Resource allocation method and device, program product and storage medium

Abstract

The invention discloses a resource allocation method and device, a program product and a storage medium, and the method comprises the steps: inputting target data into a first model, and obtaining the risk probability of the target data outputted by the first model; determining an initial risk type probability of the target product based on the risk probability and the target data; determining a target risk type probability of the target product based on the initial risk type probability and the target data; determining the risk type of the target product based on the target risk type probability; and allocating resources to events corresponding to the transaction data based on the risk types. Through application of the method and the device, the problem that the resource allocated to the event corresponding to the transaction data of the product is inaccurate due to the fact that the risk type of the product cannot be accurately judged in the related technology is solved, and the effects of accurately identifying the risk type of the product and accurately allocating the resource to the event corresponding to the transaction data of the product are achieved.

Description

Technical Field

[0001] This application relates to the field of computers, and more specifically, to a resource allocation method and apparatus, a program product, and a storage medium. Background Technology

[0002] In related technologies, traditional statistical models or simple machine learning algorithms are commonly used to assess the risk of transaction data for target products. These methods are usually limited to a single data source, such as the transaction history or basic customer information of a single merchant, and rarely consider comprehensive information from cross-departmental, cross-institutional, or even cross-industry data.

[0003] The risk assessment model for such target products mainly relies on linear analysis of historical data, thus failing to accurately determine the risk type of the product. Furthermore, the resource allocation mechanism is based on the aforementioned vague risk type classification, which in turn leads to resource waste or loss of risk control.

[0004] There is currently no solution to the above problems in the relevant technologies. Summary of the Invention

[0005] This application provides a resource allocation method and apparatus, program product, and storage medium to at least solve the problem in the related art where resources are inaccurately allocated to events corresponding to transaction data of a product due to the inability to accurately determine the risk type of the product.

[0006] According to one aspect of the embodiments of this application, a resource allocation method is provided, comprising: inputting target data into a first model to obtain a risk probability of the target data output by the first model, wherein the target data includes transaction data of a target product, the first model is used to determine the risk probability based on a target topology map, the target topology map being constructed based on data provided by N data providers, where N is an integer greater than 1; determining an initial risk type probability of the target product based on the risk probability and the target data; determining a target risk type probability of the target product based on the initial risk type probability and the target data; determining a risk type of the target product based on the target risk type probability; and allocating resources to events corresponding to the transaction data based on the risk type.

[0007] In an exemplary embodiment, the first model is trained as follows: The following objective operation is performed until the first initial model deployed on the server side uses the objective parameters and the first loss function of the first initial model satisfies a first threshold, and the currently obtained first initial model is determined as the first model: the initial parameters of the first initial model are sent to N data providers to instruct the N data providers to train their respective locally deployed second initial models based on the initial parameters, wherein each of the data providers' locally deployed second initial models is used to determine the risk probability of the product based on a first topology map, the first topology map being constructed based on the data from the data providers. The target topology map includes N first topology maps; N model parameters sent by the data providers are received to obtain N parameter differences, wherein the parameter differences are the differences between the first parameters of the second model obtained after training by the data providers and the initial parameters, and the second loss function of the trained second model satisfies a second threshold; the target parameters of the first initial model deployed on the server are determined based on the N parameter differences; when the first loss function of the first initial model deployed on the server does not satisfy the first threshold after adopting the target parameters, the first parameters are used as new initial parameters, and the target operation is performed again.

[0008] In an exemplary embodiment, determining the target parameters of the first model deployed on the server side based on N parameter differences includes: performing a first operation on the N parameter differences to obtain a second parameter, wherein the first operation is used to perform a weighted summation operation on the N parameter differences based on the amount of data provided by the N data providers.

[0009] In an exemplary embodiment, the first topology graph is constructed as follows: parsing multiple key information items in the acquired first dataset, wherein the key information items include the object's transaction behavior, asset information, basic information, the product corresponding to the object, and the industry in which the object is located; determining the business logic between the multiple key information items; and constructing the first topology graph based on the business logic and the multiple key information items.

[0010] In an exemplary embodiment, the target topology graph includes: nodes, wherein the nodes include entity nodes, derivative nodes, and auxiliary nodes; the entity nodes represent objects in a target scenario, the target scenario being a scenario using the target product, the objects including the target product; the derivative nodes represent factors influencing the risk probability of the target product; the auxiliary nodes represent the time and place of the transaction of the target product, and objects with implicit business relationships with the target product; edges connecting the nodes, wherein the edges include relationship edges, risk edges, and behavior edges; the relationship edges represent business relationships between the entity nodes; the risk edges represent risk propagation paths between the entity nodes and the derivative nodes, or risk edges represent risk propagation paths between the entity nodes and the auxiliary nodes; the behavior edges connect the entity nodes and derivative nodes generated due to the behavior of the objects represented by the entity nodes; and the weights of the edges.

[0011] In an exemplary embodiment, the weight of the edge is determined by the first model based on the initial weight of the edge and the degree of influence of the derived node or the auxiliary node connected by the edge on the risk probability of the entity node. The initial weight is obtained by calculating the product of a preset weight and a decay factor. The preset weight is set based on the type of the edge, and the decay factor is determined according to the generation time of the derived node or the auxiliary node.

[0012] In an exemplary embodiment, determining the target risk type probability of the target product based on the initial risk type probability and the target data includes: performing feature transformation operations on the initial risk type probability and the target data respectively to obtain a first feature and a second feature; fusing the first feature and the second feature to obtain a third feature; and inputting the third feature into a second model to obtain the target risk type probability output by the second model. The second model includes multiple decision trees, and is used to perform a second operation on the third feature and a third operation on multiple first values ​​obtained after performing the second operation. The second operation is used to determine the degree of matching between the third feature and the conditions represented by the leaf nodes of the decision trees using the multiple decision trees, obtaining multiple first values. The third operation is used to calculate the average of the multiple first values ​​to obtain the target risk type probability.

[0013] According to another aspect of the embodiments of this application, a resource allocation apparatus is also provided, comprising: an input module, configured to input target data into a first model to obtain a risk probability of the target data output by the first model, wherein the target data includes transaction data of a target product, and the first model is configured to determine the risk probability based on a target topology map, the target topology map being constructed based on data provided by N data providers, where N is an integer greater than 1; a first determining module, configured to determine an initial risk type probability of the target product based on the risk probability and the target data; a second determining module, configured to determine a target risk type probability of the target product based on the initial risk type probability and the target data; a third determining module, configured to determine a risk type of the target product based on the target risk type probability; and an allocation module, configured to allocate resources to events corresponding to the transaction data based on the risk type.

[0014] According to another aspect of the embodiments of this application, a computer-readable storage medium is also provided, wherein a computer program is stored therein, wherein the computer program is configured to perform the steps in any of the above method embodiments when executed by a processor.

[0015] According to another aspect of the embodiments of this application, a computer program product or computer program is provided, the computer program product or computer program including computer instructions stored in a computer-readable storage medium. A processor of a computer device reads the computer instructions from the computer-readable storage medium, and the processor executes the computer instructions, causing the computer device to perform the steps in any of the method embodiments described above.

[0016] According to another aspect of the embodiments of this application, an electronic device is also provided, including a memory and a processor, wherein the memory stores a computer program, and the processor is configured to perform the steps of any of the above method embodiments through the computer program.

[0017] This application inputs target data, including transaction data of the target product, into a first model. Then, a target topology graph is used to determine the risk probability of the target product. Because the constructed target topology graph fully utilizes information from different data providers, the first model can deeply explore the complex relationships between entities and their dynamic correlation with industry environment, macroeconomic factors, and other factors. This accurately determines the risk probability of the target data, providing a solid foundation for subsequently accurately determining the risk type of the target product. By determining the risk type probability of the target product twice, the final determined risk type of the target product is more accurate and can more accurately reflect the true risk status of the target product. Finally, based on the determined risk type, resources are allocated to the events involved in the corresponding transaction data. Through the preliminary risk assessment and type determination, it can be ensured that the resource allocation matches the risk type of the target product, avoiding resource waste and unreasonable allocation. This ensures that high-risk events receive sufficient attention and resource investment, while low-risk events reduce unnecessary resource consumption, achieving precise optimization and efficient utilization of resources. Therefore, it can solve the problem in related technologies where the inability to accurately determine the risk type of a product leads to inaccurate resource allocation for events corresponding to the product's transaction data, achieving the technical effect of accurately identifying the risk type of a product and accurately allocating resources to the events corresponding to the product's transaction data. Attached Figure Description

[0018] Figure 1 This is a schematic diagram illustrating an application scenario of a resource allocation method according to an embodiment of this application;

[0019] Figure 2 This is a flowchart illustrating a resource allocation method according to an embodiment of this application;

[0020] Figure 3 This is a schematic diagram of the structure of a target topology map constructed in a target scenario according to an embodiment of this application;

[0021] Figure 4 This is a flowchart illustrating a resource allocation method in a target scenario according to an embodiment of this application;

[0022] Figure 5 This is a structural block diagram of a resource allocation device according to an embodiment of this application. Detailed Implementation

[0023] To enable those skilled in the art to better understand the present application, the technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present application, and not all embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative effort should fall within the scope of protection of the present application. Furthermore, in the absence of conflict, the various embodiments and features in the embodiments of the present application can be arbitrarily combined with each other in principle.

[0024] It should be noted that the terms "first," "second," etc., in the specification, claims, and accompanying drawings of this application are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that the embodiments of this application described herein can be implemented in orders other than those illustrated or described herein. Furthermore, the terms "comprising" and "having," and any variations thereof, are intended to cover non-exclusive inclusion; for example, a process, method, system, product, or apparatus that comprises a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or apparatus.

[0025] According to one aspect of the embodiments of this application, a resource allocation method is provided. Optionally, in this embodiment, the above-described resource allocation method may be applied, but is not limited to, to... Figure 1 The hardware environment shown includes terminal device 102 and server 104. Server 104 can be connected to terminal device 102 via a network and can be used to provide services (e.g., application services, etc.) to terminal device 102 or clients installed on terminal device 102. A database can be set up on server 104 or independently of server 104 to provide data storage services for server 104.

[0026] The aforementioned network may include, but is not limited to, at least one of the following: wired network and wireless network. The aforementioned wired network may include, but is not limited to, at least one of the following: wide area network (WAN), metropolitan area network (MAN), and local area network (LAN). The aforementioned wireless network may include, but is not limited to, at least one of the following: Wireless Fidelity (WIFI) and Bluetooth. Terminal device 102 may be, but is not limited to, a personal computer (PC), mobile phone, tablet computer, etc. Server 104 may be, but is not limited to, a cloud server, server cluster, or other server types.

[0027] The resource allocation method of this application embodiment can be executed by server 104, terminal device 102, or jointly by server 104 and terminal device 102. Alternatively, the resource allocation method of this application embodiment can be executed by a client installed on terminal device 102.

[0028] Taking the resource allocation method in this embodiment as an example, which is executed by server 104, Figure 2 This is a flowchart illustrating a resource allocation method according to an embodiment of this application, such as... Figure 2 As shown, the process of this method may include the following steps:

[0029] Step S202: Input the target data into the first model to obtain the risk probability of the target data output by the first model. The target data includes transaction data of the target product. The first model is used to determine the risk probability based on the target topology map. The target topology map is constructed based on data provided by N data providers, where N is an integer greater than 1.

[0030] Optionally, the solution in this embodiment can be applied to business areas requiring high data consistency, such as banking, finance, insurance, and securities. Its application scope includes, but is not limited to, enterprise data center resource planning, dynamic adjustment of cloud service resources, financial transaction risk control, big data platform resource scheduling, and supply chain management optimization and upgrading. For example, when an e-commerce platform faces the pressure of processing massive orders during major shopping festivals, this method can, based on commodity transaction data and consumer behavior analysis, assess in real time the risk of stockouts, slow sales, and logistics delays for each commodity. It can then allocate corresponding processing resources to commodities and orders with different risk levels, such as increasing warehouse picking personnel, optimizing logistics routes, and strengthening after-sales service teams, ensuring efficient resource utilization and customer service experience during peak periods. As another example, in the loan approval process, this method can comprehensively analyze the borrower's financial situation, repayment history, industry background, and macroeconomic factors to provide a more accurate loan risk assessment, thereby helping financial institutions rationally allocate approval resources and reduce credit risk.

[0031] Optionally, the target data in this embodiment refers to specific data that requires risk assessment, typically including transaction data of the target product. This data can be structured (e.g., transaction amount, date, product characteristics, etc.) or unstructured (e.g., user reviews, social media mentions, etc.). In this embodiment, the target data should at least include transaction records of the target product, such as purchase volume, return records, payment delay information, etc.

[0032] Optionally, the first model in this embodiment is a risk assessment model built based on the target topology graph. It is trained using graph neural network technology and a federated learning framework, which can integrate information from multiple data providers to calculate the risk probability of the target data.

[0033] For example, the first model used in this embodiment can be a graph neural network (GNN). The target topology graph used in the first model is constructed based on the node feature matrix, adjacency matrix and initial edge weight parameters of the topology graphs constructed by N data providers. The input of the first model is the target data and the output is the risk probability of the target product represented by the target data.

[0034] Optionally, in this embodiment, the N data providers refer to multiple data sources, including but not limited to banks, credit rating agencies, e-commerce platforms, and logistics service providers. Each data provider contributes specific types of transaction data or other relevant information to jointly construct the target topology.

[0035] Optionally, the risk probability in this embodiment is the likelihood of target data under a specific risk type, determined by the first model, such as the probability of defaulting on a loan within the next year or the probability of returning a product. It is obtained based on the target topology analysis and reflects the risk level under the combined effect of multiple factors.

[0036] Optionally, the target product in this embodiment includes, but is not limited to: financial products (such as loans, insurance, and securities), physical goods (such as mobile phones, home appliances, and automobiles), and service items (such as telecommunications services, medical services, and education services). For example, in the financial field, the target product could be a loan. By analyzing the borrower's credit history, repayment behavior, and industry economic indicators, the default risk can be predicted, and collection resources can be allocated accordingly. In the retail industry, the target product could be a commodity. Based on consumer purchase history, return records, and market trends, the risk of unsold goods can be predicted, thereby optimizing inventory management and promotional strategies.

[0037] Step S204: Determine the initial risk type probability of the target product based on the above risk probabilities and the above target data;

[0038] Optionally, the initial risk type probability in this embodiment is the probability that the target product faces different risk types (such as high risk, medium risk, and low risk) based on the risk probability and target data. It serves as the basis for further determining the target risk type probability.

[0039] Optionally, in practical use, this step can be performed using a third model. That is, the aforementioned risk probabilities and target data are input into the third model to obtain the initial risk type probability of the target product output by the third model. For example, in a bank loan approval scenario, the borrower's credit data is used as the target data. The third model converts the credit data and the risk probabilities output by the first model into corresponding feature vectors. Then, the credit feature data and the risk probability vectors are concatenated and fused to obtain a fused vector. The fused vector is then used to determine the initial risk type probability of the borrower's loan (i.e., the aforementioned target product). The third model can be a deep neural network (DNN).

[0040] Step S206: Determine the target risk type probability of the target product based on the initial risk type probability and the target data.

[0041] Optionally, the target risk type probability in this embodiment is the final risk type probability obtained by adjusting the initial risk type probability based on more business rules and historical data. This probability more accurately reflects the true risk status of the target product.

[0042] Step S208: Determine the risk type of the target product based on the probability of the target risk type mentioned above;

[0043] Optionally, the risk type in this embodiment is one of several risk types that the target product may face, determined based on the probability of the target risk type. The classification of risk types helps in the formulation of subsequent resource allocation strategies. For example, in the scenario of bank loan approval, the risk type of the target product may be: very difficult to repay, relatively difficult to repay, or forgotten repayment.

[0044] Step S210: Allocate resources to the events corresponding to the above transaction data based on the above risk types.

[0045] Optionally, resource allocation in this embodiment refers to allocating appropriate resources to events corresponding to the transaction data of the target product based on the risk type. For example, in loan management, more review and collection resources are allocated to high-risk loans; in commodity sales, inventory strategies are adjusted and after-sales service support is enhanced for commodities with high return rates.

[0046] The resource allocation method in this application embodiment is explained below with reference to several optional examples. For example, in bank loan approval, the borrower's transaction data and background information are used as target data, and the probability of default risk is predicted through a first model. Based on the risk type, more stringent review resources, such as manual review, are allocated to high-risk loans, while the process is simplified and the approval speed is accelerated for low-risk loans, thereby improving the overall efficiency of post-loan management and asset security. Another example is that e-commerce platforms use commodity transaction data as target data to analyze return rates, stockout risks, etc. For example, for an electronic device, if the model predicts a high return rate, the platform can increase investment in after-sales service resources based on the probability of the target risk type, such as increasing customer service personnel and optimizing the return and exchange process to improve customer satisfaction and reduce the impact of negative reviews on the brand. Yet another example is that telecommunications operators can use users' consumption data and credit history as target data to assess the risk of account arrears. For users with low credit scores, operators can allocate collection resources in advance, such as sending reminder SMS messages and phone calls, to prevent arrears, reduce the pressure of later accounting processing, and protect good customer relationships.

[0047] In this embodiment, target data, including transaction data of the target product, is input into a first model. A target topology graph is then used to determine the risk probability of the target product. Because the constructed target topology graph fully utilizes information from different data providers, the first model can deeply explore the complex relationships between entities and their dynamic correlation with industry environment, macroeconomic factors, and other factors. This accurately determines the risk probability of the target data, providing a solid foundation for accurately determining the risk type of the target product. By determining the risk type probability of the target product twice, the final determined risk type is more accurate and can more accurately reflect the true risk status of the target product. Finally, based on the determined risk type, resources are allocated to the events involved in the corresponding transaction data. Through the preliminary risk assessment and type determination, it can be ensured that resource allocation matches the risk type of the target product, avoiding resource waste and unreasonable allocation. This ensures that high-risk events receive sufficient attention and resource investment, while low-risk events reduce unnecessary resource consumption, achieving precise optimization and efficient utilization of resources. Therefore, this solves the problem in related technologies where the inability to accurately determine the risk type of a product leads to inaccurate resource allocation for events corresponding to the product's transaction data, achieving the technical effect of accurately identifying the risk type of a product and accurately allocating resources to the events corresponding to the product's transaction data.

[0048] In an exemplary embodiment, the first model is trained as follows: The following objective operation is performed until the first initial model deployed on the server side uses the objective parameters and the first loss function of the first initial model satisfies a first threshold, and the currently obtained first initial model is determined as the first model: the initial parameters of the first initial model are sent to N data providers to instruct the N data providers to train their respective locally deployed second initial models based on the initial parameters, wherein each of the data providers' locally deployed second initial models is used to determine the risk probability of the product based on a first topology map, the first topology map being constructed based on the data from the data providers. The target topology map includes N first topology maps; N model parameters sent by the data providers are received to obtain N parameter differences, wherein the parameter differences are the differences between the first parameters of the second model obtained after training by the data providers and the initial parameters, and the second loss function of the trained second model satisfies a second threshold; the target parameters of the first initial model deployed on the server are determined based on the N parameter differences; when the first loss function of the first initial model deployed on the server does not satisfy the first threshold after adopting the target parameters, the first parameters are used as new initial parameters, and the target operation is performed again.

[0049] Optionally, in this embodiment, the first initial model deployed on the server side is an untrained model initially deployed on the central server, used to initiate the entire federated learning process. Its parameters are randomly initialized or set to preset values ​​for subsequent training and optimization.

[0050] Optionally, in this embodiment, the first loss function is used to measure the difference between the model's predicted value and the actual value, and the first threshold is a pre-set model convergence standard. When the loss function is lower than this threshold, it is considered that the model training has achieved the expected effect.

[0051] Optionally, in this embodiment, each data provider deploys a second initial model locally, trains the model using its own dataset, and the trained second initial model is the second model. Then, the parameter difference between the first parameters of the trained second model and the initial parameters is returned to the server.

[0052] Optionally, in this embodiment, the data provided by each data provider can be data collected from different levels. For example, in the scenario of bank loan approval, the borrower's transaction data and background information are used as target data, and the first model is used to predict the probability of default risk. In this case, the data provided by the data providers in this embodiment includes, but is not limited to: external data, institutional data, and loan data and basic customer information from within the bank. Among them, the aforementioned loan data helps the model understand the basic information of the loan; the basic customer information is used by the model to determine the borrower's basic attributes and socioeconomic status; the aforementioned external data includes behavioral data (such as customer consumption patterns, online activities, etc.), industry data (such as industry cyclicality, degree of competition, profitability, etc.), and collateral data (such as the value, type, liquidity, etc. of collateral); the aforementioned institutional data includes operating performance, financial statements, and macroeconomic indicators, such as debt-to-equity ratio, current ratio, operating cash flow, and other financial health indicators, which can reflect the borrower / company's debt repayment ability and operating status.

[0053] Optionally, in this embodiment, the target parameters are a new set of parameters used to update the current model, determined by weighted averaging or other methods based on the parameter differences returned by the data provider after each round of training of the first initial model deployed on the server.

[0054] The training method of the first model in this application embodiment is explained below with reference to an optional example. In this optional embodiment, the first model is applied in cross-bank credit risk assessment, and the N data providers are all banks. Each data provider constructs a locally deployed second initial model in the following way: First, the data involving personal privacy in the collected data is anonymized, such as using random customer numbers instead of ID card numbers, retaining the first and last three digits of mobile phone numbers, and anonymizing the rest; for usernames, only the surname is retained, and the rest is anonymized. Feature engineering is performed on the anonymized data to extract key attributes from the original data, such as industry profiles, income levels, loan tags, etc. Using the relationship between the institution and external data, and bank data, a first topology graph is constructed to intuitively display the customer relationship network and loan distribution, and to build a deep connection between debt and customer management. Then, the second initial model is constructed using the first topology graph. Assuming that all N data providers have already constructed their locally deployed second initial models, the server sends the initial model. The N data providers then perform multiple batches of stochastic gradient descent training based on the initial parameters until the loss function of their respective locally deployed second initial models satisfies the second threshold. Afterward, the parameter differences between the trained second model's parameters and the initial parameters are calculated and sent to the server. The server determines the target parameters based on the received N parameter differences. If the first initial model uses the target parameters and its first loss function satisfies the first threshold, the currently obtained first initial model is determined as the first model, and training ends.

[0055] In this embodiment, model parameters are iterated and updated within a federated learning framework to ensure that the first loss function of the initial model meets a first threshold. Federated learning guarantees data privacy and security by avoiding direct exchange of raw data, thus enhancing data protection. Joint training with multiple data sources enhances the model's generalization ability and improves the accuracy and stability of risk assessment. Iterative parameter optimization ensures continuous model evolution, enabling it to adapt to market changes and the addition of new data, thereby improving the timeliness and accuracy of resource allocation decisions.

[0056] In an exemplary embodiment, determining the target parameters of the first model deployed on the server side based on N parameter differences includes: performing a first operation on the N parameter differences to obtain a second parameter, wherein the first operation is used to perform a weighted summation operation on the N parameter differences based on the amount of data provided by the N data providers.

[0057] Optionally, the first operation in this embodiment is a weighted summation operation, where the weight of each data provider is determined based on the proportion of the data provided by that data provider to the total data volume. For example, there are two data providers, A and B. A provides 100,000 data entries, and B provides 50,000 data entries. One of the target parameters is equal to the difference in parameters returned by A. The parameter difference returned by 2 / 3+B 1 / 3.

[0058] In this embodiment, the parameter differences from each provider are weighted and summed based on their data volume to obtain a second parameter, which serves as the target parameter for the first model on the server side. Data volume weighting ensures the fairness of model parameter updates; providers with larger sample sizes have a greater impact on the model, improving prediction accuracy. Furthermore, by using weighted summation rather than simple averaging, the importance of each provider's data is fully considered, making model training more scientific and reasonable.

[0059] In an exemplary embodiment, the first topology graph is constructed as follows: parsing multiple key information items in the acquired first dataset, wherein the key information items include the object's transaction behavior, asset information, basic information, the product corresponding to the object, and the industry in which the object is located; determining the business logic between the multiple key information items; and constructing the first topology graph based on the business logic and the multiple key information items.

[0060] Optionally, the first topology graph in this embodiment is a network structure composed of nodes and edges, used to visually demonstrate the relationship, interaction and potential risk propagation path between the target product or entity and other entities in its environment.

[0061] Optionally, the first dataset in this embodiment includes all collected raw data, including but not limited to transaction data, asset information, personal basic information, product details, industry dynamics, etc. These data come from multiple channels, such as bank internal records, third-party credit assessments, public databases, etc., and are the cornerstone for constructing the first topology graph.

[0062] Optionally, the key information in this embodiment is the core data elements extracted from the first dataset, specifically including: the object's transaction behavior, asset information, basic information, the product corresponding to the object, and the industry in which the object operates. For example, in the scenario of bank loan approval, the borrower's transaction data and background information are used as target data, and the first model is used to predict the probability of default risk. In this case, the object's transaction behavior in this embodiment includes loan applications, repayment records, consumption patterns, transaction frequency, etc., which can reflect the financial habits and credit history of an individual or entity; asset information includes total assets, liabilities, working capital, fixed assets, etc., which helps the model assess the borrower's asset quality and debt repayment ability; basic information includes identity verification, contact information, residential address, etc., used to verify the authenticity of the individual or company; the product corresponding to the object is the relevant data or characteristics of the target product itself, such as loan type, insurance terms, product category, etc.; the industry in which the object operates includes industry attributes, market position, competitiveness analysis, etc., which helps to assess the impact of the macroeconomic environment and industry risks on the individual.

[0063] Optionally, the business logic in this embodiment is the inherent connection and causal relationship between key information, which guides the construction of the topology graph. For example, consumption patterns may affect credit scores, which in turn determine loan interest rates, while industry risks may be transmitted to companies, affecting their ability to repay loans.

[0064] This embodiment analyzes key information, including transaction behavior, asset information, and basic information, as well as the logical relationships between these key information, to construct a first topology graph. This constructed topology graph fully reflects the business relationships and risk transmission paths within the target scenario, providing a rich information foundation for the model to determine risk probabilities. Furthermore, this topology graph construction method ensures that the model can understand object attributes from multiple dimensions, thereby improving the comprehensiveness and precision of risk assessment.

[0065] In an exemplary embodiment, the target topology graph includes: nodes, wherein the nodes include entity nodes, derivative nodes, and auxiliary nodes; the entity nodes represent objects in a target scenario, the target scenario being a scenario using the target product, the objects including the target product; the derivative nodes represent factors influencing the risk probability of the target product; the auxiliary nodes represent the time and place of the transaction of the target product, and objects with implicit business relationships with the target product; edges connecting the nodes, wherein the edges include relationship edges, risk edges, and behavior edges; the relationship edges represent business relationships between the entity nodes; the risk edges represent risk propagation paths between the entity nodes and the derivative nodes, or risk edges represent risk propagation paths between the entity nodes and the auxiliary nodes; the behavior edges connect the entity nodes and derivative nodes generated due to the behavior of the objects represented by the entity nodes; and the weights of the edges.

[0066] Optionally, the nodes in this embodiment are divided into entity nodes, derivative nodes, and auxiliary nodes. Entity nodes are directly related business entities, such as customers, products, and companies; derivative nodes are various factors that affect entity risk, such as behavioral patterns and market environment; auxiliary nodes are objects that have indirect business relationships with the entity, such as time and location information.

[0067] Optionally, the weight of an edge in this embodiment reflects the closeness of the relationship between the two nodes connected by the edge or the risk propagation effect.

[0068] Figure 3 This is a schematic diagram of the target topology constructed under a target scenario according to an embodiment of this application. The target scenario of this optional embodiment is a bank loan approval scenario, using the borrower's transaction data and background information as target data, and predicting the probability of default risk through a first model. Figure 3As shown, in this scenario, the nodes in the target topology are divided into three categories: entity nodes, derived nodes, and auxiliary nodes. Entity nodes include the borrower, the loan (the specific loan product applied for by the borrower, such as a mortgage of 100,000 yuan), the collateral (which may be the borrower's car), and the enterprise (the borrower's workplace, such as a software development company). Derivative nodes include consumption behavior (the borrower's daily consumption patterns, such as expenses for dining, entertainment, and travel), repayment behavior (the borrower's previous timely repayments), industry risk score (the current overall risk level of the software development industry), and economic environment indicators (macroeconomic conditions, including inflation rate, unemployment rate, etc.). Auxiliary nodes include region (the borrower's living area), time (the time points when key events such as loan application, disbursement, and repayment occur), and related parties (other entities that have direct or indirect economic dealings with the borrower, such as the borrower's family members). Edges are categorized into three types: relationship edges, risk edges, and behavioral edges. Relationship edges represent inherent structured relationships between entities, such as borrower → loan, collateral → loan, and enterprise → borrower. Risk edges represent the propagation path of risk between nodes, such as industry risk → enterprise, economic environment indicators → borrower, related parties → borrower, region → borrower, and time → loan. Behavioral edges include borrower → consumption behavior and borrower → repayment behavior. It should be noted that... Figure 3 The weights of each edge shown in the target topology graph are for illustrative purposes only. In actual use, although the application scenarios may differ... Figure 3 The target topology graphs constructed may have the same scenario, but the weights may not be the same. Figure 3 The weights shown are not fixed but can be adjusted according to actual usage. After constructing such a target topology, banks can use graph neural network technology to comprehensively analyze multiple dimensions of information, including a borrower's personal credit, consumption behavior, loan details, business status, and the macroeconomic environment, to conduct a full-scale assessment of the borrower's loan risk.

[0069] This embodiment further clarifies the various types of nodes and edges included in the target topology graph. Such a structural design of the target topology graph enables the model to more accurately identify and quantify risk factors, thereby enhancing the accuracy of risk assessment.

[0070] In an exemplary embodiment, the weight of the edge is determined by the first model based on the initial weight of the edge and the degree of influence of the derived node or the auxiliary node connected by the edge on the risk probability of the entity node. The initial weight is obtained by calculating the product of a preset weight and a decay factor. The preset weight is set based on the type of the edge, and the decay factor is determined according to the generation time of the derived node or the auxiliary node.

[0071] Optionally, in this embodiment, the initial weight is a fixed edge weight value in the early stage of model training, calculated by multiplying the preset weight by the time decay factor, that is, the initial weight of the edge = preset weight × decay factor.

[0072] Optionally, the preset weights in this embodiment are determined based on the type of edge (such as relationship edge, risk edge, and behavior edge), reflecting the inherent importance of different types of edges in risk assessment. For example, in a loan scenario in the financial field, the preset weight of relationship edges may be higher than that of behavior edges, because the lending relationship directly determines the basic outline of loan risk; the weight of relationship edges is 0.9, the weight of risk edges is 0.5, and the weight of behavior edges is 0.3. It should be noted that the above is only an example, and in actual use, it can be adjusted according to the application scenario and usage, and is not limited here.

[0073] Optionally, in this embodiment, the time decay factor determines the dynamic change pattern of edge weights based on the timeliness of the connected node data. The closer the data is to the current time, the larger the decay factor, and the greater the edge weight; conversely, the weight decreases as the data becomes more distant. For example, the decay factor for recent data (data acquired within the last 3 months) is set to 1.0, the decay factor for mid-term data (data acquired within the last year) is set to 0.7, and the decay factor for historical data (data acquired more than 1 year ago) is set to 0.3. It should be noted that the above is only an example, and adjustments can be made according to the application scenario and usage in actual use; no limitations are imposed here.

[0074] Optionally, in this embodiment, the weight of the edge refers specifically to the relationship edge used to represent the business relationship between entity nodes, as well as the risk edge.

[0075] Optionally, the edge weights are determined by the first model over time by continuously learning the influence of the derived or auxiliary nodes connected by the edge on the risk probability of the entity node, based on the initial weights of the edges. For example, if a borrower has negative news in their economic activities in their region (such as defaulting on payments to suppliers), this will be reflected with a higher weight in the borrower's personal credit assessment through the "region → borrower" risk edge.

[0076] This embodiment further illustrates how edge weights are dynamically adjusted based on the degree of business impact between nodes, and how initial weights are calculated based on preset weights and a time decay factor. The weight adjustment strategy in this embodiment reflects business dynamics and time effects, improving the model's adaptability and predictive ability to real-world situations. The introduction of the time decay factor allows the model to place greater emphasis on recent data, enhancing its ability to capture trend changes.

[0077] In an exemplary embodiment, determining the target risk type probability of the target product based on the initial risk type probability and the target data includes: performing feature transformation operations on the initial risk type probability and the target data respectively to obtain a first feature and a second feature; fusing the first feature and the second feature to obtain a third feature; and inputting the third feature into a second model to obtain the target risk type probability output by the second model. The second model includes multiple decision trees, and is used to perform a second operation on the third feature and a third operation on multiple first values ​​obtained after performing the second operation. The second operation is used to determine the degree of matching between the third feature and the conditions represented by the leaf nodes of the decision trees using the multiple decision trees, obtaining multiple first values. The third operation is used to calculate the average of the multiple first values ​​to obtain the target risk type probability.

[0078] Optionally, the feature transformation operation in this embodiment is used to transform the initial risk type probability and target data into a form or feature set that the model can process. For example, the probability is converted into a value between 0 and 1, and the text information in the target data is converted into a numerical code.

[0079] Optionally, the second model in this embodiment is a series of integrated decision tree models used to further refine and calibrate the probabilities of risk types. In practical use, the second model can be an extreme gradient boosting (XGBOOST) model.

[0080] Optionally, the second model in this embodiment uses multiple gradient boosting decision trees to serially integrate the prediction errors of all preceding trees as the learning target, so that each new tree focuses on correcting the prediction bias of the initial risk type probability, and dynamically adjusts the separation boundary of the risk category in the feature space through node splitting of each tree, and finally integrates the prediction values ​​of all decision trees and converts them into probability outputs to achieve systematic correction of the initial risk type probability and generate the target risk type probability.

[0081] Figure 4 This is a flowchart illustrating a resource allocation method in a target scenario according to an embodiment of this application. The target scenario is a bank loan approval scenario. Specifically, a bank is using a resource allocation method, combined with graph neural networks and federated learning, to optimize the post-loan management process for a target enterprise loan, such as... Figure 4 As shown, the process of this method may include the following steps:

[0082] Step S402: Input the target company's loan data and transaction data (i.e., the aforementioned target data) into the first model to obtain the risk probability corresponding to the target company's loan output by the first model. The first model is used to determine the risk probability based on the target topology graph. The first model is constructed by the bank in conjunction with multiple data providers (such as credit rating agencies, industry analysis companies, etc.) after optimizing the model parameters according to the federated learning protocol. The bank and multiple data providers construct the target topology graph by sharing encrypted loan transaction data, borrower information, industry dynamics, macroeconomic indicators, etc., thereby enabling the first model to deeply explore the complex network relationships between enterprises and the dynamic changes in loan risk based on GNN technology, and output the risk probability of the target company's loan.

[0083] Step S404: Combine the risk probability with the target company's loan data and transaction data to determine the initial risk type probability of the loan, that is, to preliminarily determine the probability of the target company's loan being low-risk, medium-risk, or high-risk.

[0084] Step S406: Based on the initial risk type probability, further analyze the target data (e.g., events that may increase loan risk, such as declining corporate performance or supply chain disruptions), adjust the probability distribution of risk types, and derive the target risk type probability.

[0085] Step S408: Determine the risk type of the target enterprise loan based on the probability of the target risk type;

[0086] Step S410: Allocate resources for the loan to the target company based on the risk type. If the risk type is high risk, allocate more storage and software resources to record more behavioral data of the target company and use more software to monitor the target company's consumption behavior and other behaviors that may affect repayment. Increase monitoring frequency, adjust collection plans, and intervene in non-performing asset management in advance. If the risk type is low risk, reduce the corresponding computing resources, simplify regulatory procedures, reduce monitoring frequency, and analyze the data of the target company obtained from monitoring.

[0087] Through this optional example, by constructing a target topology graph and utilizing the model, banks can gain a more comprehensive understanding of the interrelationships between enterprises and the network effects of loan risks, improving the accuracy of risk warnings. The target topology graph and model parameters can be updated in real time to respond to market changes and enterprise dynamics, enhancing the flexibility and adaptability of management processes. Based on multi-dimensional risk assessment under federated learning, banks can accurately identify and prioritize high-risk loans, avoiding blind allocation of resources and saving time and costs. For enterprises assessed as low-risk, banks can appropriately relax regulations, maintain good relationships with enterprises through preferential policies, and promote future business cooperation. Therefore, this addresses the problem in related technologies where the inability to accurately determine the risk type of a product leads to inaccurate resource allocation to events corresponding to the product's transaction data, thereby achieving the effect of accurately identifying the risk type of a product and accurately allocating resources to events corresponding to the product's transaction data.

[0088] It should be noted that, for the sake of simplicity, the foregoing method embodiments are all described as a series of actions. However, those skilled in the art should understand that this application is not limited to the described order of actions, as some steps may be performed in other orders or simultaneously according to this application. Furthermore, those skilled in the art should also understand that the embodiments described in the specification are preferred embodiments, and the actions and modules involved are not necessarily essential to this application.

[0089] Through the above description of the embodiments, those skilled in the art can clearly understand that the methods according to the above embodiments can be implemented by means of software plus necessary general-purpose hardware platforms. Of course, they can also be implemented by hardware, but in many cases the former is a better implementation method. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product is stored in a storage medium (such as read-only memory (ROM) / random access memory (RAM), magnetic disk, optical disk), and includes several instructions to cause a terminal device (which may be a mobile phone, computer, server, or network device, etc.) to execute the methods described in the various embodiments of this application.

[0090] According to another aspect of the embodiments of this application, a resource allocation apparatus is also provided, which can be used to implement the resource allocation method provided in the above embodiments, and will not be repeated hereafter. As used below, the term "module" can be a combination of software and / or hardware that implements a predetermined function. Although the apparatus described in the following embodiments is preferably implemented in software, hardware implementation, or a combination of software and hardware, is also possible and contemplated.

[0091] Figure 5 This is a structural block diagram of a resource allocation device according to an embodiment of this application, such as... Figure 5 As shown, the device includes:

[0092] The input module 502 is used to input target data into the first model to obtain the risk probability of the target data output by the first model. The target data includes transaction data of the target product. The first model is used to determine the risk probability based on the target topology map. The target topology map is constructed based on data provided by N data providers, where N is an integer greater than 1.

[0093] The first determining module 504 is used to determine the initial risk type probability of the target product based on the above-mentioned risk probability and the above-mentioned target data.

[0094] The second determining module 506 is used to determine the target risk type probability of the target product based on the above-mentioned initial risk type probability and the above-mentioned target data.

[0095] The third determining module 508 is used to determine the risk type of the target product based on the probability of the target risk type mentioned above.

[0096] The allocation module 510 is used to allocate resources to the events corresponding to the above transaction data based on the above risk types.

[0097] The target topology graph includes: nodes, where the nodes include entity nodes, derivative nodes, and auxiliary nodes. Entity nodes represent objects in the target scenario, which is a scenario using the target product. The objects include the target product. Derivative nodes represent factors affecting the risk probability of the target product. Auxiliary nodes represent the time and place of trading the target product and objects with implicit business relationships with the target product. Edges connecting the nodes include relationship edges, risk edges, and behavior edges. Relationship edges represent business relationships between entity nodes. Risk edges represent risk propagation paths between entity nodes and derivative nodes, or between entity nodes and auxiliary nodes. Behavior edges connect entity nodes and derivative nodes generated due to the behavior of the objects represented by the entity nodes. The weights of the edges are also included.

[0098] The weight of the aforementioned edge is determined by the first model based on the initial weight of the aforementioned edge and the degree of influence of the aforementioned derived node or the aforementioned auxiliary node connected by the aforementioned edge on the risk probability of the aforementioned entity node. The initial weight is obtained by calculating the product of a preset weight and a decay factor. The preset weight is set based on the type of the aforementioned edge, and the decay factor is determined based on the generation time of the aforementioned derived node or the aforementioned auxiliary node.

[0099] It should be noted that the input module 502 in this embodiment can be used to execute the above step S202, the first determining module 504 in this embodiment can be used to execute the above step S204, the second determining module 506 in this embodiment can be used to execute the above step S206, the third determining module 508 in this embodiment can be used to execute the above step S208, and the allocation module 510 in this embodiment can be used to execute the above step S210.

[0100] In an exemplary embodiment, the input module 502 is further configured to train a first model by performing the following target operation until the first initial model deployed on the server side uses target parameters and the first loss function of the first initial model satisfies a first threshold, and then determining the currently obtained first initial model as the first model: sending the initial parameters of the first initial model to N data providers to instruct the N data providers to train their respective locally deployed second initial models based on the initial parameters, wherein each of the data providers' locally deployed second initial models is used to determine the risk probability of the product based on a first topology map, the first topology map being based on the data provider's... The data is constructed such that the target topology graph includes N first topology graphs; N model parameters sent by the data provider are received to obtain N parameter differences, wherein the parameter differences are the differences between the first parameters of the second model obtained after training by the data provider and the initial parameters, and the second loss function of the trained second model satisfies a second threshold; the target parameters of the first initial model deployed on the server are determined based on the N parameter differences; if the first loss function of the first initial model deployed on the server does not satisfy the first threshold after adopting the target parameters, the first parameters are used as new initial parameters, and the target operation is performed again.

[0101] In an exemplary embodiment, the input module 502 is further configured to perform a first operation on the N parameter differences to obtain a second parameter, wherein the first operation is configured to perform a weighted summation operation on the N parameter differences based on the amount of data provided by the N data providers.

[0102] In an exemplary embodiment, the input module 502 is further configured to construct the first topology graph by parsing multiple key information in the acquired first dataset, wherein the key information includes the object's transaction behavior, asset information, basic information, the product corresponding to the object, and the industry in which the object is located; determining the business logic between the multiple key information; and constructing the first topology graph based on the business logic and the multiple key information.

[0103] In an exemplary embodiment, the second determining module 506 is further configured to: perform feature transformation operations on the initial risk type probability and the target data respectively to obtain a first feature and a second feature; fuse the first feature and the second feature to obtain a third feature; input the third feature into a second model to obtain the target risk type probability output by the second model, wherein the second model includes multiple decision trees, the second model is configured to perform a second operation on the third feature and a third operation on multiple first values ​​obtained after performing the second operation, the second operation is configured to use the multiple decision trees to determine the degree of matching between the third feature and the conditions represented by the leaf nodes of the decision trees to obtain multiple first values, and the third operation is configured to calculate the average value of the multiple first values ​​to obtain the target risk type probability.

[0104] It should be noted that the above modules can be implemented by software or hardware. For the latter, they can be implemented in the following ways, but are not limited to: all the above modules are located in the same processor; or, the above modules are located in different processors in any combination.

[0105] According to another aspect of the embodiments of this application, a computer-readable storage medium is provided, the computer-readable storage medium including a stored program, wherein the program executes the steps in any of the above method embodiments when it is run.

[0106] In one exemplary embodiment, the aforementioned computer-readable storage medium may include, but is not limited to, various media capable of storing computer programs, such as USB flash drives, ROMs, RAMs, portable hard drives, magnetic disks, or optical disks.

[0107] According to another aspect of the embodiments of this application, an electronic device is provided, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor is configured to perform the steps of any of the above method embodiments by means of the computer program.

[0108] In one exemplary embodiment, the electronic device may further include a transmission device and an input / output device, wherein the transmission device is connected to the processor and the input / output device is connected to the processor.

[0109] Specific examples in this embodiment can be found in the examples described in the above embodiments and exemplary implementations, and will not be repeated here.

[0110] Embodiments of this application also provide a computer program product, which includes a computer program that, when executed by a processor, implements the steps in any of the above method embodiments.

[0111] Embodiments of this application also provide another computer program product, including a non-volatile computer-readable storage medium storing a computer program that, when executed by a processor, implements the steps in any of the above method embodiments.

[0112] The embodiments described herein also provide a computer program that includes computer instructions stored in a computer-readable storage medium; a processor of a computer device reads the computer instructions from the computer-readable storage medium and executes the computer instructions, causing the computer device to perform the steps in any of the above method embodiments.

[0113] Specific examples in this embodiment can be found in the examples described in the above embodiments and exemplary implementations, and will not be repeated here.

[0114] Obviously, those skilled in the art should understand that the modules or steps of this application described above can be implemented using general-purpose computing devices. They can be centralized on a single computing device or distributed across a network of multiple computing devices. They can be implemented using computer-executable program code, and thus can be stored in a storage device for execution by a computing device. In some cases, the steps shown or described can be performed in a different order than those presented here, or they can be fabricated as separate integrated circuit modules, or multiple modules or steps can be fabricated as a single integrated circuit module. Thus, this application is not limited to any particular combination of hardware and software.

[0115] The above description is merely a preferred embodiment of this application and is not intended to limit this application. Various modifications and variations can be made to this application by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the principles of this application should be included within the protection scope of this application.

Claims

1. A method for allocating resources, characterized in that, include: The target data is input into the first model to obtain the risk probability of the target data output by the first model. The target data includes transaction data of the target product. The first model is used to determine the risk probability based on the target topology map, which is constructed based on data provided by N data providers, where N is an integer greater than 1. Determine the initial risk type probability of the target product based on the risk probability and the target data; The target risk type probability of the target product is determined based on the initial risk type probability and the target data. The risk type of the target product is determined based on the probability of the target risk type. Resources are allocated to the events corresponding to the transaction data based on the risk type.

2. The method according to claim 1, characterized in that, The first model was trained in the following way: Perform the following objective operation until the first initial model deployed on the server uses the target parameters and the first loss function of the first initial model satisfies the first threshold, then determine the currently obtained first initial model as the first model: The initial parameters of the first initial model are sent to N data providers to instruct the N data providers to train their respective locally deployed second initial models based on the initial parameters. Each of the data providers' locally deployed second initial models is used to determine the risk probability of a product based on a first topology map, which is constructed based on the data of the data providers. The target topology map includes N of the first topology maps. Receive N model parameters sent by the data provider and obtain N parameter differences, wherein the parameter differences are the differences between the first parameters of the second model obtained after training by the data provider and the initial parameters, and the second loss function of the trained second model satisfies a second threshold; The target parameters of the first initial model deployed on the server are determined based on the N parameter differences. After the first initial model deployed on the server adopts the target parameters, if the first loss function of the first initial model does not meet the first threshold, the first parameters are used as the new initial parameters, and the target operation is executed again.

3. The method according to claim 2, characterized in that, The target parameters of the first model deployed on the server are determined based on N parameter differences, including: A first operation is performed on the N parameter differences to obtain a second parameter, wherein the first operation is used to perform a weighted summation operation on the N parameter differences based on the amount of data provided by the N data providers.

4. The method according to claim 2, characterized in that, The first topology graph is constructed in the following way: The first dataset is parsed to extract multiple key pieces of information, including the object's transaction behavior, asset information, basic information, the product corresponding to the object, and the industry in which the object belongs. Determine the business logic between the multiple key pieces of information; The first topology diagram is constructed based on the business logic and multiple key pieces of information.

5. The method according to claim 1, characterized in that, The target topology map includes: The nodes include entity nodes, derivative nodes, and auxiliary nodes. The entity nodes represent objects in the target scenario, which is a scenario in which the target product is used. The objects include the target product. The derivative nodes represent factors that affect the risk probability of the target product. The auxiliary nodes represent the time and place of the transaction of the target product and objects that have implicit business relationships with the target product. Edges used to connect the nodes, wherein the edges include relationship edges, risk edges and behavior edges, the relationship edges are used to represent the business relationship between the entity nodes, the risk edges are used to represent the risk propagation path between the entity node and the derived node, or the risk edges are used to represent the risk propagation path between the entity node and the auxiliary node, and the behavior edges are used to connect the entity node and the derived node generated due to the behavior of the object represented by the entity node; The weight of the edge.

6. The method according to claim 5, characterized in that, The weight of the edge is determined by the first model based on the initial weight of the edge and the degree of influence of the derived node or the auxiliary node connected by the edge on the risk probability of the entity node. The initial weight is obtained by calculating the product of a preset weight and a decay factor. The preset weight is set based on the type of the edge, and the decay factor is determined according to the generation time of the derived node or the auxiliary node.

7. The method according to claim 1, characterized in that, Determining the target risk type probability of the target product based on the initial risk type probability and the target data includes: Perform feature transformation operations on the initial risk type probability and the target data respectively to obtain the first feature and the second feature; By combining the first feature and the second feature, a third feature is obtained; The third feature is input into the second model to obtain the target risk type probability output by the second model. The second model includes multiple decision trees. The second model is used to perform a second operation on the third feature and a third operation on multiple first values ​​obtained after performing the second operation. The second operation is used to determine the degree of matching between the third feature and the conditions represented by the leaf nodes of the decision trees using the multiple decision trees to obtain multiple first values. The third operation is used to calculate the average value of the multiple first values ​​to obtain the target risk type probability.

8. A resource allocation device, characterized in that, include: An input module is used to input target data into a first model to obtain the risk probability of the target data output by the first model. The target data includes transaction data of a target product. The first model is used to determine the risk probability based on a target topology map, which is constructed based on data provided by N data providers, where N is an integer greater than 1. The first determining module is used to determine the initial risk type probability of the target product based on the risk probability and the target data; The second determining module is used to determine the target risk type probability of the target product based on the initial risk type probability and the target data; The third determining module is used to determine the risk type of the target product based on the probability of the target risk type. The allocation module is used to allocate resources to the events corresponding to the transaction data based on the risk type.

9. A computer program product comprising a computer program / instructions, characterized in that, When the computer program / instructions are executed by the processor, they implement the steps of the method according to any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a computer program, wherein the computer program, when executed by a processor, implements the steps of the method according to any one of claims 1 to 7.

11. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the computer program, it implements the steps of the method according to any one of claims 1 to 7.

Images