Method and system for constructing financial risk control model system based on artificial intelligence

Abstract

The invention relates to the technical field of financial risk control, and discloses a financial risk control model system construction method and system based on artificial intelligence, and the method comprises the following steps: collecting financing enterprise risk control information and historical risk cache resource configuration information; respectively generating structural data according to the acquired financing enterprise risk control information and historical risk cache resource configuration information; and inputting structural data generated by the financing enterprise risk control information into the financing enterprise risk control model, and outputting a result representing a financing enterprise risk control strategy. In the supply chain financing, the risk slow release resource configuration strategy is embedded into the financing enterprise risk control process, so that the coordination of risk slow release and risk control can be realized, the repeated work is reduced, the resource configuration is optimized, the risk control efficiency is improved, and the risk response capability is enhanced; and the loan party can realize a better risk management effect at a lower cost.

Description

Technical Field

[0001] The present invention relates to the field of financial risk control, and more specifically, it relates to a method and system for constructing a financial risk control model system based on artificial intelligence. Background Art

[0002] Compared with traditional credit, supply chain financing has the characteristics of diversified financing entities, complex transaction structures, and hidden risk transmission. This makes the credit risk assessment of supply chain financing unable to simply rely on the financial data of a single enterprise, but needs to consider the supply chain network structure and the credit risk correlation and transmission effect between upstream and downstream enterprises.

[0003] In reality, the risk transmission mechanism between enterprises may be more complex. In addition to transaction relationships, it may also be affected by various factors such as equity relationships, guarantee relationships, and industry policies, resulting in many lenders for supply chain financing being unable to correctly handle the risk management of this financing. Summary of the Invention

[0004] The present invention provides a method and system for constructing a financial risk control model system based on artificial intelligence to solve the technical problems in the related art.

[0005] According to one aspect of the present invention, a method for constructing a financial risk control model system based on artificial intelligence is provided, including the following steps: Step 100, collecting risk control information of financing enterprises and historical risk cache resource configuration information; Step 200, respectively generating structured data from the collected risk control information of financing enterprises and historical risk cache resource configuration information; Step 300, inputting the structured data generated from the risk control information of financing enterprises into the risk control model of financing enterprises, and outputting the result representing the risk control strategy of financing enterprises; Inputting the structured data generated from the historical risk cache resource configuration information and the result representing the risk control strategy of financing enterprises into the configuration update model, and outputting the result representing the updated risk mitigation resource configuration strategy; Step 400, according to the result representing the risk control strategy of financing enterprises, implementing different financing enterprise quotas, financing terms, and collateral requirements, and adjusting the risk mitigation resource configuration structure according to the result of the updated risk mitigation resource configuration strategy.

[0006] Furthermore, the risk control information of financing enterprises includes financing enterprise network structure data, real-time risk event data of financing enterprises, real-time financial and operating data of financing enterprises, and application information of financing enterprises; The financing enterprise network structure data includes the transaction relationships, equity relationships, and guarantee relationships between the financing enterprise and its upstream and downstream enterprises; the real-time risk event data of the financing enterprise includes the risk events of the financing enterprise and its upstream and downstream enterprises being in the state of overdue payment, default, or bankruptcy application; the real-time financial and operating data of the financing enterprise includes the cash flow capacity of the financing enterprise and the inventory turnover volume of the financing enterprise; the application information of the financing enterprise includes the financing amount, financing term, and collateral. The historical risk cache resource allocation information includes collateral information, guarantee information, risk transfer tool information, risk reserve information, and capital information; the collateral information includes real estate, movable property, and certificate of rights; the guarantee includes personal guarantee, enterprise guarantee, and government guarantee; the risk transfer tools include credit insurance, credit derivatives, and asset securitization; the risk reserve information includes the reserve type and the total amount of reserves; the capital information includes the type of own funds and the total amount of capital.

[0007] Further, based on the sorting of the financing enterprise risk control information, the first one-dimensional structure data is obtained. The first one-dimensional structure data includes n data items sorted by time. The t-th data item represents the first two-dimensional structure data generated from the financing enterprise risk control information collected at the t-th moment. The first two-dimensional structure data includes the first data matrix and the first relationship matrix. A first cell of the first data matrix represents the financing enterprise risk control information of an independent object. The independent objects include the financing enterprise, upstream enterprises, downstream enterprises, collateral, and lenders. A first cell of the first data matrix only contains the financing enterprise risk control information of the independent object it represents. The element in the i-th row and j-th column of the first relationship matrix represents the association between the independent objects represented by the i-th cell and the j-th cell of the first data matrix. If there is an association, the value of this element in the first relationship matrix is 1, otherwise it is 0. The existence of an association between the financing enterprise and its upstream and downstream enterprises means that there are transaction relationships, equity relationships, and guarantee relationships between the financing enterprise and its upstream and downstream enterprises. The existence of an association between the financing enterprise and the lender means that the financing enterprise applies for financing funds from the lender through the application information. The existence of an association between the collateral and the lender means that the collateral is held by the lender, and the lender uses the collateral for financing collateral to the lender. The existence of an association between the upstream and downstream enterprises and the lender means that there are direct or indirect transaction relationships or guarantee relationships between the upstream and downstream enterprises and the lender.

[0008] Further, generate the second two-dimensional structure data based on the historical risk cache resource configuration information. The second two-dimensional structure data includes a second data matrix and a second relationship matrix. A cell in the second data matrix represents the second one-dimensional structure data of an independent object. The independent objects include lenders, collateral, suretyship guarantees, risk transfer tools, risk reserves, and capital. A cell in the second data matrix only contains the historical risk cache resource configuration information of the independent object it represents; The element in the i-th row and j-th column of the second relationship matrix represents the association between the independent objects represented by the i-th cell and the j-th cell in the second data matrix. If there is an association, the value of this element in the second relationship matrix is 1; otherwise, it is 0; The existence of an association between a lender and collateral, suretyship guarantees, risk transfer tools, risk reserves, and capital means that the lender holds collateral, suretyship guarantees, risk transfer tools, risk reserves, and capital; The association between collaterals means that the collateral types can be switched with each other; The association between suretyship guarantees means that the suretyship guarantee types can be switched with each other; The association between risk transfer tools means that the risk transfer tool types can be switched with each other; The second one-dimensional structure data includes n data items sorted by time. The t-th data item represents the historical risk cache resource configuration information collected at the t-th moment.

[0009] Further, input the first two-dimensional structure data into the financing enterprise risk control model. The financing enterprise risk control model includes a first intermediate layer, a second intermediate layer, and a first output layer. The second intermediate layer outputs the second intermediate representation data to the first output layer. The first two-dimensional structure data is input into the first intermediate layer. The first intermediate layer outputs the first intermediate representation data to the second intermediate layer. The second intermediate layer outputs the second intermediate representation data to the first output layer. The first output layer outputs the result representing the financing enterprise risk control strategy; The financing enterprise risk control strategy includes the financing enterprise grade, financing enterprise quota, financing term, and collateral requirements; Input the second one-dimensional structure data and the result representing the risk control strategy of the financing enterprise into the configuration update model. The configuration update model includes a third intermediate layer, a fourth intermediate layer, a first intermediate representation fusion layer, a fifth intermediate layer, and a second output layer. Input the second one-dimensional structure data into the third intermediate layer, and the third intermediate layer outputs the third intermediate representation data to the first intermediate representation fusion layer. Input the result of the risk control strategy of the financing enterprise into the fourth intermediate layer to output the fourth intermediate representation data, and then input the fourth intermediate representation data into the first intermediate representation fusion layer as well. Output the first fusion representation data to the fifth intermediate layer, and the fifth intermediate layer outputs the fifth intermediate representation data to the second output layer. The second output layer outputs the result representing the updated risk mitigation resource allocation strategy; The updated risk mitigation resource allocation strategy includes collateral information, guarantee information, risk transfer tool information, risk reserve information, and capital information.

[0010] Further, the calculation formula of the first intermediate layer is as follows: ; ; ; Where represents the first intermediate representation data of the v-th first unit of the first data matrix of the t-th data item in the first one-dimensional structure data, and respectively represent the risk control information of the financing enterprise of the independent objects represented by the v-th and u-th first units of the first data matrix of the t-th data item in the first one-dimensional structure data, represents the attention coefficient of the v-th and u-th first units of the first data matrix of the t-th data item in the first one-dimensional structure data, represents the attention coefficient of the v-th and s-th first units of the first data matrix of the t-th data item in the first one-dimensional structure data, represents the normalized attention coefficient of the v-th and u-th units of the first data matrix of the t-th data item in the first one-dimensional structure data, represents the set of units of the first data matrix associated with the v-th unit of the first data matrix of the t-th data item in the first one-dimensional structure data, represents the exponential function with the natural constant as the base, represents the trainable attention vector parameter, represents the concatenation operation, represents the transpose, represents the two-dimensional recognition weight parameter of the first intermediate layer, is the LeakyReLU activation function; The calculation formula of the second intermediate layer is as follows: ; where represents the t-th second intermediate representation data, represents the (t - 1)-th second intermediate representation data, = , represents the set of all first units of the first data matrix of the t-th data item in the first one-dimensional structure data, and are the first and second weight parameters of the second intermediate layer, is the bias parameter of the second intermediate layer, and tanh is the hyperbolic tangent function; The calculation formula of the first output layer is as follows: ; where represents the first output vector, and the c-th component value thereof represents the probability value of the c-th strategy. The strategy group with the largest probability value is selected as the output. The strategy group contains all executable strategies. A strategy includes a financing enterprise level not lower than the level requirement and a set of financing enterprise amount, financing term, and collateral requirement for the loan direction to the financing enterprise; For a more complex strategy, it can be described using a vector representation: A strategy is represented as a vector, and the j-th component of the vector represents the financing enterprise amount, financing term, and collateral requirement that the financing enterprise can obtain for the j-th one; where represents the n-th second intermediate representation data, is the weight parameter in the second output layer, is the bias parameter in the second output layer, represents the sigmoid function.

[0011] Furthermore, the calculation formula of the third intermediate layer is as follows: ; where represents the g-th third intermediate representation data of the v-th second unit in the second two-dimensional structure data's second data matrix, represents the (g - 1)-th third intermediate representation data of the v-th second unit in the second two-dimensional structure data's second data matrix, represents the g-th data item of the second one-dimensional structure data corresponding to the v-th second unit in the second two-dimensional structure data's second data matrix, and are the first and second weight parameters of the third intermediate layer, is the bias parameter of the third intermediate layer, and tanh is the hyperbolic tangent function; The calculation formula for the fourth intermediate layer is as follows: ; Where represents the fourth intermediate representation data output from the fourth intermediate layer with the result input of the risk control strategy of the financing enterprise, represents the convolution function.

[0012] Furthermore, the calculation formula for the first intermediate representation fusion layer is as follows: ; Where, represents the first fusion representation data of the v-th second unit of the second data matrix, represents the last third intermediate representation data of the v-th second unit of the second data matrix, represents the fusion function, represents the summation weight matrix, represents the summation bias parameter; The calculation formula for the fifth intermediate layer is as follows: ; Where represents the fifth intermediate representation data of the v-th second unit of the second data matrix, is the set of second units of the second data matrix that are associated with the v-th second unit of the second data matrix, = , , represents the first fusion representation data of the u-th second unit of the second data matrix, is a learnable scalar parameter, represents the two-dimensional recognition weight parameter of the fifth intermediate layer, represents the multi-layer perceptron; The calculation formula for the second output layer is as follows: ; Where represents the second output vector, and the q-th component value thereof represents the probability value of the q-th strategy. The strategy group with the largest probability value is selected as the output. The strategy group includes all executable strategies. A strategy includes a set of collateral information, guarantee information, risk transfer tool information, risk reserve information, and capital information; For relatively complex strategies, vector representation can be used for description: A strategy is represented as a vector, and the j-th component of the vector represents the j-th risk buffer resource allocation combination held by the lender, represents the fifth intermediate representation data of the v-th second unit of the second data matrix, A set of second units of the second data matrix related to the recognition result is the weight parameter in the second input layer is the bias parameter in the second input layer represents the sigmoid function

[0013] The present invention also proposes a construction system of a financial risk control model system based on artificial intelligence, including: Data acquisition module: acquiring the risk control information of financing enterprises and the configuration information of historical risk cache resources Data encoding module: generating structured data from the acquired risk control information of financing enterprises and the configuration information of historical risk cache resources respectively Data processing module: inputting the structured data into the risk control model of financing enterprises and the configuration update model respectively, and outputting the result representing the risk control strategy of financing enterprises and the result representing the updated risk mitigation resource allocation strategy Risk control execution module: lending to financing enterprises according to the result representing the risk control strategy of financing enterprises, implementing different financing enterprise quotas, financing periods, and collateral requirements, and adjusting the risk mitigation resource allocation structure of the lender after lending according to the result of the updated risk mitigation resource allocation strategy

[0014] The present invention also provides a storage medium storing non-temporary computer-readable instructions for executing one or more steps in the foregoing method for constructing a financial risk control model system based on artificial intelligence

[0015] The beneficial effects of the present invention are as follows In the supply chain financing, by embedding the risk mitigation resource allocation strategy into the risk control process of financing enterprises, the present invention can achieve the coordination of risk mitigation and risk control, reduce duplicate work, optimize resource allocation, improve risk control efficiency, enhance the risk response ability, and enable the lender to achieve better risk management effects at a lower cost Description of the Drawings

[0016] Figure 1 is a flowchart of the method for constructing a financial risk control model system based on artificial intelligence of the present invention Figure 2 is a structural block diagram of the construction system of the financial risk control model system based on artificial intelligence of the present invention

[0017] In the figure: 101, data acquisition module; 102, data encoding module; 103, data processing module; 104, risk control execution module Detailed Embodiments

[0018] Reference will now be made to example embodiments to discuss the subject matter described herein. It should be understood that discussing these embodiments is only to enable those skilled in the art to better understand and thus implement the subject matter described herein. Without departing from the scope of protection of the content of this specification, changes can be made to the functions and arrangements of the elements discussed. Each example can omit, substitute, or add various processes or components as needed. Additionally, the features described relative to some examples can also be combined in other examples.

[0019] As Figure 1 shown, a method for constructing an artificial intelligence-based financial risk control model system includes the following steps: Step 100, collect risk control information of financing enterprises and historical risk cache resource configuration information; The risk control information of financing enterprises includes the network structure data of financing enterprises, real-time risk event data of financing enterprises, real-time financial and operating data of financing enterprises, and application information of financing enterprises; The network structure data of financing enterprises includes transaction relationships, equity relationships, and guarantee relationships between financing enterprises and upstream and downstream enterprises; The real-time risk event data of financing enterprises includes risk events in which financing enterprises and their upstream and downstream enterprises are in an overdue, default, or bankruptcy application state; The real-time financial and operating data of financing enterprises includes the cash flow capacity of financing enterprises and the inventory turnover volume of financing enterprises; The application information of financing enterprises includes the financing amount, financing term, and collateral; The historical risk cache resource configuration information includes collateral information, guarantee information, risk transfer tool information, risk reserve information, and capital information; The collateral information includes real estate, movable property, and rights certificates; The guarantee includes personal guarantee, enterprise guarantee, and government guarantee; The risk transfer tools include credit insurance, credit derivatives, and asset securitization; The risk reserve information includes the reserve type and the total reserve amount; The capital information includes the type of own funds and the total amount of capital; Step 200, respectively generate structured data from the collected risk control information of financing enterprises and historical risk cache resource configuration information; Obtain the first one-dimensional structured data based on the sorting of the risk control information of financing enterprises. The first one-dimensional structured data includes n data items sorted by time, and the t-th data item represents the first two-dimensional structured data generated from the risk control information of financing enterprises collected at the t-th moment; The first two-dimensional structure data includes a first data matrix and a first relationship matrix. A first cell of the first data matrix represents the risk control information of a financing enterprise of an independent object. The independent objects include financing enterprises, upstream enterprises, downstream enterprises, collateral, and lenders. A first cell of the first data matrix only contains the risk control information of the financing enterprise of the independent object it represents; The element in the i-th row and j-th column of the first relationship matrix represents the association between the independent objects represented by the i-th cell and the j-th cell of the first data matrix. If there is an association, the value of this element in the first relationship matrix is 1, otherwise it is 0; The existence of an association between a financing enterprise and an upstream enterprise or a downstream enterprise means that: there are transaction relationships, equity relationships, and guarantee relationships between the financing enterprise and the upstream enterprise or the downstream enterprise; The existence of an association between a financing enterprise and a lender means that: the financing enterprise applies for financing funds from the lender through application information; The existence of an association between collateral and a lender means that: the collateral is held by the lender, and the lender uses the collateral for financing collateral against the lender; The existence of an association between an upstream enterprise, a downstream enterprise and a lender means that: there is a direct or indirect transaction relationship or guarantee relationship between the upstream enterprise, the downstream enterprise and the lender; The second two-dimensional structure data is generated based on historical risk cache resource allocation information. The second two-dimensional structure data includes a second data matrix and a second relationship matrix. A cell of the second data matrix represents the second one-dimensional structure data of an independent object. The independent objects include lenders, collateral, guarantee guarantees, risk transfer tools, risk reserves, and capital funds. A cell of the second data matrix only contains the historical risk cache resource allocation information of the independent object it represents; The element in the i-th row and j-th column of the second relationship matrix represents the association between the independent objects represented by the i-th cell and the j-th cell of the second data matrix. If there is an association, the value of this element in the second relationship matrix is 1, otherwise it is 0; The existence of an association between a lender and collateral, guarantee guarantees, risk transfer tools, risk reserves, and capital funds means that: the lender holds collateral, guarantee guarantees, risk transfer tools, risk reserves, and capital funds; The association between collaterals means that: the types of collaterals are switched with each other; The association between guarantee guarantees means that: the types of guarantee guarantees are switched with each other; The association between risk transfer tools means that: the types of risk transfer tools are switched with each other; The second one-dimensional structure data includes n data items sorted by time. The t-th data item represents the historical risk cache resource allocation information collected at the t-th moment; Step 300: Input the first two-dimensional structure data into the risk control model of the financing enterprise. The risk control model of the financing enterprise includes a first intermediate layer, a second intermediate layer, and a first output layer. The second intermediate layer outputs the second intermediate representation data to the first output layer. The first two-dimensional structure data is input into the first intermediate layer, and the first intermediate layer outputs the first intermediate representation data to the second intermediate layer. The second intermediate layer outputs the second intermediate representation data to the first output layer, and the first output layer outputs the result representing the risk control strategy of the financing enterprise. The risk control strategy of the financing enterprise includes the financing enterprise level, the financing enterprise quota, the financing period, and the collateral requirements. Input the second one-dimensional structure data and the result representing the risk control strategy of the financing enterprise into the configuration update model. The configuration update model includes a third intermediate layer, a fourth intermediate layer, a first intermediate representation fusion layer, a fifth intermediate layer, and a second output layer. Input the second one-dimensional structure data into the third intermediate layer, and the third intermediate layer outputs the third intermediate representation data to the first intermediate representation fusion layer. Input the result of the risk control strategy of the financing enterprise into the fourth intermediate layer to output the fourth intermediate representation data, and then input the fourth intermediate representation data into the first intermediate representation fusion layer as well. Output the first fusion representation data to the fifth intermediate layer, and the fifth intermediate layer outputs the fifth intermediate representation data to the second output layer. The second output layer outputs the result representing the updated risk mitigation resource allocation strategy. The updated risk mitigation resource allocation strategy includes collateral information, guarantee information, risk transfer tool information, risk reserve information, and capital information. Step 400: According to the result representing the risk control strategy of the financing enterprise, if the level of the financing enterprise meets the standard, then lend to the financing enterprise. During the lending process, implement different financing enterprise quotas, financing periods, and collateral requirements. After the lending, implement the result of the updated risk mitigation resource allocation strategy to adjust the risk mitigation resource allocation structure of the lender.

[0020] In an embodiment of the present invention, the calculation formula of the first intermediate layer is as follows: ; ; ; Where represents the first intermediate representation data of the v-th first unit of the first data matrix of the t-th data item in the first one-dimensional structure data, and respectively represent the risk control information of the financing enterprise of the independent objects represented by the v-th and u-th first units of the first data matrix of the t-th data item in the first one-dimensional structure data, represents the attention coefficient of the v-th and u-th first units of the first data matrix of the t-th data item in the first one-dimensional structure data. The attention coefficient of the v-th and s-th first-order cells of the first-order data matrix representing the t-th data item in the first-order one-dimensional structure data, The normalized attention coefficient of the v-th and u-th cells of the first-order data matrix representing the t-th data item in the first-order one-dimensional structure data, The set of cells of the first-order data matrix associated with the v-th cell of the first-order data matrix representing the t-th data item in the first-order one-dimensional structure data, Denotes the exponential function with the natural constant as the base, Denotes the trainable attention vector parameter, Denotes the concatenation operation, Denotes the transpose, Denotes the two-dimensional recognition weight parameter of the first intermediate layer, Is the LeakyReLU activation function; In an embodiment of the present invention, the calculation formula of the second intermediate layer is as follows: ; Where Denotes the t-th second intermediate representation data, Denotes the (t - 1)-th second intermediate representation data, = , Denotes the set of all first cells of the first-order data matrix representing the t-th data item in the first-order one-dimensional structure data, And Are the first and second weight parameters of the second intermediate layer, Is the bias parameter of the second intermediate layer, and tanh is the hyperbolic tangent function.

[0021] In an embodiment of the present invention, the calculation formula of the first output layer is as follows: ; Where Denotes the first output vector, and the c-th component value thereof represents the probability value of the c-th strategy. The strategy group with the largest probability value is selected as the output. The strategy group includes all executable strategies. A strategy includes a financing enterprise level not lower than the grade requirement and a set of financing enterprise amounts, financing periods, and collateral requirements for lending to financing enterprises in a loan direction; For relatively complex strategies, vector representation can be used for description: A strategy is represented as a vector, and the j-th component of the vector represents the financing enterprise amount, financing period, and collateral requirement that the j-th financing enterprise can obtain; Where Denotes the n-th second intermediate representation data, Is the weight parameter in the second output layer, is the bias parameter in the second input layer, represents the sigmoid function; The steps for training the risk control model of the financing enterprise include: Step 011, initialize the parameters of the risk control model of the financing enterprise; Step 012, observe the risk control information of the financing enterprise at time e , the strategy executed at time e , the risk control information of the financing enterprise at time e+1 , the executed strategy and the obtained reward ; Step 013, then calculate the policy error:

[0022] where represents the policy error at time e, represents the discount factor, , represents the maximum probability value in the first output vector output by the risk control model of the financing enterprise when the input is , represents the probability value corresponding to the strategy in the first output vector output by the risk control model of the financing enterprise when the input is ; ; represents the degree of risk exposure of the financing enterprise, with a value range of [0, 1], where 0 indicates no risk and 1 indicates extremely high risk.

[0023] represents the effectiveness of the risk mitigation measure, with a value range of [0, 1], where 0 indicates ineffective and 1 indicates completely effective.

[0024] represents the cost of implementing the risk control strategy, with a value range of [0, 1], where 0 indicates no cost and 1 indicates extremely high cost.

[0025] , , respectively represent the weight coefficients of risk exposure, risk mitigation effectiveness, and strategy cost in the reward function, with a value range of [0, 1].

[0026] Step 014, update the risk control model of the financing enterprise, and the update formula is as follows: ; , represents the step size of deep learning, Indicates the transfer of updates; Step 015, iterate steps 012 - 014 until the risk control model of the financing enterprise converges or the number of iterations reaches the set value. The default value of this value is 50.

[0027] In an embodiment of the present invention, the calculation formula of the third intermediate layer is as follows: ; Where represents the g-th third intermediate representation data of the v-th second unit in the second data matrix of the second two-dimensional structure data, represents the (g - 1)-th third intermediate representation data of the v-th second unit in the second data matrix of the second two-dimensional structure data, represents the g-th data item of the second one-dimensional structure data corresponding to the v-th second unit in the second data matrix of the second two-dimensional structure data, and are the first and second weight parameters of the third intermediate layer, is the bias parameter of the third intermediate layer, and tanh is the hyperbolic tangent function.

[0028] In an embodiment of the present invention, the calculation formula of the fourth intermediate layer is as follows: ; Where represents the fourth intermediate representation data output from the input of the risk control strategy of the financing enterprise to the fourth intermediate layer, represents the convolution function.

[0029] In an embodiment of the present invention, the calculation formula of the first intermediate representation fusion layer is as follows: ; Where, represents the first fusion representation data of the v-th second unit in the second data matrix, represents the last third intermediate representation data of the v-th second unit in the second data matrix, represents the fusion function (concatenation function or summation function), represents the summation weight matrix, represents the summation bias parameter.

[0030] In an embodiment of the present invention, the calculation formula of the fifth intermediate layer is as follows: ; Where represents the fifth intermediate representation data of the v-th second unit in the second data matrix, is a set of second data matrix cells that are associated with the v-th second cell of the second data matrix, = , , represents the first fusion representation data of the u-th second cell of the second data matrix, is a learnable scalar parameter, represents the two-dimensional recognition weight parameter of the fifth intermediate layer, represents a multi-layer perceptron.

[0031] In an embodiment of the present invention, the calculation formula of the second output layer is as follows: ; where represents the second output vector, and the q-th component value thereof represents the probability value of the q-th strategy. The strategy group with the largest probability value is selected as the output. The strategy group includes all executable strategies. A strategy includes a set of collateral information, guarantee information, risk transfer tool information, risk reserve information, and capital information; For relatively complex strategies, vector representation can be used to illustrate: A strategy is represented as a vector, and the j-th component of the vector represents the j-th risk cache resource allocation combination held by the lender, represents the fifth intermediate representation data of the v-th second cell of the second data matrix, represents a set of second data matrix cells related to the recognition result, is the weight parameter in the second output layer, is the bias parameter in the second output layer, represents the sigmoid function.

[0032] The steps for training the configuration update model include: Step 021, initialize the parameters of the configuration update model; Step 022, observe the historical risk cache resource allocation information at time u , the strategy executed at time u , the historical risk cache resource allocation information at time u+1 , the reward obtained by executing the strategy ; Step 023, then calculate the policy error: ; where represents the policy error at time u, represents the second discount factor, , represents the input of the configuration update model​ The maximum probability value in the first output vector output at represents the input for updating the configuration update model when the first output vector output corresponds to the policy of the probability value; ; where represents the degree of risk exposure, with a value range of [0, 1], where 0 represents no risk and 1 represents extremely high risk, represents the value of the collateral, with a value range of [0, 1], where 0 represents no collateral and 1 represents extremely high collateral value, represents the reliability of the guarantee, with a value range of [0, 1], where 0 represents no guarantee and 1 represents extremely high guarantee reliability, represents the effectiveness of the risk transfer tool, with a value range of [0, 1], where 0 represents no risk transfer and 1 represents excellent risk transfer effect, represents the adequacy of the risk reserve, with a value range of [0, 1], where 0 represents no reserve and 1 represents extremely adequate reserve, represents the adequacy of the capital, with a value range of [0, 1], where 0 represents no capital and 1 represents extremely adequate capital; , , , , respectively represent the weight coefficients of risk exposure, collateral value, guarantee reliability, risk transfer effectiveness, risk reserve adequacy, and capital adequacy in the reward, with a value range of [0, 1].

[0033] Step 024, update the configuration update model, and the update formula is as follows: ; , represents the step size of the second deep learning, represents the transfer update; Step 025, iterate steps 022 - 024 until the configuration update model converges or the number of iterations reaches the set value. The default value of this value is 30.

[0034] As Figure 2 shown, in an embodiment of the present invention, a construction system for a financial risk control model system based on artificial intelligence is also proposed, including: Data acquisition module 101: Collect the risk control information of the financing enterprise and the historical risk cache resource configuration information; Data Encoding Module 102: Generate structured data from the collected risk control information of the financing enterprise and the historical risk cache resource allocation information respectively; Data Processing Module 103: Input the structured data into the risk control model of the financing enterprise and the configuration update model respectively, and output the result representing the risk control strategy of the financing enterprise and the result representing the updated risk mitigation resource allocation strategy; Risk Control Execution Module 104: According to the result representing the risk control strategy of the financing enterprise, provide loans to the financing enterprise, execute different financing enterprise quotas, financing periods, and collateral requirements, and after lending, execute the result of the updated risk mitigation resource allocation strategy to adjust the risk mitigation resource allocation structure of the lender.

[0035] At least one embodiment of the present disclosure provides a storage medium storing non-temporary computer-readable instructions for performing one or more steps in the foregoing method for constructing an artificial intelligence-based financial risk control model system.

[0036] The computer program may be stored / distributed on a suitable medium, such as an optical storage medium or a solid-state medium supplied together with or as part of other hardware, but may also be distributed in other forms, such as via the Internet or other wired or wireless telecommunication systems. Any reference signs in the claims should not be construed as limiting the scope.

[0037] The above has described the embodiments of this embodiment, but this embodiment is not limited to the above specific implementation manners. The above specific implementation manners are merely illustrative and not restrictive. Under the inspiration of this embodiment, those of ordinary skill in the art can also make many forms, all of which fall within the protection scope of this embodiment.

Claims

1. A method for constructing a financial risk control model system based on artificial intelligence, characterized in that: The following steps are involved: Step 100, collecting financing enterprise risk control information and historical risk cache resource configuration information; Step 200, generating structure data from the collected financing enterprise risk control information and historical risk cache resource configuration information; Step 300, inputting the structured data generated by the financing enterprise risk control information into the financing enterprise risk control model, and outputting the result representing the financing enterprise risk control strategy; Input the structure data generated by the historical risk cache resource configuration information and the result representing the risk control strategy of the financing enterprise into the configuration update model, and output the result representing the updated risk mitigation resource configuration strategy; Step 400, according to the result representing the risk control strategy of the financing enterprise, different financing enterprise quotas, financing terms, and collateral requirements are implemented, and according to the result of the updated risk mitigation resource allocation strategy, the risk mitigation resource allocation structure is adjusted.

2. The method for constructing a financial risk control model system based on artificial intelligence according to claim 1, characterized in that: Financing enterprise risk control information includes financing enterprise network structure data, financing enterprise real-time risk event data, financing enterprise real-time financial and operating data, and financing enterprise application information; The network structure data of financing enterprises include the transaction relationship, equity relationship and guarantee relationship between financing enterprises and upstream and downstream enterprises; the real-time risk event data of financing enterprises include risk events of overdue, default and bankruptcy application of financing enterprises and their upstream and downstream enterprises; the real-time financial and operating data of financing enterprises include the cash flow capacity and inventory turnover of financing enterprises; the application information of financing enterprises includes the financing amount, financing period and collateral; Historical risk cache resource allocation information includes collateral information, guarantee information, risk transfer tool information, risk reserve information, and capital information; collateral information includes real estate, movable property, and rights certificates; guarantees include personal guarantees, corporate guarantees, and government guarantees; risk transfer tools include credit insurance, credit derivatives, and asset securitization; risk reserve information includes reserve type and total reserve amount; capital information includes own fund type and total capital amount.

3. The method for constructing a financial risk control model system based on artificial intelligence according to claim 2, characterized in that: Based on the sorting of the risk control information of the financing enterprise, the one-dimensional structure data No. 1 is obtained, and the one-dimensional structure data No. 1 includes n data items sorted by time, and the t-th data item represents the one-dimensional structure data No. 1 generated by the risk control information of the financing enterprise collected at the t-th moment; The No. 1 two-dimensional structure data includes a No. 1 data matrix and a No. 1 relationship matrix. A No. 1 unit of the No. 1 data matrix represents the risk control information of an independent object of the financing enterprise. The independent objects include the financing enterprise, the upstream enterprise, the downstream enterprise, the collateral, and the lender. A No. 1 unit of the No. 1 data matrix only contains the risk control information of the financing enterprise of the independent object it represents. The element in the i-th row and j-th column of the No. 1 relationship matrix represents the association between the i-th unit of the No. 1 data matrix and the independent object represented by the j-th unit. If there is an association, the value of this element of the No. 1 relationship matrix is ​​1, otherwise it is 0; The relationship between the financing enterprise and the upstream and downstream enterprises refers to: the financing enterprise has transaction relationship, equity relationship and guarantee relationship with the upstream and downstream enterprises; The relationship between the financing enterprise and the lender means that the financing enterprise applies for financing funds from the lender through application information; The relationship between the collateral and the lender means that the collateral is held by the lender, and the lender uses the collateral to provide the lender with financing collateral; The existence of a connection between upstream enterprises, downstream enterprises and lenders means that there is a direct or indirect transaction relationship or guarantee relationship between upstream enterprises, downstream enterprises and lenders.

4. The method for constructing a financial risk control model system based on artificial intelligence according to claim 3, characterized in that: Generate No. 2 two-dimensional structure data based on historical risk cache resource configuration information, the No. 2 two-dimensional structure data includes No. 2 data matrix and No. 2 relationship matrix, one unit of the No. 2 data matrix represents No. 2 one-dimensional structure data of an independent object, the independent object includes a lender, collateral, guarantee, risk transfer tool, risk reserve, capital, and one unit of the No. 2 data matrix only includes historical risk cache resource configuration information of the independent object represented by it; The element in the i-th row and j-th column of the second relationship matrix represents the association between the i-th unit and the independent object represented by the j-th unit of the second data matrix. If there is an association, the value of this element of the second relationship matrix is ​​1, otherwise it is 0; The relationship between the lender and the collateral, guarantee, risk transfer instrument, risk reserve and capital refers to: the lender holds the collateral, guarantee, risk transfer instrument, risk reserve and capital; The association between collaterals refers to: switching between collateral types; The association between guarantees refers to: switching between guarantee types; The linkage between risk transfer instruments refers to: switching between risk transfer instrument types; The second one-dimensional structure data includes n data items sorted by time, and the t-th data item represents the historical risk cache resource configuration information collected at the t-th moment.

5. The method for constructing a financial risk control model system based on artificial intelligence according to claim 4, characterized in that: Input the No. 1 two-dimensional structure data into the risk control model of the financing enterprise. The risk control model of the financing enterprise includes a first intermediate layer, a second intermediate layer and a first output layer. The second intermediate layer outputs the second intermediate representation data to the first output layer. The No. 1 two-dimensional structure data is input into the first intermediate layer. The first intermediate layer outputs the first intermediate representation data to the second intermediate layer. The second intermediate layer outputs the second intermediate representation data to the first output layer. The first output layer outputs the result representing the risk control strategy of the financing enterprise. The risk control strategy of the financing enterprise includes the financing enterprise level, financing enterprise quota, financing period, and collateral requirements; Input the No. 2 one-dimensional structure data and the result representing the risk control strategy of the financing enterprise into the configuration update model, which includes a third intermediate layer, a fourth intermediate layer, a first intermediate representation fusion layer, a fifth intermediate layer and a second output layer, input the No. 2 one-dimensional structure data into the third intermediate layer, the third intermediate layer outputs the third intermediate representation data to the first intermediate representation fusion layer, input the result of the risk control strategy of the financing enterprise into the fourth intermediate layer, outputs the fourth intermediate representation data, and then inputs the fourth intermediate representation data into the first intermediate representation fusion layer, outputs the first fusion representation data to the fifth intermediate layer, the fifth intermediate layer outputs the fifth intermediate representation data to the second output layer, and the second output layer outputs the result representing the updated risk mitigation resource allocation strategy; The updated risk mitigation resource allocation strategy includes collateral information, guarantee information, risk transfer tool information, risk reserve information, and capital information.

6. The method for constructing a financial risk control model system based on artificial intelligence according to claim 5, characterized in that: The calculation formula for the first intermediate layer is as follows: ； ； ； in The first intermediate representation data of the vth unit No. 1 of the No. 1 data matrix No. 1 representing the tth data item in the No. 1 one-dimensional structure data, and The risk control information of the financing enterprise of the independent object represented by the vth and uth No. 1 units of the No. 1 data matrix of the tth data item in the No. 1 one-dimensional structure data is respectively represented. represents the attention coefficients of the vth and uth No. 1 units of the No. 1 data matrix of the No. 1 data item in the No. 1 one-dimensional structure data, represents the attention coefficients of the vth and sth No. 1 units of the No. 1 data matrix of the No. 1 data item in the No. 1 one-dimensional structure data, represents the normalized attention coefficients of the vth and uth units of the data matrix No. 1 for the tth data item in the one-dimensional structured data No. 1, represents a set of cells of data matrix No. 1 that are associated with the v-th cell of data matrix No. 1 of the t-th data item in one-dimensional structured data No. 1, represents an exponential function with a natural constant as base, represents the trainable attention vector parameters, Represents a splicing operation, represents transpose, represents the two-dimensional recognition weight parameter of the first intermediate layer, is the LeakyReLU activation function; The calculation formula for the second intermediate layer is as follows: ； in represents the t-th second intermediate representation data, represents the t-1th second intermediate representation data, = , represents the set of all first units of the No. 1 data matrix of the t-th data item in the No. 1 one-dimensional structure data, and are the first and second weight parameters of the second intermediate layer, is the bias parameter of the second intermediate layer, tanh is the hyperbolic tangent function; The calculation formula of the first output layer is as follows: ； in represents the first output vector, and its cth component value represents the probability value of the cth strategy. The strategy group with the largest probability value is selected as the output. The strategy group contains all executable strategies. A strategy includes a financing enterprise grade that is not lower than the grade requirement and a group of financing enterprise quotas, financing terms, and collateral requirements for the lending party to provide loans to the financing enterprise; For more complex strategies, vector representation can be used: a strategy is represented by a vector, and the jth component of the vector represents the financing amount, financing period, and collateral requirements that the jth financing enterprise can obtain; in represents the nth second intermediate representation data, is the weight parameter in the second output layer, is the bias parameter in the second input layer, Represents the sigmoid function.

7. The method for constructing a financial risk control model system based on artificial intelligence according to claim 6, characterized in that: The calculation formula for the third intermediate layer is as follows: ； in represents the g-th third intermediate representation data of the v-th second unit of the second data matrix in the second two-dimensional structure data, represents the g-1th third intermediate representation data of the vth second unit of the second data matrix in the second two-dimensional structure data, represents the g-th data item of the second one-dimensional structure data corresponding to the v-th second unit of the second data matrix in the second two-dimensional structure data, and are the first and second weight parameters of the third intermediate layer, is the bias parameter of the third intermediate layer, tanh is the hyperbolic tangent function; The calculation formula for the fourth intermediate layer is as follows: ； in The fourth intermediate representation data representing the result of the risk control strategy of the financing enterprise is input to the output of the fourth intermediate layer, Represents the convolution function.

8. The method for constructing a financial risk control model system based on artificial intelligence according to claim 7, characterized in that: The calculation formula of the first intermediate representation fusion layer is as follows: ； in, The first fusion representation data of the vth No. 2 unit of the No. 2 data matrix, Represents the last third intermediate representation data of the vth unit of the second data matrix, represents the fusion function, represents the sum weight matrix, represents the summation bias parameter; The calculation formula for the fifth intermediate layer is as follows: ； in The fifth intermediate representation data of the vth unit No. 2 of the No. 2 data matrix, is the set of No. 2 cells of the No. 2 data matrix that are associated with the v-th No. 2 cell of the No. 2 data matrix, = , , The first fusion representation data of the u-th unit of the second data matrix, is a learnable scalar parameter, represents the two-dimensional recognition weight parameter of the fifth intermediate layer, represents a multi-layer perceptron; The calculation formula of the second output layer is as follows: ； in represents the second output vector, whose qth component value represents the probability value of the qth strategy. The strategy group with the largest probability value is selected as the output. The strategy group contains all executable strategies. A strategy includes a set of collateral information, guarantee information, risk transfer tool information, risk reserve information, and capital information. For more complex strategies, vector representation can be used: a strategy is represented by a vector, and the jth component of the vector represents the jth risk buffer resource allocation combination held by the lender. The fifth intermediate representation data of the vth unit No. 2 of the No. 2 data matrix, represents the set of the second unit of the second data matrix related to the recognition result, is the weight parameter in the second output layer, is the bias parameter in the second input layer, Represents the sigmoid function.

9. A system for constructing a financial risk control model system based on artificial intelligence, used to execute one or more steps in the method for constructing a financial risk control model system based on artificial intelligence as claimed in any one of claims 1 to 8, characterized in that: include: Data collection module: collects financing enterprise risk control information and historical risk cache resource configuration information; Data encoding module: Generates structured data from the collected financing enterprise risk control information and historical risk cache resource configuration information; Data processing module: inputs the structural data into the financing enterprise risk control model and the configuration update model respectively, and outputs the results representing the financing enterprise risk control strategy and the results representing the updated risk mitigation resource allocation strategy; Risk control execution module: based on the results of the risk control strategy of the financing enterprise, loans are made to the financing enterprise, and different financing enterprise quotas, financing terms, and collateral requirements are implemented. After the loan is made, the updated risk mitigation resource allocation strategy is implemented to adjust the risk mitigation resource allocation structure of the lender.

10. A storage medium, characterized in that: Non-temporary computer-readable instructions are stored for executing one or more steps in the method for constructing a financial risk control model system based on artificial intelligence as described in any one of claims 1-8.

Images