Data processing method and device, equipment, storage medium and program product

Abstract

This application provides a data processing method, apparatus, device, storage medium, and program product, relating to fintech and other fields. The method includes: obtaining a target financial qualification scoring model based on a first financial qualification scoring model and a second financial qualification scoring model. The first financial qualification scoring model is trained based on merchant profile data of a first participant, and the second financial qualification scoring model is trained based on merchant operating data of a second participant. The first and second financial qualification scoring models have the same model structure. The target financial qualification scoring model is then distributed to the target participant. This method improves the accuracy of banks' assessment of merchants' financial qualifications.

Description

Technical Field

[0001] This application relates to financial technology and other fields, and in particular to a data processing method, apparatus, device, storage medium and program product. Background Technology

[0002] In the digital economy era, small and micro enterprises, especially individual business owners such as small restaurants and retailers in cities, have long faced difficulties in obtaining financing due to a lack of traditional collateral and credit records. Their transaction records, customer behavior, and cash flow data are scattered across third-party payment platforms and bank systems, forming "data silos." Due to high data privacy protection requirements, banks cannot directly access the raw data from third-party platforms, while these platforms lack customer profile data from banks. This results in banks having difficulty accurately assessing the financial qualifications of merchants.

[0003] Therefore, improving the accuracy of banks' assessment of merchants' financial qualifications is an urgent problem to be solved. Summary of the Invention

[0004] This application provides a data processing method, apparatus, device, storage medium, and program product to solve the technical problem of poor accuracy in banks' assessment of merchants' financial qualifications.

[0005] In a first aspect, this application provides a data processing method applied to a server, comprising:

[0006] Based on the first financial qualification scoring model and the second financial qualification scoring model, a target financial qualification scoring model is obtained. The first financial qualification scoring model is trained based on the merchant profile data of the first participant, and the second financial qualification scoring model is trained based on the merchant operation data of the second participant. The model structures of the first financial qualification scoring model and the second financial qualification scoring model are the same.

[0007] The target financial qualification scoring model will be distributed to the target participants.

[0008] Optionally, the step of obtaining the target financial qualification scoring model based on the first financial qualification scoring model trained on the merchant profile data of the first participant and the second financial qualification scoring model trained on the merchant operating data of the second participant includes:

[0009] Obtain the first model parameters of the first financial qualification scoring model and the second model parameters of the second financial qualification scoring model;

[0010] The target financial qualification scoring model is generated based on the first model parameters and the second model parameters.

[0011] Optionally, generating the target financial qualification scoring model based on the first model parameters and the second model parameters includes:

[0012] Federated learning is performed based on the first model parameters and the second model parameters to generate the target financial qualification scoring model.

[0013] Optionally, the first model parameters and the second model parameters are data encrypted using the target encryption algorithm.

[0014] In a first aspect, this application provides a data processing method applied to a first participating party, comprising:

[0015] The target financial qualification scoring model is obtained from the server. The target financial qualification scoring model is generated based on the first financial qualification scoring model trained on the merchant profile data of the first participant and the second financial qualification scoring model trained on the merchant operation data of the second participant.

[0016] The target financial qualification scoring model is used to select target merchants that meet the target financial qualifications from the candidate merchants.

[0017] Optional, also includes:

[0018] Obtain the merchant profile data;

[0019] The first financial qualification scoring model is generated based on the merchant profile data.

[0020] Optionally, generating the first financial qualification scoring model based on the merchant profile data includes:

[0021] Information gain calculation and weight filtering are performed on the features in the merchant profile data to obtain a high-weight merchant profile feature set.

[0022] The high-weight merchant profile feature set is input into the initial model framework built based on the XGBoost algorithm for ensemble tree training, so as to segment features and calculate training residuals through multiple decision trees in the initial model framework.

[0023] The parameters of the initial model framework are adjusted based on the training residuals until the training results meet the preset stopping conditions, thereby obtaining the first financial qualification scoring model.

[0024] Optionally, obtaining the merchant profile data includes:

[0025] Obtain initial merchant profile data;

[0026] The initial merchant profile data is filtered based on the key features corresponding to the target financial qualification to obtain the merchant profile data. The feature filtering is used to remove redundant and noisy features from the initial merchant profile data.

[0027] Thirdly, this application provides a data processing apparatus for use on a server, comprising:

[0028] The processing module is used to obtain a target financial qualification scoring model based on the first financial qualification scoring model and the second financial qualification scoring model. The first financial qualification scoring model is trained based on the merchant profile data of the first participant, and the second financial qualification scoring model is trained based on the merchant operation data of the second participant. The model structures of the first financial qualification scoring model and the second financial qualification scoring model are the same.

[0029] The sending module is used to distribute the target financial qualification scoring model to the target participants.

[0030] Fourthly, this application provides a data processing apparatus applied to a first participant, comprising:

[0031] The acquisition module is used to acquire the target financial qualification scoring model from the server. The target financial qualification scoring model is generated based on the first financial qualification scoring model trained on the merchant profile data of the first participant and the second financial qualification scoring model trained on the merchant operation data of the second participant.

[0032] The processing module is used to select target merchants that meet the target financial qualifications from the candidate merchants using the target financial qualification scoring model.

[0033] Fifthly, this application provides an electronic device, including: a processor and a memory; the processor and the memory are communicatively connected;

[0034] The memory stores computer-executed instructions;

[0035] The processor executes computer execution instructions stored in the memory to implement the method as described in any one of the first or second aspects.

[0036] In a sixth aspect, this application provides a computer-readable storage medium storing computer-executable instructions, which, when executed by a processor, are used to implement the data processing method as described in any one of the first or second aspects.

[0037] In a seventh aspect, this application provides a computer program product, which, when executed by a processor, is used to implement the data processing method as described in any one of the first or second aspects.

[0038] The data processing method, apparatus, equipment, storage medium, and program products provided in this application obtain a target financial qualification scoring model through a server based on a first financial qualification scoring model and a second financial qualification scoring model. The first financial qualification scoring model is trained based on merchant profile data from a first participant, and the second financial qualification scoring model is trained based on merchant operating data from a second participant. The model structures of the first and second financial qualification scoring models are identical. The target financial qualification scoring model is then distributed to the first and / or second participants, enabling them to use the model to select target merchants that meet the target financial qualifications from candidate merchants. This allows for in-depth analysis of the merchants' financial attribute characteristics through multi-source data, resulting in more accurate analysis results. Based on these more accurate results, the accuracy of banks' assessments of merchants' financial qualifications is improved. Attached Figure Description

[0039] 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.

[0040] Figure 1 This application provides a schematic diagram of a data processing scenario.

[0041] Figure 2 A flowchart illustrating a data processing method provided in an embodiment of this application;

[0042] Figure 3 A flowchart illustrating another data processing method provided in an embodiment of this application;

[0043] Figure 4 A flowchart illustrating another data processing method provided in an embodiment of this application;

[0044] Figure 5 A flowchart illustrating another data processing method provided in an embodiment of this application;

[0045] Figure 6 This is a schematic diagram of the structure of a data processing device provided in an embodiment of this application;

[0046] Figure 7 This is a schematic diagram of another data processing apparatus provided in an embodiment of this application;

[0047] Figure 8This is a schematic diagram of the structure of an electronic device provided in an embodiment of this application.

[0048] 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

[0049] 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.

[0050] 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.

[0051] 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.

[0052] It should be noted that the data processing methods, apparatus, devices, storage media, and program products provided in this application can be used in financial technology and other fields, as well as in any field other than financial technology and other fields. The application fields of the data processing methods, apparatus, devices, storage media, and program products in this application are not limited.

[0053] In today's rapidly advancing fintech landscape, data integration between banks and third-party platforms is crucial for improving the accuracy of financial services and optimizing risk management. Currently, data integration between banks and third-party platforms primarily relies on the following two methods:

[0054] Method 1: Banks acquire raw data held by third-party platforms through agreements. This raw data covers various customer information on the platforms, such as spending records and social behavior data. After successfully acquiring the data, banks centrally process this raw data from different channels. For example, banks can use data analytics to uncover the potential value within the data, thereby building scoring models for assessing customer credit, predicting business risks, or developing marketing strategies.

[0055] However, banks encounter a large amount of sensitive customer information, such as personal identification details and financial status, when acquiring, storing, and processing raw data from third-party platforms. Vulnerabilities in data security management can easily lead to data breaches, severely damaging customers' privacy and rights. Furthermore, third-party platforms are extremely cautious about providing raw data due to compliance requirements. With increasingly stringent data protection laws and regulations, third-party platforms often refuse to provide raw data to banks to avoid legal liability for data breaches. This makes it difficult for banks to obtain sufficient quantity and quality of raw data in practice, thus affecting the effectiveness of data fusion and the accuracy of scoring models.

[0056] Method 2: Achieve multi-source data fusion through joint modeling using data anonymization or desensitization. Data anonymization involves transforming the original data, for example, replacing customers' real names with random codes using substitution techniques; and encrypting sensitive financial information using encryption techniques. Anonymization completely removes personally identifiable information from the data, making it impossible to directly associate the data with any specific individual. After this processing, banks and third-party platforms can jointly analyze and model the anonymized or desensitized data without directly exchanging the original data. Both parties use specific algorithms and models to uncover correlations and patterns in the data to construct scoring models suitable for business needs, thereby achieving data fusion and utilization.

[0057] However, data anonymized or de-identified may lose some key characteristic information. For example, transaction time distribution information reflects the level of customer spending activity at different times, which is of great reference value for assessing customer credit risk and spending habits; customer behavior path information can reveal customers' operating processes and preferences on the platform, helping to gain a deeper understanding of customer needs and behavioral patterns. However, after anonymization or de-identification, this key characteristic information may be destroyed or lost, causing models built on this data to fail to accurately capture customers' true characteristics and behavioral patterns, thus significantly reducing the effectiveness of the models.

[0058] In view of this, this application provides a data processing method in which a server obtains a target financial qualification scoring model based on a first financial qualification scoring model and a second financial qualification scoring model. The first financial qualification scoring model is trained based on merchant profile data from a first participant, and the second financial qualification scoring model is trained based on merchant operating data from a second participant. The model structures of the first and second financial qualification scoring models are identical. The target financial qualification scoring model is then distributed to the first and / or second participants, enabling them to use it to select target merchants that meet the target financial qualifications from candidate merchants. This allows for in-depth analysis of the financial attribute characteristics of merchants through multi-source data, resulting in more accurate analysis results. Based on these more accurate results, the accuracy of banks' assessments of merchants' financial qualifications is improved.

[0059] Figure 1 This is a schematic diagram illustrating a data processing scenario provided in this application. Figure 1 As shown, the scenario includes: a server, a first participant, and a second participant.

[0060] In this embodiment, the first participant can be, for example, a bank or financial institution, and the second participant can be, for example, a third-party trading platform. Optionally, the first participant can also be a third-party trading platform, and the second participant can be, for example, a bank or financial institution. Subsequent embodiments of this application will be described using the example of a bank as the first participant and a third-party trading platform as the second participant.

[0061] The server is used to integrate the first financial qualification scoring model of the first participant and the second financial qualification scoring model of the second participant to obtain a target financial qualification scoring model that comprehensively considers multi-dimensional information such as merchant profile data and operational data, possesses more comprehensive evaluation capabilities, and can accurately reflect the merchant's true financial qualification status. This server can be deployed at the location of the first or second participant, or it can be deployed outside of the first or second participant. The server can be a local server or a cloud server.

[0062] The server can distribute the target financial qualification scoring model to the first participant and / or the second participant, so that the first participant and / or the second participant can obtain a more accurate merchant financial qualification score based on the target financial qualification scoring model.

[0063] Below, with Figure 1Taking the illustrated scenario architecture as an example, the technical solution of this application and how the technical solution of this application solves the above-mentioned technical problems will be described in detail through specific embodiments. The following 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 be described below with reference to the accompanying drawings.

[0064] Figure 2 This is a flowchart illustrating a data processing method provided in an embodiment of this application. Figure 2 As shown, the method may include:

[0065] S201. The server obtains the target financial qualification scoring model based on the first financial qualification scoring model and the second financial qualification scoring model.

[0066] The first financial qualification scoring model is trained based on the merchant profile data of the first participant, and the second financial qualification scoring model is trained based on the merchant operation data of the second participant. The model structures of the first financial qualification scoring model and the second financial qualification scoring model are the same.

[0067] In this embodiment, within the financial services scenario, the first and second participants each have their own focus regarding data resources. The first participant (i.e., the bank) possesses multi-dimensional customer data, enabling it to construct a comprehensive, accurate, and dynamic customer profile, including basic information (such as age, gender, and occupation), transaction records, behavioral preferences, and online banking login frequency. Therefore, the first participant also possesses merchant profile data, which may include, for example, merchant registration information, credit history, and market positioning.

[0068] Since the second participant is a third-party transaction platform, it has a large amount of merchant operating data, which may include: data reflecting the degree of tightness of funds (such as the percentage of full withdrawals in the previous month out of all withdrawals in the previous month, the number of days the loan homepage was viewed in the past 30 days), data reflecting the operating income status (such as the average amount of successful transactions in the past 3 months), and data reflecting the authenticity of transactions (such as the percentage of transactions in different time periods), etc.

[0069] Based on merchant profile data, the first participant can use machine learning algorithms to train a first-level financial qualification scoring model. The trained model can then output a score for the merchant at a specific financial qualification level, such as a loan score, credit risk score, or development potential score, based on the input merchant profile data.

[0070] The second participant uses the same model structure as the first participant, and trains the model based on merchant operating data to obtain a second financial qualification scoring model. The trained second financial qualification scoring model can output a score for a merchant under a specific financial qualification evaluation level based on the input merchant operating data.

[0071] After obtaining the first and second financial qualification scoring models, the server can fuse these two models to obtain the target financial qualification scoring model. For example, the target financial qualification scoring model can be obtained by fusing the model parameters of the first and second financial qualification scoring models; alternatively, model transfer learning techniques in deep learning can be used to use one model as the basic architecture and transfer the key features or knowledge of the other model to that architecture. After fine-tuning and training, a target financial qualification scoring model that combines the advantages of both models can be obtained.

[0072] S202. The server distributes the target financial qualification scoring model to the target participants.

[0073] The target participants include at least one of the first participant and the second participant. The server distributes the target financial qualification scoring model to the target participants, enabling them to use the model to select target merchants that meet the target financial qualifications from the candidate merchants.

[0074] After successfully obtaining the target financial qualification scoring model, the server can distribute it to the target participant. Assume the target participant is the first participant. The first participant can input the merchant profile data and / or operational data of these candidate merchants into the target financial qualification scoring model. The target financial qualification scoring model processes the input data and outputs the financial qualification score result for each candidate merchant.

[0075] Then, the first participating party can screen the financial qualification scores according to the preset target financial qualification standards, and select the candidate merchants whose financial qualification scores meet the preset target financial qualification standards as target merchants that meet the target financial qualifications.

[0076] For example, suppose the target financial qualification scoring model is used by the first participant to determine the loan eligibility, and the results generated are shown in Table 1 below:

[0077] Table 1

[0078]

[0079] The primary participant can determine its preset target financial qualification standards based on the sample proportions of each segment. For example, the preset target financial qualification standard can be set to only consider merchants with financial qualification scores <0.399 when making loans. Therefore, merchants with financial qualification scores <0.399 can be selected from the candidate merchants as target merchants, thereby controlling the probability of loan default.

[0080] The method provided in this application embodiment obtains a target financial qualification scoring model through a server based on a first financial qualification scoring model and a second financial qualification scoring model. The first financial qualification scoring model is trained based on merchant profile data from a first participant, and the second financial qualification scoring model is trained based on merchant operating data from a second participant. The model structures of the first and second financial qualification scoring models are identical. The target financial qualification scoring model is then distributed to the first and / or second participants, enabling them to use the target financial qualification scoring model to filter target merchants that meet the target financial qualifications from candidate merchants. This allows for in-depth analysis of the financial attribute characteristics of merchants through multi-source data, resulting in more accurate analysis results of merchant financial attribute characteristics. Based on these more accurate results, the accuracy of banks' judgments on the financial qualifications of merchants is improved.

[0081] The following section will take the method of fusion through model parameters as an example to explain in detail how the server obtains the target financial qualification scoring model in the aforementioned step S201 based on the first financial qualification scoring model trained on the merchant profile data of the first participant and the second financial qualification scoring model trained on the merchant operation data of the second participant.

[0082] Figure 3 This is a flowchart illustrating another data processing method provided in an embodiment of this application. Figure 3 As shown, the aforementioned step S201 may specifically include:

[0083] S301. The server obtains the first model parameters of the first financial qualification scoring model and the second model parameters of the second financial qualification scoring model.

[0084] In this step, the first model parameter of the first financial qualification scoring model can be actively sent to the server by the first participant, or it can be fed back to the server by the first participant after the server sends a request to the first participant.

[0085] Similarly, the second model parameters of the second financial qualification scoring model can be actively sent to the server by the second participant, or they can be returned to the server by the second participant after the server sends a request to the second participant.

[0086] S302. The server generates the target financial qualification scoring model based on the first model parameters and the second model parameters.

[0087] One possible implementation is that the server can perform a weighted average of the first and second model parameters. After obtaining the first and second model parameters, the server assigns a weight to each parameter. Assume the weight of the first model parameter is α, and the weight of the second model parameter is β, with α + β = 1. This weight allocation can be adjusted according to different needs and data characteristics. For example, if merchant profile data is considered more important for assessing financial qualifications, α can be set larger; conversely, if operational data is more important, β can be set larger. The server can perform a weighted average calculation on each corresponding parameter in the first and second model parameters to ultimately obtain the parameters of the target financial qualification scoring model, thereby generating the target financial qualification scoring model.

[0088] Another possible implementation is that the server can perform federated learning based on the first model parameters and the second model parameters to generate a target financial qualification scoring model.

[0089] Specifically, federated learning is a technique for training models without sharing the original data. After obtaining the first and second model parameters, the server can first initialize these parameters to a suitable initial state for federated learning.

[0090] The server then sends these initial parameters back to both the first and second participants. The first participant uses its local merchant profile data, combined with the received initial parameters, to perform local training on the first financial qualification scoring model. During this local training, the first participant updates the model's parameters, obtaining a new set of local model parameters. Similarly, the second participant uses its local merchant operating data to perform local training on the second financial qualification scoring model, also obtaining a new set of local model parameters.

[0091] The first and second participants send the updated local model parameters back to the server. Upon receiving these parameters, the server performs an aggregation operation. Aggregation can be a simple average or a weighted average based on different strategies. For example, different weights can be assigned based on factors such as the quality and quantity of each participant's data. After aggregation, the server obtains a new set of global model parameters.

[0092] The server sends the new global model parameters back to the first and second participants, allowing them to continue local training. This process iterates until a preset stopping condition is met, such as the model's performance stabilizing on the validation set or reaching a preset number of iterations. Ultimately, the global model parameters obtained by the server become the parameters of the target financial literacy scoring model, thus generating the target financial literacy scoring model.

[0093] The method provided in this application, by processing and fusing different model parameters, fully integrates the data advantages of the first and second participants, thereby obtaining a more accurate and comprehensive target financial qualification scoring model, and thus providing a more reliable basis for the target participants to screen target merchants.

[0094] In one possible implementation, the first model parameters and the second model parameters can be encrypted using the target encryption algorithm and then sent to the server.

[0095] In this implementation, to protect data security and privacy, the first and second participants can encrypt the first and second model parameters using a target encryption algorithm when sending them to the server. Assuming the target encryption algorithm is a symmetric encryption algorithm, the first participant and the server can share an encryption key beforehand. After obtaining the first model parameters, the first participant uses this key to encrypt the parameters and sends the encrypted parameters to the server. Upon receiving the encrypted parameters, the server decrypts them using the same key, thereby obtaining the original first model parameters.

[0096] Similarly, the second participant can also use the same encryption key to encrypt the second model parameters and send them to the server. The server then decrypts them to obtain the second model parameters. This encrypted transmission method can prevent the model parameters from being stolen or tampered with during transmission, ensuring the security and integrity of the data.

[0097] Optionally, noise can be added to the first and second model parameters sent to further improve the security and privacy of the model parameters.

[0098] The method provided in this application employs a target encryption algorithm to encrypt and transmit the first model parameters and the second model parameters, ensuring the security and integrity of the parameters during transmission. At the same time, noise is added to the parameters to further enhance their security and privacy, so as to achieve secure sharing and fusion of model parameters under the premise of ensuring data security and privacy, thereby laying the foundation for obtaining an accurate and reliable target financial qualification scoring model.

[0099] The following section provides a detailed explanation of how the first participating party generates the first financial qualification scoring model. Figure 4This is a flowchart illustrating another data processing method provided in an embodiment of this application. Figure 4 As shown, the method may further include:

[0100] S401, The first participant obtains merchant profile data.

[0101] The first participating party can obtain merchant profile data from merchant registration information systems, credit rating agency data, market research data, and other sources. This merchant profile data may include, for example, the aforementioned basic information (such as age, gender, and occupation), transaction records, behavioral preferences, and number of online banking logins.

[0102] Optionally, to improve data quality and validity, the primary participant can preprocess the merchant profile data after obtaining it. For example, the data can be cleaned to remove duplicates, errors, and missing values. For missing values, interpolation methods (such as mean interpolation or median interpolation) can be used to fill in the gaps. Then, the merchant profile data can be standardized to unify data with different features onto the same scale, preventing certain features from having an excessive impact on model training due to their large numerical range. After preprocessing, merchant profile data suitable for training the first financial qualification scoring model can be obtained.

[0103] S402. The first participant generates the first financial qualification scoring model based on the merchant profile data.

[0104] One possible implementation is that the first participant can use traditional machine learning algorithms, such as logistic regression, to generate a first financial qualification scoring model based on merchant profile data. Logistic regression is a simple and effective classification algorithm that linearly combines input feature data and maps the results to the [0, 1] interval using a logistic function, outputting a probability value representing the likelihood that a merchant possesses a certain financial qualification.

[0105] Another possible implementation is that the first participant can train a first financial qualification scoring model based on the XGBoost algorithm and merchant profile data. For example, this implementation can be achieved through the following sub-steps:

[0106] S4021. The first participant performs information gain calculation and weight filtering on the features in the merchant profile data to obtain a set of high-weight merchant profile features.

[0107] In this step, before training the model using the XGBoost algorithm, the first participant can filter the features in the merchant profile data to improve the training efficiency and accuracy of the model. Information gain is an indicator that measures the contribution of a feature to the classification result; it represents the degree to which the uncertainty of the classification result is reduced when a certain feature is known.

[0108] The first participant can begin by calculating the information gain value of each feature in the merchant profile data. For each feature, the information gain is calculated by comparing the impact on the financial qualification classification result with and without the feature. For example, for the feature "credit history," the information gain of this feature is obtained by calculating the change in uncertainty regarding the merchant's financial qualification classification when credit history information is known and unknown.

[0109] Based on the calculated information gain value, the first participant can assign a weight to each feature. A higher information gain value indicates a greater contribution of the feature to the financial qualification classification, and thus a higher weight is assigned. Then, the first participant uses a preset weight threshold to filter out features with weights higher than that threshold, forming a high-weight merchant profile feature set. The features in this high-weight merchant profile feature set are those with high importance for assessing financial qualifications.

[0110] S4022. The first participant inputs the high-weight merchant profile feature set into the initial model framework built based on the XGBoost algorithm for ensemble tree training, so as to segment features and calculate training residuals through multiple decision trees in the initial model framework.

[0111] XGBoost is an ensemble learning algorithm based on decision trees. It can perform classification and regression tasks by constructing multiple decision trees. The first participant first builds an initial model framework based on the XGBoost algorithm. This framework includes the basic structure of the decision tree and some initial parameters.

[0112] Then, the first participant inputs the set of high-weight merchant profile features into this initial model framework for ensemble tree training. During training, multiple decision trees are constructed sequentially. Each decision tree segments the data based on the input features, dividing the data into different nodes. For example, a single decision tree can use a discriminative feature as the root node. Based on the sample's performance on that feature, an optimal threshold is selected to divide the data into left and right parts. The left and right parts then select appropriate features according to the specific situation of the samples to continue splitting. After sufficient depth is reached, the splitting stops, and finally, the data in each leaf node has the same score and the same category, and the feature combination of each leaf node is consistent.

[0113] After each decision tree is built, the training residuals are calculated. The training residuals are the differences between the model's predictions and the actual financial literacy scores. For example, for a given merchant, the difference between the model's predicted financial literacy score and the actual financial literacy score is the training residual. These training residuals are then used as input to the next decision tree to further optimize the model's predictions.

[0114] S4023. The first participant adjusts the parameters of the initial model framework based on the training residuals until the training results meet the preset stopping conditions, thus obtaining the first financial qualification scoring model.

[0115] The first participant adjusts the parameters of the initial model framework based on the calculated training residuals. In the XGBoost algorithm, parameter adjustment is mainly achieved by optimizing the structure of the decision tree and the splitting rules of the nodes. For example, based on the magnitude of the training residuals, the splitting threshold of the decision tree on a certain feature is adjusted so that the next decision tree can better fit the training residuals.

[0116] This training process iterates continuously, building a new decision tree in each iteration and adjusting the model parameters based on the training residuals. The first participant sets a preset stopping condition, such as the sum of the training residuals being less than a certain threshold, or reaching a preset number of decision trees. When the training results meet the preset stopping condition, the training process ends, and the resulting model is the first financial qualification scoring model.

[0117] The method provided in this application involves a first participant acquiring raw merchant profile data from multiple channels and performing preprocessing and feature filtering to obtain high-quality training data. Then, a model is trained on the filtered data using the XGBoost algorithm. In the XGBoost algorithm, high-weight features are filtered through information gain, an ensemble tree is trained, and parameters are adjusted based on the training residuals to accurately generate a first financial qualification scoring model that can effectively assess a merchant's financial qualifications. This provides a reliable foundation for subsequent fusion with the second participant's model to comprehensively and accurately determine the merchant's financial qualifications.

[0118] Figure 5 This is a flowchart illustrating another data processing method provided in an embodiment of this application. Figure 5 As shown, the aforementioned step S401 may further include:

[0119] S501, The first participant obtains the initial merchant profile data.

[0120] The first participant can obtain initial merchant profile data through the method described in step S401 above. The only difference is that the data obtained through step S401 is called merchant profile data, while the data obtained in this step is called initial merchant profile data. The initial profile data obtained in this step still needs to undergo further feature filtering processing before it can be used to train the first financial qualification scoring model.

[0121] S502. The first participating party performs feature filtering on the initial merchant profile data based on the key characteristics corresponding to the target financial qualifications, and obtains the merchant profile data.

[0122] Feature filtering is used to remove redundant and noisy features from the initial merchant profile data.

[0123] After obtaining the initial merchant profile data, the primary participant can further filter out features closely related to the target financial qualifications. These key features are determined based on professional financial knowledge, industry experience, and an understanding of the target financial business. For example, when assessing a merchant's creditworthiness, key features might include their credit history, balance sheet, and operating cash flow; while when assessing a merchant's investment potential, key features might focus on market share, innovation capabilities, and industry trends.

[0124] To identify these key characteristics, the primary stakeholders can organize discussions with financial experts and data analysts to develop a list of key characteristics or a rule base, drawing on past business cases and statistical data analysis results. This list or rule base clarifies which characteristics are highly important for assessing the target financial qualifications and which can be disregarded.

[0125] Then, the first participating party can perform feature filtering on the initial merchant profile data based on a list of key features or a rule base. During the filtering process, each feature in the merchant profile data is iterated through to check if it is in the list of key features. If a feature is not in the list, it is considered a redundant or noisy feature and will be removed from the data.

[0126] Through feature filtering, the first participant can remove irrelevant information from the initial data, making the data more concise and effective. This not only helps improve the efficiency of subsequent model training and reduce the waste of computing resources, but also avoids the interference of redundant and noisy features on model performance, thereby improving the accuracy and reliability of the model. The final merchant profile data is a dataset containing only features closely related to the target financial qualifications. Each merchant in this dataset has a set of concise feature vectors describing their financial qualification-related information, providing a high-quality data foundation for the subsequent generation of the first financial qualification scoring model.

[0127] The method provided in this application removes redundant and noisy information by performing feature filtering on initial merchant profile data, thereby obtaining more targeted and effective merchant profile data. This lays a solid data foundation for accurately generating the first financial qualification scoring model. Furthermore, the entire data processing follows strict logic and standards. From data collection and preprocessing to feature filtering, each step is closely centered on the target financial qualification assessment, ensuring that the final model accurately reflects the merchant's true financial qualification status and provides reliable support for financial business decisions.

[0128] Correspondingly, the second participant can also generate a second financial qualification scoring model in the same way as the first participant. The only difference is that the data used is merchant operating data, which will not be elaborated on in this application.

[0129] In practical applications, the method provided in this application can be widely used in various fields such as credit approval, investment decision-making, and risk assessment of financial institutions, helping financial institutions to more accurately screen out qualified merchants, reduce financial risks, and improve business efficiency.

[0130] Figure 6 This is a schematic diagram of the structure of a data processing device provided in an embodiment of this application. Figure 6 As shown, the data processing device is applied to a server and may include: a processing module 11 and a sending module 12.

[0131] Processing module 11 is used to obtain a target financial qualification scoring model based on the first financial qualification scoring model and the second financial qualification scoring model. The first financial qualification scoring model is trained based on the merchant profile data of the first participant, and the second financial qualification scoring model is trained based on the merchant operation data of the second participant. The model structures of the first financial qualification scoring model and the second financial qualification scoring model are the same.

[0132] The sending module 12 is used to send the target financial qualification scoring model to the target participants.

[0133] Optionally, processing module 11 is specifically used to obtain the first model parameters of the first financial qualification scoring model and the second model parameters of the second financial qualification scoring model. Based on the first model parameters and the second model parameters, a target financial qualification scoring model is generated.

[0134] Optionally, the processing module 11 is specifically used to perform federated learning based on the first model parameters and the second model parameters to generate a target financial qualification scoring model.

[0135] Optionally, the first model parameter and the second model parameter are data encrypted using the target encryption algorithm.

[0136] The data processing apparatus provided in this application embodiment can execute the data processing method executed by the server in the above method embodiment. Its implementation principle and technical effect are similar, and will not be described again here.

[0137] Figure 7 This is a schematic diagram of another data processing apparatus provided in an embodiment of this application. Figure 7 As shown, the data processing device is applied to the first participant and may include: an acquisition module 21 and a processing module 22.

[0138] The acquisition module 21 is used to obtain the target financial qualification scoring model from the server. The target financial qualification scoring model is generated based on the first financial qualification scoring model trained on the merchant profile data of the first participant and the second financial qualification scoring model trained on the merchant operation data of the second participant.

[0139] Processing module 22 is used to select target merchants that meet the target financial qualifications from the candidate merchants using the target financial qualification scoring model.

[0140] Optionally, the acquisition module 21 is also used to acquire merchant profile data. The processing module 22 is also used to generate a first financial qualification scoring model based on the merchant profile data.

[0141] Optionally, processing module 22 is specifically used to calculate information gain and perform weight filtering on the features in the merchant profile data to obtain a high-weight merchant profile feature set. This high-weight merchant profile feature set is then input into an initial model framework built based on the XGBoost algorithm for ensemble tree training. This allows for feature segmentation using multiple decision trees within the initial model framework, and the calculation of training residuals. The parameters of the initial model framework are adjusted based on the training residuals until the training results meet a preset stopping condition, resulting in the first financial qualification scoring model.

[0142] Optionally, module 21 is used to acquire initial merchant profile data. Based on the key characteristics corresponding to the target financial qualifications, the initial merchant profile data is filtered to obtain merchant profile data. Feature filtering is used to remove redundant and noisy features from the initial merchant profile data.

[0143] The data processing apparatus provided in this application embodiment can execute the data processing method executed by the first participant in the above method embodiment. Its implementation principle and technical effect are similar, and will not be described again here.

[0144] Figure 8 This is a schematic diagram of an electronic device provided in an embodiment of this application. The electronic device is used to execute the aforementioned data processing method. Figure 8As shown, the electronic device 800 may include at least one processor 801, a memory 802, and a communication interface 803.

[0145] The memory 802 is used to store programs. Specifically, the program may include program code, which includes computer operation instructions.

[0146] The memory 802 may include high-speed RAM memory, and may also include non-volatile memory, such as at least one disk storage device.

[0147] The processor 801 is used to execute computer execution instructions stored in the memory 802 to implement the method described in the foregoing method embodiments. The processor 801 may be a CPU, an Application Specific Integrated Circuit (ASIC), or one or more integrated circuits configured to implement the embodiments of this application.

[0148] The processor 801 can communicate and interact with external devices through the communication interface 803. In specific implementations, if the communication interface 803, memory 802, and processor 801 are implemented independently, they can be interconnected via a bus to complete communication. The bus can be an Industry Standard Architecture (ISA) bus, a Peripheral Component Interconnect (PCI) bus, or an Extended Industry Standard Architecture (EISA) bus, etc. Buses can be categorized as address buses, data buses, control buses, etc., but this does not imply that there is only one bus or one type of bus.

[0149] Optionally, in a specific implementation, if the communication interface 803, memory 802, and processor 801 are integrated on a single chip, then the communication interface 803, memory 802, and processor 801 can communicate through an internal interface.

[0150] This application also provides a computer-readable storage medium, which may include various media capable of storing program code, such as a USB flash drive, a portable hard drive, a read-only memory (ROM), a random access memory (RAM), a magnetic disk, or an optical disk. Specifically, the computer-readable storage medium stores program instructions, which are used in the methods described in the above embodiments.

[0151] This application also provides a program product including executable instructions stored in a readable storage medium. At least one processor of a computing device can read the executable instructions from the readable storage medium, and the at least one processor executes the executable instructions to cause the computing device to perform the data processing method described above.

[0152] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of this application, and are not intended to limit them. Although this application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some or all of the technical features therein. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of this application.

Claims

1. A data processing method, characterized in that, Applied to servers, including: Based on the first financial qualification scoring model and the second financial qualification scoring model, a target financial qualification scoring model is obtained. The first financial qualification scoring model is trained based on the merchant profile data of the first participant, and the second financial qualification scoring model is trained based on the merchant operation data of the second participant. The model structures of the first financial qualification scoring model and the second financial qualification scoring model are the same. The target financial qualification scoring model will be distributed to the target participants.

2. The method according to claim 1, characterized in that, The process of obtaining the target financial qualification scoring model based on the first financial qualification scoring model trained on merchant profile data from the first participant and the second financial qualification scoring model trained on merchant operating data from the second participant includes: Obtain the first model parameters of the first financial qualification scoring model and the second model parameters of the second financial qualification scoring model; The target financial qualification scoring model is generated based on the first model parameters and the second model parameters.

3. The method according to claim 2, characterized in that, The step of generating the target financial qualification scoring model based on the first model parameters and the second model parameters includes: Federated learning is performed based on the first model parameters and the second model parameters to generate the target financial qualification scoring model.

4. The method according to claim 2, characterized in that, The first model parameter and the second model parameter are data encrypted using the target encryption algorithm.

5. A data processing method, characterized in that, Applied to the first participant, including: The target financial qualification scoring model is obtained from the server. The target financial qualification scoring model is generated based on the first financial qualification scoring model trained on the merchant profile data of the first participant and the second financial qualification scoring model trained on the merchant operation data of the second participant. The target financial qualification scoring model is used to select target merchants that meet the target financial qualifications from the candidate merchants.

6. The method according to claim 5, characterized in that, Also includes: Obtain the merchant profile data; The first financial qualification scoring model is generated based on the merchant profile data.

7. The method according to claim 6, characterized in that, The step of generating the first financial qualification scoring model based on the merchant profile data includes: Information gain calculation and weight filtering are performed on the features in the merchant profile data to obtain a high-weight merchant profile feature set. The high-weight merchant profile feature set is input into the initial model framework built based on the XGBoost algorithm for ensemble tree training, so as to segment features and calculate training residuals through multiple decision trees in the initial model framework. The parameters of the initial model framework are adjusted based on the training residuals until the training results meet the preset stopping conditions, thereby obtaining the first financial qualification scoring model.

8. The method according to claim 6, characterized in that, The process of obtaining the merchant profile data includes: Obtain initial merchant profile data; The initial merchant profile data is filtered based on the key features corresponding to the target financial qualification to obtain the merchant profile data. The feature filtering is used to remove redundant and noisy features from the initial merchant profile data.

9. A data processing apparatus, characterized in that, Applied to servers, including: The processing module is used to obtain a target financial qualification scoring model based on the first financial qualification scoring model and the second financial qualification scoring model. The first financial qualification scoring model is trained based on the merchant profile data of the first participant, and the second financial qualification scoring model is trained based on the merchant operation data of the second participant. The model structures of the first financial qualification scoring model and the second financial qualification scoring model are the same. The sending module is used to distribute the target financial qualification scoring model to the target participants.

10. A data processing apparatus, characterized in that, Applied to the first participant, including: The acquisition module is used to acquire the target financial qualification scoring model from the server. The target financial qualification scoring model is generated based on the first financial qualification scoring model trained on the merchant profile data of the first participant and the second financial qualification scoring model trained on the merchant operation data of the second participant. The processing module is used to select target merchants that meet the target financial qualifications from the candidate merchants using the target financial qualification scoring model.

11. 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 8.

12. 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 8.

13. 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 8.

Images