Risk control assessment method and system based on multi-dimensional perception and enterprise data quantification

Abstract

The invention relates to the technical field of risk analysis, and discloses a risk control assessment method and system based on multi-dimensional perception and enterprise data quantification, and the method comprises the following steps: collecting enterprise data; constructing graph structure data; inputting the graph structure data into a propagation model, wherein the propagation model outputs a result representing an enterprise risk factor infection path; inputting the graph structure data into a decision model, and outputting a result representing a risk coping strategy by the decision model; by analyzing the historical data and the real-time data of the enterprise and comprehensively considering the real-time external data, the propagation path and the coping strategy of the internal and external risks of the enterprise in the whole enterprise are obtained, and compared with only providing the risk propagation path, the system can provide the risk coping strategy for reducing risks and saving cost and resources, and the user experience is improved.

Description

Technical Field

[0001] The present invention relates to the technical field of risk analysis, and more specifically, to a risk control assessment method and system based on multi-dimensional perception and enterprise data quantification. Background Art

[0002] In today's highly dynamic and complex business environment, enterprises face a variety of potential risks, which may not only cause damage to themselves, but also spread throughout the entire enterprise system through various channels. Using deep learning technology and integrating information from multiple data sources such as different departments, suppliers, customers or markets of the enterprise can automatically capture and model risk factors inside and outside the enterprise and their possible transmission paths. However, in a market environment with increasing uncertainty, past data may not accurately reflect future trends, affecting the reliability of the prediction model. At the same time, when institutions identify the risk factors of an enterprise and their transmission paths, they should strengthen supervision of high-risk industries or enterprises to ensure that they adopt appropriate risk management strategies to avoid the spread of risks. Summary of the invention

[0003] The purpose of the present invention is to provide a risk control assessment method and system based on multi-dimensional perception and enterprise data quantification in order to solve the above problems.

[0004] The present invention provides a risk control assessment method based on multi-dimensional perception and enterprise data quantification, comprising the following steps: Step 100, collecting enterprise data, which is collected once at each time sampling point, including department information, supplier information, customer information and external environment information; Step 200, constructing graph structure data, the graph structure data includes nodes and edges connecting the nodes, a node represents a department, supplier, customer or external environment; there are edges between a department and its departments with which it has project and process associations, there are edges between a department and its subordinate departments, there are edges between a department and its suppliers, there are edges between a department and its customers, and there are edges between all departments and the external environment; The node characteristics of the node representing the department are equal to the department characteristics, the node characteristics of the node representing the supplier are equal to the supplier characteristics, the node characteristics of the node representing the customer are equal to the customer characteristics, and the node characteristics of the node representing the external environment are equal to the external environment characteristics; Department characteristics, supplier characteristics, customer characteristics and external environment characteristics are obtained through feature engineering; Encode the enterprise data of each node into first-order sequence data, the first-order sequence data includes n first-order sequence units, and the t-th first-order sequence unit represents the enterprise data of a node at the t-th time sampling point; Step 300, inputting the graph structure data into a propagation model, the propagation model comprising a first hidden layer, a second hidden layer and a first output layer, the propagation model outputting a result representing a transmission path of enterprise risk factors; Step 400, encode the enterprise risk factor transmission path into No. 2 sequence data; the No. 2 sequence data includes B No. 2 sequence units sorted by time, the zth No. 2 sequence unit represents the risk path collected at the zth moment, and the graph structure data is input into the decision model. The decision model includes a third hidden layer, a feature fusion layer, a fourth hidden layer, a sixth hidden layer and a second output layer. The decision model output represents the result of the risk response strategy.

[0005] Furthermore, the time sampling point refers to the moment at which data is collected at a predetermined time interval, and the default value of the predetermined time interval is 24 hours.

[0006] Furthermore, department information includes basic information of the department, subordinate departments, financial information, human resource information, and department process information; Supplier information includes basic information of suppliers and supplier's supply information; Customer information includes basic customer information and purchase information; External environmental information includes market intelligence information, social media trend information, and news report information.

[0007] Furthermore, in the propagation model, sequence data No. 1 is input into the first hidden layer, the first hidden layer outputs the first hidden state to the second hidden layer, the second hidden layer also inputs graph structure data, the second hidden layer outputs the second hidden state to the first output layer and the feature fusion layer, and the first output layer outputs the result representing the transmission path of the enterprise risk factor.

[0008] Furthermore, the loss function calculation method of the propagation model is: ; in, represents the propagation loss value, is the number of propagation samples, is the path length of each propagation sample, is the total number of department nodes, Indicates The propagation sample is in The actual propagation node is If it is a department node, its value is 1, otherwise it is 0; The propagation model predicts The propagation sample is in The propagation node is The probability of a department node, Represents the logarithmic function with the natural constant e as the base.

[0009] Furthermore, in the decision model, the first sequence data is input into the third hidden layer, the third hidden layer outputs the third hidden state to the feature fusion layer, the second sequence data is input into the fourth hidden layer, the fourth hidden layer outputs the fourth hidden state to the feature fusion layer, the feature fusion layer outputs the first fusion state to the sixth hidden layer, the sixth hidden layer also inputs the graph structure data, the sixth hidden layer outputs the sixth hidden state to the second output layer, and the second output layer outputs the result representing the risk response strategy.

[0010] Furthermore, the risk response strategy represents the action plan, resource allocation, and responsibility allocation implemented in response to the transmission path of enterprise risk factors.

[0011] Furthermore, the steps of training the decision model include: Step 401, building a target network, and replicating a decision model of the target network; Step 402: randomly obtain an experience from the experience pool, the experience includes the state at time e , the action performed at time e , the state at time e+1 , perform the action Rewards ; The states at time e and e+1 represent the enterprise data at the e-th and e+1-th time input into the decision model respectively; Then calculate the state The target network output corresponds to The probability value ; ; express The maximum probability value of the target network output; Step 403, according to and Calculate the loss value and update the decision model; ; express The output of the decision model corresponds to The probability value of is the discount factor, a value between 0 and 1; Step 404, updating the target network every fixed number of network updates so that its parameters are the same as those of the current decision model; Step 405, until the number of collected experiences reaches a set threshold or the decision model converges, the step is terminated.

[0012] Furthermore, rewards The calculation formula is as follows: ; in is the expected return, is the cost of implementing the risk response strategy, is the effect of the risk response strategy. If the implementation of the risk response strategy successfully reduces the risk, then is 1, otherwise it is 0. , , are the first, second, and third weight coefficients of the reward, It is a comprehensive indicator of various resources consumed when implementing risk response strategies. The calculation method is as follows: ; in is the time required to execute the strategy, It is the product of the number of people involved and the number of working hours. is the actual amount of funds consumed, is the equipment and technical resources used, , , , It is the first, second, third and fourth weight coefficient of the comprehensive indicator.

[0013] The present invention provides a risk control assessment system based on multi-dimensional perception and enterprise data quantification, which is used to store computer-readable instructions. When the computer-readable instructions are read, the aforementioned risk control assessment method based on multi-dimensional perception and enterprise data quantification can be executed.

[0014] The beneficial effects of the present invention are as follows: the present invention analyzes the enterprise's historical data and real-time data, comprehensively considers real-time external data, and obtains the transmission path and response strategy of internal and external risks in the entire enterprise. Compared with only providing risk transmission paths, the system can provide risk response strategies that reduce risks, save costs and resources, and improve user experience. BRIEF DESCRIPTION OF THE DRAWINGS

[0015] Figure 1 It is a flow chart of a risk control assessment method based on multi-dimensional perception and enterprise data quantification of the present invention. DETAILED DESCRIPTION

[0016] The subject matter described herein will now be discussed with reference to example embodiments. It should be understood that the discussion of these embodiments is only to enable those skilled in the art to better understand and implement the subject matter described herein, and the functions and arrangements of the elements discussed may be changed without departing from the scope of protection of the contents of this specification. Each example may omit, replace or add various processes or components as needed. In addition, the features described relative to some examples may also be combined in other examples.

[0017] At least one embodiment of the present invention discloses a risk control assessment method based on multi-dimensional perception and enterprise data quantification, such as Figure 1 As shown, the following steps are included: Step 100, collecting enterprise data, which is collected once at each time sampling point, including department information, supplier information, customer information and external environment information; Department information includes basic department information (department name, functions, important KPIs, current project overview), subordinate departments, financial information, human resources information (employee list, position, work record, skill matrix, performance record), and department process information; Supplier information includes basic information of suppliers and supplier's supply information; Customer information includes basic customer information and purchase information; External environmental information includes market intelligence information, social media trend information, and news report information; The time sampling point refers to the moment when data is collected at a predetermined time interval; In one embodiment of the present invention, the default value of the predetermined time interval is 24 hours.

[0018] In one embodiment of the present invention, a service agreement will be signed with the enterprise before collecting enterprise data. The service agreement includes the collection and use permission of information and data, which complies with relevant laws and regulations.

[0019] Step 200, constructing graph structure data, the graph structure data includes nodes and edges connecting the nodes, and a node represents a department, supplier, customer or external environment; There are edges between departments and departments with which they have project and process associations, between departments and their subordinate departments, between departments and their suppliers, between departments and their customers, and between all departments and the external environment; The node characteristics of the node representing the department are equal to the department characteristics, the node characteristics of the node representing the supplier are equal to the supplier characteristics, the node characteristics of the node representing the customer are equal to the customer characteristics, and the node characteristics of the node representing the external environment are equal to the external environment characteristics; Department characteristics, supplier characteristics, customer characteristics and external environment characteristics are obtained through feature engineering; Encode the enterprise data of each node into first-order sequence data, the first-order sequence data includes n first-order sequence units, and the t-th first-order sequence unit represents the enterprise data of a node at the t-th time sampling point; In one embodiment of the present invention, the enterprise data of the department is processed by feature engineering to obtain department features; The enterprise data of the department is in text mode, and the Word Embedding algorithm is used as a feature engineering method; In one embodiment of the present invention, the enterprise data of the supplier is processed by feature engineering to obtain supplier features; The supplier's enterprise data is in text mode, and the Word Embedding algorithm is used as a feature engineering method; In one embodiment of the present invention, the customer's enterprise data is processed by feature engineering to obtain customer features; The customer's corporate data is in text mode, and the Word Embedding algorithm is used as a feature engineering method; In one embodiment of the present invention, the enterprise data of the external environment is processed by feature engineering to obtain external environment features; The enterprise data in the external environment is in text mode, and the Word Embedding algorithm is used as a feature engineering method; Step 300, inputting graph structure data into a propagation model, the propagation model comprising a first hidden layer, a second hidden layer and a first output layer, wherein the first sequence data is input into the first hidden layer, the first hidden layer outputs a first hidden state to the second hidden layer, the second hidden layer also inputs graph structure data, the second hidden layer outputs a second hidden state to the first output layer and the feature fusion layer, and the first output layer outputs a result representing the transmission path of the enterprise risk factor; In one embodiment of the present invention, the formula of the first hidden layer is as follows:

[0020] in Represents the tth first hidden state of the vth node of the graph structure data, Represents the t-1th first hidden state of the vth node of the graph structure data, Represents the tth first-order sequence unit of the first-order sequence data corresponding to the vth node of the graph structure data, and are the first and second sequence weight parameters of the first hidden layer, is the first sequence bias parameter of the first hidden layer, and tanh is the hyperbolic tangent function.

[0021] In one embodiment of the present invention, the formula of the second hidden layer is as follows:

[0022] in represents the second hidden state of the vth node, represents the weight parameter of the first graph of the second hidden layer, , , and Represents the nth first hidden state of the vth and uth nodes of the graph structure data, is the set of nodes that have edges with the node v of the graph structure data. express The total number of nodes in ;

[0023] in represents the first aggregation coefficient of the second hidden layer between the vth and uth nodes, tanh represents the hyperbolic tangent function, T represents the transpose, Indicates adjustable parameters; The default value is 1 / N, where N is the dimension of the node features.

[0024] In one embodiment of the present invention, the formula of the first output layer is as follows:

[0025] in represents the risk path vector of the vth department node. A component of the risk path vector represents the probability that the vth department node belongs to a path of the enterprise risk factor transmission path. If the probability is greater than 0.5, it means that it belongs to the enterprise risk factor transmission path, otherwise it does not belong to the enterprise risk factor transmission path. represents the second hidden state of the v-th department node, is the first weight parameter of the first output layer, is the first bias parameter of the first output layer, Represents a sigmoid function.

[0026] In one embodiment of the present invention, the loss function calculation method of the propagation model is:

[0027] in, represents the propagation loss value, is the number of propagation samples, is the path length of each propagation sample, is the total number of department nodes, Indicates The propagation sample is in The actual propagation node is If it is a department node, its value is 1, otherwise it is 0; The propagation model predicts The propagation sample is in The propagation node is The probability of a department node, Represents the logarithmic function with the natural constant e as the base.

[0028] Step 400, inputting graph structure data into a decision model, the decision model comprising a third hidden layer, a feature fusion layer, a fourth hidden layer, a sixth hidden layer, and a second output layer, wherein the first sequence data is input into the third hidden layer, the third hidden layer outputs a third hidden state to the feature fusion layer, the second sequence data is input into the fourth hidden layer, the fourth hidden layer outputs a fourth hidden state to the feature fusion layer, the feature fusion layer outputs a first fusion state to the sixth hidden layer, the sixth hidden layer also inputs graph structure data, the sixth hidden layer outputs a sixth hidden state to the second output layer, and the second output layer outputs a result representing a risk response strategy; In one embodiment of the present invention, the enterprise risk factor transmission path is encoded as second sequence data; the second sequence data includes B second sequence units sorted by time, and the zth second sequence unit represents the risk path collected at the zth moment.

[0029] In one embodiment of the present invention, the formula of the third hidden layer is as follows:

[0030]

[0031]

[0032]

[0033] in, , , represents the first, second, and third sequence weight parameters of the third hidden layer, , , represents the first, second, and third sequence bias parameters of the third hidden layer, represents the dot product, , and They represent the first, second, and third intermediate states of the vth node of the graph structure data, respectively. Represents the tth first-order sequence unit of the first-order sequence data corresponding to the vth node of the graph structure data, and They represent the t-th and t-1-th third hidden states of the v-th node of the graph structure data, n≥t≥1, n represents the total number of the first sequence units of the first sequence data, and t=1 , tanh is the hyperbolic tangent function, Represents a sigmoid function.

[0034] In one embodiment of the present invention, the formula of the fourth hidden layer is as follows:

[0035] in Indicates A fourth hidden state, Indicates A fourth hidden state, Indicates the first The second sequence unit, and are the first and second sequence weight parameters of the fourth hidden layer, is the first sequence bias parameter of the fourth hidden layer, and tanh is the hyperbolic tangent function.

[0036] In one embodiment of the present invention, the formula of the feature fusion layer is as follows:

[0037] in, Represents the first fusion state of the vth node of the graph structure data, Represents the nth third hidden state of the vth node of the graph structure data, Indicates A fourth hidden state, represents the concatenation function, represents the sum weight matrix, Represents the summation bias parameter.

[0038] In one embodiment of the present invention, the formula of the sixth hidden layer is as follows:

[0039] in Represents the sixth hidden state of the vth node of the graph structure data, The first aggregation coefficient of the sixth hidden layer of the vth node representing the graph structure data, represents the activation function, represents the first graph weight parameter of the sixth hidden layer;

[0040]

[0041]

[0042] in , , and Respectively represent the first fusion state of the vth and uth nodes of the graph structure data, and Respectively represent the linear transformation features of the vth and uth nodes of the graph structure data, represents the aggregation weight coefficient of the sixth hidden layer, represents the concatenation weight coefficient of the sixth hidden layer, T represents transposition, represents an exponential function with a natural constant as the base, LeakyRelu represents a modified linear unit function, Represents the set of nodes that have edges with the vth node.

[0043] In one embodiment of the present invention, the formula of the second output layer is as follows:

[0044] in Represents a risk response vector. A component value of the risk response vector represents the probability value of a risk response strategy. The risk response strategy with the largest probability value is selected as the output. The risk response strategy group contains all executable risk response strategies. A risk response strategy represents the action plan, resource allocation, and responsibility allocation executed for the transmission path of enterprise risk factors. represents the concatenation of the sixth hidden states of all nodes, M represents the set of all nodes, is the first weight parameter of the second output layer, is the first bias parameter of the second output layer, Represents a sigmoid function.

[0045] In one embodiment of the present invention, the step of training the decision model includes: Step 401, building a target network, and replicating a decision model of the target network; Step 402: randomly obtain an experience from the experience pool, the experience includes the state at time e , the action performed at time e , the state at time e+1 , perform the action Rewards ; The states at time e and e+1 represent the enterprise data at the e-th and e+1-th time input into the decision model respectively; Then calculate the state The target network output corresponds to The probability value ;

[0046] express The maximum probability value of the target network output; Step 403, according to and Calculate the loss value and update the decision model;

[0047] express The output of the decision model corresponds to The probability value of is the discount factor, a value between 0 and 1; Step 404, updating the target network every fixed number of network updates so that its parameters are the same as those of the current decision model; Step 405, until the number of collected experiences reaches a set threshold or the decision model converges, the step is terminated.

[0048] In one embodiment of the present invention, the experience is derived from the record of historical action execution.

[0049] In one embodiment of the present invention, the reward The calculation formula is as follows:

[0050] in is the expected return, is the cost of implementing the risk response strategy, is the effect of the risk response strategy. If the implementation of the risk response strategy successfully reduces the risk, then is 1, otherwise it is 0. The amount of resources consumed in implementing risk response strategies, , , It is the first, second, and third weight coefficient of the reward.

[0051] In one embodiment of the present invention, It is a comprehensive indicator of various resources consumed, and the calculation method is as follows:

[0052] in is the time required to execute the strategy, It is the product of the number of people involved and the number of working hours. is the actual amount of funds consumed, is the equipment and technical resources used, , , , It is the first, second, third and fourth weight coefficient of the comprehensive indicator.

[0053] In at least one embodiment of the present invention, a risk control assessment system based on multi-dimensional perception and enterprise data quantification is provided, which is used to store computer-readable instructions, and when the computer-readable instructions are read, it can execute the aforementioned risk control assessment method based on multi-dimensional perception and enterprise data quantification.

[0054] The above describes an embodiment of the present invention, but this embodiment is not limited to the above-mentioned specific implementation mode. The above-mentioned specific implementation mode is merely illustrative and not restrictive. Under the guidance of this embodiment, ordinary technicians in this field can also make more forms of equivalent embodiments, all of which are within the protection of this embodiment.

Claims

1. A risk control assessment method based on multi-dimensional perception and enterprise data quantification, characterized in that: include: Step 100, collecting enterprise data, which is collected once at each time sampling point, including department information, supplier information, customer information and external environment information; Step 200, constructing graph structure data, the graph structure data includes nodes and edges connecting the nodes, a node represents a department, supplier, customer or external environment; there are edges between a department and its departments with which it has project and process associations, there are edges between a department and its subordinate departments, there are edges between a department and its suppliers, there are edges between a department and its customers, and there are edges between all departments and the external environment; The node characteristics of the node representing the department are equal to the department characteristics, the node characteristics of the node representing the supplier are equal to the supplier characteristics, the node characteristics of the node representing the customer are equal to the customer characteristics, and the node characteristics of the node representing the external environment are equal to the external environment characteristics; Department characteristics, supplier characteristics, customer characteristics and external environment characteristics are obtained through feature engineering; Encode the enterprise data of each node into first-order sequence data, the first-order sequence data includes n first-order sequence units, and the t-th first-order sequence unit represents the enterprise data of a node at the t-th time sampling point; Step 300, inputting the graph structure data into a propagation model, the propagation model comprising a first hidden layer, a second hidden layer and a first output layer, the propagation model outputting a result representing a transmission path of enterprise risk factors; Step 400, encode the enterprise risk factor transmission path into No. 2 sequence data; the No. 2 sequence data includes B No. 2 sequence units sorted by time, the zth No. 2 sequence unit represents the risk path collected at the zth moment, and the graph structure data is input into the decision model. The decision model includes a third hidden layer, a feature fusion layer, a fourth hidden layer, a sixth hidden layer and a second output layer. The decision model output represents the result of the risk response strategy.

2. According to claim 1, a risk control assessment method based on multi-dimensional perception and enterprise data quantification is characterized in that: The time sampling point refers to the moment when data is collected at a predetermined time interval. The default value of the predetermined time interval is 24 hours.

3. The risk control assessment method based on multi-dimensional perception and enterprise data quantification according to claim 1 is characterized in that: Department information includes basic department information, subordinate departments, financial information, human resources information, and department process information; Supplier information includes basic information of suppliers and supplier's supply information; Customer information includes basic customer information and purchase information; External environmental information includes market intelligence information, social media trend information, and news report information.

4. The risk control assessment method based on multi-dimensional perception and enterprise data quantification according to claim 1 is characterized in that: In the propagation model, the first hidden layer is input with sequence data No. 1, the first hidden layer outputs the first hidden state to the second hidden layer, the second hidden layer also inputs graph structure data, the second hidden layer outputs the second hidden state to the first output layer and the feature fusion layer, and the first output layer outputs the result representing the transmission path of the enterprise risk factors.

5. The risk control assessment method based on multi-dimensional perception and enterprise data quantification according to claim 4 is characterized in that: The loss function calculation method of the propagation model is: ； in, represents the propagation loss value, is the number of propagation samples, is the path length of each propagation sample, is the total number of department nodes, Indicates The propagation sample is in The actual propagation node in step 1 is If it is a department node, its value is 1, otherwise it is 0; The propagation model predicts The propagation sample is in The propagation node is The probability of a department node, Represents the logarithmic function with the natural constant e as the base.

6. The risk control assessment method based on multi-dimensional perception and enterprise data quantification according to claim 1 is characterized in that: In the decision model, the first sequence data is input into the third hidden layer, the third hidden layer outputs the third hidden state to the feature fusion layer, the second sequence data is input into the fourth hidden layer, the fourth hidden layer outputs the fourth hidden state to the feature fusion layer, the feature fusion layer outputs the first fusion state to the sixth hidden layer, the sixth hidden layer also inputs the graph structure data, the sixth hidden layer outputs the sixth hidden state to the second output layer, and the second output layer outputs the result representing the risk response strategy.

7. The risk control assessment method based on multi-dimensional perception and enterprise data quantification according to claim 6 is characterized in that: Risk response strategy represents the action plan, resource allocation and responsibility allocation implemented in response to the transmission path of enterprise risk factors.

8. The risk control assessment method based on multi-dimensional perception and enterprise data quantification according to claim 7 is characterized in that: The steps of training the decision model include: Step 401, building a target network, and replicating a decision model of the target network; Step 402: randomly obtain an experience from the experience pool, the experience includes the state at time e , the action performed at time e , the state at time e+1 , perform the action Rewards ; The states at time e and e+1 represent the enterprise data at the e-th and e+1-th time input into the decision model respectively; Then calculate the state The target network output corresponds to The probability value ; ； express The maximum probability value of the target network output; Step 403, according to and Calculate the loss value and update the decision model; ； express The output of the decision model corresponds to The probability value of is the discount factor, a value between 0 and 1; Step 404, updating the target network every fixed number of network updates so that its parameters are the same as those of the current decision model; Step 405, until the number of collected experiences reaches a set threshold or the decision model converges, the step is terminated.

9. The risk control assessment method based on multi-dimensional perception and enterprise data quantification according to claim 8 is characterized in that: award The calculation formula is as follows: ； in is the expected return, is the cost of implementing the risk response strategy, is the effect of the risk response strategy. If the implementation of the risk response strategy successfully reduces the risk, then is 1, otherwise it is 0. , , are the first, second, and third weight coefficients of the reward, It is a comprehensive indicator of various resources consumed when implementing risk response strategies. The calculation method is as follows: ； in is the time required to execute the strategy, It is the product of the number of people involved and the number of working hours. is the actual amount of funds consumed, is the equipment and technical resources used, , , , It is the first, second, third and fourth weight coefficient of the comprehensive indicator.

10. A risk control assessment system based on multi-dimensional perception and enterprise data quantification, characterized in that: It is used to store computer-readable instructions, which, when read, can execute a risk control assessment method based on multi-dimensional perception and enterprise data quantification as described in any one of claims 1-9.

Images