Loan business processing method and device

Abstract

The present invention provides a loan business processing method and device, which belongs to the field of artificial intelligence technology. The loan business processing method includes: receiving a loan business, obtaining a loan enterprise public opinion text, enterprise characteristic data and a loan amount according to the loan business; inputting the loan enterprise public opinion text into a public opinion classification model created based on a public opinion training text to obtain enterprise public opinion risk data; inputting the enterprise public opinion risk data and enterprise characteristic data into an enterprise risk identification model created based on enterprise characteristic training data to obtain enterprise risk data; and processing the loan business according to the comparison result between the loan threshold corresponding to the enterprise risk data and the loan amount. The present invention can efficiently and accurately process loan business, reduce loan risks and bank profit losses.

Description

Technical Field

[0001] The present invention relates to the field of artificial intelligence technology, and in particular to a loan business processing method and device. Background Art

[0002] In order to ensure the quality of bank loan business and bank profitability, banks will conduct a lot of risk assessment work on enterprises when lending to enterprises. In the existing technology, the approval personnel can use enterprise risk identification tools or other means to assess the anti-fraud risk of enterprises to save manpower and reduce time costs. In order to protect the interests of banks and reduce the risks of loan business, the establishment of a loan business system that can accurately process loan business in real time is the key to the bank's risk control system.

[0003] Existing technologies generally use enterprise industry chain information to establish specified rule models or use machine learning algorithms such as logistic regression, decision trees or random forests to predict enterprise risks in order to handle enterprise loan business. Although this has solved the problem of information islands to a certain extent, it lacks the ability to mine relational data. Since there may be a relationship of risk superposition, contagion and diffusion between enterprises, the accuracy of existing enterprise risk identification is not high. In addition, the types and content of data sources of existing technologies are not rich enough, and they will be affected by uncertain or unpredictable factors in the context of a constantly changing market environment. For example, the potential risks brought about by online public opinion hotspots may cause the company's situation to deviate seriously from expectations. Existing technologies cannot obtain the potential impact of online public opinion hotspots on the company, and thus cannot efficiently and accurately handle corporate loan business. Summary of the invention

[0004] The main purpose of the embodiments of the present invention is to provide a loan business processing method and device to efficiently and accurately process loan business and reduce loan risks and bank profit losses.

[0005] In order to achieve the above object, an embodiment of the present invention provides a loan business processing method, including:

[0006] Receive loan business, and obtain loan enterprise public opinion text, enterprise characteristic data and loan amount based on the loan business;

[0007] Input the loan enterprise public opinion text into the public opinion classification model created based on the public opinion training text to obtain the enterprise public opinion risk data;

[0008] Input the enterprise public opinion risk data and enterprise characteristic data into the enterprise risk identification model created based on the enterprise characteristic training data to obtain the enterprise risk data;

[0009] The loan business is processed based on the comparison results between the loan threshold corresponding to the enterprise risk data and the loan amount.

[0010] The embodiment of the present invention further provides a loan business processing device, including:

[0011] The acquisition module is used to receive loan business and obtain the loan enterprise public opinion text, enterprise characteristic data and loan amount according to the loan business;

[0012] The public opinion classification module is used to input the public opinion text of the loan enterprise into the public opinion classification model created based on the public opinion training text to obtain the enterprise public opinion risk data;

[0013] An enterprise risk identification module is used to input enterprise public opinion risk data and enterprise characteristic data into an enterprise risk identification model created based on enterprise characteristic training data to obtain enterprise risk data;

[0014] The loan business processing module is used to process the loan business based on the comparison result between the loan threshold corresponding to the enterprise risk data and the loan amount.

[0015] An embodiment of the present invention also provides a computer device, including a memory, a processor, and a computer program stored in the memory and running on the processor. When the processor executes the computer program, the steps of the loan business processing method are implemented.

[0016] An embodiment of the present invention further provides a computer-readable storage medium on which a computer program is stored. When the computer program is executed by a processor, the steps of the loan business processing method are implemented.

[0017] The loan business processing method and device of the embodiment of the present invention first obtain the loan enterprise public opinion text, enterprise characteristic data and loan amount according to the loan business, then input the loan enterprise public opinion text into the public opinion classification model to obtain the enterprise public opinion risk data, input the enterprise public opinion risk data and enterprise characteristic data into the enterprise risk identification model to obtain the enterprise risk data, and finally process the loan business efficiently and accurately according to the loan threshold and loan amount corresponding to the enterprise risk data, thereby reducing loan risks and bank profit losses. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0019] Figure 1 is a flow chart of a loan business processing method in one embodiment of the present invention;

[0020] Figure 2 is a flow chart of a loan business processing method in another embodiment of the present invention;

[0021] Figure 3 is a flow chart of creating a public opinion classification model in one embodiment of the present invention;

[0022] Figure 4 is a flow chart of creating a public opinion classification model in another embodiment of the present invention;

[0023] Figure 5 is a schematic diagram of a public opinion classification model in an embodiment of the present invention;

[0024] Figure 6 is a flow chart of S201 in an embodiment of the present invention;

[0025] Figure 7 is a flow chart of creating an enterprise risk identification model in one embodiment of the present invention;

[0026] Figure 8 is a flow chart of creating an enterprise risk identification model in another embodiment of the present invention;

[0027] Fig. 9 is a schematic diagram of an enterprise risk identification model in an embodiment of the present invention;

[0028] Fig.10 is a structural block diagram of a loan business processing device in an embodiment of the present invention;

[0029] Fig.11 It is a structural block diagram of a computer device in an embodiment of the present invention. DETAILED DESCRIPTION

[0030] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0031] Those skilled in the art will appreciate that the embodiments of the present invention may be implemented as a system, device, apparatus, method or computer program product. Therefore, the present disclosure may be specifically implemented in the following forms, namely: complete hardware, complete software (including firmware, resident software, microcode, etc.), or a combination of hardware and software.

[0032] In view of the fact that the existing technology lacks the ability to mine relational data, it will be affected by uncertain or unpredictable factors in the case of a constantly changing market environment, and thus cannot efficiently and accurately process corporate loan business. The embodiment of the present invention provides a loan business processing method to efficiently and accurately process loan business, reduce loan risks and bank interest losses. The present invention is described in detail below with reference to the accompanying drawings.

[0033] Figure 1 It is a flow chart of a loan business processing method in one embodiment of the present invention. Figure 2 FIG. 1 is a flow chart of a loan business processing method in another embodiment of the present invention. Figure 1 and Figure 2 As shown, the loan business processing methods include:

[0034] S101: Receive loan business, and obtain loan enterprise public opinion text, enterprise characteristic data and loan amount according to the loan business.

[0035] Among them, the loan enterprise public opinion text, enterprise characteristic data and loan amount in S101 are all real-time data. Enterprise characteristic data include enterprise growth capacity, enterprise quarterly profit, number of enterprise shareholders and equity proportion of the largest shareholder.

[0036] In the specific implementation, first obtain the loan enterprise information data (including the loan enterprise name), the loan enterprise's affiliated enterprise information data (including the affiliated enterprise name), shareholders and industrial chain from the bank database to determine the enterprise keyword information, and then enter the enterprise keyword information, loan enterprise name and affiliated enterprise name into each monitoring website, and use the crawling tool to crawl the articles searched by the enterprise keyword information, loan enterprise name and affiliated enterprise name (a specified number of articles can be crawled according to demand or the influence of each public website), and obtain the abstract, main text, and loan enterprise public opinion text such as user click-throughs, number of replies and number of comments in the article.

[0037] S102: Input the loan enterprise public opinion text into the public opinion classification model created based on the public opinion training text to obtain the enterprise public opinion risk data.

[0038] Among them, corporate public opinion risk data includes positive, neutral, relatively negative and severely negative.

[0039] S103: Inputting the enterprise public opinion risk data and the enterprise characteristic data into an enterprise risk identification model created based on the enterprise characteristic training data to obtain the enterprise risk data.

[0040] In specific implementation, the enterprise's growth ability, quarterly profit and number of shareholders can be used as node features, the equity ratio of the largest shareholder and the enterprise's public opinion risk data can be used as features of each edge, and the above data can be normalized using Z-score standardization. The normalized data can be input into the enterprise risk identification model to obtain enterprise risk data.

[0041] S104: Processing the loan business according to the comparison result between the loan threshold corresponding to the enterprise risk data and the loan amount.

[0042] Each enterprise risk data corresponds to a different loan threshold. When the loan threshold corresponding to the enterprise risk data of the enterprise applying for a loan is less than the loan amount applied by the enterprise, the enterprise is not allowed to apply for a loan, and a loan business processing failure message is returned. When the loan threshold corresponding to the enterprise risk data of the enterprise applying for a loan is greater than or equal to the loan amount applied by the enterprise, the enterprise is allowed to apply for a loan, and a loan business processing success message is returned.

[0043] Figure 1 The execution subject of the loan business processing method shown can be a computer. Figure 1 As can be seen from the process shown, the loan business processing method of the embodiment of the present invention first obtains the loan enterprise public opinion text, enterprise characteristic data and loan amount according to the loan business, and then inputs the loan enterprise public opinion text into the public opinion classification model to obtain the enterprise public opinion risk data, and inputs the enterprise public opinion risk data and the enterprise characteristic data into the enterprise risk identification model to obtain the enterprise risk data, and finally processes the loan business efficiently and accurately according to the loan threshold and loan amount corresponding to the enterprise risk data, thereby reducing loan risks and bank profit losses.

[0044] Figure 3 It is a flow chart of creating a public opinion classification model in one embodiment of the present invention. Figure 4 It is a flowchart of creating a public opinion classification model in another embodiment of the present invention. Figure 5 Schematic diagram of the public opinion classification model in the embodiment of the present invention. Figure 3-Figure 5 As shown, the present invention builds a public opinion classification model based on a bidirectional recurrent neural network, and effectively obtains deep and high-quality network public opinion information by combining the information of the input sequence in both the forward and backward directions, thereby making the public opinion sentiment classification more accurate. Creating a public opinion classification model based on public opinion training text includes:

[0045] Perform the following iterative processing:

[0046] S201: Determine public opinion prediction data based on public opinion training text and public opinion classification model parameters.

[0047] The public opinion training text is the annotated historical corporate public opinion text in the training set. The present invention divides the sentiment category of the text into 4 labels: positive, neutral, relatively negative and severely negative. All annotated data can be randomly divided into a training set and a validation set in proportion for subsequent public opinion classification model training and validation.

[0048] Among them, the public opinion classification model parameters include the first public opinion classification model parameters, the second public opinion classification model parameters and the third public opinion classification model parameters.

[0049] Figure 6 is a flow chart of S201 in an embodiment of the present invention. Figure 6 As shown, S201 includes:

[0050] S301: Generate first public opinion classification output data according to the public opinion training text and the first public opinion classification model parameters.

[0051] Among them, the first public opinion classification model parameters include a first input weight matrix, a second input weight matrix and a first bias, and the first public opinion classification output data is the output data of the first layer of the bidirectional recurrent neural network.

[0052] In specific implementation, the first public opinion classification output data can be obtained through the following formula:

[0053] c t =f(Wc t-1 +Ux t +b c );

[0054] Among them, c t is the first public opinion classification output data at time t (tth iteration), f is the first activation function, W is the first input weight matrix, U is the second input weight matrix, c t-1 is the first public opinion classification output data at time t-1 (t-1th iteration), x t is the public opinion training text at time t (tth iteration), b c is the first bias.

[0055] S302: Generate second public opinion classification output data according to the public opinion training text and the second public opinion classification model parameters.

[0056] Among them, the second public opinion classification model parameters include a third input weight matrix, a fourth input weight matrix and a second bias, and the second public opinion classification output data is the output data of the second layer of the bidirectional recurrent neural network.

[0057] In specific implementation, the second public opinion classification output data can be obtained through the following formula:

[0058] c′ t =f(W′c′t+1 +U′x t +b′ c );

[0059] Among them, c′ t is the second public opinion classification output data at time t (tth iteration), W′ is the third input weight matrix, U′ is the fourth input weight matrix, c′ t+1 is the second public opinion classification output data at time t+1 (t+1th iteration), b′ c is the second bias.

[0060] S303: Determine public opinion prediction data according to the first public opinion classification output data, the first public opinion classification output data and the third public opinion classification model parameters.

[0061] Among them, the third public opinion classification model parameters include the first output weight matrix, the second output weight matrix and the third bias, and the public opinion prediction data is the output data of the output layer of the bidirectional recurrent neural network.

[0062] In specific implementation, the public opinion prediction data can be determined by the following formula:

[0063] y t =g(Vc t +V′c′ t +b y );

[0064] Among them, y t is the public opinion prediction data at time t (tth iteration), g is the second activation function, V is the first output weight matrix, V′ is the second output weight matrix, b y is the third bias.

[0065] S202: Determine a public opinion loss function based on the public opinion prediction data and the actual public opinion data corresponding to the public opinion training text.

[0066] Among them, the actual public opinion data are the labels corresponding to the public opinion training texts, including positive, neutral, relatively negative and severely negative.

[0067] S203: Determine whether the public opinion loss function is less than a preset public opinion loss function threshold.

[0068] S204: When the public opinion loss function is less than a preset public opinion loss function threshold, a public opinion classification model is created according to the public opinion classification model parameters.

[0069] In one embodiment, when the public opinion loss function is less than a preset public opinion loss function threshold, the public opinion classification model created according to the public opinion classification model parameters in S204 can be verified based on the annotated historical corporate public opinion texts in the verification set, and S102 uses the verified public opinion classification model to obtain corporate public opinion risk data.

[0070] S205: When the public opinion loss function is greater than or equal to a preset public opinion loss function threshold, the public opinion classification model parameters are updated according to the public opinion loss function, and the iterative process continues.

[0071] In specific implementation, the parameters of the public opinion classification model can be optimized through Nadam gradient descent algorithm and public opinion loss function training.

[0072] Figure 7 It is a flow chart of creating an enterprise risk identification model in one embodiment of the present invention. Figure 8 It is a flow chart of creating an enterprise risk identification model in another embodiment of the present invention. Fig. 9 Schematic diagram of the enterprise risk identification model in the embodiment of the present invention. Figure 7-Figure 9 As shown, the present invention builds an enterprise risk identification model based on a graph neural network, and fully mines enterprise relationship data through graph data processing and topological connection relationship modeling. Creating an enterprise risk identification model based on enterprise feature training data includes:

[0073] Perform the following iterative processing:

[0074] S401: Determine risk prediction data based on enterprise characteristic training data and neural network model parameters.

[0075] Table 1

[0076] feature definition Q Enterprise growth capability P Quarterly profit of the company V Number of shareholders Y Equity share of the largest shareholder of the enterprise F Actual public opinion data

[0077] Since the enterprise risk identification model uses a graph neural network, the input of this type of network model is graph data, so it is necessary to determine the characteristics of each node and edge in the graph data. Table 1 is a schematic table of enterprise feature training data in an embodiment of the present invention. As shown in Table 1, the enterprise feature training data is the historical enterprise feature data and public opinion actual data F in the training set. The enterprise feature data includes the enterprise growth ability Q, the enterprise quarterly profit P, the number of enterprise shareholders V, and the equity ratio Y of the largest shareholder of the enterprise.

[0078] Each enterprise in the enterprise feature training data constitutes a node in the graph data, and the characteristics of the node correspond to the enterprise feature data. The present invention uses the enterprise growth ability Q, the enterprise quarterly profit P and the number of shareholders V as the characteristics of the node, so the characteristic vector of each node in the graph data is a three-dimensional vector [Q, P, V]. The present invention also uses the equity ratio Y of the largest shareholder of the enterprise and the actual public opinion data F as the characteristics of each edge, so the characteristic of each edge in the graph data is a two-dimensional vector [Y, F]. In order to avoid the problem of difficult convergence caused by excessive numerical differences between features, the enterprise feature training data is normalized using Z-score standardization. The normalized enterprise feature training data is combined with the topological information represented by the adjacency matrix to form the graph data input during the training of the enterprise risk identification model. All graph data are randomly divided into training sets and verification sets in proportion for subsequent enterprise risk identification model training and verification.

[0079] S401 specifically includes:

[0080] a. Use FCN (fully connected neural network) to perform feature transformation on the input graph data (enterprise feature training data), and transform the 3D node input features and 2D edge input features into 64 dimensions through two FCNs respectively.

[0081] b. The above 64-dimensional features are sequentially input into two GNN (graph convolutional neural network) hidden layers, and ReLU is used as the activation function between each hidden layer. The output of the second GNN hidden layer can obtain a graph embedding vector of length 128 after passing through the graph pooling layer.

[0082] c. Input the graph embedding vector into two FCN layers in sequence. The number of neurons in the two FCN layers is 128 and 4 respectively. The output of the last FCN layer is passed through a softmax classifier to classify the input graph data and obtain risk prediction data.

[0083] S402: Determine a risk loss function based on the risk prediction data and the actual risk data corresponding to the enterprise characteristic training data.

[0084] The actual risk data includes high risk, medium risk, low risk and no risk.

[0085] S403: Determine whether the risk loss function is less than a preset risk loss function threshold.

[0086] S404: When the risk loss function is less than a preset risk loss function threshold, an enterprise risk identification model is created according to the neural network model parameters.

[0087] In one embodiment, when the risk loss function is less than a preset risk loss function threshold, the enterprise risk identification model created according to the neural network model parameters in S404 can be verified based on the historical enterprise feature data and actual public opinion data in the verification set, and S103 uses the verified enterprise risk identification model to obtain the enterprise risk data.

[0088] S405: When the risk loss function is greater than or equal to a preset risk loss function threshold, the neural network model parameters are updated according to the risk loss function, and the iterative process continues.

[0089] In specific implementation, the neural network model parameters can be optimized through Nadam gradient descent algorithm and risk loss function training.

[0090] The specific process of the embodiment of the present invention is as follows:

[0091] 1. Generate first public opinion classification output data according to the public opinion training text and the first public opinion classification model parameters.

[0092] 2. Generate second public opinion classification output data based on the public opinion training text and the second public opinion classification model parameters.

[0093] 3. Determine the public opinion prediction data based on the first public opinion classification output data, the second public opinion classification output data and the third public opinion classification model parameters.

[0094] 4. Determine the public opinion loss function based on the public opinion prediction data and the actual public opinion data corresponding to the public opinion training text.

[0095] 5. When the public opinion loss function is less than the preset public opinion loss function threshold, create a public opinion classification model according to the public opinion classification model parameters; otherwise, update the public opinion classification model parameters according to the public opinion loss function and return to step 1.

[0096] 6. Determine risk prediction data based on enterprise characteristic training data and neural network model parameters.

[0097] 7. Determine the risk loss function based on the actual risk data corresponding to the risk prediction data and the enterprise characteristic training data.

[0098] 8. When the risk loss function is less than the preset risk loss function threshold, create an enterprise risk identification model based on the neural network model parameters; otherwise, update the neural network model parameters based on the risk loss function and return to step 6.

[0099] 9. Receive loan business, and obtain the loan enterprise public opinion text, enterprise characteristic data and loan amount based on the loan business.

[0100] 10. Input the loan enterprise public opinion text into the public opinion classification model to obtain the enterprise public opinion risk data.

[0101] 11. Input the enterprise public opinion risk data and enterprise characteristic data into the enterprise risk identification model to obtain enterprise risk data.

[0102] 12. Process loan business based on the comparison results between the loan threshold corresponding to the enterprise risk data and the loan amount.

[0103] In summary, the loan business processing provided by the embodiment of the present invention has the following beneficial effects:

[0104] (1) Using graph computing technology, we construct enterprise relationship graph data based on the enterprise characteristic data and enterprise public opinion risk data of the loan enterprises, and build a graph neural network model to fully mine the enterprise relationship data and improve the accuracy of enterprise risk identification.

[0105] (2) Combined with the hot information of online public opinion, a public opinion classification model for loan enterprises is established to dynamically predict the potential risks of uncertainties and unpredictable factors to enterprises in real time.

[0106] (3) Efficiently and accurately process loan business based on the loan thresholds and loan amounts corresponding to the enterprise risk data, thereby reducing loan risks and bank profit losses.

[0107] Based on the same inventive concept, an embodiment of the present invention also provides a loan business processing device. Since the principle of solving the problem by the device is similar to that of the loan business processing method, the implementation of the device can refer to the implementation of the method, and the repeated parts will not be repeated.

[0108] Fig.10 : is a structural block diagram of a loan business processing device in an embodiment of the present invention. Fig.10 As shown, the loan business processing device includes:

[0109] The acquisition module is used to receive loan business and obtain the loan enterprise public opinion text, enterprise characteristic data and loan amount according to the loan business;

[0110] The public opinion classification module is used to input the public opinion text of the loan enterprise into the public opinion classification model created based on the public opinion training text to obtain the enterprise public opinion risk data;

[0111] An enterprise risk identification module is used to input enterprise public opinion risk data and enterprise characteristic data into an enterprise risk identification model created based on enterprise characteristic training data to obtain enterprise risk data;

[0112] The loan business processing module is used to process the loan business based on the comparison result between the loan threshold corresponding to the enterprise risk data and the loan amount.

[0113] In one embodiment, the public opinion classification module is specifically used to perform the following iterative processing:

[0114] Determine public opinion prediction data based on public opinion training text and public opinion classification model parameters;

[0115] Determine the public opinion loss function based on the public opinion prediction data and the actual public opinion data corresponding to the public opinion training text;

[0116] When the public opinion loss function is less than the preset public opinion loss function threshold, a public opinion classification model is created according to the public opinion classification model parameters; otherwise, the public opinion classification model parameters are updated according to the public opinion loss function, and the iterative processing continues.

[0117] In one embodiment, the public opinion classification model parameters include a first public opinion classification model parameter, a second public opinion classification model parameter, and a third public opinion classification model parameter;

[0118] The public opinion classification module is specifically used for:

[0119] Generate first public opinion classification output data according to the public opinion training text and the first public opinion classification model parameters;

[0120] Generate second public opinion classification output data according to the public opinion training text and the second public opinion classification model parameters;

[0121] The public opinion prediction data is determined according to the first public opinion classification output data, the first public opinion classification output data and the third public opinion classification model parameters.

[0122] In one embodiment, the enterprise risk identification module is specifically configured to perform the following iterative processing:

[0123] Determine risk prediction data based on enterprise characteristic training data and neural network model parameters;

[0124] Determine the risk loss function based on the actual risk data corresponding to the risk prediction data and the enterprise characteristic training data;

[0125] When the risk loss function is less than the preset risk loss function threshold, an enterprise risk identification model is created according to the neural network model parameters, otherwise the neural network model parameters are updated according to the risk loss function and the iterative processing continues.

[0126] In summary, the loan business processing device of the embodiment of the present invention first obtains the loan enterprise public opinion text, enterprise characteristic data and loan amount according to the loan business, and then inputs the loan enterprise public opinion text into the public opinion classification model to obtain the enterprise public opinion risk data, and inputs the enterprise public opinion risk data and the enterprise characteristic data into the enterprise risk identification model to obtain the enterprise risk data, and finally processes the loan business efficiently and accurately according to the loan threshold and loan amount corresponding to the enterprise risk data, thereby reducing loan risks and bank profit losses.

[0127] The embodiment of the present invention also provides a specific implementation of a computer device that can implement all the steps in the loan business processing method in the above embodiment. Fig.11 is a block diagram of a computer device in an embodiment of the present invention, see Fig.11 , the computer equipment specifically includes the following contents:

[0128] Processor (processor) 1101 and memory (memory) 1102.

[0129] The processor 1101 is used to call the computer program in the memory 1102. When the processor executes the computer program, all steps in the loan business processing method in the above embodiment are implemented. For example, when the processor executes the computer program, the following steps are implemented:

[0130] Receive loan business, and obtain loan enterprise public opinion text, enterprise characteristic data and loan amount based on the loan business;

[0131] Input the loan enterprise public opinion text into the public opinion classification model created based on the public opinion training text to obtain the enterprise public opinion risk data;

[0132] Input the enterprise public opinion risk data and enterprise characteristic data into the enterprise risk identification model created based on the enterprise characteristic training data to obtain the enterprise risk data;

[0133] The loan business is processed based on the comparison results between the loan threshold corresponding to the enterprise risk data and the loan amount.

[0134] In summary, the computer device of the embodiment of the present invention first obtains the loan enterprise public opinion text, enterprise characteristic data and loan amount according to the loan business, and then inputs the loan enterprise public opinion text into the public opinion classification model to obtain the enterprise public opinion risk data, and inputs the enterprise public opinion risk data and the enterprise characteristic data into the enterprise risk identification model to obtain the enterprise risk data, and finally processes the loan business efficiently and accurately according to the loan threshold and loan amount corresponding to the enterprise risk data, thereby reducing loan risks and bank profit losses.

[0135] The embodiment of the present invention also provides a computer-readable storage medium capable of implementing all the steps in the loan business processing method in the above embodiment. The computer-readable storage medium stores a computer program. When the computer program is executed by a processor, all the steps in the loan business processing method in the above embodiment are implemented. For example, when the processor executes the computer program, the following steps are implemented:

[0136] Receive loan business, and obtain loan enterprise public opinion text, enterprise characteristic data and loan amount based on the loan business;

[0137] Input the loan enterprise public opinion text into the public opinion classification model created based on the public opinion training text to obtain the enterprise public opinion risk data;

[0138] Input the enterprise public opinion risk data and enterprise characteristic data into the enterprise risk identification model created based on the enterprise characteristic training data to obtain the enterprise risk data;

[0139] The loan business is processed based on the comparison results between the loan threshold corresponding to the enterprise risk data and the loan amount.

[0140] In summary, the computer-readable storage medium of the embodiment of the present invention first obtains the loan enterprise public opinion text, enterprise characteristic data and loan amount according to the loan business, and then inputs the loan enterprise public opinion text into the public opinion classification model to obtain the enterprise public opinion risk data, and inputs the enterprise public opinion risk data and the enterprise characteristic data into the enterprise risk identification model to obtain the enterprise risk data, and finally processes the loan business efficiently and accurately according to the loan threshold and loan amount corresponding to the enterprise risk data, thereby reducing loan risks and bank profit losses.

[0141] The specific embodiments described above further illustrate the objectives, technical solutions and beneficial effects of the present invention in detail. It should be understood that the above description is only a specific embodiment of the present invention and is not intended to limit the scope of protection of the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the scope of protection of the present invention.

[0142] Those skilled in the art may also understand that the various illustrative logical blocks, units, and steps listed in the embodiments of the present invention may be implemented by electronic hardware, computer software, or a combination of the two. In order to clearly demonstrate the interchangeability of hardware and software, the various illustrative components, units, and steps described above have generally described their functions. Whether such functions are implemented by hardware or software depends on the specific application and the design requirements of the entire system. Those skilled in the art may use various methods to implement the described functions for each specific application, but such implementation should not be understood as exceeding the scope of protection of the embodiments of the present invention.

[0143] The various illustrative logic blocks, or units, or devices described in the embodiments of the present invention can be implemented or operated by a general-purpose processor, a digital signal processor, an application-specific integrated circuit (ASIC), a field programmable gate array or other programmable logic device, a discrete gate or transistor logic, a discrete hardware component, or any combination of the above. The general-purpose processor can be a microprocessor, and optionally, the general-purpose processor can also be any conventional processor, controller, microcontroller or state machine. The processor can also be implemented by a combination of computing devices, such as a digital signal processor and a microprocessor, a plurality of microprocessors, one or more microprocessors combined with a digital signal processor core, or any other similar configuration.

[0144] The steps of the method or algorithm described in the embodiments of the present invention can be directly embedded in hardware, a software module executed by a processor, or a combination of the two. The software module can be stored in a RAM memory, a flash memory, a ROM memory, an EPROM memory, an EEPROM memory, a register, a hard disk, a removable disk, a CD-ROM, or other storage media of any form in the art. Exemplarily, the storage medium can be connected to the processor so that the processor can read information from the storage medium and can write information to the storage medium. Optionally, the storage medium can also be integrated into the processor. The processor and the storage medium can be arranged in an ASIC, and the ASIC can be arranged in a user terminal. Optionally, the processor and the storage medium can also be arranged in different components in the user terminal.

[0145] In one or more exemplary designs, the above functions described in the embodiments of the present invention can be implemented in hardware, software, firmware or any combination of the three. If implemented in software, these functions can be stored on a computer-readable medium, or transmitted in the form of one or more instructions or codes on a computer-readable medium. Computer-readable media include computer storage media and communication media that facilitate the transfer of computer programs from one place to another. The storage medium can be any available medium that can be accessed by any general or special computer. For example, such computer-readable media can include but are not limited to RAM, ROM, EEPROM, CD-ROM or other optical disk storage, disk storage or other magnetic storage devices, or any other medium that can be used to carry or store program codes in the form of instructions or data structures and other forms that can be read by general or special computers, or general or special processors. In addition, any connection can be appropriately defined as a computer-readable medium, for example, if the software is transmitted from a website site, server or other remote resource through a coaxial cable, fiber optic cable, twisted pair, digital subscriber line (DSL) or wirelessly, such as infrared, wireless and microwave, it is also included in the defined computer-readable medium. The disk and disc include compact disk, laser disk, optical disk, DVD, floppy disk and blue-ray disk. Disks usually copy data magnetically, while discs usually copy data optically with lasers. The above combination can also be included in computer readable media.

Claims

1. A loan business processing method, characterized in that: include: Receive loan business, and obtain loan enterprise public opinion text, enterprise characteristic data and loan amount according to the loan business; Inputting the loan enterprise public opinion text into a public opinion classification model created based on the public opinion training text to obtain enterprise public opinion risk data; Inputting the enterprise public opinion risk data and the enterprise characteristic data into an enterprise risk identification model created based on enterprise characteristic training data to obtain enterprise risk data; Processing the loan business according to a comparison result between the loan threshold corresponding to the enterprise risk data and the loan amount; The enterprise risk identification model is a graph neural network, and the enterprise feature training data is the historical enterprise feature data and actual public opinion data in the training set. The enterprise feature data includes the enterprise growth capacity, the enterprise's quarterly profit, the number of enterprise shareholders, and the equity ratio of the largest shareholder of the enterprise; Each enterprise in the enterprise feature training data constitutes a node in the graph data, and the node's features correspond to the enterprise feature data; the feature vector of each node in the graph data is a three-dimensional vector including the enterprise's growth capacity, the enterprise's quarterly profit, and the number of shareholders of the enterprise; the feature of each edge in the graph data is a two-dimensional vector including the equity ratio of the largest shareholder of the enterprise and the actual data of public opinion; Creating a public opinion classification model based on public opinion training text includes: Perform the following iterative processing: Determine public opinion prediction data according to the public opinion training text and public opinion classification model parameters; Determine a public opinion loss function according to the public opinion prediction data and the actual public opinion data corresponding to the public opinion training text; When the public opinion loss function is less than a preset public opinion loss function threshold, a public opinion classification model is created according to the public opinion classification model parameters; otherwise, the public opinion classification model parameters are updated according to the public opinion loss function, and the iterative process is continued; The public opinion classification model parameters include a first public opinion classification model parameter, a second public opinion classification model parameter and a third public opinion classification model parameter; Determining the public opinion prediction data according to the public opinion training text and the public opinion classification model parameters includes: Generate first public opinion classification output data according to the public opinion training text and the first public opinion classification model parameters; Generate second public opinion classification output data according to the public opinion training text and the second public opinion classification model parameters; Public opinion prediction data is determined according to the first public opinion classification output data, the second public opinion classification output data and the third public opinion classification model parameters.

2. The loan business processing method according to claim 1, characterized in that: Creating an enterprise risk identification model based on enterprise feature training data includes: Perform the following iterative processing: Determine risk prediction data based on the enterprise characteristic training data and neural network model parameters; Determine a risk loss function based on the risk prediction data and the actual risk data corresponding to the enterprise characteristic training data; When the risk loss function is less than a preset risk loss function threshold, an enterprise risk identification model is created according to the neural network model parameters; otherwise, the neural network model parameters are updated according to the risk loss function and the iterative process is continued.

3. A loan business processing device, characterized in that: include: An acquisition module, used to receive loan business, and acquire loan enterprise public opinion text, enterprise characteristic data and loan amount according to the loan business; A public opinion classification module, used for inputting the loan enterprise public opinion text into a public opinion classification model created based on the public opinion training text to obtain enterprise public opinion risk data; An enterprise risk identification module, used for inputting the enterprise public opinion risk data and the enterprise characteristic data into an enterprise risk identification model created based on enterprise characteristic training data to obtain enterprise risk data; A loan business processing module, used for processing the loan business according to the comparison result between the loan threshold corresponding to the enterprise risk data and the loan amount; The enterprise risk identification model is a graph neural network, and the enterprise feature training data is the historical enterprise feature data and actual public opinion data in the training set. The enterprise feature data includes the enterprise growth capacity, the enterprise's quarterly profit, the number of enterprise shareholders, and the equity ratio of the largest shareholder of the enterprise; Each enterprise in the enterprise feature training data constitutes a node in the graph data, and the node's features correspond to the enterprise feature data; the feature vector of each node in the graph data is a three-dimensional vector including the enterprise's growth capacity, the enterprise's quarterly profit, and the number of shareholders of the enterprise; the feature of each edge in the graph data is a two-dimensional vector including the equity ratio of the largest shareholder of the enterprise and the actual data of public opinion; The public opinion classification module is specifically used to perform the following iterative processing: Determine public opinion prediction data according to the public opinion training text and public opinion classification model parameters; Determine a public opinion loss function according to the public opinion prediction data and the actual public opinion data corresponding to the public opinion training text; When the public opinion loss function is less than a preset public opinion loss function threshold, a public opinion classification model is created according to the public opinion classification model parameters; otherwise, the public opinion classification model parameters are updated according to the public opinion loss function, and the iterative process is continued; The public opinion classification model parameters include a first public opinion classification model parameter, a second public opinion classification model parameter and a third public opinion classification model parameter; The public opinion classification module is specifically used for: Generate first public opinion classification output data according to the public opinion training text and the first public opinion classification model parameters; Generate second public opinion classification output data according to the public opinion training text and the second public opinion classification model parameters; Public opinion prediction data is determined according to the first public opinion classification output data, the second public opinion classification output data and the third public opinion classification model parameters.

4. The loan business processing device according to claim 3, characterized in that: The enterprise risk identification module is specifically used to perform the following iterative processing: Determine risk prediction data based on the enterprise characteristic training data and neural network model parameters; Determine a risk loss function based on the risk prediction data and the actual risk data corresponding to the enterprise characteristic training data; When the risk loss function is less than a preset risk loss function threshold, an enterprise risk identification model is created according to the neural network model parameters; otherwise, the neural network model parameters are updated according to the risk loss function and the iterative process is continued.

5. A computer device comprising a memory, a processor, and a computer program stored in the memory and running on the processor, characterized in that: When the processor executes the computer program, the steps of the loan business processing method described in any one of claims 1 to 2 are implemented.

6. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the loan business processing method described in any one of claims 1 to 2 are implemented.

Images