Enterprise Transaction Data Access Method and Device

Abstract

The present invention provides an enterprise transaction data access method and device, which relates to the field of artificial intelligence technology and can be applied to the field of financial technology or other technical fields. The method includes: converting the string data in the current enterprise transaction feature data corresponding to the transaction access request into numerical data, encrypting the converted current enterprise transaction feature data with the bank public key from the federated learning platform, and then sending it to the data center side; inputting the encrypted current enterprise transaction feature data and the encrypted current enterprise association feature data returned by the data center side into the bank transaction coefficient model to obtain the bank transaction coefficient; determining the credit limit according to the bank transaction coefficient, the data center transaction coefficient, and the total enterprise amount; and processing the transaction data access request according to the credit limit, which can effectively improve the accuracy and interpretability of transaction data access, further expand the customer scope, and reasonably increase the credit limit of high-quality customers.

Description

Technical Field

[0001] The present invention relates to the technical field of artificial intelligence, and can be applied to the financial field or other fields. Specifically, it relates to a method and device for enterprise transaction data access. Background Art

[0002] Cross-border transactions require banks to cooperate with data centers. Using big data technology, through front-end model deployment and data screening, batch credit granting and online handling are actively carried out for import and export enterprises. Currently, cross-border transaction cooperation models and control rules are used to screen and filter customers applying for cross-border transactions, and high-quality customers are selected for cross-border transactions.

[0003] Cross-border transactions face several pain points. One is that the cross-border transaction amount is relatively low, and the advantage is not obvious compared with other scenarios, resulting in weak attraction to customers; the second is that some high-quality cross-border transaction enterprises fail to pass the model access and have no credit limit; the third is that the data source completely depends on the data center, lacking relevant indicators of the bank side for enterprise customers, such as turnover and assets, etc., resulting in general data access effect. Summary of the Invention

[0004] The main purpose of the embodiments of the present invention is to provide a method and device for enterprise transaction data access, so as to effectively improve the accuracy and interpretability of transaction data access, further expand the customer scope, and reasonably increase the credit limit for high-quality customers.

[0005] To achieve the above purpose, the embodiments of the present invention provide a method for enterprise transaction data access, including:

[0006] Receiving a transaction data access request, and obtaining the current enterprise transaction feature data corresponding to the transaction access request;

[0007] Converting the string data in the current enterprise transaction feature data into numerical data, encrypting the converted current enterprise transaction feature data according to the bank public key from the federated learning platform, and sending the encrypted current enterprise transaction feature data to the data center end so that the data center end returns the encrypted current enterprise association feature data and the data center transaction coefficient based on the encrypted current enterprise transaction feature data;

[0008] Inputting the encrypted current enterprise transaction feature data and the encrypted current enterprise association feature data into a bank transaction coefficient model to obtain a bank transaction coefficient; wherein, the bank transaction coefficient model is trained by historical enterprise transaction feature data, historical enterprise association feature data and historical access results with overlapping users and encrypted;

[0009] Determining the credit limit according to the bank transaction coefficient, the data center transaction coefficient and the total enterprise amount;

[0010] Process the transaction data access request according to the credit limit.

[0011] In one embodiment, it further includes:

[0012] Convert the string data in the historical enterprise transaction feature data into numerical data;

[0013] Encrypt the converted historical enterprise transaction feature data according to the bank public key and send it to the data center side, so that the data center side performs a collision operation based on the encrypted historical enterprise association feature data and the encrypted historical enterprise transaction feature data to obtain the encrypted historical enterprise transaction feature data and historical enterprise association feature data with overlapping users;

[0014] Receive the encrypted data center transaction model coefficients, the encrypted historical enterprise transaction feature data with overlapping users, and the historical enterprise association feature data from the data center side.

[0015] In one embodiment, it further includes:

[0016] Perform the following iterative processing:

[0017] Encrypt the bank transaction model coefficients according to the bank public key;

[0018] Determine the encrypted bank gradient data according to the encrypted bank transaction model coefficients, the encrypted data center transaction model coefficients, the encrypted historical enterprise transaction feature data with overlapping users, the historical enterprise association feature data, and the historical access results;

[0019] Send the encrypted bank gradient data to the federated learning platform, so that the federated learning platform decrypts the encrypted bank gradient data based on the bank private key to obtain the bank gradient data;

[0020] Determine the bank loss function according to the bank gradient data from the federated learning platform;

[0021] When the bank loss function converges, create a bank transaction coefficient model according to the bank transaction model coefficients, otherwise update the bank transaction model coefficients according to the bank gradient data.

[0022] In one embodiment, determining the credit limit according to the bank transaction coefficients, the data center transaction coefficients, and the total enterprise amount includes:

[0023] Determine the target transaction coefficient according to the bank transaction coefficients and the data center transaction coefficients;

[0024] Determine the credit limit according to the target transaction coefficient and the total enterprise amount.

[0025] In one embodiment, processing the transaction data access request according to the credit limit includes:

[0026] Processing the transaction data access request according to the comparison result between the credit limit and the transaction amount in the transaction data access request.

[0027] An embodiment of the present invention further provides an enterprise transaction data access device, including:

[0028] A feature data acquisition module, configured to receive a transaction data access request and acquire current enterprise transaction feature data corresponding to the transaction access request;

[0029] A conversion and encryption module, configured to convert the string data in the current enterprise transaction feature data into numerical data, encrypt the converted current enterprise transaction feature data according to the bank public key from the federated learning platform, and send the encrypted current enterprise transaction feature data to the data center side so that the data center side returns the encrypted current enterprise associated feature data and the data center transaction coefficient based on the encrypted current enterprise transaction feature data;

[0030] A bank transaction coefficient module, configured to input the encrypted current enterprise transaction feature data and the encrypted current enterprise associated feature data into a bank transaction coefficient model to obtain a bank transaction coefficient; wherein, the bank transaction coefficient model is trained by historical enterprise transaction feature data, historical enterprise associated feature data with overlapping users and historical access results that have been encrypted;

[0031] A credit limit module, configured to determine a credit limit according to the bank transaction coefficient, the data center transaction coefficient and the total enterprise amount;

[0032] A transaction data access module, configured to process the transaction data access request according to the credit limit.

[0033] In one embodiment, it further includes:

[0034] A conversion module, configured to convert the string data in the historical enterprise transaction feature data into numerical data;

[0035] A feature encryption module, configured to encrypt the converted historical enterprise transaction feature data according to the bank public key and send it to the data center side, so that the data center side performs a collision detection operation based on the encrypted historical enterprise associated feature data and the encrypted historical enterprise transaction feature data to obtain the encrypted historical enterprise transaction feature data and historical enterprise associated feature data with overlapping users;

[0036] A receiving module, configured to receive the encrypted data center transaction model coefficient from the data center side, as well as the encrypted historical enterprise transaction feature data and historical enterprise associated feature data with overlapping users.

[0037] In one embodiment, it further includes:

[0038] A coefficient encryption module, configured to encrypt the bank transaction model coefficients according to the bank public key;

[0039] An encrypted bank gradient data module, configured to determine the encrypted bank gradient data according to the encrypted bank transaction model coefficients, the encrypted data center transaction model coefficients, the encrypted historical enterprise transaction feature data with overlapping users, the historical enterprise association feature data, and the historical access results;

[0040] A sending module, configured to send the encrypted bank gradient data to the federated learning platform, so that the federated learning platform decrypts the encrypted bank gradient data based on the bank private key to obtain the bank gradient data;

[0041] A bank loss function module, configured to determine the bank loss function according to the bank gradient data from the federated learning platform;

[0042] An iteration module, configured to create a bank transaction coefficient model according to the bank transaction model coefficients when the bank loss function converges, otherwise update the bank transaction model coefficients according to the bank gradient data.

[0043] In one embodiment, the credit limit module includes:

[0044] A target transaction coefficient unit, configured to determine the target transaction coefficient according to the bank transaction coefficient and the data center transaction coefficient;

[0045] A credit limit unit, configured to determine the credit limit according to the target transaction coefficient and the total enterprise amount.

[0046] In one embodiment, the transaction data access module is specifically configured to:

[0047] Process the transaction data access request according to the comparison result between the credit limit and the transaction amount in the transaction data access request.

[0048] An embodiment of the present invention further provides a computer device, including a memory, a processor, and a computer program stored on the memory and running on the processor. When the processor executes the computer program, the steps of the enterprise transaction data access method are implemented.

[0049] An embodiment of the present invention further provides a computer-readable storage medium, on which a computer program is stored. When the computer program is executed by a processor, the steps of the enterprise transaction data access method are implemented.

[0050] An embodiment of the present invention also provides a computer program product, including computer programs / instructions, which implement the steps of the enterprise transaction data access method when executed by a processor.

[0051] An embodiment of the present invention also provides an enterprise transaction data access system, including:

[0052] The enterprise transaction data access device as described above;

[0053] A federated learning platform for sending the bank public key to the enterprise transaction data access device.

[0054] The data center side is used to return the encrypted current enterprise association feature data and the data center transaction coefficient based on the encrypted current enterprise transaction feature data.

[0055] The enterprise transaction data access method, device and system of the embodiment of the present invention first obtain the current enterprise transaction feature data corresponding to the transaction access request, then convert the string data in the current enterprise transaction feature data into numerical data, encrypt it according to the bank public key from the federated learning platform and send it to the data center side so that the data center side returns the encrypted current enterprise association feature data and the data center transaction coefficient, and then input the encrypted current enterprise transaction feature data and the encrypted current enterprise association feature data into the bank transaction coefficient model to obtain the bank transaction coefficient, and determine the credit limit according to the bank transaction coefficient, the data center transaction coefficient and the total enterprise amount. Finally, process the transaction data access request according to the credit limit, which can effectively improve the accuracy and interpretability of the transaction data access, further expand the customer scope, and reasonably increase the credit limit of high-quality customers. Description of the Drawings

[0056] In order to more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the following will briefly introduce the drawings required for the description of the embodiments. Obviously, the following drawings are only some embodiments of the present invention. For those of ordinary skill in the art, other drawings can be obtained based on these drawings without creative efforts.

[0057] Figure 1 is the flowchart of the enterprise transaction data access method in the embodiment of the present invention;

[0058] Figure 2 is the flowchart of S104 in the embodiment of the present invention;

[0059] Figure 3 is the flowchart of obtaining the data center transaction model coefficient and historical feature data in the embodiment of the present invention;

[0060] Figure 4It is a schematic diagram of sample construction in an embodiment of the present invention;

[0061] Figure 5 It is a flowchart of training a bank transaction coefficient model in an embodiment of the present invention;

[0062] Figure 6 It is a structural block diagram of an enterprise transaction data access device in an embodiment of the present invention;

[0063] Figure 7 It is a structural block diagram of a computer device in an embodiment of the present invention;

[0064] Figure 8 It is a structural block diagram of an enterprise transaction data access system in an embodiment of the present invention. Detailed implementation manners

[0065] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the protection scope of the present invention.

[0066] Those skilled in the art know that the implementation manners of the present invention can be realized as a system, a device, a device, a method, or a computer program product. Therefore, the present disclosure can be specifically realized in the following forms, that is: complete hardware, complete software (including firmware, resident software, microcode, etc.), or a combination of hardware and software.

[0067] In the technical solution of this application, the acquisition, storage, use, processing, etc. of data all comply with the relevant regulations of national laws and regulations.

[0068] In view of the general data access effect of the existing technology and the weak attraction to customers, the embodiments of the present invention provide an enterprise transaction data access method, which comprehensively considers the index characteristics of enterprise customers on the bank side and data such as customs declarations in the data center to build a model, comprehensively evaluates the business capabilities and potential default risks of customers, effectively improves the accuracy and interpretability of transaction data access, further expands the customer scope, and reasonably increases the credit limit of high-quality customers. The present invention will be described in detail below with reference to the accompanying drawings.

[0069] Figure 1 It is a flowchart of the enterprise transaction data access method in an embodiment of the present invention. As Figure 1 shown, the enterprise transaction data access method includes:

[0070] S101: Receive a transaction data access request and obtain the current enterprise transaction feature data corresponding to the transaction access request.

[0071] S102: Convert the string data in the current enterprise transaction feature data into numerical data, encrypt the converted current enterprise transaction feature data according to the bank public key from the federated learning platform, and send the encrypted current enterprise transaction feature data to the data center side so that the data center side returns the encrypted current enterprise association feature data and the data center transaction coefficient based on the encrypted current enterprise transaction feature data.

[0072] S103: Input the encrypted current enterprise transaction feature data and the encrypted current enterprise association feature data into the bank transaction coefficient model to obtain the bank transaction coefficient.

[0073] Among them, the bank transaction coefficient model is trained by the historical enterprise transaction feature data, historical enterprise association feature data and historical access results with overlapping users and encrypted.

[0074] The historical enterprise transaction feature data can be located on the bank side, and the historical enterprise association feature data can be located on the data center side. They can both be divided into good samples and bad samples. In the bad samples, the enterprises default (such as enterprise credit default and enterprise loan default), and the transactions are overdue; in the good samples, the enterprises and transactions are normal (enterprise normality can include normal enterprise credit and normal enterprise loan).

[0075] Table 1

[0076] Serial number Major categories of enterprise transaction characteristic data Examples of minor categories of enterprise transaction characteristic data 1 Legal person's bank statement Average balance of legal person's bank statement in the recent N months 2 Individual's bank statement Ratio of standard deviation of individual's bank statement's inflow amount in the recent N months 3 Individual assets Total individual assets 4 Legal person's assets Legal person's asset balance 5 Individual credit report Number of outstanding consumer credit lines of individual credit report in the recent N months 6 Legal person's credit report Number of institutional inquiries of legal person's credit report in the recent N months

[0077] Table 2

[0078]

[0079]

[0080] Table 1 is the historical enterprise transaction feature data table, and Table 2 is the historical enterprise association feature data table. As shown in Table 1, the historical enterprise transaction feature data can include enterprise turnover, assets and credit, etc.; as shown in Table 2, the historical enterprise association feature data table can include enterprise basic information, number of import and export trading countries, number of import and export months and annual operating income of the previous year, etc.

[0081] Figure 3 It is the flowchart for obtaining the data center transaction model coefficient and historical feature data in the embodiments of the present invention. As Figure 3 shown, the enterprise transaction data access method further includes:

[0082] S301: Convert the string data in the historical enterprise transaction feature data into numerical data.

[0083] The present invention needs to process numerical data, and the feature data used in creating the model also includes string data, such as the region where the customer company is located, the household registration location of the customer, etc. In the present invention, the encrypted Paillier algorithm cannot directly encrypt string data, so it is necessary to first perform manual encoding on string data such as the region where the customer company is located. Taking the region as an example. Manual encoding can be performed according to the economic situation of each region. Larger encoded values indicate stronger economic strength, and smaller encoded values indicate weaker economic strength. The encoding can be divided between [1, 10].

[0084] When transmitting data between the two parties, not only the feature data used by the model is required, but the bank side also needs some fields for customer profiling, such as information like the customer company name and the company registration time. The data center side needs to transmit information such as customer authorization instructions. Therefore, this part of the data does not need to be encrypted. The formula involved in data transmission is as follows:

[0085] ToA←concat(Paillier(Float_Fea B ),Paillier(encoding(Str_Fea B )),Note B );

[0086] ToB←concat(Paillier(Float_Fea A ),Paillier(encoding(Str_Fea A )),Note A );

[0087] Among them, concat is for concatenation, representing the process of concatenating the data to be transmitted. encoding is the process of encoding character data. Float_Fea A is the numerical feature of the first party (bank side), Str_Fea A is the character feature of the first party (bank side), Note A is the postscript field of the first party (bank side), Float_Fea B is the numerical feature of the second party (data center side), Str_Fea B is the character feature of the second party (data center side), Note B is the postscript field of the second party (data center side).

[0088] S302: Encrypt the converted historical enterprise transaction feature data using the bank's public key and send it to the data center side, so that the data center side can perform a database collision operation based on the encrypted historical enterprise association feature data and the encrypted historical enterprise transaction feature data to obtain the encrypted historical enterprise transaction feature data and historical enterprise association feature data with overlapping users.

[0089] Figure 4 is the schematic diagram of sample construction in the embodiment of the present invention. As Figure 4 shown, in order to effectively combine the data of both parties, a vertical federated modeling scheme is adopted to implement a logistic regression model to judge whether to admit the customer group applying for cross-border transactions. The sample numbers are uniformly specified by both parties as the unified social authentication numbers of enterprises. The historical enterprise transaction feature data (bank data) is mainly shown in Table 1, and the historical enterprise association feature data (data center data) is mainly shown in Table 2. The labels of the samples (historical admission results) include good samples and bad samples, which are provided by the bank side. The bank side sends the labeled samples to the database of the data center for a database collision operation, and the result is the customer group overlapping between the two parties. For this part of the customers, a training set and a validation set are divided for modeling.

[0090] S303: Receive the encrypted data center transaction model coefficients from the data center side, as well as the encrypted historical enterprise transaction feature data and historical enterprise association feature data with overlapping users.

[0091] Among them, the encrypted data center transaction model coefficients are generated by the data center side according to the data center public key encrypted by the federated learning platform.

[0092] Figure 5 is the flowchart of training the bank transaction coefficient model in the embodiment of the present invention. As Figure 5 shown, the enterprise transaction data admission method further includes:

[0093] Perform the following iterative processing:

[0094] S401: Encrypt the bank transaction model coefficients using the bank's public key.

[0095] For data security considerations, both parties perform homomorphic encryption on numerical data; homomorphic encryption is a special encryption method, that is, directly processing the plaintext, which is the same as the result obtained by processing the plaintext and then encrypting the processing result. The present invention uses the Paillier algorithm to implement homomorphic encryption, which supports addition and multiplication by constants on the encrypted data. Since there is an exponential calculation in the gradient calculation formula of logistic regression, in order to meet the requirement that homomorphic encryption cannot perform exponential operations, it is necessary to approximate the gradient formula through Taylor expansion.

[0096] S402: Determine the encrypted bank gradient data based on the encrypted bank transaction model coefficients, the encrypted data center transaction model coefficients, the encrypted historical enterprise transaction feature data with overlapping users, the historical enterprise association feature data, and the historical admission results.

[0097] During specific implementation, the encrypted bank gradient data can be determined through the following formula:

[0098]

[0099] Where, is the encrypted bank gradient data, X A is the encrypted historical enterprise transaction feature data with overlapping users, X B is the encrypted historical enterprise association feature data with overlapping users, θ A is the encrypted bank transaction model coefficient, θ B is the encrypted data center transaction model coefficient, y is the historical admission result, that is, the label information, with a value of 0 or 1. 1 indicates that the sample corresponding to the feature data is a bad sample, and 0 indicates that the sample corresponding to the feature data is a good sample. λ is the speed parameter, used to control the proportion of the original model coefficients in the model and the update speed of the model coefficients. n is the number of historical enterprise transaction feature data.

[0100] S403: Send the encrypted bank gradient data to the federated learning platform so that the federated learning platform decrypts the encrypted bank gradient data based on the bank private key to obtain the bank gradient data.

[0101] S404: Determine the bank loss function based on the bank gradient data from the federated learning platform.

[0102] Among them, the above formula is obtained through and after approximate calculation. Among them, J s (θ) is the loss function of logistic regression, θ is the parameter to be solved in the model, yi is the label (good sample or bad sample) of the i-th sample, and xi is the feature data of the i-th sample.

[0103] S405: Determine whether the bank loss function converges.

[0104] S406: When the bank loss function converges, create a bank transaction coefficient model based on the bank transaction model coefficients.

[0105] S407: When the bank loss function does not converge, update the bank transaction model coefficients according to the bank gradient data.

[0106] S104: Determine the credit limit based on the bank transaction coefficients, the data center transaction coefficients, and the total enterprise amount.

[0107] Among them, the process for the data center side to obtain the data center transaction coefficient is similar to the process for the bank side to obtain the bank transaction coefficient.

[0108] Specifically in implementation, the data center side obtains the corresponding previous enterprise association feature data based on the encrypted current enterprise transaction feature data, converts the string data in the current enterprise association feature data into numerical data, encrypts the converted current enterprise association feature data according to the data center public key of the federated learning platform, and inputs the encrypted current enterprise transaction feature data and the encrypted current enterprise association feature data into the data center transaction coefficient model to obtain the data center transaction coefficient.

[0109] The process for creating the data center transaction coefficient model is as follows:

[0110] Receive the encrypted bank transaction model coefficient from the bank side and perform the following iterative process:

[0111] 1. Encrypt the data center transaction model coefficient according to the data center public key.

[0112] 2. Determine the encrypted data center gradient data according to the encrypted bank transaction model coefficient, the encrypted data center transaction model coefficient, the encrypted historical enterprise transaction feature data with overlapping users, the historical enterprise association feature data, and the historical admission results.

[0113] Specifically in implementation, the encrypted data center gradient data can be determined by the following formula:

[0114]

[0115] Among them, is the encrypted data center gradient data, and the number of historical enterprise association feature data is equal to the number of historical enterprise transaction feature data.

[0116] 3. Send the encrypted data center gradient data to the federated learning platform so that the federated learning platform decrypts the encrypted data center gradient data based on the data center private key to obtain the data center gradient data.

[0117] 4. Determine the data center loss function according to the data center gradient data from the federated learning platform. When the data center loss function converges, create the data center transaction coefficient model according to the data center transaction model coefficient; otherwise, update the data center transaction model coefficient according to the data center gradient data.

[0118] Figure 2 is the flowchart of S104 in the embodiments of the present invention. As Figure 2 shown, S104 includes:

[0119] S201: Determine the target transaction coefficient based on the bank transaction coefficient and the data center transaction coefficient.

[0120] In specific implementation, first compare and verify the bank transaction coefficient model and the data center transaction coefficient model to ensure the reliability and stability of the models. For example, calculate various stability indicators and availability indicators (such as KS, AUC, ROC, and confusion matrix, etc.) of each model, and determine the first factor and the second factor according to the indicators. In theory, the indicators are the same for both, but if both parties have personalized customization of the models, there will be differences. Therefore, the comparison of statistical indicators needs to be carried out on the premise of ensuring the same environment and configuration for both parties. When the indicators of the bank transaction coefficient model are better than those of the data center transaction coefficient model, the first factor corresponding to the bank transaction coefficient model is greater than the second factor corresponding to the data center transaction coefficient model; otherwise, the first factor is less than the second factor.

[0121] Specifically, S201 includes: determining the target transaction coefficient based on the bank transaction coefficient, the first factor, the data center transaction coefficient, and the second factor. Among them, the range of the target transaction coefficient is 0 to 1. The lower the target transaction coefficient, the smaller the probability of the enterprise being overdue after obtaining a loan.

[0122] S202: Determine the credit limit according to the target transaction coefficient and the total amount of the enterprise.

[0123] In specific implementation, first determine the credit coefficient according to the target transaction coefficient, and then determine the credit limit according to the credit coefficient and the total amount of the enterprise.

[0124] For example, when the range where the target transaction coefficient is located is (0, 0.05], the credit coefficient is M3; when the range where the target transaction coefficient is located is (0.05, 0.08], the credit coefficient is M2; when the range where the target transaction coefficient is located is (0.08, 0.12], the credit coefficient is M1; the lower the target transaction coefficient, the greater the credit coefficient, and customers with a target transaction coefficient greater than 0.12 are not allowed to enter. By dividing customers according to the credit coefficient, the number of customers, the proportion, the average amount per household, and the total sample credit amount corresponding to each credit coefficient can be obtained. The personalized credit limit allocation for different customers in the present invention can reduce transaction risks, give a greater credit limit to high-quality customers, and bring mutual benefits to enterprises and banks, meeting the quickness, convenience, and demand of customers for financial services.

[0125] In specific implementation, the credit limit can be determined through the following formula:

[0126] Credit limit = min(annual average value of the import and export declaration amount in the past two years of the enterprise total amount × 10% × β), where β is the credit coefficient. The enterprise total amount can adopt the annual average value of the import and export declaration amount in the past two years.

[0127] S105: Process the transaction data access request according to the credit limit.

[0128] In one embodiment, S105 includes: processing the transaction data access request according to the comparison result between the transaction amount in the transaction data access request and the credit limit. Specifically, when the transaction amount is less than or equal to the credit limit, the transaction data access request is processed.

[0129] Figure 1 The execution subject of the enterprise transaction data access method shown can be a computer on the bank side. Figure 1 As can be seen from the shown process, the enterprise transaction data access method of the embodiment of the present invention first obtains the current enterprise transaction feature data corresponding to the transaction access request, then converts the string data in the current enterprise transaction feature data into numerical data, encrypts it according to the bank public key from the federated learning platform and sends it to the data center side so that the data center side returns the encrypted current enterprise association feature data and the data center transaction coefficient, and then inputs the encrypted current enterprise transaction feature data and the encrypted current enterprise association feature data into the bank transaction coefficient model to obtain the bank transaction coefficient, and determines the credit limit according to the bank transaction coefficient, the data center transaction coefficient and the total enterprise amount. Finally, the transaction data access request is processed according to the credit limit, which can effectively improve the accuracy and interpretability of transaction data access, further expand the customer range, and reasonably increase the credit limit of high-quality customers.

[0130] Based on the same inventive concept, the embodiment of the present invention also provides an enterprise transaction data access device. Since the principle of the device to solve the problem is similar to that of the enterprise transaction data access method, the implementation of the device can refer to the implementation of the method, and the repeated parts will not be described again.

[0131] Figure 6 It is the structural block diagram of the enterprise transaction data access device in the embodiment of the present invention. As Figure 6 shown, the enterprise transaction data access device includes:

[0132] A feature data acquisition module, configured to receive a transaction data access request and acquire the current enterprise transaction feature data corresponding to the transaction access request;

[0133] A conversion and encryption module, configured to convert the string data in the current enterprise transaction feature data into numerical data, encrypt the converted current enterprise transaction feature data according to the bank public key from the federated learning platform, and send the encrypted current enterprise transaction feature data to the data center side so that the data center side returns the encrypted current enterprise association feature data and the data center transaction coefficient based on the encrypted current enterprise transaction feature data;

[0134] A bank transaction coefficient module, which is used to input the encrypted current enterprise transaction feature data and the encrypted current enterprise associated feature data into a bank transaction coefficient model to obtain a bank transaction coefficient; wherein, the bank transaction coefficient model is trained by the encrypted historical enterprise transaction feature data, historical enterprise associated feature data, and historical access results with overlapping users.

[0135] A credit limit module, which is used to determine the credit limit according to the bank transaction coefficient, the data center transaction coefficient, and the total enterprise amount.

[0136] A transaction data access module, which is used to process a transaction data access request according to the credit limit.

[0137] In one embodiment, it further includes:

[0138] A conversion module, which is used to convert the string data in the historical enterprise transaction feature data into numerical data.

[0139] A feature encryption module, which is used to encrypt the converted historical enterprise transaction feature data according to the bank public key and send it to the data center side, so that the data center side performs a collision operation based on the encrypted historical enterprise associated feature data and the encrypted historical enterprise transaction feature data to obtain the encrypted historical enterprise transaction feature data and historical enterprise associated feature data with overlapping users.

[0140] A receiving module, which is used to receive the encrypted data center transaction model coefficient, the encrypted historical enterprise transaction feature data, and the historical enterprise associated feature data with overlapping users from the data center side.

[0141] In one embodiment, it further includes:

[0142] A coefficient encryption module, which is used to encrypt the bank transaction model coefficient according to the bank public key.

[0143] An encrypted bank gradient data module, which is used to determine the encrypted bank gradient data according to the encrypted bank transaction model coefficient, the encrypted data center transaction model coefficient, the encrypted historical enterprise transaction feature data with overlapping users, the historical enterprise associated feature data, and the historical access results.

[0144] A sending module, which is used to send the encrypted bank gradient data to the federated learning platform, so that the federated learning platform decrypts the encrypted bank gradient data based on the bank private key to obtain the bank gradient data.

[0145] A bank loss function module, which is used to determine the bank loss function according to the bank gradient data from the federated learning platform.

[0146] An iterative module, configured to create a bank transaction coefficient model based on bank transaction model coefficients when the bank loss function converges, or update the bank transaction model coefficients based on bank gradient data otherwise.

[0147] In one embodiment, the credit line module includes:

[0148] A target transaction coefficient unit, configured to determine a target transaction coefficient based on bank transaction coefficients and data center transaction coefficients;

[0149] A credit line unit, configured to determine a credit line based on the target transaction coefficient and the total enterprise amount.

[0150] In one embodiment, the transaction data access module is specifically configured to:

[0151] Process the transaction data access request according to the comparison result between the credit line and the transaction amount in the transaction data access request.

[0152] In summary, the enterprise transaction data access device according to the embodiments of the present invention first obtains the current enterprise transaction feature data corresponding to the transaction access request, then converts the string data in the current enterprise transaction feature data into numerical data, encrypts it using the bank public key from the federated learning platform and sends it to the data center side so that the data center side returns the encrypted current enterprise association feature data and data center transaction coefficients, and then inputs the encrypted current enterprise transaction feature data and the encrypted current enterprise association feature data into the bank transaction coefficient model to obtain bank transaction coefficients, determines the credit line based on the bank transaction coefficients, data center transaction coefficients and the total enterprise amount, and finally processes the transaction data access request according to the credit line, which can effectively improve the accuracy and interpretability of transaction data access, further expand the customer scope, and reasonably increase the credit line for high-quality customers.

[0153] The embodiments of the present invention also provide a specific implementation manner of a computer device capable of implementing all steps in the enterprise transaction data access method in the above embodiments. Figure 7 It is a structural block diagram of the computer device in the embodiments of the present invention. Refer to Figure 7 and the computer device specifically includes the following:

[0154] A processor 701 and a memory 702.

[0155] The processor 701 is configured to call a computer program in the memory 702. When the processor executes the computer program, it implements all steps in the enterprise transaction data access method in the above embodiments. For example, when the processor executes the computer program, it implements the following steps:

[0156] Receive a transaction data access request and obtain the current enterprise transaction feature data corresponding to the transaction access request;

[0157] Convert the string data in the current enterprise transaction feature data into numerical data, encrypt the converted current enterprise transaction feature data according to the bank public key from the federated learning platform, and send the encrypted current enterprise transaction feature data to the data center side so that the data center side returns the encrypted current enterprise association feature data and the data center transaction coefficient based on the encrypted current enterprise transaction feature data;

[0158] Input the encrypted current enterprise transaction feature data and the encrypted current enterprise association feature data into the bank transaction coefficient model to obtain the bank transaction coefficient; wherein, the bank transaction coefficient model is trained by the historical enterprise transaction feature data, historical enterprise association feature data and historical access results with overlapping users and encrypted;

[0159] Determine the credit limit according to the bank transaction coefficient, the data center transaction coefficient and the total enterprise amount;

[0160] Process the transaction data access request according to the credit limit.

[0161] In summary, the computer device in the embodiment of the present invention first obtains the current enterprise transaction feature data corresponding to the transaction access request, then converts the string data in the current enterprise transaction feature data into numerical data, encrypts it according to the bank public key from the federated learning platform and sends it to the data center side so that the data center side returns the encrypted current enterprise association feature data and the data center transaction coefficient, and then inputs the encrypted current enterprise transaction feature data and the encrypted current enterprise association feature data into the bank transaction coefficient model to obtain the bank transaction coefficient to determine the credit limit according to the bank transaction coefficient, the data center transaction coefficient and the total enterprise amount, and finally processes the transaction data access request according to the credit limit, which can effectively improve the accuracy and interpretability of transaction data access, further expand the customer range, and reasonably increase the credit limit of high-quality customers.

[0162] The embodiment of the present invention also provides a computer-readable storage medium capable of implementing all steps in the enterprise transaction data access method in the above embodiment. A computer program is stored on the computer-readable storage medium. When the computer program is executed by a processor, all steps in the enterprise transaction data access method in the above embodiment are implemented. For example, when the processor executes the computer program, the following steps are implemented:

[0163] Receive a transaction data access request and obtain the current enterprise transaction feature data corresponding to the transaction access request;

[0164] Convert the string data in the current enterprise transaction feature data into numerical data, encrypt the converted current enterprise transaction feature data according to the bank public key from the federated learning platform, and send the encrypted current enterprise transaction feature data to the data center end so that the data center end returns the encrypted current enterprise association feature data and the data center transaction coefficient based on the encrypted current enterprise transaction feature data;

[0165] Input the encrypted current enterprise transaction feature data and the encrypted current enterprise association feature data into the bank transaction coefficient model to obtain the bank transaction coefficient; wherein, the bank transaction coefficient model is trained by the historical enterprise transaction feature data, historical enterprise association feature data and historical access results with overlapping users and encrypted;

[0166] Determine the credit limit according to the bank transaction coefficient, the data center transaction coefficient and the total enterprise amount;

[0167] Process the transaction data access request according to the credit limit.

[0168] In summary, the computer-readable storage medium of the embodiment of the present invention first obtains the current enterprise transaction feature data corresponding to the transaction access request, then converts the string data in the current enterprise transaction feature data into numerical data, encrypts it according to the bank public key from the federated learning platform and sends it to the data center end so that the data center end returns the encrypted current enterprise association feature data and the data center transaction coefficient, and then inputs the encrypted current enterprise transaction feature data and the encrypted current enterprise association feature data into the bank transaction coefficient model to obtain the bank transaction coefficient to determine the credit limit according to the bank transaction coefficient, the data center transaction coefficient and the total enterprise amount, and finally processes the transaction data access request according to the credit limit, which can effectively improve the accuracy and interpretability of the transaction data access, further expand the customer range, and reasonably increase the credit limit of high-quality customers.

[0169] The embodiment of the present invention also provides a computer program product capable of implementing all the steps in the enterprise transaction data access method in the above embodiment. The computer program product includes computer programs / instructions, and when the computer programs / instructions are executed by a processor, all the steps in the enterprise transaction data access method in the above embodiment are implemented. For example, when the processor executes the computer program, the following steps are implemented:

[0170] Receive a transaction data access request and obtain the current enterprise transaction feature data corresponding to the transaction access request;

[0171] Convert the string data in the current enterprise transaction feature data into numerical data, encrypt the converted current enterprise transaction feature data according to the bank public key from the federated learning platform, and send the encrypted current enterprise transaction feature data to the data center side so that the data center side returns the encrypted current enterprise association feature data and the data center transaction coefficient based on the encrypted current enterprise transaction feature data;

[0172] Input the encrypted current enterprise transaction feature data and the encrypted current enterprise association feature data into the bank transaction coefficient model to obtain the bank transaction coefficient; wherein, the bank transaction coefficient model is trained by the historical enterprise transaction feature data, historical enterprise association feature data and historical access results with overlapping users and encrypted;

[0173] Determine the credit limit according to the bank transaction coefficient, the data center transaction coefficient and the total enterprise amount;

[0174] Process the transaction data access request according to the credit limit.

[0175] In summary, the computer program product of the embodiment of the present invention first obtains the current enterprise transaction feature data corresponding to the transaction access request, then converts the string data in the current enterprise transaction feature data into numerical data, encrypts it according to the bank public key from the federated learning platform and sends it to the data center side so that the data center side returns the encrypted current enterprise association feature data and the data center transaction coefficient, and then inputs the encrypted current enterprise transaction feature data and the encrypted current enterprise association feature data into the bank transaction coefficient model to obtain the bank transaction coefficient to determine the credit limit according to the bank transaction coefficient, the data center transaction coefficient and the total enterprise amount, and finally processes the transaction data access request according to the credit limit, which can effectively improve the accuracy and interpretability of the transaction data access, further expand the customer range, and reasonably improve the credit limit of high-quality customers.

[0176] Based on the same inventive concept, the embodiment of the present invention also provides an enterprise transaction data access system. Since the principle of the system to solve the problem is similar to that of the enterprise transaction data access method, the implementation of the system can refer to the implementation of the method, and the repeated parts will not be described again.

[0177] Figure 8 It is a schematic diagram of the enterprise transaction data access system in the embodiment of the present invention. As Figure 8 shown, the enterprise transaction data access system includes:

[0178] The above-mentioned enterprise transaction data access device, applied to the bank side;

[0179] The federated learning platform is used to send the bank public key to the enterprise transaction data access device.

[0180] The data center side is used to return the encrypted current enterprise associated feature data and the data center transaction coefficient based on the encrypted current enterprise transaction feature data.

[0181] The specific process of the enterprise transaction data access system according to the embodiments of the present invention is as follows:

[0182] 1. The federated learning platform sends the bank public key to the bank side and sends the data center public key to the data center side.

[0183] 2. The bank side converts the string data in the historical enterprise transaction feature data into numerical data, encrypts the converted historical enterprise transaction feature data according to the bank public key, and sends it to the data center side.

[0184] 3. The data center side performs a collision operation based on the encrypted historical enterprise associated feature data and the encrypted historical enterprise transaction feature data to obtain the encrypted historical enterprise transaction feature data and historical enterprise associated feature data with overlapping users, encrypts the data center transaction model coefficient according to the data center public key, and sends the encrypted data center transaction model coefficient, the encrypted historical enterprise transaction feature data and historical enterprise associated feature data with overlapping users to the bank side.

[0185] 4. The bank side encrypts the bank transaction model coefficient according to the bank public key.

[0186] 5. The bank side determines the encrypted bank gradient data according to the encrypted bank transaction model coefficient, the encrypted data center transaction model coefficient, the encrypted historical enterprise transaction feature data with overlapping users, the historical enterprise associated feature data and the historical access result, and sends the encrypted bank gradient data to the federated learning platform.

[0187] 6. The federated learning platform decrypts the encrypted bank gradient data based on the bank private key to obtain the bank gradient data.

[0188] 7. The bank side determines the bank loss function according to the bank gradient data from the federated learning platform; when the bank loss function converges, a bank transaction coefficient model is created according to the bank transaction model coefficient, otherwise the bank transaction model coefficient is updated according to the bank gradient data, and step 4 is returned.

[0189] 8. The bank side receives a transaction data access request, obtains the current enterprise transaction feature data corresponding to the transaction access request, converts the string data in the current enterprise transaction feature data into numerical data, encrypts the converted current enterprise transaction feature data according to the bank public key, and sends the encrypted current enterprise transaction feature data to the data center side.

[0190] 9. The data center returns the encrypted current enterprise associated feature data and the data center transaction coefficient to the bank based on the encrypted current enterprise transaction feature data.

[0191] 10. The bank inputs the encrypted current enterprise transaction feature data and the encrypted current enterprise associated feature data into the bank transaction coefficient model to obtain the bank transaction coefficient, and determines the target transaction coefficient based on the bank transaction coefficient and the data center transaction coefficient.

[0192] 11. The bank determines the credit limit based on the target transaction coefficient and the total enterprise amount, and processes the transaction data access request according to the comparison result between the credit limit and the transaction amount in the transaction data access request.

[0193] In summary, the enterprise transaction data access system according to the embodiment of the present invention first obtains the current enterprise transaction feature data corresponding to the transaction access request, then converts the string data in the current enterprise transaction feature data into numerical data, encrypts it with the bank public key from the federated learning platform and sends it to the data center to make the data center return the encrypted current enterprise associated feature data and the data center transaction coefficient, and then inputs the encrypted current enterprise transaction feature data and the encrypted current enterprise associated feature data into the bank transaction coefficient model to obtain the bank transaction coefficient to determine the credit limit based on the bank transaction coefficient, the data center transaction coefficient and the total enterprise amount, and finally processes the transaction data access request according to the credit limit, which can effectively improve the accuracy and interpretability of transaction data access, further expand the customer range, and reasonably increase the credit limit for high-quality customers.

[0194] In the specific embodiments described above, the purpose, technical solutions and beneficial effects of the present invention are further described in detail. It should be understood that the above are only specific embodiments of the present invention and are not used to limit the protection scope of the present invention. Any modifications, equivalent replacements, improvements, etc. made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.

[0195] Those skilled in the art can also understand that the various illustrative logical blocks, units, and steps listed in the embodiments of the present invention can be implemented by electronic hardware, computer software, or a combination of both. To clearly show the interchangeability of hardware and software, the above various illustrative components, units, and steps have been generally described in terms of their functions. Whether such functions are implemented by hardware or software depends on the specific application and the design requirements of the entire system. Those skilled in the art can use various methods to implement the described functions for each specific application, but such implementation should not be construed as exceeding the scope of protection of the embodiments of the present invention.

[0196] In the embodiments of the present invention, the various illustrative logical blocks, or units, or devices can be implemented or operate the described functions through a general-purpose processor, a digital signal processor, an application-specific integrated circuit (ASIC), a field-programmable gate array or other programmable logic devices, discrete gate or transistor logic, discrete hardware components, or any combination of the above designs. The general-purpose processor can be a microprocessor. Optionally, the general-purpose processor can also be any conventional processor, controller, microcontroller, or state machine. The processor can also be implemented by a combination of computing devices, such as a digital signal processor and a microprocessor, multiple microprocessors, one or more microprocessors combined with a digital signal processor core, or any other similar configuration.

[0197] The steps of the methods or algorithms described in the embodiments of the present invention can be directly embedded in hardware, software modules executed by a processor, or a combination of both. The software modules can be stored in a RAM memory, a flash memory, a ROM memory, an EPROM memory, an EEPROM memory, a register, a hard disk, a removable disk, a CD-ROM, or any other form of storage medium in the art. Exemplarily, the storage medium can be connected to the processor so that the processor can read information from the storage medium and write information to the storage medium. Optionally, the storage medium can also be integrated into the processor. The processor and the storage medium can be disposed in an ASIC, and the ASIC can be disposed in a user terminal. Optionally, the processor and the storage medium can also be disposed in different components of the user terminal.

[0198] In one or more exemplary designs, the functions described above in embodiments of the present invention may be implemented in hardware, software, firmware, or any combination of the three. If implemented in software, the functions may be stored on a computer-readable medium or transmitted as one or more instructions or code on a computer-readable medium. A computer-readable medium includes both computer storage media and communication media that facilitate transfer of a computer program from one place to another. Storage media may be any available media that is accessible by a general purpose or special purpose computer. By way of example, and not limitation, such computer-readable media may include RAM, ROM, EEPROM, CD-ROM or other optical disk storage, magnetic disk storage or other magnetic storage devices, or any other medium that can be used to carry or store program code in the form of instructions or data structures and that can be accessed by a general purpose or special purpose computer, or a general purpose or special purpose processor. In addition, any connection is properly termed a computer-readable medium, such as if software is transmitted from a website, server, or other remote source using a coaxial cable, fiber optic cable, twisted pair, DSL, or wireless technologies such as infrared, radio, and microwave. Disk and disc include compact disk, laser disk, optical disk, DVD, floppy disk, and Blu-ray disk, where disks usually reproduce data magnetically, while discs usually reproduce data optically with lasers. Combinations of the above should also be included within the scope of computer-readable media.

Claims

1. An enterprise transaction data access method, characterized in that, Including: Receiving a transaction data access request and obtaining the current enterprise transaction feature data corresponding to the transaction access request; Converting the string data in the current enterprise transaction feature data into numerical data, encrypting the converted current enterprise transaction feature data according to the bank public key from the federated learning platform, and sending the encrypted current enterprise transaction feature data to the data center end so that the data center end returns the encrypted current enterprise associated feature data and the data center transaction coefficient based on the encrypted current enterprise transaction feature data. Among them, the data center end obtains the current enterprise associated feature data based on the encrypted current enterprise transaction feature data, converts the string data in the current enterprise associated feature data into numerical data, encrypts the converted current enterprise associated feature data according to the data center public key from the federated learning platform, and inputs the encrypted current enterprise transaction feature data and the encrypted current enterprise associated feature data into the data center transaction coefficient model to obtain the data center transaction coefficient; Inputting the encrypted current enterprise transaction feature data and the encrypted current enterprise associated feature data into the bank transaction coefficient model to obtain the bank transaction coefficient; among them, the bank transaction coefficient model is trained by the historical enterprise transaction feature data, historical enterprise associated feature data, and historical access results with overlapping users and encrypted; Determining the credit limit according to the bank transaction coefficient, the data center transaction coefficient, and the total enterprise amount; Processing the transaction data access request according to the credit limit.

2. The enterprise transaction data access method according to claim 1, characterized in that Also including: Converting the string data in the historical enterprise transaction feature data into numerical data; Encrypting the converted historical enterprise transaction feature data according to the bank public key and sending it to the data center end so that the data center end performs a database collision operation based on the encrypted historical enterprise associated feature data and encrypted historical enterprise transaction feature data to obtain the historical enterprise transaction feature data and historical enterprise associated feature data with overlapping users and encrypted; Receiving the encrypted data center transaction model coefficient and the historical enterprise transaction feature data and historical enterprise associated feature data with overlapping users and encrypted from the data center end.

3. The enterprise transaction data access method according to claim 2, wherein Also including: Performing the following iterative processing: Encrypting the bank transaction model coefficient according to the bank public key; Determining the encrypted bank gradient data according to the encrypted bank transaction model coefficient, the encrypted data center transaction model coefficient, the historical enterprise transaction feature data with overlapping users and encrypted, the historical enterprise associated feature data, and the historical access results; Sending the encrypted bank gradient data to the federated learning platform so that the federated learning platform decrypts the encrypted bank gradient data based on the bank private key to obtain the bank gradient data; Determining the bank loss function according to the bank gradient data from the federated learning platform; When the bank loss function converges, creating the bank transaction coefficient model according to the bank transaction model coefficient, otherwise updating the bank transaction model coefficient according to the bank gradient data.

4. The enterprise transaction data access method according to claim 1, wherein Determining the credit limit based on the bank transaction coefficient, the data center transaction coefficient, and the total enterprise amount includes: Determining a target transaction coefficient based on the bank transaction coefficient and the data center transaction coefficient; Determining the credit limit based on the target transaction coefficient and the total enterprise amount.

5. The enterprise transaction data access method according to claim 1, characterized in that Processing the transaction data access request according to the credit limit includes: Processing the transaction data access request according to the comparison result between the credit limit and the transaction amount in the transaction data access request.

6. An enterprise transaction data access device, which is applied to the bank side, is characterized in that Including: A feature data acquisition module, configured to receive a transaction data access request and acquire current enterprise transaction feature data corresponding to the transaction access request; A conversion and encryption module, configured to convert the string data in the current enterprise transaction feature data into numerical data, encrypt the converted current enterprise transaction feature data according to the bank public key from the federated learning platform, and send the encrypted current enterprise transaction feature data to the data center side so that the data center side returns the encrypted current enterprise associated feature data and the data center transaction coefficient based on the encrypted current enterprise transaction feature data, wherein the data center side obtains the current enterprise associated feature data based on the encrypted current enterprise transaction feature data, converts the string data in the current enterprise associated feature data into numerical data, encrypts the converted current enterprise associated feature data according to the data center public key from the federated learning platform, and inputs the encrypted current enterprise transaction feature data and the encrypted current enterprise associated feature data into a data center transaction coefficient model to obtain the data center transaction coefficient; A bank transaction coefficient module, configured to input the encrypted current enterprise transaction feature data and the encrypted current enterprise associated feature data into a bank transaction coefficient model to obtain the bank transaction coefficient; wherein, the bank transaction coefficient model is trained by historical enterprise transaction feature data, historical enterprise associated feature data, and historical access results with overlapping users and encrypted; A credit limit module, configured to determine the credit limit according to the bank transaction coefficient, the data center transaction coefficient, and the total enterprise amount; A transaction data access module, configured to process the transaction data access request according to the credit limit.

7. A computer device, comprising a memory, a processor, and a computer program stored on the memory and running on the processor, characterized in that, When the processor executes the computer program, the steps of the enterprise transaction data access method according to any one of claims 1 to 5 are implemented.

8. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, the steps of the enterprise transaction data access method according to any one of claims 1 to 5 are implemented.

9. A computer program product, comprising a computer program / instructions, characterized in that, When the computer program / instructions are executed by the processor, the steps of the enterprise transaction data access method according to any one of claims 1 to 5 are implemented.

10. An enterprise transaction data access system, characterized in that, Including: The enterprise transaction data access device according to claim 6; A federated learning platform, configured to send the bank public key to the enterprise transaction data access device; A data center side, configured to return the encrypted current enterprise associated feature data and the data center transaction coefficient based on the encrypted current enterprise transaction feature data.

Images