations of flows and/or blocks in the flow diagrams and/or block diagrams, can be implemented by computer program instructions. These computer program instructions may be provided to a processor of a general purpose computer, special purpose computer, embedded computer, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, create means for implementing the functions specified in the flowchart flow or flows and/or block diagram block or blocks.
These computer program instructions may also be stored in a computer-readable memory that can direct a computer or other programmable data processing apparatus to function in a particular manner, such that the instructions stored in the computer-readable memory produce an article of manufacture including instruction means which implement the function specified in the flowchart flow or flows and/or block diagram block or blocks.
These computer program instructions may also be loaded onto a computer or other programmable data processing apparatus to cause a series of operational steps to be performed on the computer or other programmable apparatus to produce a computer implemented process such that the instructions which execute on the computer or other programmable apparatus provide steps for implementing the functions specified in the flowchart flow or flows and/or block diagram block or blocks.
While the preferred embodiments of the present invention have been described, additional variations and modifications in those embodiments may occur to those skilled in the art once they learn of the basic inventive concepts. Therefore, it is intended that the appended claims be interpreted as including preferred embodiments and all such alterations and modifications as fall within the scope of the invention.
It will be apparent to those skilled in the art that various changes and modifications may be made in the present invention without departing from the spirit and scope of the invention. Thus, if such modifications and variations of the present invention fall within the scope of the claims of the present invention and their equivalents, the present invention is also intended to include such modifications and variations.
Claims (10)
1. A model training method for enterprise business risk prediction, the method comprising:
constructing a training set; the training set comprises enterprise business data of at least two enterprises and business relation data of the at least two enterprises;
constructing an initial prediction model; wherein the initial prediction model comprises a graph convolution network layer and a fully connected network layer; the graph volume network layer comprises a characteristic input end used for inputting the enterprise business data, an adjacent matrix input end used for inputting the business relationship data and at least one graph volume network; the graph convolution network includes a first graph convolution layer and a second graph convolution layer;
and iteratively training the initial prediction model according to the training set to obtain a target prediction model for enterprise operation risk prediction.
2. The model training method of claim 1, wherein the constructing a training set comprises:
removing non-numerical characteristics in the original enterprise operation data of the at least two enterprises to obtain numerical enterprise operation data of the at least two enterprises; wherein the original enterprise business data comprises one or more of enterprise business age, enterprise registered capital, enterprise related litigation information, and enterprise recruitment information;
unifying the dimension and the characteristic vacancy value of the numerical enterprise operation data to obtain enterprise operation data of the at least two enterprises;
constructing an adjacency matrix according to the enterprise business relationship information between the at least two enterprises, and acquiring business relationship data of the at least two enterprises; wherein the enterprise business relationship information comprises one or more of investment relationship information, project participation relationship information and infringement litigation relationship information between the at least two enterprises.
3. The model training method of claim 1, wherein the constructing an initial predictive model comprises:
constructing an activation function of a first graph convolution layer of the graph convolution network by using a linear rectification function ReLU;
constructing an activation function of a second graph convolution layer of the graph convolution network by using a Softmax function;
and constructing a Loss function of the initial prediction model by utilizing a Focal local function.
4. The model training method of claim 1, wherein iteratively training the initial prediction model according to the training set to obtain a target prediction model for enterprise business risk prediction comprises:
training the initial prediction model by using the training set, obtaining a plurality of prediction models, and constructing a first model set;
obtaining a verification result of each model in the first model set by using a verification set;
screening one or more prediction models according to the verification result of each model to construct a second model set;
testing the models in the second model set by using a test set to obtain the change rate of the loss function value of each model in the second model set;
and taking the model with the second model centralized loss function value change rate smaller than a set threshold value as the target prediction model.
5. A model training apparatus for enterprise business risk prediction, the apparatus comprising:
the first construction module is used for constructing a training set; the training set comprises enterprise business data of at least two enterprises and business relation data of the at least two enterprises;
the second construction model is used for constructing the initial prediction model; wherein the initial prediction model comprises a graph convolution network layer and a fully connected network layer; the graph volume network layer comprises a characteristic input end used for inputting the enterprise business data, an adjacent matrix input end used for inputting the business relationship data and at least one graph volume network; the graph convolution network includes a first graph convolution layer and a second graph convolution layer;
and the first acquisition module is used for iteratively training the initial prediction model according to the training set to acquire a target prediction model for enterprise operation risk prediction.
6. The model training apparatus of claim 5, wherein the first building block comprises:
the first processing module is used for eliminating non-numerical characteristics in the original enterprise operation data of the at least two enterprises and acquiring numerical enterprise operation data of the at least two enterprises; wherein the original enterprise business data comprises one or more of enterprise business age, enterprise registered capital, enterprise related litigation information, and enterprise recruitment information;
the second processing module is used for unifying the dimension and the characteristic vacancy value of the numerical enterprise operation data and acquiring enterprise operation data of at least two enterprises;
the third construction module is used for constructing an adjacency matrix according to the enterprise business relationship information between the at least two enterprises and acquiring the business relationship data of the at least two enterprises; wherein the enterprise business relationship information comprises one or more of investment relationship information, project participation relationship information and infringement litigation relationship information between the at least two enterprises.
7. An enterprise operation risk prediction method is characterized by comprising the following steps:
acquiring enterprise operation data of at least two enterprises to be predicted and operation relation data of the at least two enterprises to be detected;
and acquiring enterprise operation risk prediction results of the at least two enterprises to be predicted by using the target prediction model obtained by the model training method according to any one of claims 1 to 4.
8. An enterprise operation risk prediction device, comprising:
the fifth acquisition module is used for acquiring enterprise operation data of at least two enterprises to be predicted and operation relation data of the at least two enterprises to be detected;
a sixth obtaining module, configured to obtain the enterprise operation risk prediction results of the at least two enterprises to be predicted by using the target prediction model obtained by the model training method according to any one of claims 1 to 4.
9. An electronic device, comprising:
a memory for storing a computer program;
a processor for executing the computer program to carry out the steps of the method of claim 1, 2, 3, 4 or 7.
10. A computer-readable storage medium, on which a computer program is stored which, when being executed by a processor, carries out the steps of the method of claim 1, 2, 3, 4 or 7.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110218834.XA CN114970926A (en) | 2021-02-26 | 2021-02-26 | Model training method, enterprise operation risk prediction method and device |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110218834.XA CN114970926A (en) | 2021-02-26 | 2021-02-26 | Model training method, enterprise operation risk prediction method and device |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN114970926A true CN114970926A (en) | 2022-08-30 |
Family
ID=82974043
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202110218834.XA Pending CN114970926A (en) | 2021-02-26 | 2021-02-26 | Model training method, enterprise operation risk prediction method and device |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN114970926A (en) |
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN116109195A (en) * | 2023-02-23 | 2023-05-12 | 深圳市迪博企业风险管理技术有限公司 | Performance evaluation method and system based on graph convolution neural network |
| CN116796909A (en) * | 2023-08-16 | 2023-09-22 | 浙江同信企业征信服务有限公司 | Judicial litigation risk prediction method, device, equipment and storage medium |
| CN117557106A (en) * | 2024-01-02 | 2024-02-13 | 广东浩迪智云技术有限公司 | Enterprise operation health degree detection method, device, equipment and medium |
-
2021
- 2021-02-26 CN CN202110218834.XA patent/CN114970926A/en active Pending
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN116109195A (en) * | 2023-02-23 | 2023-05-12 | 深圳市迪博企业风险管理技术有限公司 | Performance evaluation method and system based on graph convolution neural network |
| CN116796909A (en) * | 2023-08-16 | 2023-09-22 | 浙江同信企业征信服务有限公司 | Judicial litigation risk prediction method, device, equipment and storage medium |
| CN117557106A (en) * | 2024-01-02 | 2024-02-13 | 广东浩迪智云技术有限公司 | Enterprise operation health degree detection method, device, equipment and medium |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN114970926A (en) | Model training method, enterprise operation risk prediction method and device | |
| CN110703057B (en) | Power equipment partial discharge diagnosis method based on data enhancement and neural network | |
| CN111127246A (en) | Intelligent prediction method for transmission line engineering cost | |
| CN114357594B (en) | Bridge abnormity monitoring method, system, equipment and storage medium based on SCA-GRU | |
| CN110716512A (en) | Environmental protection equipment performance prediction method based on coal-fired power plant operation data | |
| CN111639783A (en) | Line loss prediction method and system based on LSTM neural network | |
| KR101968449B1 (en) | Automatic inspection system for label type data based on Artificial Intelligence Learning to improve data productivity, and method thereof | |
| CN113449919B (en) | Power consumption prediction method and system based on feature and trend perception | |
| CN115542429A (en) | XGboost-based ozone quality prediction method and system | |
| CN113327037A (en) | Model-based risk identification method and device, computer equipment and storage medium | |
| CN114818353A (en) | Train control vehicle-mounted equipment fault prediction method based on fault characteristic relation map | |
| CN114662793A (en) | Business process remaining time prediction method and system based on interpretable hierarchical model | |
| CN117318033B (en) | Power grid data management method and system combining data twinning | |
| CN116932523B (en) | Platform for integrating and supervising third party environment detection mechanism | |
| CN115438190B (en) | Power distribution network fault auxiliary decision knowledge extraction method and system | |
| CN117217020A (en) | Industrial model construction method and system based on digital twin | |
| Friederich et al. | A Framework for Validating Data-Driven Discrete-Event Simulation Models of Cyber-Physical Production Systems | |
| CN113139332A (en) | Automatic model construction method, device and equipment | |
| CN114970674A (en) | Time sequence data concept drift adaptation method based on relevance alignment | |
| Raj et al. | On the Impact of ML use cases on Industrial Data Pipelines | |
| CN114818460A (en) | Laboratory equipment residual service life prediction method based on automatic machine learning | |
| Wirawan et al. | Application of data mining to prediction of timeliness graduation of students (a case study) | |
| CN113807704A (en) | Intelligent algorithm platform construction method for urban rail transit data | |
| CN117076454B (en) | Engineering quality acceptance form data structured storage method and system | |
| CN118014006A (en) | Method and system for predicting multitasking business process of sewage treatment plant |
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 |