Credit scoring simulation method and device based on large model

Abstract

This specification provides one or more embodiments of a credit scoring simulation method and apparatus based on a large model. The method uses a semantic large model to semantically parse target information input by a target user to obtain the target user's intent features. Through a pre-generated credit scoring knowledge graph and the target user's historical credit data, multiple credit simulation paths corresponding to the intent features are generated. After generating multiple credit simulation paths that meet the user's intent, a target credit simulation path determined by the target user based on the multiple credit simulation paths is used in conjunction with the credit scoring knowledge graph in a large inference model to perform scoring simulation, generating first simulated change trend information of the credit score corresponding to the target credit simulation path over a future time period.

Description

Technical Field

[0001] This specification relates to the technical field of credit scoring simulation based on large models, and more particularly to a method and apparatus for credit scoring simulation based on large models. Background Technology

[0002] Currently, various credit scoring systems mainly use preset static rules or machine learning models to score users' existing behavioral data. However, for users, this function can only provide their current credit score and cannot simulate future credit scores, thus preventing users from obtaining effective advice when planning for future credit scores.

[0003] While some related technologies offer suggestions for improving credit scores by asking questions of the model, these suggestions are often too generalized, offering little direct relevance to users. Furthermore, users cannot accurately predict future trends in their credit scores from these suggestions. Therefore, providing users with a more meaningful and referential credit scoring simulation method is a pressing technical challenge. Summary of the Invention

[0004] In view of this, one or more embodiments of this specification provide a credit scoring simulation method and apparatus based on a large model.

[0005] To achieve the above objectives, one or more embodiments of this specification provide the following technical solutions:

[0006] According to a first aspect of one or more embodiments of this specification, a credit scoring simulation method based on a large model is proposed, comprising:

[0007] The system acquires target information input by the target user and performs semantic parsing on the target information based on a semantic big data model to obtain the target user's intent features; the intent features are used to characterize the target user's credit scoring goals in the future time period.

[0008] Based on a pre-generated credit scoring knowledge graph and the target user's historical credit data, multiple credit simulation paths corresponding to the intent features are generated; each credit simulation path includes one or more behavioral nodes; the credit scoring knowledge graph includes the association between behavioral nodes and credit scores.

[0009] Based on the target credit simulation path determined by the target user according to the multiple credit simulation paths, and combined with the credit scoring knowledge graph, a scoring simulation is performed in the inference big model to generate the first simulated change trend information of the credit score corresponding to the target credit simulation path in the future time period.

[0010] According to a second aspect of one or more embodiments of this specification, a credit scoring simulation apparatus based on a large model is proposed, comprising:

[0011] The acquisition module acquires target information input by the target user and performs semantic parsing on the target information based on a semantic big data model to obtain the target user's intent features; the intent features are used to characterize the target user's credit scoring target in the future time period.

[0012] The simulation path module generates multiple credit simulation paths corresponding to the intent features based on a pre-generated credit scoring knowledge graph and the target user's historical credit data. Each credit simulation path includes one or more behavioral nodes. The credit scoring knowledge graph includes a first association between a single behavioral node and a credit score, and a second association between the arrangement of multiple behavioral nodes and the credit score.

[0013] The reasoning module, based on the target credit simulation path determined by the target user according to the multiple credit simulation paths, and combined with the credit scoring knowledge graph, performs scoring simulation in the large reasoning model to generate the first simulated change trend information of the credit score corresponding to the target credit simulation path in the future time period.

[0014] According to a third aspect of one or more embodiments of this specification, an electronic device is provided, comprising:

[0015] processor;

[0016] Memory used to store processor-executable instructions;

[0017] The processor implements the method as described in the first aspect by running the executable instructions.

[0018] According to a fourth aspect of one or more embodiments of this specification, a computer-readable storage medium is provided that stores computer instructions thereon, which, when executed by a processor, implement the steps of the method as described in the first aspect.

[0019] According to a fifth aspect of one or more embodiments of this specification, a computer program product is provided, comprising: a computer program / instructions that, when executed by a processor, implement the method as described in the first aspect.

[0020] As can be seen from the above embodiments, the credit scoring simulation method and apparatus based on a large model provided in one or more embodiments of this specification perform semantic parsing on the target information input by the target user through a semantic large model to obtain the target user's intent features. This allows the model to obtain the target user's true intent, laying the foundation for improving the accuracy of subsequent credit scoring simulations. Then, based on a pre-generated credit scoring knowledge graph and the target user's historical credit data, multiple credit simulation paths corresponding to the intent features are generated. Each credit simulation path includes one or more behavioral nodes, and the credit scoring knowledge graph includes the association between behavioral nodes and credit scores. After generating multiple credit simulation paths that meet the user's intent, the target credit simulation path is further determined through the target user, allowing the user to edit or select multiple credit simulation paths, further ensuring that the target credit simulation path better meets the user's practical expectations. Finally, based on the credit scoring knowledge graph and the target credit simulation path, a scoring simulation is performed in the inference model to generate a first simulated trend information of the credit score corresponding to the target credit simulation path in the future time period. This first simulated trend information is then sent to the target user, allowing the user to intuitively perceive the trend of the credit score corresponding to the target credit simulation path in the future time period and evaluate the target simulation path based on this trend. In addition, when generating the trend of the credit score corresponding to the target credit simulation path, the inference model can also refer to the relationship between the behavioral nodes and credit scores contained in the credit scoring knowledge graph, further improving the accuracy of the credit score trend. Attached Figure Description

[0021] Figure 1 This is an exemplary embodiment of the architecture diagram of an application scenario for a credit scoring simulation method based on a large model.

[0022] Figure 2 This is a flowchart of a credit scoring simulation method based on a large model, provided as an exemplary embodiment.

[0023] Figure 3 This is a schematic diagram of a credit scoring knowledge graph provided in an exemplary embodiment.

[0024] Figure 4 This is a flowchart of another credit scoring simulation method based on a large model, provided in an exemplary embodiment.

[0025] Figure 5 This is a schematic diagram of the structure of a device provided in an exemplary embodiment.

[0026] Figure 6 This is a block diagram of a credit scoring simulation device based on a large model, provided as an exemplary embodiment. Detailed Implementation

[0027] To enable those skilled in the art to better understand the technical solutions in this specification, the technical solutions in the embodiments of this specification will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this specification, and not all embodiments. Based on the embodiments in this specification, all other embodiments obtained by those skilled in the art without creative effort should fall within the scope of protection of this specification.

[0028] The organizational information (including but not limited to organizational equipment information, organizational personal information, etc.) and data (including but not limited to data used for analysis, stored data, and displayed data) involved in this manual are all information and data authorized by the organization or fully authorized by all parties. Furthermore, the collection, use, and processing of such data must comply with the relevant laws, regulations, and standards of the relevant countries and regions, and corresponding operation portals are provided for the organization to choose to authorize or refuse.

[0029] As described in the background section, current credit scoring systems primarily use preset static rules or machine learning models to score users' existing behavioral data. Therefore, they cannot simulate a user's future credit score. Furthermore, while some related technologies offer suggestions for improving credit scores by asking the model questions, these suggestions are often too general, such as advising users to repay on time or avoid defaults. Because these responses lack specific analysis tailored to the user's individual circumstances, users cannot directly glean valuable information from them. Additionally, these technologies cannot predict the future trend of a credit score along a particular simulation path, meaning that even if a user receives a suggestion to improve their credit score, they cannot predict the corresponding future trend of that score.

[0030] The credit scoring simulation method based on a large model provided in this specification uses a semantic large model to semantically parse the target information input by the target user, obtaining the target user's intent features. This allows the model to understand the target user's true intent, providing a foundation for the accuracy of subsequent credit scoring simulations. Then, based on a pre-generated credit scoring knowledge graph and the target user's historical credit data, multiple credit simulation paths corresponding to the intent features are generated. Each credit simulation path includes one or more behavioral nodes, and the credit scoring knowledge graph includes the relationships between behavioral nodes and credit scores. After generating multiple credit simulation paths that meet the user's intent, the target user further determines the target credit simulation path based on these paths. This allows the user to edit or select from multiple credit simulation paths, further ensuring that the target credit simulation path better matches the user's practical expectations. Finally, based on the credit scoring knowledge graph and the target credit simulation path, a scoring simulation is performed in the inference big model to generate first simulated change trend information of the credit score corresponding to the target credit simulation path in the future time period. This first simulated change trend information is then sent to the target user, allowing the user to intuitively perceive the change trend of the credit score corresponding to the target credit simulation path in the future time period and take corresponding measures based on this change trend. Furthermore, when generating the change trend of the credit score corresponding to the target credit simulation path, the inference big model can also refer to the association between behavioral nodes and credit scores contained in the credit scoring knowledge graph, further improving the accuracy of the credit score change trend. The credit scoring simulation method based on a big model provided in this specification offers users a method for simulating credit paths and deriving the change trend of the credit score corresponding to the simulated path.

[0031] Figure 1 This is a schematic diagram illustrating the architecture of an application scenario for a credit scoring simulation method based on a large model, provided as an exemplary embodiment. For example... Figure 1 As shown, the method may include a server 11, a network 12, and several electronic devices, such as a PC (Personal Computer) 13, a mobile phone 14, etc.

[0032] Server 11 can be a physical server containing a single host, or it can be a virtual server hosted in a host cluster. During operation, server 11 can run server-side programs for a specific application to implement the relevant functions of that application. For example, when server 11 runs a credit scoring simulation program based on a large model, it can implement the credit scoring simulation method based on the large model.

[0033] PC13 and mobile phone 14 are just some of the types of electronic devices that organizations can use. In reality, organizations can obviously also use electronic devices such as tablets, laptops, PDAs (Personal Digital Assistants), wearable devices (such as smart glasses, smartwatches, etc.), etc., and one or more embodiments in this specification do not limit this. During operation, the electronic device can run a client-side program of an application to achieve the relevant functions of that application. For example, when the electronic device runs a program related to the aforementioned credit scoring simulation method based on a large model, it can be implemented as a client of that related program. In some embodiments, the user can send target information to server 11 through the client and receive the first simulated trend information returned by server 11. The client application of the aforementioned program service can be launched and run on the electronic device. The client-side program can be a native application installed on the electronic device, or it can be a mini-program, quick app, or other similar form. Of course, when using web technologies such as HTML5 or similar, the relevant functions can be achieved through a page displayed by a browser. Here, the browser can be a standalone browser application or a browser module embedded in some applications.

[0034] As for the network 12 that enables interaction between electronic devices such as PC13 and mobile phone 14 and server 11, wired or wireless networks can be selected for communication based on the communication methods supported by the respective electronic devices. This specification does not impose any restrictions on this. For example, PC13 can support both wired and wireless communication, so it can use either wired or wireless networks as needed. Mobile phone 14 typically only supports wireless communication, so it can use a wireless network for communication.

[0035] refer to Figure 2 The flowchart below illustrates a credit scoring simulation method based on a large model, which includes the following steps:

[0036] S202, obtain the target information input by the target user, and perform semantic parsing on the target information based on the semantic big data model to obtain the intent features of the target user; the intent features are used to characterize the credit scoring target of the target user in the future time period.

[0037] The target user is the user who will be simulated for credit scoring. In some embodiments, the target user can be any user using the credit scoring simulation method based on a large model according to the embodiments of this specification. In some embodiments, the target information input by the user can be the target user's expectation of future credit scores. For example, a target information could be: "I hope to improve my credit score to 650 within the next three months." To accurately obtain the target user's intent features, the target information is semantically parsed using a semantic large model to better understand the user's true needs and avoid misunderstandings of intent. It should be noted that the intent features can be used to characterize the target user's credit scoring goals for a future time period. These credit scoring goals can be goals such as improving, maintaining, or lowering the target user's credit score. The aforementioned semantic large model can be any large language model (LLM) used for semantic parsing in related technologies, and is not limited thereto.

[0038] To more accurately obtain the intent characteristics of the target user, when it is determined that the target information is missing key information, a semantic big data model can proactively initiate a dialogue with the target user to obtain the missing key information. For example, the target information entered by a user may be "I want to improve my credit score", and the corresponding dialogue initiated by the semantic big data model to the user may be "How long do you hope to complete the credit score improvement? What is the specific range of credit score improvement?"

[0039] S204, based on the pre-generated credit scoring knowledge graph and the target user's historical credit data, generate multiple credit simulation paths corresponding to the intent features; the credit simulation path includes one or more behavioral nodes; the credit scoring knowledge graph includes the association between behavioral nodes and credit scores.

[0040] The target user's historical credit data can be historical data related to the user's credit that the target user has already generated. In some embodiments, this historical credit data may include historical credit behavior data and historical credit status data. The credit scoring knowledge graph can be a knowledge graph used to represent the association between various behavioral nodes and credit scores. In some embodiments of this specification, the credit scoring knowledge graph not only includes a first association between a single behavioral node and a credit score, but also a second association between the arrangement of multiple behavioral nodes and the credit score. This ensures that when generating a credit simulation path, not only can the combination of behavioral nodes in the credit simulation path be considered, but also the possible arrangements of the multiple behavioral nodes that make up the credit simulation path. For example, a credit simulation path consists of two behavioral nodes: borrowing and repayment. Through the aforementioned credit scoring knowledge graph, the inference model can distinguish between two different credit simulation paths: borrowing first and then repaying, and repaying first and then borrowing. (Reference) Figure 3 This is a credit scoring knowledge graph provided in the embodiments of this specification. Figure 3 One type includes three behavioral nodes: paying social security, borrowing money, and full repayment. Figure 3 The corresponding credit scoring knowledge graph not only includes the relationship between the three behavioral nodes of paying social security, borrowing, and full repayment and the credit score, but also the relationship between the different orders of these three behavioral nodes and the credit score. Therefore, in Figure 3 In this system, the three behavioral nodes of paying social security, borrowing money, and repaying the full amount can form 6 possible permutations and combinations. Each permutation and combination corresponds to a permutation order, and each permutation and combination is independently associated with the credit score.

[0041] After obtaining the credit scoring knowledge graph, the target user's intent features, and the target user's historical credit data, multiple credit simulation paths corresponding to the intent features can be generated based on the pre-generated credit scoring knowledge graph and the target user's historical credit data. Each credit simulation path can be obtained by combining one or more behavioral nodes. In some embodiments, prompt words can be generated based on the credit scoring knowledge graph, the intent features, and the target user's historical credit data and input into the inference model to generate multiple credit simulation paths corresponding to the intent features. It should be noted that the aforementioned inference model can be any large language model (LLM) used for logical reasoning in related technologies, and there is no limitation thereto. In some embodiments, multiple candidate behavioral nodes matching the target user's historical credit data can be determined from the credit scoring knowledge graph first, and then multiple credit simulation paths corresponding to the intent features can be generated based on the multiple candidate behavioral nodes and the credit scoring knowledge graph.

[0042] In some embodiments of this specification, the credit scoring simulation method based on a large model further includes:

[0043] Graph data is generated based on the historical credit data of multiple users; the graph data includes the initial association between behavioral nodes and credit scores;

[0044] The graph data is processed by the trained graph neural network to obtain an initial knowledge graph;

[0045] The initial knowledge graph is normalized by pre-defined rules to obtain the credit scoring knowledge graph that conforms to the pre-defined rules.

[0046] To accurately obtain a credit scoring knowledge graph, historical credit data from multiple users is first acquired. In some embodiments, this historical credit data can be obtained from various credit scoring-related systems. After acquiring the historical credit data from multiple users, graph data is generated based on this data. This graph data includes the initial association between behavioral nodes and credit scores. In some embodiments, the graph data may also include the initial association between the permutation of multiple behavioral nodes and their respective credit scores. In some embodiments, prompt words can be generated based on the historical credit data from multiple users, and then these prompt words are input into a large language model to generate graph data. In some embodiments, graph data can be generated based on an expert experience base and the historical credit data from multiple users. The expert experience base may include various empirical rules for generating graph data. To ensure greater accuracy in the final credit scoring knowledge graph, the first association between a single behavior node and the credit score, and the second association between the order of multiple behavior nodes and the credit score, a trained graph neural network is used to process the graph data after it is obtained, resulting in an initial knowledge graph. This graph neural network, through processing the graph data, can improve the accuracy of the associations between a single behavior node and the credit score, as well as the order of multiple behavior nodes and the credit score. In some embodiments, any type of graph neural network (GNN) can be selected to process the graph data; there is no limitation on this.

[0047] To further improve the reliability of the credit scoring knowledge graph and avoid overfitting of some relationships within the knowledge graph by the graph neural network, after obtaining the initial knowledge graph through the graph neural network, it can be further normalized using preset rules to obtain the credit scoring knowledge graph that conforms to the preset rules. For example, a preset rule could be set such that the increase in credit score by a single behavior node cannot exceed a certain threshold, to prevent unreasonable relationships from appearing in the initial knowledge graph. It should be noted that the specific preset rules can be set as needed and are not limited thereto.

[0048] In some embodiments, when training a graph neural network, graph data generated from the historical credit data of multiple users can be input as graph sample data into the graph neural network. The graph neural network outputs a corresponding sample knowledge graph, from which various relationships are extracted. These extracted relationships are then validated using multiple methods, and the graph neural network is adjusted based on the validation results until the relationships in the output sample knowledge graph meet the training requirements. In some embodiments, the various methods for validating the extracted relationships can be selected as needed. For example, some predefined constraints can be set, and these constraints can be used to determine whether the extracted relationships meet these constraints. Alternatively, the extracted relationships, along with the historical data used to generate the graph sample data, can be input into a large language model, which can then determine whether the extracted relationships are reasonable and accurate. Alternatively, counterfactual reasoning can be applied to the extracted relationships, and then the graph neural network can determine whether the results are reasonable. For example, if a behavioral node indicates a loan default, and its corresponding credit score is a decrease, then the counterfactual reasoning would be that a behavioral node indicates a loan not defaulted. If, in this case, the graph neural network outputs that the credit score corresponding to the behavioral node with no loan default is a decrease, it indicates that the graph neural network has not truly learned the relationship between the behavioral node and the credit score, and further training is needed. To improve the efficiency of training the graph neural network, residual analysis can be used to identify multiple target relationships from each relationship where the residual between the predicted credit change and the actual change is greater than a preset threshold. Then, the common behaviors among these multiple target relationships can be analyzed to find the causes of errors, allowing for targeted adjustments to the graph neural network.

[0049] In some embodiments of this specification, based on a pre-generated credit scoring knowledge graph and the target user's historical credit data, multiple credit simulation paths corresponding to the intent features are generated, including:

[0050] Based on the credit scoring knowledge graph and the target user's historical credit data, prompt words are generated and input into the reasoning model to determine multiple candidate behavior nodes that match the target user's historical credit data from the credit scoring knowledge graph.

[0051] Based on the candidate behavior nodes, the intent features, and the credit scoring knowledge graph, prompt words are generated and input into the reasoning model to generate multiple credit simulation paths corresponding to the intent features.

[0052] To more accurately generate multiple credit simulation paths corresponding to the intent features, prompt words can first be generated based on the credit scoring knowledge graph and the target user's historical credit data, and then input into the inference model. This allows for the identification of multiple candidate behavioral nodes from the credit scoring knowledge graph that match the target user's historical credit data. These candidate behavioral nodes, because they match the target user's historical credit data, are considered behavioral nodes that are more easily implemented by the target user. After determining the candidate behavioral nodes, prompt words are further generated based on the candidate behavioral nodes, the intent features, and the credit scoring knowledge graph, and then input into the inference model to generate multiple credit simulation paths corresponding to the intent features. It should be noted that the candidate behavioral nodes help the inference model determine which behavioral nodes can be used in the generated multiple credit simulation paths, and the intent features and the credit scoring knowledge graph help the inference model determine how to combine and arrange the candidate behavioral nodes to satisfy the target user's final intent.

[0053] In some embodiments of this specification, prompt words are generated based on the credit scoring knowledge graph and the target user's historical credit data, and then input into a large-scale inference model to determine multiple candidate behavioral nodes from the credit scoring knowledge graph that match the target user's historical credit data, including:

[0054] Determine the target user's historical credit behavior data from the target user's historical credit data;

[0055] By performing counterfactual deduction on the historical credit behavior data, reverse behavior data corresponding to the historical credit behavior data is obtained;

[0056] Based on the reverse behavior data, the credit scoring knowledge graph, and the target user's historical credit data, prompt words are generated and input into the reasoning model to determine multiple candidate behavior nodes that match the target user's historical credit data from the credit scoring knowledge graph.

[0057] Considering that for a target user, to change their current credit status, besides increasing the intensity of currently taken actions, they can also change their current credit status by avoiding certain actions or taking actions contrary to their current actions. Therefore, to more comprehensively identify multiple alternative behavioral nodes matching the target user's historical credit data, the historical credit behavior data of the target user can be determined first from the target user's historical credit data. Then, counterfactual deduction can be performed on the historical credit behavior data to obtain the reverse behavior data corresponding to the historical credit behavior data. For example, if a historical credit behavior data is a loan delinquency, then the corresponding reverse behavior data could be a loan not being delinquent. In some embodiments, a large language model can be used to perform counterfactual deduction on the historical credit behavior data, or a counterfactual deduction lookup table can be pre-set. This counterfactual deduction lookup table includes various historical credit behavior data and the corresponding reverse behavior data for each type of historical credit behavior data. When performing counterfactual deduction on a certain historical credit behavior data, the counterfactual deduction lookup table can be directly referenced to find the corresponding reverse behavior data. After obtaining the reverse behavior data corresponding to the historical credit behavior data, prompt words can be generated based on the reverse behavior data, the credit scoring knowledge graph, and the target user's historical credit data, and then input into the inference model to determine multiple candidate behavior nodes that match the target user's historical credit data from the credit scoring knowledge graph. It should be noted that, in the embodiments of this specification, the multiple candidate behavior nodes that match the target user's historical credit data include not only behavior nodes that match the target user's historical credit behavior data, but also behavior nodes that match the target user's reverse behavior data.

[0058] In some embodiments of this specification, the credit scoring knowledge graph also includes the credit delay time, credit peak time, and credit half-life corresponding to each behavioral node.

[0059] Considering that many knowledge graphs are currently static, and given that a user's action does not immediately affect their credit score, the impact of certain user actions on credit scores has a certain delay. For example, a user's repayment of a loan typically reflects in their credit score after 7 days. Furthermore, the impact of certain user actions on credit scores changes over time, and may even diminish. For instance, a one-time payment only increases the credit score in the short term, and if the user does not update their credit score for an extended period, the score will revert to its pre-payment level. Therefore, in this embodiment of the specification, the credit scoring knowledge graph includes parameters such as credit delay time, credit peak time, and credit half-life for each behavior node. This allows the generation of a credit simulation path based on the knowledge graph to fully consider the impact of each behavior node's changes over time on the final credit score, further improving the matching degree between the credit simulation path and the target user's intent characteristics.

[0060] In some embodiments of this specification, the credit scoring simulation method based on a large model further includes:

[0061] In response to the determination that the amount of historical credit data of the target user is less than a preset amount;

[0062] Obtain the attribute information of the target user, and determine the alternative user that matches the target user from multiple users based on the attribute information;

[0063] The target user's historical credit data is obtained based on the historical credit data of the alternative user.

[0064] Considering that some users, such as new employees, have limited credit history and historical credit data, it is impossible to accurately recommend a credit simulation path to them. To improve the accuracy of recommending credit simulation paths to new users, this embodiment of the specification uses the target user's attribute information to determine a suitable substitute user from multiple users. This substitute user shares the same attribute information as the target user, such as user type, tags, or user profile. After determining the substitute user, the target user's historical credit data can be obtained from the substitute user's historical credit data; that is, the substitute user's historical credit data is used as the target user's historical credit data. It should be noted that the aforementioned preset data volume can be set as needed and is not limited thereto.

[0065] In some embodiments of this specification, the credit scoring simulation method based on a large model further includes:

[0066] The system recommends the multiple credit simulation paths to the target user and, in response to the target user's target operation on the multiple credit simulation paths, determines the target credit simulation path based on the target operation.

[0067] After generating multiple credit simulation paths corresponding to the intent features, these paths can be further recommended to the target user so that the target user can select a suitable one. The target user can determine the target credit simulation path through a target operation on each of the multiple credit simulation paths. That is, when a target user's target operation on any of the multiple credit simulation paths is received, the target credit simulation path can be determined based on that operation. It should be noted that, considering that in some cases, the multiple credit simulation paths recommended to the target user may not be satisfactory for the current target user, the target operation is not limited to a selection operation from multiple credit simulation paths. In some embodiments of this specification, the target operation includes a selection operation to determine a credit simulation path from the multiple credit simulation paths, and an adjustment operation to modify the credit simulation path determined by the selection operation. The adjustment operation includes deleting or modifying any behavioral node included in the credit simulation path determined by the selection operation, or adding a new behavioral node to the credit simulation path determined by the selection operation.

[0068] In some embodiments of this specification, the credit scoring simulation method based on a large model further includes:

[0069] The priority of each credit simulation path is determined based on the target user's historical credit data and preset priority rules.

[0070] Based on the priority of each credit simulation path, the multiple credit simulation paths are recommended to the target user.

[0071] Considering that different credit simulation paths can be used to achieve the intended characteristics of the target user, and that the execution difficulty and cost-benefit ratio of different credit simulation paths vary for the target user, when recommending multiple credit simulation paths to the target user, the priority of each credit simulation path can be determined first based on the target user's historical credit data and preset priority rules. Then, the order in which the multiple credit simulation paths are recommended to the target user can be adjusted according to the priority of each path. In some embodiments, higher-priority credit simulation paths can be placed earlier in the order. It should be noted that the preset priority rules can be set as needed. For example, the preset priority rules can be set with reference to the user's execution difficulty and / or cost-benefit ratio. In one example, the preset priority rules may include the priorities of each credit simulation path corresponding to different historical credit data.

[0072] S206, based on the target credit simulation path determined by the target user according to the multiple credit simulation paths, and combined with the credit scoring knowledge graph, a scoring simulation is performed in the inference big model to generate the first simulated change trend information of the credit score corresponding to the target credit simulation path in the future time period.

[0073] After determining the target credit simulation path for the target user, to allow the target user to more intuitively understand the future trend of the credit score corresponding to the target credit simulation path, a scoring simulation can be performed in the inference model based on the credit scoring knowledge graph and the target credit simulation path to generate first simulated trend information of the credit score corresponding to the target credit simulation path in the future time period. In some embodiments, the first simulated trend information can also be sent to the target user for reference. It should be noted that the aforementioned future time period can be a future time period understood through target intent features, such as the next three months. The aforementioned first simulated trend information can be represented in the form of tables, text, or graphs, and there is no limitation on this.

[0074] In some embodiments of this specification, the credit scoring simulation method based on a large model further includes:

[0075] Obtain a preset knowledge base; the preset knowledge base includes multiple simulated path samples and the principle explanation results corresponding to each simulated path sample;

[0076] Based on the preset knowledge base and the target credit simulation path, prompt words are generated and input into the semantic big model to generate principle explanation information corresponding to the first simulated change trend information.

[0077] To better assist target users in understanding the correlation between the target credit simulation path and the first simulated trend information, in this embodiment, while sending the first simulated trend information to the target user, prompt words can also be generated based on the preset knowledge base and the target credit simulation path and input into the semantic big model to generate principle explanation information corresponding to the first simulated trend information. It should be noted that the preset knowledge base can be pre-generated, including multiple simulation path samples and a knowledge base of principle explanation results corresponding to each simulation path sample. This preset knowledge base helps the semantic big model better understand how to generate principle explanation information corresponding to the first simulated trend information, and the specific format and style of the generated principle explanation information. In some embodiments, after generating the principle explanation information corresponding to the first simulated trend information, the principle explanation information can be directly sent to the target user, or it can be sent to the target user after receiving an explanation request; there is no limitation on this.

[0078] In some embodiments of this specification, the method further includes:

[0079] The principle explanation information is standardized by using preset rules to obtain principle explanation information that conforms to the preset rules.

[0080] To further standardize the generated principle explanation information corresponding to the first simulated trend information and avoid obvious logical errors that could lead target users to question the credibility of the first simulated trend information, this embodiment of the specification, after generating the principle explanation information corresponding to the first simulated trend information through a semantic big model, further standardizes the principle explanation information using preset rules to obtain principle explanation information that conforms to the preset rules. It should be noted that in some embodiments, the preset rules can also be used to standardize a preset knowledge base, further improving the standardization of the principle explanation information generated by the semantic big model corresponding to the first simulated trend information.

[0081] In some embodiments of this specification, the credit scoring simulation method based on a large model further includes:

[0082] Identify the recommended credit simulation path with the highest priority from multiple credit simulation paths;

[0083] When the target credit simulation path is different from the suggested credit simulation path, a scoring simulation is performed in the inference model based on the credit scoring knowledge graph and the suggested credit simulation path to generate a second simulated change trend information of the credit score corresponding to the suggested credit simulation path in the future time period.

[0084] The second simulated trend information is sent to the target user.

[0085] Considering that, under the influence of various factors, the target credit simulation path chosen by the target user may differ from the suggested credit simulation path with the highest priority obtained through priority comparison, this specification, in an embodiment where the target credit simulation path differs from the suggested credit simulation path, performs a scoring simulation in the inference model based on the credit scoring knowledge graph and the suggested credit simulation path. This generates a second simulated trend information of the credit score corresponding to the suggested credit simulation path over a future time period, and sends this second simulated trend information to the target user so that the target user can directly compare the first and second simulated trend information. In some embodiments, to help the target user intuitively understand the impact of the target credit simulation path on their original credit status, the inference model can further process the target user's historical credit data and credit scoring knowledge graph to generate a third simulated trend information of the credit score corresponding to the target user's original credit path over a future time period, and then sends this third simulated trend information to the target user.

[0086] In some embodiments of this specification, the credit scoring simulation method based on a large model further includes:

[0087] The target preset behavior interface corresponding to the target credit simulation path is determined from a plurality of preset behavior interfaces; wherein, each preset behavior interface corresponds to a behavior node;

[0088] The system recommends the target preset behavior interface to the target user and, in response to the triggering of the target preset behavior interface, jumps to the behavior execution platform corresponding to the target preset behavior interface; the behavior execution platform corresponding to the target preset behavior interface is used to execute the behavior node corresponding to the target preset behavior interface.

[0089] To facilitate direct implementation of the target credit simulation path by target users after the credit scoring simulation, multiple preset behavior interfaces can be pre-set. Each preset behavior interface corresponds to a behavior node, and these interfaces help users achieve their corresponding behavior nodes. Once the target credit simulation path is determined, the behavior nodes included in that path are identified, allowing the selection of the target preset behavior interface from among the multiple preset interfaces. This target preset behavior interface is then recommended to the target user. Upon receiving the interface, the user can choose to implement the corresponding behavior node. When the interface is triggered, the user's current interface is redirected to the corresponding behavior execution platform, which executes the behavior node. It should be noted that the specific display format of these preset behavior interfaces can be customized as needed; for example, they can be links or touch-sensitive components, without limitation.

[0090] In some embodiments of this specification, the credit scoring simulation method based on a large model further includes:

[0091] Obtain the target execution progress of the target user's behavior node in the behavior execution platform corresponding to the target preset behavior interface;

[0092] In response to determining that the target execution progress is less than the expected execution progress corresponding to the target credit simulation path, a credit progress reminder message is sent to the target user.

[0093] To remind the target user to complete the target credit simulation path on time, in this embodiment, the target execution progress of the target user's behavior node in the behavior execution platform corresponding to the target preset behavior interface can be further obtained. When the target execution progress is less than the expected execution progress corresponding to the target credit simulation path, a credit progress reminder message can be sent to the target user so that the target user can complete the target credit simulation path as soon as possible according to the expected execution progress based on the reminder message. It should be noted that in some implementations, the expected execution progress can be set in advance for different credit simulation paths as needed, thereby determining the expected execution progress corresponding to the target credit simulation path. In some embodiments, the expected execution progress corresponding to the target credit simulation path can also be determined by the inference big model based on the target user's intent characteristics, which is not limited thereto.

[0094] In some embodiments of this specification, the credit scoring simulation method based on a large model further includes:

[0095] Obtain the target execution progress of the target user's behavior node in the behavior execution platform corresponding to the target preset behavior interface, and the credit score of the target user corresponding to the target execution progress;

[0096] The credit score knowledge graph is adjusted based on the target execution progress and the credit score of the target user corresponding to the target execution progress.

[0097] To further improve the accuracy of the credit scoring knowledge graph, the actual credit scores obtained by target users when they actually implement the target credit simulation path can be collected. That is, the target execution progress of the target user in the behavior execution platform corresponding to the target preset behavior interface is obtained, as well as the target user's credit score corresponding to the target execution progress. Then, the credit scoring knowledge graph is adjusted according to the target execution progress and the target user's credit score corresponding to the target execution progress, thereby further improving the accuracy of the credit scoring knowledge graph.

[0098] refer to Figure 4 This is another credit scoring simulation method based on a large model provided in the embodiments of this specification. The method includes the following steps:

[0099] S402, Obtain target information input by the target user.

[0100] In some embodiments, the target information may include the target user's credit score target for a future time period. For example, a target information may be: "I want to improve my credit score to 600 within the next three months."

[0101] S404 identifies the intent features of the target user through a semantic big data model.

[0102] A large semantic model can more accurately identify the true intent of target users, providing a foundation for accurate subsequent recommendation and credit simulation paths. It should be noted that this large semantic model can be any large language model used for semantic parsing in related technologies.

[0103] S406, Obtain the target user's historical credit data.

[0104] Historical credit data can include historical credit behavior data and historical credit status data. To ensure the accuracy of the historical credit data obtained for a target user, it is generally obtained directly through various credit scoring systems, rather than directly from the user's input information.

[0105] S408 generates multiple credit simulation paths through a large inference model.

[0106] When generating multiple credit simulation paths using a large-scale inference model, prompts can be generated based on the credit scoring knowledge graph, the target user's historical credit data, and the intent features, and then input into the large-scale inference model to generate multiple credit simulation paths corresponding to the intent features. In some embodiments, the generated prompts primarily serve to instruct the large-scale inference model to generate multiple credit simulation paths, as well as specifying the conditions and requirements for generation. For example, a generated prompt might be: "Please generate multiple credit simulation paths matching the intent features based on the following credit scoring knowledge graph, the target user's historical credit data, and intent features."

[0107] S410 determines the target credit simulation path based on the target user's target operation.

[0108] The target user's target operation may include a selection operation to determine a credit simulation path from the multiple credit simulation paths, and an adjustment operation to modify the credit simulation path determined by the selection operation; the adjustment operation may include deleting or modifying any behavior node contained in the credit simulation path determined by the selection operation, or adding a new behavior node to the credit simulation path determined by the selection operation.

[0109] S412, determine the credit score change trend of the target credit simulation path through the inference big model.

[0110] After determining the target credit simulation path, prompt words can be generated based on the target credit simulation path and the credit scoring knowledge graph and input into the reasoning model to generate the credit score change trend of the target credit simulation path. In some embodiments, the credit score change trend can be displayed through one or more combinations of text, tables, and line graphs.

[0111] S414 uses a semantic big data model to explain the changing trends in credit scores.

[0112] To better help users understand the causal relationship between credit score change trends and the target credit simulation path, prompt words can be generated based on the target credit simulation path and input into the semantic big data model to generate explanations of the underlying principles of credit score change trends.

[0113] The credit scoring simulation method based on a large model provided in this specification uses a semantic large model to semantically parse the target information input by the target user, obtaining the target user's intent features. This allows the model to understand the target user's true intent, providing a foundation for the accuracy of subsequent credit scoring simulations. Then, an inference large model processes the credit scoring knowledge graph, the intent features, and the target user's historical credit data to generate multiple credit simulation paths corresponding to the intent features. Each credit simulation path includes one or more behavioral nodes. The credit scoring knowledge graph includes a first association between a single behavioral node and the credit score, and a second association between the order of multiple behavioral nodes and the credit score. This ensures that the inference large model, when generating credit simulation paths, considers not only the impact of a single behavioral node on the credit score but also the impact of the order of multiple behavioral nodes, further improving the accuracy of the generated credit simulation paths. After generating multiple credit simulation paths that match the user's intent, these paths are recommended to the target user. A target credit simulation path is then determined based on the user's intended actions on these paths, allowing the user to edit or select the recommended path, further ensuring the target path aligns with the user's practical expectations. Finally, based on the credit scoring knowledge graph and the target credit simulation path, a scoring simulation is performed in the inference model. This generates a first simulated trend of credit score change corresponding to the target path over a future time period, which is then sent to the target user. This allows the user to intuitively perceive the future trend of credit score change corresponding to the target path and evaluate the path accordingly. Furthermore, when generating the trend of credit score change corresponding to the target path, the inference model can also refer to the second correlation between the arrangement of multiple behavioral nodes in the credit scoring knowledge graph and the credit score, further improving the accuracy of the credit score trend.

[0114] Figure 5 This is a schematic structural diagram of a device provided in an exemplary embodiment. For example... Figure 5As shown, device 500 mainly consists of a communication interface 502, a mechanism interface 504, a processor 506, and a data storage 508. These components are interconnected and communicate with each other via a method bus, network, or other connection mechanism 510. The communication interface 502 enables device 500 to communicate with other devices, access networks, and transmission networks via analog or digital modulation. For example, the communication interface 502 may include a chipset and antenna for wireless communication with a radio access network or access point. Furthermore, the communication interface 502 can be a wired interface such as Ethernet, Token Ring, or a USB port, or a wireless interface such as Wi-Fi, Bluetooth, Global Positioning System (GPS), or a wide-area wireless interface (e.g., WiMAX or LTE). Of course, the communication interface 502 can also support other forms of physical layer interfaces and standard or proprietary communication protocols. The communication interface 502 may also include multiple physical communication interfaces, such as Wi-Fi, Bluetooth, and wide-area wireless interfaces.

[0115] Mechanism interface 504 includes receiving mechanism input and providing output to the mechanism. Therefore, mechanism interface 504 may include input components such as a keypad, keyboard, touch-sensitive or presence-sensitive panel, computer mouse, trackball, joystick, microphone, still camera, and video camera, and output components such as a display screen (which may be combined with a touch-sensitive panel), CRT, LCD, LED, display using DLP technology, printer, and other similar devices known or developed in the future. Mechanism interface 504 may also generate auditory output via speakers, speaker jacks, audio output ports, audio output devices, headphones, and other similar devices known or developed in the future. In some embodiments, mechanism interface 504 may include software, circuitry, or other forms of logic capable of transmitting and receiving data from external mechanism input / output devices. Additionally or alternatively, device 500 may support remote access from other devices via communication interface 502 or another physical interface (not shown). Mechanism interface 504 may be configured to receive mechanism input, the position and movement of which may be indicated by an indicator or cursor described herein. Mechanism interface 504 may also be configured as a display device for rendering or displaying text fragments.

[0116] Processor 506 may contain one or more general-purpose processors and / or special-purpose processors.

[0117] Data storage 508 may include one or more volatile and / or non-volatile storage components and may be integrated wholly or partially with processor 506. Data storage 508 may include removable and non-removable components.

[0118] Processor 506 is capable of executing program instructions 518 (e.g., compiled or uncompiled program logic and / or machine code) stored in data storage 508 to perform the various functions described herein. Data storage 508 may contain a non-transitory computer-readable medium on which program instructions are stored, which, when executed by device 500, enable device 500 to perform any methods, processes, or functions disclosed in this specification and / or the accompanying drawings. Execution of program instructions 518 by processor 506 may result in processor 506 using data 512.

[0119] For example, program instructions 518 may include operating methods 522 (e.g., operating method kernels, device drivers, and / or other modules) installed on device 500 and one or more applications 520 (e.g., browsers, social applications, or game applications). Similarly, data 512 may include operating method data 516 and application data 514. Operating method data 516 is primarily accessible to operating method 522, while application data 514 is primarily accessible to one or more applications 520. Application data 514 may reside in file methods visible or hidden from the device 500.

[0120] Application 520 can communicate with operation method 522 through one or more application programming interfaces (APIs). These APIs help application 520 read and / or write application data 514, transmit or receive information via communication interface 502, receive or display information on mechanism interface 504, etc.

[0121] In some terminology, application 520 may be simply referred to as "app". Furthermore, application 520 can be downloaded to device 500 through one or more online app stores or app markets. However, applications can also be installed on device 500 in other ways, such as through a web browser or a physical interface on device 500 (e.g., a USB port).

[0122] Please refer to Figure 6 Credit scoring simulation devices based on large models can be applied to, for example... Figure 5 The device shown implements the technical solution described in this specification. This credit scoring simulation device based on a large model may include:

[0123] The acquisition module 602 acquires the target information input by the target user, and performs semantic parsing on the target information based on the semantic big data model to obtain the intent features of the target user; the intent features are used to characterize the credit scoring target of the target user in the future time period.

[0124] The simulation path module 604 generates multiple credit simulation paths corresponding to the intent features based on a pre-generated credit scoring knowledge graph and the target user's historical credit data; each credit simulation path includes one or more behavioral nodes; the credit scoring knowledge graph includes the association between behavioral nodes and credit scores.

[0125] The reasoning module 606, based on the target credit simulation path determined by the target user according to the multiple credit simulation paths, and combined with the credit scoring knowledge graph, performs scoring simulation in the large reasoning model to generate the first simulated change trend information of the credit score corresponding to the target credit simulation path in the future time period.

[0126] In some embodiments of this specification, the large-model-based credit scoring simulation device further includes a knowledge graph module for:

[0127] Graph data is generated based on the historical credit data of multiple users; the graph data includes the initial association between behavioral nodes and credit scores;

[0128] The graph data is processed by the trained graph neural network to obtain an initial knowledge graph;

[0129] The initial knowledge graph is normalized by pre-defined rules to obtain the credit scoring knowledge graph that conforms to the pre-defined rules.

[0130] In some embodiments of this specification, the simulated path module includes:

[0131] The alternative behavior node unit generates prompt words based on the credit scoring knowledge graph and the target user's historical credit data, and inputs them into the reasoning model to determine multiple alternative behavior nodes that match the target user's historical credit data from the credit scoring knowledge graph.

[0132] The generation unit generates prompt words based on the candidate behavior nodes, the intent features, and the credit scoring knowledge graph, and inputs them into the reasoning model to generate multiple credit simulation paths corresponding to the intent features.

[0133] In some embodiments of this specification, the alternative behavior node unit is specifically used for:

[0134] Determine the target user's historical credit behavior data from the target user's historical credit data;

[0135] By performing counterfactual deduction on the historical credit behavior data, reverse behavior data corresponding to the historical credit behavior data is obtained;

[0136] Based on the reverse behavior data, the credit scoring knowledge graph, and the target user's historical credit data, prompt words are generated and input into the reasoning model to determine multiple candidate behavior nodes that match the target user's historical credit data from the credit scoring knowledge graph.

[0137] In some embodiments of this specification, the device further includes a principle explanation module for:

[0138] Obtain a preset knowledge base; the preset knowledge base includes multiple simulated path samples and the principle explanation results corresponding to each simulated path sample;

[0139] Based on the preset knowledge base and the target credit simulation path, prompt words are generated and input into the semantic big model to generate principle explanation information corresponding to the first simulated change trend information.

[0140] In some embodiments of this specification, the apparatus further includes a specification module for:

[0141] The principle explanation information is standardized by using preset rules to obtain principle explanation information that conforms to the preset rules.

[0142] In some embodiments of this specification, the association between the behavior node and the credit score includes a first association between a single behavior node and the credit score, and a second association between the order of multiple behavior nodes and the credit score.

[0143] In some embodiments of this specification, the credit scoring knowledge graph also includes the credit delay time, credit peak time, and credit half-life corresponding to each behavioral node.

[0144] In some embodiments of this specification, the large-model-based credit scoring simulation device further includes a response module for:

[0145] In response to the determination that the amount of historical credit data of the target user is less than a preset amount;

[0146] Obtain the attribute information of the target user, and determine the alternative user that matches the target user from multiple users based on the attribute information;

[0147] The target user's historical credit data is obtained based on the historical credit data of the alternative user.

[0148] In some embodiments of this specification, the apparatus further includes a sorting module for:

[0149] The priority of each credit simulation path is determined based on the target user's historical credit data and preset priority rules.

[0150] Based on the priority of each credit simulation path, the multiple credit simulation paths are recommended to the target user.

[0151] In some embodiments of this specification, the apparatus further includes a suggestion module for:

[0152] Identify the recommended credit simulation path with the highest priority from multiple credit simulation paths;

[0153] When the target credit simulation path is different from the suggested credit simulation path, a scoring simulation is performed in the inference model based on the credit scoring knowledge graph and the suggested credit simulation path to generate a second simulated change trend information of the credit score corresponding to the suggested credit simulation path in the future time period.

[0154] The second simulated trend information is sent to the target user.

[0155] In some embodiments of this specification, the target user's operation of determining a target credit simulation path based on the multiple credit simulation paths includes a selection operation of determining one credit simulation path from the multiple credit simulation paths, and an adjustment operation of modifying the credit simulation path determined by the selection operation; the adjustment operation includes deleting or modifying any behavior node contained in the credit simulation path determined by the selection operation, or adding a new behavior node to the credit simulation path determined by the selection operation.

[0156] In some embodiments of this specification, the apparatus further includes an interface module for:

[0157] The target preset behavior interface corresponding to the target credit simulation path is determined from a plurality of preset behavior interfaces; wherein, each preset behavior interface corresponds to a behavior node;

[0158] The system recommends the target preset behavior interface to the target user and, in response to the triggering of the target preset behavior interface, jumps to the behavior execution platform corresponding to the target preset behavior interface; the behavior execution platform corresponding to the target preset behavior interface is used to execute the behavior node corresponding to the target preset behavior interface.

[0159] In some embodiments of this specification, the device further includes a progress indication module for:

[0160] Obtain the target execution progress of the target user's behavior node in the behavior execution platform corresponding to the target preset behavior interface;

[0161] In response to determining that the target execution progress is less than the expected execution progress corresponding to the target credit simulation path, a credit progress reminder message is sent to the target user.

[0162] In some embodiments of this specification, the device further includes an adjustment module for:

[0163] Obtain the target execution progress of the target user's behavior node in the behavior execution platform corresponding to the target preset behavior interface, and the credit score of the target user corresponding to the target execution progress;

[0164] The credit score knowledge graph is adjusted based on the target execution progress and the credit score of the target user corresponding to the target execution progress.

[0165] For ease of description, the above devices are described by dividing them into various modules or units based on their functions. Of course, when implementing one or more of these specifications, the functions of each module or unit can be implemented in the same or different software and / or hardware, or a module that performs the same function can be implemented by a combination of multiple sub-modules or sub-units, etc. The device embodiments described above are merely illustrative. For example, the division of units is only a logical functional division; in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another method, or some features may be ignored or not executed.

[0166] Based on the same concept as the above method, this specification also provides an electronic device, including: a processor; a memory for storing processor-executable instructions; wherein the processor executes the executable instructions to implement the steps of the credit scoring simulation method based on a large model as described in any of the above embodiments.

[0167] Based on the same concept as the methods described above, this specification also provides a computer-readable storage medium having computer instructions stored thereon that, when executed by a processor, implement the steps of the credit scoring simulation method based on a large model as described in any of the above embodiments.

[0168] Based on the same concept as the methods described above, this specification also provides a computer program product, including a computer program / instructions that, when executed by a processor, implement the steps of the credit scoring simulation method based on a large model as described in any of the above embodiments.

[0169] What those skilled in the art will understand is:

[0170] In this specification, the terms "comprising," "including," or any other variations thereof are intended to cover a non-exclusive inclusion, such that a process, method, product, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, product, or apparatus. Without further limitation, the presence of additional identical or equivalent elements in a process, method, product, or apparatus that includes said elements is not excluded.

[0171] In this specification, “a,” “an,” and “the” do not specifically refer to the singular, but may also include the plural.

[0172] In this specification, ordinal numbers such as "first," "second," etc., do not necessarily indicate order; they are often used to distinguish between objects. For example, "first server" and "second server" usually refer to two servers. To differentiate between these two servers, they are described as "first server" and "second server." Of course, sometimes these two servers may be the same server.

[0173] In this specification, unless explicitly stated otherwise, "receiving and sending data" does not necessarily mean direct receiving and sending; it can also mean indirect receiving and sending. For example, A receiving data sent by B can be understood as A directly receiving the data sent by B, or it can be understood as A indirectly receiving the data sent by B through other entities such as C. Similarly, B sending data to A can be understood as B sending the data directly to A, or it can be understood as B indirectly sending the data to A through other entities such as C. Here, C can be one entity, or it can be two or more entities.

[0174] In this specification, unless explicitly stated otherwise, the relationships between structures can be direct or indirect. For example, when describing "A is connected to B," unless it is explicitly stated that A and B are directly connected, it should be understood that A can be directly connected to B or indirectly connected to B. Similarly, when describing "A is on top of B," unless it is explicitly stated that A is directly above B (AB is adjacent and A is above B), it should be understood that A can be directly above B or indirectly above B (AB is separated by other elements, and A is above B). And so on.

[0175] This specification uses specific terms to describe embodiments thereof. Terms such as "an embodiment," "one embodiment," and / or "some embodiments" refer to a particular feature, structure, or characteristic associated with at least one embodiment of this specification. Therefore, it should be emphasized and noted that references to "an embodiment," "one embodiment," or "an alternative embodiment" in different locations throughout this specification do not necessarily refer to the same embodiment. Furthermore, those skilled in the art can combine and integrate the different embodiments or examples described herein, as well as the features of those different embodiments or examples, without contradiction.

[0176] Although one or more embodiments of this specification provide method steps as described in the embodiments or flowcharts, it is understood that the order of steps listed in the embodiments or flowcharts is only one of many possible execution orders and does not represent the only execution order. Therefore, when the claims involve method steps, any changes or adjustments to the order of such steps, or the parallelism between steps, are also within the scope of protection of the claims.

Claims

1. A credit scoring simulation method based on a large model, comprising: Obtain target information input by the target user, and perform semantic parsing on the target information based on a semantic big data model to obtain the target user's intent features; The intent feature is used to characterize the target user's credit score target in a future time period; Based on a pre-generated credit scoring knowledge graph and the target user's historical credit data, multiple credit simulation paths corresponding to the intent features are generated; each credit simulation path includes one or more behavioral nodes; the credit scoring knowledge graph includes the association between behavioral nodes and credit scores. Based on the target credit simulation path determined by the target user according to the multiple credit simulation paths, and combined with the credit scoring knowledge graph, a scoring simulation is performed in the inference big model to generate the first simulated change trend information of the credit score corresponding to the target credit simulation path in the future time period. Specifically, based on a pre-generated credit scoring knowledge graph and the target user's historical credit data, multiple credit simulation paths corresponding to the intent features are generated, including: Based on the credit scoring knowledge graph and the target user's historical credit data, prompt words are generated and input into the reasoning model to determine multiple candidate behavior nodes that match the target user's historical credit data from the credit scoring knowledge graph. Based on the candidate behavior nodes, the intent features, and the credit scoring knowledge graph, prompt words are generated and input into the reasoning model to generate multiple credit simulation paths corresponding to the intent features; Specifically, prompt words are generated based on the credit scoring knowledge graph and the target user's historical credit data, and then input into the inference model to determine multiple candidate behavioral nodes that match the target user's historical credit data from the credit scoring knowledge graph, including: Determine the target user's historical credit behavior data from the target user's historical credit data; By performing counterfactual deduction on the historical credit behavior data, reverse behavior data corresponding to the historical credit behavior data is obtained; Based on the reverse behavior data, the credit scoring knowledge graph, and the target user's historical credit data, prompt words are generated and input into the reasoning model to determine multiple candidate behavior nodes that match the target user's historical credit data from the credit scoring knowledge graph.

2. The method according to claim 1, further comprising: Graph data is generated based on the historical credit data of multiple users; the graph data includes the initial association between behavioral nodes and credit scores; The graph data is processed by the trained graph neural network to obtain an initial knowledge graph; The initial knowledge graph is normalized by pre-defined rules to obtain the credit scoring knowledge graph that conforms to the pre-defined rules.

3. The method according to claim 1, further comprising: Retrieve the preset knowledge base; The preset knowledge base includes multiple simulated path samples and the principle explanation results corresponding to each simulated path sample; Based on the preset knowledge base and the target credit simulation path, prompt words are generated and input into the semantic big model to generate principle explanation information corresponding to the first simulated change trend information.

4. The method according to claim 3, further comprising: The principle explanation information is standardized by using preset rules to obtain principle explanation information that conforms to the preset rules.

5. The method according to claim 1, wherein the association between the behavior node and the credit score includes a first association between a single behavior node and the credit score, and a second association between the order of multiple behavior nodes and the credit score.

6. The method according to claim 1, wherein the credit scoring knowledge graph further includes the credit delay time, credit peak time and credit half-life corresponding to each behavioral node.

7. The method according to claim 1, further comprising: The response is that the amount of historical credit data of the target user is less than a preset amount; Obtain the attribute information of the target user, and determine the alternative user that matches the target user from multiple users based on the attribute information; The target user's historical credit data is obtained based on the historical credit data of the alternative user.

8. The method according to claim 1, further comprising: The priority of each credit simulation path is determined based on the target user's historical credit data and preset priority rules. Based on the priority of each credit simulation path, the multiple credit simulation paths are recommended to the target user.

9. The method according to claim 8, further comprising: Identify the recommended credit simulation path with the highest priority from multiple credit simulation paths; When the target credit simulation path is different from the suggested credit simulation path, a scoring simulation is performed in the inference model based on the credit scoring knowledge graph and the suggested credit simulation path to generate a second simulated change trend information of the credit score corresponding to the suggested credit simulation path in the future time period. The second simulated trend information is sent to the target user.

10. The method according to claim 1, wherein the target user's target operation of determining a target credit simulation path based on the plurality of credit simulation paths includes a selection operation of determining a credit simulation path from the plurality of credit simulation paths, and an adjustment operation of modifying the credit simulation path determined by the selection operation; the adjustment operation includes deleting or modifying any behavior node contained in the credit simulation path determined by the selection operation, or adding a new behavior node to the credit simulation path determined by the selection operation.

11. The method according to claim 1, further comprising: The target preset behavior interface corresponding to the target credit simulation path is determined from a plurality of preset behavior interfaces; wherein, each preset behavior interface corresponds to a behavior node; The system recommends the target preset behavior interface to the target user and, in response to the triggering of the target preset behavior interface, jumps to the behavior execution platform corresponding to the target preset behavior interface; the behavior execution platform corresponding to the target preset behavior interface is used to execute the behavior node corresponding to the target preset behavior interface.

12. The method according to claim 11, further comprising: Obtain the target execution progress of the target user's behavior node in the behavior execution platform corresponding to the target preset behavior interface; In response to determining that the target execution progress is less than the expected execution progress corresponding to the target credit simulation path, a credit progress reminder message is sent to the target user.

13. The method according to claim 11, further comprising: Obtain the target execution progress of the target user's behavior node in the behavior execution platform corresponding to the target preset behavior interface, and the credit score of the target user corresponding to the target execution progress; The credit score knowledge graph is adjusted based on the target execution progress and the credit score of the target user corresponding to the target execution progress.

14. A credit scoring simulation device based on a large model, comprising: The acquisition module acquires target information input by the target user and performs semantic parsing on the target information based on a semantic big data model to obtain the intent features of the target user. The intent feature is used to characterize the target user's credit score target in a future time period; The simulation path module generates multiple credit simulation paths corresponding to the intent features based on a pre-generated credit scoring knowledge graph and the target user's historical credit data; each credit simulation path includes one or more behavioral nodes; the credit scoring knowledge graph includes the association between behavioral nodes and credit scores. The reasoning module, based on the target credit simulation path determined by the target user according to the multiple credit simulation paths, and combined with the credit scoring knowledge graph, performs scoring simulation in the reasoning big model to generate the first simulated change trend information of the credit score corresponding to the target credit simulation path in the future time period. The simulated path module includes: The alternative behavior node unit generates prompt words based on the credit scoring knowledge graph and the target user's historical credit data, and inputs them into the reasoning model to determine multiple alternative behavior nodes that match the target user's historical credit data from the credit scoring knowledge graph. The generation unit generates prompt words based on the candidate behavior nodes, the intent features, and the credit scoring knowledge graph, and inputs them into the reasoning model to generate multiple credit simulation paths corresponding to the intent features; Specifically, the candidate behavior node unit is used for: Determine the target user's historical credit behavior data from the target user's historical credit data; By performing counterfactual deduction on the historical credit behavior data, reverse behavior data corresponding to the historical credit behavior data is obtained; Based on the reverse behavior data, the credit scoring knowledge graph, and the target user's historical credit data, prompt words are generated and input into the reasoning model to determine multiple candidate behavior nodes that match the target user's historical credit data from the credit scoring knowledge graph.

15. An electronic device comprising: processor; Memory used to store processor-executable instructions; The processor executes the executable instructions to implement the method as described in any one of claims 1-13.

16. A computer-readable storage medium having stored thereon computer instructions that, when executed by a processor, implement the steps of the method as claimed in any one of claims 1-13.

17. A computer program product comprising: A computer program / instruction that, when executed by a processor, implements the method as described in any one of claims 1-13.

Images