Method, device, equipment, storage medium and program product for processing multi-source data
By characterizing banking transaction and environmental data, and combining risk prediction models and multimodal rule engines, the problem of insufficient adaptability of traditional economic capital management systems to sudden risks is solved, and efficient and accurate resource management strategy generation and self-optimization updates are achieved.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- INDUSTRIAL AND COMMERCIAL BANK OF CHINA
- Filing Date
- 2026-01-16
- Publication Date
- 2026-05-29
AI Technical Summary
Traditional economic capital management systems rely heavily on static models and historical data, making it difficult to cope with sudden risks. They are unable to analyze unstructured data in real time, resulting in insufficient risk coverage. Furthermore, the fragmentation of internal bank systems leads to high data integration costs and long model parameter update cycles, making it difficult to respond to high-frequency market changes.
By acquiring business transaction data and environmental data, performing feature extraction and splicing, using a risk prediction model set and a multimodal rule engine to determine resource management strategies, and updating the training model through a multi-objective optimization algorithm, dynamic parsing and integration of multi-source data is achieved.
It expands the coverage of risk prediction data, avoids the limitations of a single model, improves the accuracy and response speed of resource management strategies, reduces human intervention errors, and forms a self-optimizing iterative closed-loop architecture.
Smart Images
Figure CN122115100A_ABST
Abstract
Description
Technical Field
[0001] This application relates to the field of financial technology, and in particular to a method, apparatus, device, storage medium, and program product for processing multi-source data. Background Technology
[0002] In the modern financial system, banks need to dynamically assess and allocate capital through economic capital management tools to cope with complex and ever-changing market environments and risk factors. Economic capital management is a core component of bank risk management, and its core objective is to ensure that banks have sufficient capital buffers when facing credit risk, market risk, and operational risk by scientifically quantifying risk exposure.
[0003] However, traditional economic capital management systems rely heavily on static models and historical data, making them ill-equipped to handle sudden risks and lacking the ability to analyze unstructured data. For example, when black swan events occur in the market, traditional systems, lacking real-time analysis capabilities for unstructured data, cannot adjust capital allocation strategies in a timely manner, resulting in insufficient risk coverage. Furthermore, fragmented internal bank systems lead to high data integration costs and long model parameter update cycles, making it difficult to respond to high-frequency market changes.
[0004] Therefore, there is an urgent need for an efficient and accurate solution for processing multi-source data. Summary of the Invention
[0005] This application provides a method, apparatus, device, storage medium, and program product for processing multi-source data, in order to solve the technical problem that existing management strategy generation schemes are difficult to integrate multi-source heterogeneous data and dynamically analyze risk factors, resulting in low accuracy of management strategy generation.
[0006] Firstly, this application provides a method for processing multi-source data, including:
[0007] Acquire business transaction data and business environment data;
[0008] Feature extraction processing is performed on business transaction data and business environment data to obtain feature vectors;
[0009] Based on the feature vectors and the set of risk prediction models, at least one target risk prediction model is determined; the feature vectors are then input into the at least one target risk prediction model to obtain at least one risk prediction result.
[0010] At least one risk prediction result is integrated and processed to obtain the target risk prediction result; based on the target risk prediction result and the preset resource management rules, a resource management strategy is determined.
[0011] In one possible implementation, feature extraction processing is performed on business transaction data and business environment data to obtain a feature vector, including:
[0012] The business environment data is parsed and processed to obtain the keyword features corresponding to the business environment data; the business transaction data is processed to extract features to obtain the transaction features corresponding to the business transaction data.
[0013] The keyword features and transaction features are concatenated to obtain a feature vector.
[0014] In one possible implementation, based on feature vectors and a set of risk prediction models, at least one target risk prediction model is determined, including:
[0015] The matching degree between the feature vector and each risk prediction model in the risk prediction model set is calculated through a gated routing mechanism.
[0016] Risk prediction models with a matching degree greater than the preset matching degree are selected as candidate risk prediction models.
[0017] Determine whether the number of candidate risk prediction models exceeds a preset threshold; if the number of candidate risk prediction models does not exceed the preset threshold, then multiple candidate risk prediction models are used as the target risk prediction model.
[0018] If the number of candidate risk prediction models exceeds a preset threshold, the candidate risk prediction model with the highest matching degree (the one that matches the preset threshold) will be selected as the target risk prediction model.
[0019] In one possible implementation, a resource management strategy is determined based on the target risk prediction results and preset resource management rules, including:
[0020] Using a multimodal rule engine, based on the target risk prediction results and preset resource management rules, at least one resource management rule corresponding to the target risk prediction results is determined; wherein, the preset resource management rules refer to the processing strategies for different target risk prediction results.
[0021] Generate a resource management strategy based on at least one resource management rule and economic resources.
[0022] In one possible implementation, the method further includes:
[0023] The resource management strategy is transmitted to the execution system so that the execution system can execute the resource management strategy and report the strategy loss data.
[0024] The target risk prediction model is updated and trained based on the strategy loss data.
[0025] In one possible implementation, the target risk prediction model is updated and trained based on the policy loss data, including:
[0026] The policy loss data is decomposed into multiple sub-objective data through a multi-objective optimization algorithm;
[0027] The target risk prediction model is updated and trained using a Pareto front optimization strategy based on multiple sub-target data and the aforementioned feature vectors.
[0028] Secondly, this application provides a multi-source data processing apparatus, comprising:
[0029] The acquisition module is used to acquire business transaction data and business environment data;
[0030] The processing module is used to perform feature extraction processing on business transaction data and business environment data to obtain feature vectors;
[0031] The processing module is also used to determine at least one target risk prediction model based on the feature vector and the set of risk prediction models; and to input the feature vector into at least one target risk prediction model to obtain at least one risk prediction result.
[0032] The generation module is used to integrate and process at least one risk prediction result to obtain the target risk prediction result; based on the target risk prediction result and the preset resource management rules, the resource management strategy is determined.
[0033] In one possible implementation, the processing module is further configured to:
[0034] The business environment data is parsed and processed to obtain the keyword features corresponding to the business environment data; the business transaction data is processed to extract features to obtain the transaction features corresponding to the business transaction data.
[0035] The keyword features and transaction features are concatenated to obtain a feature vector.
[0036] In one possible implementation, the processing module is further configured to:
[0037] The matching degree between the feature vector and each risk prediction model in the risk prediction model set is calculated through a gated routing mechanism.
[0038] Risk prediction models with a matching degree greater than the preset matching degree are selected as candidate risk prediction models.
[0039] Determine whether the number of candidate risk prediction models exceeds a preset threshold; if the number of candidate risk prediction models does not exceed the preset threshold, then multiple candidate risk prediction models are used as the target risk prediction model.
[0040] If the number of candidate risk prediction models exceeds a preset threshold, the candidate risk prediction model with the highest matching degree (the one that matches the preset threshold) will be selected as the target risk prediction model.
[0041] In one possible implementation, the generation module is further configured to:
[0042] Using a multimodal rule engine, based on the target risk prediction results and preset resource management rules, at least one resource management rule corresponding to the target risk prediction results is determined; wherein, the preset resource management rules refer to the processing strategies for different target risk prediction results.
[0043] Generate a resource management strategy based on at least one resource management rule and economic resources.
[0044] In one possible implementation, the processing module is further configured to:
[0045] The resource management strategy is transmitted to the execution system so that the execution system can execute the resource management strategy and report the strategy loss data.
[0046] The target risk prediction model is updated and trained based on the strategy loss data.
[0047] In one possible implementation, the processing module is further configured to:
[0048] The policy loss data is decomposed into multiple sub-objective data through a multi-objective optimization algorithm;
[0049] The target risk prediction model is updated and trained using a Pareto front optimization strategy based on multiple sub-target data and the aforementioned feature vectors.
[0050] Thirdly, this application provides an electronic device, including: a processor, and a memory communicatively connected to the processor; the memory stores computer-executable instructions; the processor executes the computer-executable instructions stored in the memory to implement the aforementioned method.
[0051] Fourthly, this application provides a computer-readable storage medium storing computer-executable instructions, which, when executed by a processor, are used to implement the aforementioned method.
[0052] Fifthly, embodiments of this application provide a computer program product, including a computer program that, when executed by a processor, implements the first aspect and / or various possible implementations of the first aspect.
[0053] The multi-source data processing method, apparatus, equipment, storage medium, and program product provided in this application acquire business transaction data and business environment data; perform feature extraction processing on the business transaction data and business environment data to obtain feature vectors; determine at least one target risk prediction model based on the feature vectors and a set of risk prediction models; input the feature vectors into the at least one target risk prediction model to obtain at least one risk prediction result; integrate the at least one risk prediction result to obtain a target risk prediction result; and determine a resource management strategy based on the target risk prediction result and preset resource management rules. This application expands the coverage of risk prediction data by characterizing business transaction data and business environment data to obtain feature vectors; simultaneously, by determining multiple risk prediction models based on feature vectors, it avoids the limitations of a single model for complex risk scenarios; and by determining a resource management strategy based on the integrated target risk prediction result and preset resource management rules, it not only avoids reliance on manual experience in allocating rules but also improves the accuracy of strategy generation. Attached Figure Description
[0054] The accompanying drawings, which are incorporated in and form part of this specification, illustrate embodiments consistent with this application and, together with the description, serve to explain the principles of this application.
[0055] Figure 1 A flowchart illustrating a multi-source data processing method provided in this application embodiment. Figure 1 ;
[0056] Figure 2 A flowchart illustrating a multi-source data processing method provided in this application. Figure 2 ;
[0057] Figure 3 A flowchart illustrating a multi-source data processing method provided in this application. Figure 3 ;
[0058] Figure 4 This application provides a schematic diagram of the structure of a multi-source data processing device;
[0059] Figure 5 This is a schematic diagram of the structure of an electronic device provided in this application.
[0060] The accompanying drawings illustrate specific embodiments of this application, which will be described in more detail below. These drawings and descriptions are not intended to limit the scope of the concept in any way, but rather to illustrate the concept of this application to those skilled in the art through reference to particular embodiments. Detailed Implementation
[0061] Exemplary embodiments will now be described in detail, examples of which are illustrated in the accompanying drawings. When the following description relates to the drawings, unless otherwise indicated, the same numbers in different drawings denote the same or similar elements. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with this application. Rather, they are merely examples of apparatuses and methods consistent with some aspects of this application as detailed in the appended claims.
[0062] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, data stored, data displayed, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties. Furthermore, the collection, storage, use, processing, transmission, provision, disclosure, and application of the relevant data all comply with the relevant laws, regulations, and standards of the relevant countries and regions, have taken necessary confidentiality measures, do not violate public order and good morals, and provide corresponding operation access points for users to choose to authorize or refuse.
[0063] Furthermore, the technical solution involved in this application, which involves big data analysis of user information (including but not limited to personal biometrics, identity data, consumption data, asset data, electronic terminal operation data, etc.) and the use of artificial intelligence technology for automated decision-making, and makes decisions that have a significant impact on personal rights based on the results of automated decision-making, provides users with corresponding operation entry points for users to choose to agree to or reject the results of automated decision-making; if the user chooses to reject, the process will proceed to the expert decision-making process.
[0064] It should be noted that the multi-source data processing methods, apparatus, devices, storage media, and program products provided in this application can be used in the field of fintech, as well as in any other field. The application fields of the multi-source data processing methods, apparatus, devices, storage media, and program products in this application are not limited.
[0065] In the modern financial system, banks need to dynamically assess and allocate capital through economic capital management tools to cope with complex and ever-changing market environments and risk factors. Economic capital management is a core component of bank risk management, and its core objective is to ensure that banks have sufficient capital buffers when facing credit risk, market risk, and operational risk by scientifically quantifying risk exposure.
[0066] However, traditional economic capital management systems rely heavily on static models and historical data, making them ill-equipped to handle sudden risks and lacking the ability to analyze unstructured data. For example, when black swan events occur in the market, traditional systems, lacking real-time analysis capabilities for unstructured data, cannot adjust capital allocation strategies in a timely manner, resulting in insufficient risk coverage. Furthermore, fragmented internal bank systems lead to high data integration costs and long model parameter update cycles, making it difficult to respond to high-frequency market changes.
[0067] Therefore, there is an urgent need for an efficient and accurate solution for processing multi-source data.
[0068] The multi-source data processing method, apparatus, equipment, storage medium, and program product provided in this application acquire business transaction data and business environment data; perform feature extraction processing on the business transaction data and business environment data to obtain feature vectors; determine at least one target risk prediction model based on the feature vectors and a set of risk prediction models; input the feature vectors into the at least one target risk prediction model to obtain at least one risk prediction result; integrate the at least one risk prediction result to obtain a target risk prediction result; and determine a resource management strategy based on the target risk prediction result and preset resource management rules. This application expands the coverage of risk prediction data by characterizing business transaction data and business environment data to obtain feature vectors; simultaneously, by determining multiple risk prediction models based on feature vectors, it avoids the limitations of a single model for complex risk scenarios; and by determining a resource management strategy based on the integrated target risk prediction result and preset resource management rules, it not only avoids reliance on manual experience in allocating rules but also improves the accuracy of strategy generation.
[0069] The technical solution of this application and how the technical solution of this application solves the above-mentioned technical problems are described in detail below with specific embodiments. These specific embodiments can be combined with each other, and the same or similar concepts or processes may not be described again in some embodiments. The embodiments of this application will now be described with reference to the accompanying drawings.
[0070] Figure 1 A flowchart illustrating a multi-source data processing method provided in this application. Figure 1 ,like Figure 1 As shown in the embodiments of this application, the multi-source data processing method includes:
[0071] S101. Obtain business transaction data and business environment data;
[0072] The business transaction data includes structured data such as internal risk measurement indicators, historical transaction data, and real-time transaction data; the business environment data includes unstructured data such as economic indicator data, management information, and business transaction comments.
[0073] Specifically, business transaction data is obtained from the bank's transaction system through a data interface; business environment data is obtained from external data sources through a gateway interface.
[0074] S102. Perform feature extraction processing on business transaction data and business environment data to obtain feature vectors;
[0075] To address the problem that existing strategy generation technologies rely solely on internal structured data within banks and fail to integrate external unstructured data, resulting in insufficient risk prediction coverage, this application improves the data coverage of risk prediction by performing feature processing on business transaction data and business environment data to obtain a unified feature vector.
[0076] In one possible implementation, feature extraction processing is performed on business transaction data and business environment data to obtain feature vectors. The specific process is as follows:
[0077] The business environment data is parsed and processed to obtain the keyword features corresponding to the business environment data; the business transaction data is processed to extract features to obtain the transaction features corresponding to the business transaction data.
[0078] The keyword features and transaction features are concatenated to obtain a feature vector.
[0079] In conjunction with the above, for example, the business environment data "economic indicator data" is parsed and processed to obtain the keyword features corresponding to the "economic indicator data", such as "keyword feature 1, keyword feature 2"; the business transaction data "real-time transaction data" is processed to extract features to obtain the transaction features corresponding to the "real-time transaction data", such as "transaction feature 1 (the transaction frequency increases at 12 noon)", etc.
[0080] The obtained keyword feature 1, keyword feature 2, and transaction feature 1 are concatenated into a predefined multi-dimensional vector to obtain feature vector 1. The dimensions of the vector can be set according to requirements, for example, a 10-dimensional vector. Each dimension corresponds to different business environment data and business transaction data.
[0081] This step involves combining business environment data and business transaction data into a unified feature vector. This ensures that the data input into the risk prediction model includes both long-term trends, such as historical risk indicators, and short-term unforeseen factors, such as changes in management information. This enhances the comprehensiveness and real-time nature of risk prediction.
[0082] S103. Based on the feature vector and the set of risk prediction models, determine at least one target risk prediction model; input the feature vector into the at least one target risk prediction model respectively to obtain at least one risk prediction result;
[0083] For example, the risk prediction model set includes risk prediction model 1, risk prediction model 2, risk prediction model 3, risk prediction model 4, risk prediction model 5, and risk prediction model 6. Based on feature vector 1, at least one target risk prediction model is determined; the feature vector is then input into the obtained target prediction model to obtain the risk prediction result.
[0084] In one possible implementation, Figure 2 A flowchart illustrating a multi-source data processing method provided in this application. Figure 2 ,like Figure 2 As shown, based on the feature vectors and the set of risk prediction models, at least one target risk prediction model is determined. The specific process is as follows:
[0085] S201. Calculate the matching degree between the feature vector and each risk prediction model in the risk prediction model set through the gating routing mechanism.
[0086] Among them, the gating routing mechanism is a technology of "intelligent diversion + switch control". The core is to use "gating" (similar to a smart switch) to determine which path information / data should take and which modules should be activated for processing. It is commonly found in hybrid expert models such as MoE and financial risk control systems. The purpose is to allow the appropriate "expert" model to process the appropriate data, thereby improving efficiency and accuracy.
[0087] Based on the above, the matching degree between feature vector 1 and each risk prediction model in the risk prediction model set is calculated through the gating routing mechanism; such as "risk prediction model 1-0.75, risk prediction model 2-0.7, risk prediction model 3-0.85, risk prediction model 4-0.6, risk prediction model 5-0.65, risk prediction model 6-0.8".
[0088] S202. Select the risk prediction model with a matching degree greater than the preset matching degree as the candidate risk prediction model.
[0089] The preset matching degree is set to 0.75. Therefore, the risk prediction models with a matching degree greater than the preset matching degree include: Risk Prediction Model 3 and Risk Prediction Model 6. Therefore, Risk Prediction Model 3 and Risk Prediction Model 6 are selected as candidate risk prediction models.
[0090] S203. Determine whether the number of candidate risk prediction models exceeds the preset number threshold; if the number of candidate risk prediction models does not exceed the preset number threshold, then multiple candidate risk prediction models are used as target risk prediction models.
[0091] The preset quantity threshold refers to the maximum number of risk prediction models that can be used for risk prediction. It can be set according to needs, for example, the quantity threshold can be set to 3. If the number of candidate risk prediction models does not exceed the preset quantity threshold, then "risk prediction model 3 and risk prediction model 6" will be used as the target risk prediction models.
[0092] S204. If the number of candidate risk prediction models exceeds a preset threshold, the candidate risk prediction model with the highest matching degree among the preset thresholds shall be selected as the target risk prediction model.
[0093] Similarly, if the matching degrees of feature vector 1 with each risk prediction model in the risk prediction model set are "Risk prediction model 1 - 0.85, Risk prediction model 2 - 0.7, Risk prediction model 3 - 0.65, Risk prediction model 4 - 0.8, Risk prediction model 5 - 0.82, Risk prediction model 6 - 0.86", then the candidate risk prediction models are "Risk prediction model 1, Risk prediction model 4, Risk prediction model 5, Risk prediction model 6". Comparison shows that the number of candidate risk prediction models exceeds a preset threshold, so "Risk prediction model 1 - 0.85, Risk prediction model 5 - 0.82, Risk prediction model 6 - 0.86", which have higher matching degrees, are selected as the target risk prediction model.
[0094] This step calculates the matching degree between feature vectors and various risk prediction models through a gating routing mechanism to improve the accuracy of risk prediction model selection, ensuring that the prediction results reflect historical patterns and the impact of changes in management information. At the same time, the preset quantity threshold limits reduce computational overhead and enhance the robustness of the prediction results.
[0095] S104. Integrate at least one risk prediction result to obtain the target risk prediction result; based on the target risk prediction result and the preset resource management rules, determine the resource management strategy.
[0096] Based on the foregoing, for example, if the target risk prediction model is determined to be "risk prediction model 3 and risk prediction model 6", feature vector 1 is input into risk prediction model 3 to obtain risk prediction result 1; similarly, feature vector 2 is input into risk prediction model 6 to obtain risk prediction result 2; risk prediction result 1 and risk prediction result 2 are integrated and processed to obtain target risk prediction result a; based on target dividing line prediction result a and preset resource management rules, resource management strategy is determined.
[0097] In one possible implementation, a resource management strategy is determined based on the target risk prediction results and preset resource management rules. The specific process is as follows:
[0098] Using a multimodal rule engine, based on the target risk prediction results and preset resource management rules, at least one resource management rule corresponding to the target risk prediction results is determined; wherein, the preset resource management rules refer to the processing strategies for different target risk prediction results.
[0099] Generate a resource management strategy based on at least one resource management rule and economic resources.
[0100] Among them, a multimodal rule engine refers to an intelligent decision-making system that combines structured and unstructured rules. For example, through a multimodal rule engine, the resource management rules corresponding to the target risk prediction result a are determined from the preset resource management rules as "processing strategy 1 and processing strategy 2". Based on the current economic resources, a resource management strategy a that conforms to the current economic resource situation is generated based on the obtained processing strategy 1 and processing strategy 2.
[0101] This step improves the real-time performance and accuracy of resource management rule selection by introducing a multimodal rule engine, thereby reducing the interference of human error on resource management.
[0102] This embodiment provides a method for processing multi-source data. It involves acquiring business transaction data and business environment data; performing feature extraction on the business transaction data and business environment data to obtain feature vectors; determining at least one target risk prediction model based on the feature vectors and a set of risk prediction models; inputting the feature vectors into each of the at least one target risk prediction model to obtain at least one risk prediction result; integrating the at least one risk prediction result to obtain a target risk prediction result; and determining a resource management strategy based on the target risk prediction result and preset resource management rules. This application expands the coverage of risk prediction data by characterizing business transaction data and business environment data to obtain feature vectors. Simultaneously, determining multiple risk prediction models based on feature vectors avoids the limitations of a single model for complex risk scenarios. Determining a resource management strategy based on the integrated target risk prediction result and preset resource management rules not only avoids reliance on manual experience in allocating rules but also improves the accuracy of strategy generation.
[0103] In one possible implementation, the method further includes:
[0104] The resource management strategy is transmitted to the execution system so that the execution system can execute the resource management strategy and report the strategy loss data.
[0105] The target risk prediction model is updated and trained based on the strategy loss data.
[0106] In existing strategy generation schemes, the lack of a feedback mechanism for actual loss data prevents risk prediction models from self-optimizing, leading to a decline in prediction accuracy over time. To address this issue, this application establishes an execution feedback mechanism that receives strategy loss data from resource management strategy execution and uses this data to backward optimize and train the corresponding risk prediction model.
[0107] In one possible implementation, the target risk prediction model is updated and trained based on the policy loss data, including:
[0108] The policy loss data is decomposed into multiple sub-objective data through a multi-objective optimization algorithm;
[0109] By employing a Pareto front optimization strategy, the target risk prediction model is updated and trained based on multiple sub-target data and feature vectors.
[0110] Among them, multi-objective optimization algorithms refer to breaking down policy loss data into multiple quantifiable sub-objectives. For example, policy loss data can be broken down into "sub-objective 1 - minimizing default loss" and "sub-objective 2 - minimizing opportunity loss". Pareto front optimization strategies are used to determine the optimal equilibrium solution corresponding to the sub-objectives.
[0111] Based on the foregoing, for example, the execution system executes resource management strategy a and returns strategy loss data as "strategy loss data A". Using a multi-objective optimization algorithm, strategy loss data A is broken down into "sub-objective 1 - minimizing default loss" and "sub-objective 2 - minimizing opportunity loss". The feature vector is input into risk prediction model 3, and the multi-objective optimization algorithm iteratively calculates "sub-objective 1 - minimizing default loss" and "sub-objective 2 - minimizing opportunity loss" to generate a Pareto front curve. Based on specific business transaction requirements, the optimal solution is selected from the Pareto front curve. Based on the "sub-objective data corresponding to the optimal solution" selected by the Pareto front, risk prediction model 3 is retrained to adapt to the "dual-objective balance". Similarly, risk prediction model 6 is updated and trained based on the above method.
[0112] This step, through a multi-objective optimization algorithm, avoids the model from neglecting other risk dimensions due to overfitting to a single risk indicator; it significantly enhances the comprehensive risk coverage capability of the risk prediction model, adapting to complex and ever-changing market environments. For example, in market volatility, by balancing the optimization objectives of credit risk and market risk, it not only improves the predictive ability of default risk but also avoids capital allocation imbalances caused by market risk prediction biases.
[0113] Figure 3 A flowchart illustrating a multi-source data processing method provided in this application. Figure 3 ,like Figure 3As shown, multi-source data is acquired, including business transaction data and business environment data. The business transaction data and business environment data are then processed to obtain feature vectors. A gating routing mechanism is used to calculate the matching degree between the feature vectors and each risk prediction model. Risk prediction models with matching degrees higher than a preset matching degree are selected as candidate risk prediction models. If the number of candidate risk prediction models does not exceed a preset threshold, the candidate risk prediction model is used as the target prediction model. If the number of candidate risk prediction models exceeds the preset threshold, the candidate risk prediction model with the highest matching degree (the preset threshold) is selected as the target prediction model.
[0114] Feature vectors are input into a target risk prediction model to obtain at least one risk prediction result. These risk prediction results are then integrated to obtain the target risk prediction result. Based on the target risk prediction result and preset resource management rules, a resource management strategy is determined. The resource management strategy is transmitted to the execution system, which executes the strategy and provides feedback on strategy loss data. Based on the strategy loss data, the target risk prediction model is updated and trained. This application achieves this by mechanically fusing multi-source data, making the factors covered by the feature vectors more comprehensive. Simultaneously, dynamically determining the target risk prediction model based on the feature vectors improves the comprehensiveness of risk coverage and avoids dependence on a single model. A multimodal rule engine ensures consistency and accuracy between the processing strategy and the risk prediction results. Data feedback from the execution system drives model self-optimization, reducing human intervention errors and forming an iterative self-learning closed-loop architecture, effectively and comprehensively improving the accuracy, timeliness, and reliability of economic resource management.
[0115] Figure 4 This application provides a schematic diagram of the structure of a multi-source data processing device, as shown below. Figure 4 As shown, the multi-source data processing apparatus 400 provided in this application embodiment includes:
[0116] Module 401 is used to acquire business transaction data and business environment data;
[0117] Processing module 402 is used to perform feature extraction processing on business transaction data and business environment data to obtain feature vectors;
[0118] The processing module 402 is also used to determine at least one target risk prediction model based on the feature vector and the set of risk prediction models; and to input the feature vector into the at least one target risk prediction model to obtain at least one risk prediction result.
[0119] The generation module 403 is used to integrate and process at least one risk prediction result to obtain a target risk prediction result; and to determine a resource management strategy based on the target risk prediction result and preset resource management rules.
[0120] In one possible implementation, the processing module 402 is further configured to:
[0121] The business environment data is parsed and processed to obtain the keyword features corresponding to the business environment data; the business transaction data is processed to extract features to obtain the transaction features corresponding to the business transaction data.
[0122] The keyword features and transaction features are concatenated to obtain a feature vector.
[0123] In one possible implementation, the processing module 402 is further configured to:
[0124] The matching degree between the feature vector and each risk prediction model in the risk prediction model set is calculated through a gated routing mechanism.
[0125] Risk prediction models with a matching degree greater than the preset matching degree are selected as candidate risk prediction models.
[0126] Determine whether the number of candidate risk prediction models exceeds a preset threshold; if the number of candidate risk prediction models does not exceed the preset threshold, then multiple candidate risk prediction models are used as the target risk prediction model.
[0127] If the number of candidate risk prediction models exceeds a preset threshold, the candidate risk prediction model with the highest matching degree (the one that matches the preset threshold) will be selected as the target risk prediction model.
[0128] In one possible implementation, the generation module 403 is further configured to:
[0129] Using a multimodal rule engine, based on the target risk prediction results and preset resource management rules, at least one resource management rule corresponding to the target risk prediction results is determined; wherein, the preset resource management rules refer to the processing strategies for different target risk prediction results.
[0130] Generate a resource management strategy based on at least one resource management rule and economic resources.
[0131] In one possible implementation, the processing module 402 is further configured to:
[0132] The resource management strategy is transmitted to the execution system so that the execution system can execute the resource management strategy and report the strategy loss data.
[0133] The target risk prediction model is updated and trained based on the strategy loss data.
[0134] In one possible implementation, the processing module 402 is further configured to:
[0135] The policy loss data is decomposed into multiple sub-objective data through a multi-objective optimization algorithm;
[0136] By employing a Pareto front optimization strategy, the target risk prediction model is updated and trained based on multiple sub-target data and feature vectors.
[0137] The multi-source data processing device provided in this embodiment can execute the method provided in the above method embodiment. Its implementation principle and technical effect are similar, and will not be described in detail here.
[0138] Figure 5 This is a schematic diagram of the structure of an electronic device provided in this application. Figure 5 As shown, the electronic device 50 provided in this embodiment includes at least one processor 501 and a memory 502. Optionally, the device 50 further includes a communication component 503. The processor 501, memory 502, and communication component 503 are connected via a bus 504.
[0139] In a specific implementation, at least one processor 501 executes computer execution instructions stored in memory 502, causing at least one processor 501 to perform the above-described method.
[0140] The specific implementation process of processor 501 can be found in the above method embodiments, and its implementation principle and technical effect are similar. It will not be repeated here.
[0141] This application also provides a computer-readable storage medium storing computer-executable instructions, which, when executed by a processor, implement the above-described method; its implementation principle and technical effect are similar and will not be described in detail here.
[0142] This application also provides a computer program product, including a computer program that, when executed by a processor, implements the technical solution of the above method embodiments. Its implementation principle and technical effects are similar, and will not be repeated here.
[0143] 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 all optional embodiments, and the actions and modules involved are not necessarily essential to this application.
[0144] It should be further noted that although the steps in the flowchart are shown sequentially according to the arrows, these steps are not necessarily executed in the order indicated by the arrows. Unless explicitly stated herein, there is no strict order restriction on the execution of these steps, and they can be executed in other orders. Moreover, at least some steps in the flowchart may include multiple sub-steps or multiple stages. These sub-steps or stages are not necessarily completed at the same time, but can be executed at different times. The execution order of these sub-steps or stages is not necessarily sequential, but can be performed alternately or in turn with other steps or at least some of the sub-steps or stages of other steps.
[0145] It should be understood that the above-described device embodiments are merely illustrative, and the device of this application can also be implemented in other ways. For example, the division of units / modules in the above embodiments is only a logical functional division, and there may be other division methods in actual implementation. For example, multiple units, modules, or components may be combined, or integrated into another system, or some features may be ignored or not executed.
[0146] Furthermore, unless otherwise specified, the functional units / modules in the various embodiments of this application can be integrated into one unit / module, or each unit / module can exist physically separately, or two or more units / modules can be integrated together. The integrated units / modules described above can be implemented in hardware or as software program modules.
[0147] When integrated units / modules are implemented in hardware, the hardware can be digital circuits, analog circuits, etc. The physical implementation of the hardware structure includes, but is not limited to, transistors, memristors, etc. Unless otherwise specified, the processor can be any suitable hardware processor, such as a CPU, GPU, FPGA, DSP, and ASIC, etc. Unless otherwise specified, the storage unit can be any suitable magnetic or magneto-optical storage medium, such as Resistive Random Access Memory (RRAM), Dynamic Random Access Memory (DRAM), Static Random Access Memory (SRAM), Enhanced Dynamic Random Access Memory (EDRAM), High-Bandwidth Memory (HBM), Hybrid Memory Cube (HMC), etc.
[0148] If the integrated unit / module is implemented as a software program module and sold or used as an independent product, it can be stored in a computer-readable storage device (CMD). Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a memory and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods of the various embodiments of this application. The aforementioned memory includes various media capable of storing program code, such as a USB flash drive, read-only memory (ROM), random access memory (RAM), portable hard drive, magnetic disk, or optical disk.
[0149] In the above embodiments, the descriptions of each embodiment have their own emphasis. For parts not described in detail in a certain embodiment, please refer to the relevant descriptions of other embodiments. The technical features of the above embodiments can be combined arbitrarily. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as the combination of these technical features does not contradict each other, it should be considered within the scope of this specification.
[0150] Other embodiments of this application will readily occur to those skilled in the art upon consideration of the specification and practice of the invention disclosed herein. This application is intended to cover any variations, uses, or adaptations of this application that follow the general principles of this application and include common knowledge or customary techniques in the art not disclosed herein. The specification and examples are to be considered exemplary only, and the true scope and spirit of this application are indicated by the following claims.
[0151] It should be understood that this application is not limited to the precise structure described above and shown in the accompanying drawings, and various modifications and changes can be made without departing from its scope. The scope of this application is limited only by the appended claims.
Claims
1. A method for processing multi-source data, characterized in that, The method includes: Acquire business transaction data and business environment data; Feature extraction processing is performed on the business transaction data and the business environment data to obtain feature vectors; Based on the feature vectors and the set of risk prediction models, at least one target risk prediction model is determined; the feature vectors are then input into the at least one target risk prediction model to obtain at least one risk prediction result. The at least one risk prediction result is integrated and processed to obtain the target risk prediction result; based on the target risk prediction result and the preset resource management rules, a resource management strategy is determined.
2. The method according to claim 1, characterized in that, The step of performing feature extraction processing on the business transaction data and the business environment data to obtain a feature vector includes: The business environment data is parsed to obtain the keyword features corresponding to the business environment data; the business transaction data is subjected to feature extraction to obtain the transaction features corresponding to the business transaction data. The keyword features and the transaction features are concatenated to obtain the feature vector.
3. The method according to claim 1, characterized in that, The step of determining at least one target risk prediction model based on the feature vector and the set of risk prediction models includes: The matching degree between the feature vector and each risk prediction model in the risk prediction model set is calculated through a gating routing mechanism. The risk prediction model with a matching degree greater than the preset matching degree is used as the candidate risk prediction model; Determine whether the number of candidate risk prediction models exceeds a preset threshold; if the number of candidate risk prediction models does not exceed the preset threshold, then multiple candidate risk prediction models are used as target risk prediction models. If the number of candidate risk prediction models exceeds a preset threshold, the candidate risk prediction model with the highest matching degree is selected as the target risk prediction model.
4. The method according to claim 1, characterized in that, The step of determining a resource management strategy based on the target risk prediction results and preset resource management rules includes: Using a multimodal rule engine, based on the target risk prediction result and preset resource management rules, at least one resource management rule corresponding to the target risk prediction result is determined; wherein, the preset resource management rule refers to the processing strategy for different target risk prediction results; The resource management strategy is generated based on the at least one resource management rule and economic resources.
5. The method according to any one of claims 1-4, characterized in that, The method further includes: The resource management strategy is transmitted to the execution system, so that the execution system executes the resource management strategy and feeds back the strategy loss data; Based on the strategy loss data, the target risk prediction model is updated and trained.
6. The method according to claim 5, characterized in that, The step of updating and training the target risk prediction model based on the strategy loss data includes: The strategy loss data is decomposed into multiple sub-objective data using a multi-objective optimization algorithm; The target risk prediction model is updated and trained using a Pareto front optimization strategy based on the multiple sub-target data and the feature vector.
7. A multi-source data processing device, characterized in that, include: The acquisition module is used to acquire business transaction data and business environment data; The processing module is used to perform feature extraction processing on the business transaction data and the business environment data to obtain feature vectors; The processing module is further configured to determine at least one target risk prediction model based on the feature vector and the set of risk prediction models; and to input the feature vector into the at least one target risk prediction model to obtain at least one risk prediction result. A generation module is used to integrate and process the at least one risk prediction result to obtain a target risk prediction result; Based on the target risk prediction results and the preset resource management rules, a resource management strategy is determined.
8. An electronic device, characterized in that, include: A processor, and a memory communicatively connected to the processor; The memory stores computer-executed instructions; The processor executes computer execution instructions stored in the memory to implement the method as described in any one of claims 1 to 6.
9. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores computer-executable instructions, which, when executed by a processor, are used to implement the method as described in any one of claims 1 to 6.
10. A computer program product, characterized in that, Includes a computer program that, when executed by a processor, implements the method of any one of claims 1 to 6.