Transaction account management and control setting processing method and device and client terminal

Abstract

The invention provides a transaction account management and control setting processing method and device and a client terminal, and is applied to the technical field of artificial intelligence. Based on the method, a first client terminal obtains historical transaction data of a first transaction account and a target transaction account and associated data related to the first transaction account and the target transaction account; processing the transaction data and the associated data by using a preset account management and control strategy prediction model to obtain a corresponding target processing result; wherein the preset account management and control strategy prediction model is obtained by the first client terminal through federal learning training with the first server and the second client terminal based on an alliance chain according to a preset protocol rule in advance; determining a target management and control setting strategy of the first transaction account for the target transaction account according to the target processing result; and according to the target management and control setting strategy, performing corresponding management and control setting on the first transaction account. Therefore, the account management and control setting of the first transaction account can be automatically and accurately completed.

Description

Technical Field

[0001] This specification belongs to the field of artificial intelligence technology, and in particular to a method, device and client terminal for managing and setting up a trading account. Background Art

[0002] In financial business scenarios, when companies, institutions and other customers use the trading accounts (or trading account numbers) of relevant banks, they often need to set up control settings for counterparty restrictions on the trading accounts based on their own business management conditions, business development needs and other factors. For example, the trading account can only interact with certain accounts for trading data; or the trading account is prohibited from interacting with other accounts for trading data.

[0003] Based on the existing methods, most of them need to rely on the staff of the above-mentioned companies, institutions and other clients to manually set up specific account control for the relevant trading accounts based on their personal experience. However, when the above methods are implemented, there are often large workloads for the staff and cumbersome settings; and the settings are easily affected by the subjective factors of the staff, which is prone to omissions and other errors.

[0004] To address the above problems, no effective solution has been proposed yet. Summary of the invention

[0005] This specification provides a method, device and client terminal for processing the management and control settings of a trading account, which can automatically and accurately complete the account management and control settings of a first trading account.

[0006] This specification provides a method for processing the management and control settings of a trading account, which is applied to a first client terminal, where the first client terminal is deployed on the first client side, and the method includes:

[0007] Acquire historical transaction data of a first transaction account and a target transaction account, and associated data related to the first transaction account and the target transaction account; wherein the first transaction account is a transaction account held by a first customer, and the target transaction account is a transaction account held by a target customer;

[0008] The historical transaction data and associated data are processed using a preset account control strategy prediction model to obtain a corresponding target processing result; wherein the preset account control strategy prediction model is a model based on an extreme gradient boosting tree structure obtained by the first client terminal in advance in cooperation with the first server and the second client terminal based on the alliance chain through federated learning training according to preset protocol rules; the first server is deployed on the side of the financial service institution; the second client terminal is deployed on the side of an associated client that has a cooperative relationship with the first client;

[0009] According to the target processing result, a target control setting strategy of the first trading account for the target trading account is determined; and according to the target control setting strategy, corresponding account control settings are performed on the first trading account.

[0010] In one embodiment, the associated data includes at least one of the following: business transaction records between the first customer and the target customer, attribute parameters of the first transaction account, attribute parameters of the target transaction account, business transaction records between the target customer and associated customers, and associated files related to the first customer.

[0011] In one embodiment, the target transaction account includes a plurality of target transaction accounts;

[0012] Accordingly, according to the target control setting strategy, corresponding account control settings are performed on the first trading account, including at least one of the following:

[0013] According to the target control setting strategy, the first trading account is set to support the transaction data interaction with the target trading account;

[0014] According to the target control setting strategy, the first trading account is prohibited from interacting with the target trading account in terms of trading data;

[0015] According to the target control setting strategy, the first trading account is set to support receiving transaction data initiated by the target trading account, but is prohibited from initiating transaction data to the target trading account;

[0016] According to the target control setting strategy, the first trading account is set to support initiating transaction data to the target trading account, but is prohibited from receiving transaction data initiated by the target trading account.

[0017] In one embodiment, the historical transaction data and associated data are processed using a preset account control strategy prediction model to obtain corresponding target processing results, including:

[0018] Extracting corresponding transaction features from the historical transaction data; extracting corresponding association features from the association data;

[0019] According to a preset splicing rule, the transaction feature and the associated feature are spliced ​​to obtain a joint feature;

[0020] The joint features are input into a preset account management and control strategy prediction model for processing to obtain the corresponding target processing results.

[0021] In one embodiment, the method further comprises:

[0022] According to the preset protocol rules, a model training request for a preset account control policy prediction model is initiated in the alliance chain; wherein the first server responds to the model training request, generates an initial prediction model based on an extreme gradient boosting tree structure; and uploads the initial prediction model to the alliance chain;

[0023] Download the initial prediction model from the alliance chain; and train the initial prediction model locally using the sample data of the first customer to obtain a first intermediate model; wherein the second customer terminal downloads the initial prediction model; and trains the initial prediction model locally using the sample data of the associated customer to obtain a second intermediate model;

[0024] The model parameters of the first intermediate model are uploaded to the alliance chain; wherein the second client terminal uploads the model parameters of the second intermediate model to the alliance chain; the first server uses the model parameters of the first intermediate model and the model parameters of the second intermediate model in combination to perform model aggregation to obtain a corresponding intermediate global model;

[0025] Downloading the intermediate global model from the alliance chain; and detecting whether the intermediate global model meets the preset model requirements;

[0026] When it is determined that the intermediate global model meets the preset model requirements, the intermediate global model is determined as the preset account control policy prediction model; and an end instruction for the end of the training of the preset account control policy prediction model is initiated in the alliance chain.

[0027] In one embodiment, after detecting whether the intermediate global model meets the preset model requirements, the method further includes:

[0028] When it is determined that the intermediate global model does not meet the preset model requirements, local training is performed locally using the sample training data of the first customer and the intermediate global model to obtain an updated first intermediate model; wherein the second customer terminal locally performs local training using the sample data of the associated customer and the intermediate global model to obtain an updated second intermediate model;

[0029] The model parameters of the updated first intermediate model are uploaded to the alliance chain; wherein the second client terminal uploads the model parameters of the updated second intermediate model to the alliance chain; the first server obtains the corresponding updated intermediate global model through model aggregation based on the model parameters of the updated first intermediate model and the model parameters of the updated second intermediate model.

[0030] In one embodiment, uploading the model parameters of the first intermediate model to the alliance chain includes:

[0031] Encrypting the model parameters of the first intermediate model using a preset encryption algorithm to obtain ciphertext data of the model parameters of the first intermediate model; wherein the preset encryption algorithm includes: a Paillier semi-homomorphic encryption algorithm;

[0032] Upload the ciphertext data of the model parameters of the first intermediate model to the alliance chain;

[0033] Correspondingly, the first server uses a combination of model parameters of the first intermediate model and model parameters of the second intermediate model to perform model aggregation according to a preset encryption algorithm to obtain ciphertext data of the corresponding intermediate global model.

[0034] In one embodiment, the model parameters of the first intermediate model uploaded to the alliance chain also carry the identity certificate of the first client terminal; wherein the identity certificate of the first client terminal is generated according to the first private key of the first client terminal;

[0035] Correspondingly, before the first server combines the model parameters of the first intermediate model and the model parameters of the second intermediate model for model aggregation, the first server also authenticates the first client terminal based on the identity credentials of the first client terminal carried by the model parameters of the first intermediate model; and generates a first type of error prompt when it is determined that the identity authentication of the first client terminal has failed.

[0036] In one embodiment, the first server uses the model parameters of the first intermediate model and the model parameters of the second intermediate model in combination to perform model aggregation to obtain a corresponding intermediate global model, including:

[0037] Perform model aggregation according to the following formula to obtain the corresponding intermediate global model:

[0038]

[0039] Wherein, W1 is the model parameter of the intermediate global model, W0 is the model parameter of the initial prediction model, η is the update learning rate of the global model, n is the total amount of sample data used in this round of joint training, K is the number of participants participating in this round of joint training of the preset account control strategy prediction model, k is the number of participants, The model parameters of the intermediate model uploaded by participant numbered k.

[0040] In one embodiment, the first server uses the model parameters of the first intermediate model and the model parameters of the second intermediate model in combination to perform model aggregation to obtain a corresponding intermediate global model, including:

[0041] The first server obtains noise data that obeys the Laplace distribution; and combines the model parameters of the first intermediate model, the model parameters of the second intermediate model, and the noise data to perform model aggregation to obtain the corresponding intermediate global model.

[0042] In one embodiment, downloading the intermediate global model from the consortium chain includes:

[0043] Generate a download request for the intermediate global model; wherein the download request carries at least the identity certificate of the first client terminal;

[0044] Initiate the download request in the alliance chain; wherein the alliance chain also deploys a smart contract; the smart contract responds to the download request and verifies the authority of the first client terminal according to the identity certificate of the first client terminal; when it is determined that the authority verification passes, the smart contract sends the download certificate to the first client terminal;

[0045] Download the intermediate global model from the consortium chain based on the download credentials.

[0046] In one embodiment, after determining the intermediate global model as a preset account management and control strategy prediction model, the method further includes:

[0047] At every preset time period, a model update request for the preset account control strategy prediction model is initiated in the consortium chain;

[0048] In response to the model update request, according to the preset protocol rules, cooperate with the first server and the second client terminal respectively using the updated sample data of the preset time period to update the preset account management and control strategy prediction model based on the alliance chain.

[0049] This specification also provides a transaction account management and control setting processing device, which is applied to a first client terminal, and the first client terminal is deployed on the first client side, including:

[0050] an acquisition module, configured to acquire historical transaction data of a first transaction account and a target transaction account, and associated data related to the first transaction account and the target transaction account; wherein the first transaction account is a transaction account held by a first customer, and the target transaction account is a transaction account held by a target customer;

[0051] A prediction module, used to process the historical transaction data and associated data using a preset account control strategy prediction model to obtain a corresponding target processing result; wherein the preset account control strategy prediction model is a model based on an extreme gradient boosting tree structure obtained by the first client terminal in advance in cooperation with the first server and the second client terminal based on the alliance chain through federated learning training according to preset protocol rules; the first server is deployed on the side of the financial service institution; the second client terminal is deployed on the side of an associated client that has a cooperative relationship with the first client;

[0052] The processing module is used to determine the target control setting strategy of the first trading account for the target trading account according to the target processing result; and perform corresponding account control settings on the first trading account according to the target control setting strategy.

[0053] The present specification also provides a client terminal, comprising a processor and a memory for storing processor executable instructions, wherein the processor implements the relevant steps of the transaction account management and control setting processing method when executing the instructions.

[0054] The present specification also provides a computer-readable storage medium having computer instructions stored thereon, which, when executed by a processor, implement the following steps: obtaining historical transaction data of a first transaction account and a target transaction account, and associated data related to the first transaction account and the target transaction account; wherein the first transaction account is a transaction account held by a first customer, and the target transaction account is a transaction account held by a target customer; processing the historical transaction data and associated data using a preset account control strategy prediction model to obtain a corresponding target processing result; wherein the preset account control strategy prediction model is a model based on an extreme gradient boosting tree structure obtained by the first customer terminal in advance according to preset protocol rules in cooperation with the first server and the second customer terminal based on a consortium chain through federated learning training; the first server is deployed on the side of a financial service institution; the second customer terminal is deployed on the side of an associated customer that has a cooperative relationship with the first customer; according to the target processing result, determining a target control setting strategy for the first transaction account with respect to the target transaction account; and performing corresponding account control settings on the first transaction account according to the target control setting strategy.

[0055] Based on the transaction account control setting processing method, device and client terminal provided in this specification, before specific implementation, the first client terminal deployed on the first client side can cooperate with the first server deployed on the financial service institution and the second client terminal deployed on the second client side with which the first client has a cooperative relationship according to the preset protocol rules, respectively using the sample data held by each party locally, based on the alliance chain through federated learning, without leaking the data privacy of the participants, and jointly training to obtain a relatively comprehensive and effective preset account control strategy prediction model based on the extreme gradient boosting tree structure. During specific implementation, the first client terminal obtains the historical transaction data of the first transaction account and the target transaction account, as well as the associated data related to the first transaction account and the target transaction account; and uses the preset account control strategy prediction model to process the transaction data and the associated data to obtain the corresponding target processing result; then, according to the target processing result, determine the target control setting strategy of the first transaction account for the target transaction account; and according to the target control setting strategy, perform corresponding account control settings on the first transaction account. Thereby, the workload of the staff of the first client can be effectively simplified, and the account control settings of the first transaction account can be automatically and accurately completed. BRIEF DESCRIPTION OF THE DRAWINGS

[0056] In order to more clearly illustrate the embodiments of this specification, the drawings required for use in the embodiments will be briefly introduced below. The drawings described below are only some embodiments recorded in this specification. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0057] Figure 1 It is a flowchart of a method for controlling and setting a transaction account provided by an embodiment of this specification;

[0058] Figure 2 It is a schematic diagram of an embodiment of a method for controlling and setting a transaction account provided by an embodiment of this specification in a scenario example;

[0059] Figure 3 It is a schematic diagram of an embodiment of a method for controlling and setting a transaction account provided by an embodiment of this specification in a scenario example;

[0060] Figure 4 It is a schematic diagram of an embodiment of a method for controlling and setting a transaction account provided by an embodiment of this specification in a scenario example;

[0061] Figure 5 It is a schematic diagram of an embodiment of a method for controlling and setting a transaction account provided by an embodiment of this specification in a scenario example;

[0062] Figure 6 It is a schematic diagram of an embodiment of a method for controlling and setting a transaction account provided by an embodiment of this specification in a scenario example;

[0063] Figure 7 is a schematic diagram of the structure of a client terminal provided by an embodiment of this specification;

[0064] Figure 8 It is a schematic diagram of the structural composition of a transaction account management and control setting processing device provided by an embodiment of the present specification. DETAILED DESCRIPTION

[0065] In order to enable those skilled in the art to better understand the technical solutions in this specification, the technical solutions in the embodiments of this specification will be clearly and completely described below in conjunction with the drawings in the embodiments of this specification. Obviously, the described embodiments are only part of the embodiments of this specification, not all of the embodiments. Based on the embodiments in this specification, all other embodiments obtained by ordinary technicians in this field without creative work should fall within the scope of protection of this specification.

[0066] See also Figure 1 As shown, the embodiment of this specification provides a method for controlling and setting a transaction account. The method is specifically applied to the first client terminal. When implemented, the method may include the following contents:

[0067] S101: Acquire historical transaction data of a first transaction account and a target transaction account, and associated data related to the first transaction account and the target transaction account; wherein the first transaction account is a transaction account held by a first customer, and the target transaction account is a transaction account held by a target customer;

[0068] S102: Process the historical transaction data and associated data using a preset account control strategy prediction model to obtain a corresponding target processing result; wherein the preset account control strategy prediction model is a model based on an extreme gradient boosting tree structure obtained by the first client terminal in advance in cooperation with the first server and the second client terminal based on the alliance chain through federated learning training according to preset protocol rules; the first server is deployed on the side of the financial service institution; the second client terminal is deployed on the side of an associated client that has a cooperative relationship with the first client;

[0069] S103: Determine a target control setting strategy for the first trading account with respect to the target trading account according to the target processing result; and perform corresponding account control settings on the first trading account according to the target control setting strategy.

[0070] Based on the above embodiment, the first customer terminal can use the pre-trained preset account control strategy prediction model to process the acquired historical transaction data of the first transaction account and the target transaction account, as well as the associated data related to the first transaction account and the target transaction account, to automatically and accurately complete the account control settings of the first transaction account, thereby effectively simplifying the workload of the first customer's staff and reducing setting errors such as omissions in settings.

[0071] In some embodiments, the preset account control strategy prediction model can be specifically understood as an algorithm model based on an extreme gradient boosting tree structure obtained by the cooperation of the first client terminal, the first server, and the second client terminal through federated learning training based on the alliance chain. Based on the preset account control strategy prediction model, the historical transaction data of the first trading account and the target trading account, as well as the associated data related to the first trading account and the target trading account can be input, and through comprehensive analysis of multiple factors, combined with past and current business development conditions, as well as future business development trends, automatically determine and output the control setting strategy of the first trading account with respect to the target trading account.

[0072] The control setting strategy at least includes the account identifier of the target trading account to be restricted or prohibited. Furthermore, the control setting strategy may also include information such as the restriction method and degree of restriction for the restricted target trading account.

[0073] The first client terminal may be deployed on the first client side. The first transaction account may be a transaction account held by the first client. The first client may be a company, institution or other client.

[0074] The first server may be deployed on the side of a financial service institution. The financial service institution may include at least a bank that supports the transaction data processing of the first transaction account. Specifically, the first server may include multiple first servers, each of which corresponds to a financial service institution. Accordingly, the financial service institution may also include: an insurance company, an exchange, a securities company, a transaction regulatory agency, a transaction platform, other relevant banks, and the like.

[0075] The second client terminal may be deployed on the associated client side, wherein the associated client may be understood as other clients having a cooperative relationship with the first client.

[0076] The target transaction account may be a transaction account held by the target customer. The target transaction account may be a transaction account that may interact with the first transaction account held by the first customer, but it is yet to be determined whether to be subject to control restrictions. The target transaction account may be one or more target transaction accounts.

[0077] In specific implementation, the first client terminal may obtain and determine the target transaction account of the first transaction account based on the transaction data processing records involving the first transaction account in the historical time period and the first client's past and current cooperation plans.

[0078] During the specific implementation, depending on the specific situation, other participants can also be introduced based on the corresponding agreement, such as rival customers that are in competition with the first customer, etc., to jointly train with the first customer, related customers, and financial service institutions to obtain a preset account management strategy prediction model with relatively better results.

[0079] In some embodiments, the historical transaction data of the transaction account and the target transaction account may be historical transaction data involving the transaction account and the target transaction account occurring within a specified time period (e.g., the most recent year). The historical transaction data may include basic information related to the historical transaction data, such as transaction date, transaction amount, transaction account (or account number), name of the account holder, etc. Furthermore, the historical transaction data may also include additional information related to the historical transaction data, such as the name and number of the bank where the transaction account is opened, contract information, transaction notes, and related bidding contract identification.

[0080] In some embodiments, the associated data may specifically include at least one of the following: business transaction records between the first customer and the target customer, attribute parameters of the first transaction account, attribute parameters of the target transaction account, business transaction records between the target customer and associated customers, associated files related to the first customer, etc.

[0081] The business transaction records between the first customer and the target customer can be specifically understood as other business interaction records between the first customer and the target customer except for the transaction data. For example, the correspondence emails between the first customer and the target customer. Another example is the business cooperation records between the first customer and the target customer. The business transaction records between the target customer and the associated customers are similar and will not be elaborated in this specification.

[0082] The above attribute parameters may specifically include: the name of the trading account holder, the name of the bank where the trading account is opened, the user level of the trading account holder, the account type of the trading account, the account identifier of the trading account, etc.

[0083] The above-mentioned associated documents related to the first customer may specifically include: a letter of commitment signed by the first customer, a business contract in which the first customer participates, an announcement issued by the first customer, a business development plan of the first customer, a letter of intent for cooperation signed by the first customer and other customers, etc.

[0084] Furthermore, the above-mentioned related data may also include risky customer lists, transaction supervision reports, transaction credit whitelists issued by financial service institutions, cooperation agreements and prohibited matters between the opening bank of the first transaction account and other banks, etc.

[0085] Based on the above embodiments, relatively rich and diverse associated data can be obtained and utilized, and combined with relevant historical transaction data, so that the preset account management strategy model can perform relatively more comprehensive analysis and processing based on the above data, and output and obtain relatively more accurate target processing results.

[0086] In some embodiments, the target transaction account may specifically include multiple target transaction accounts;

[0087] Accordingly, the above-mentioned corresponding account control settings for the first trading account according to the target control setting strategy may specifically include at least one of the following:

[0088] According to the target control setting strategy, the first trading account is set to support the transaction data interaction with the target trading account;

[0089] According to the target control setting strategy, the first trading account is prohibited from interacting with the target trading account in terms of trading data;

[0090] According to the target control setting strategy, the first trading account is set to support receiving transaction data initiated by the target trading account, but is prohibited from initiating transaction data to the target trading account;

[0091] According to the target control setting strategy, the first trading account is set to support initiating transaction data to the target trading account, but is prohibited from receiving transaction data initiated by the target trading account.

[0092] Based on the above embodiment, the first client terminal can perform a variety of refined account control setting operations according to the target control setting strategy to meet the diverse account control needs of the customer.

[0093] In some embodiments, see Figure 2 As shown, the above-mentioned use of the preset account control strategy prediction model to process the historical transaction data and the associated data to obtain the corresponding target processing results may include the following contents during specific implementation:

[0094] S1: extracting corresponding transaction features from the historical transaction data; extracting corresponding association features from the association data;

[0095] S2: According to a preset splicing rule, splice the transaction feature and the associated feature to obtain a joint feature;

[0096] S3: Input the joint features into a preset account management and control strategy prediction model for processing to obtain a corresponding target processing result.

[0097] Based on the above embodiments, the preset account management and control strategy prediction model can be used to comprehensively and effectively utilize the characteristic information in the historical transaction data and related data to obtain a target processing result with good accuracy and high reference value.

[0098] In specific implementation, the transaction feature template corresponding to the transaction data can be used to extract relevant transaction features from the historical transaction data. The preset semantic recognition model corresponding to the associated data can be selected from multiple preset semantic recognition models as the target semantic recognition model; the associated data can be processed using the target semantic recognition model to obtain a semantic recognition result; and then the above-mentioned associated features can be extracted from the semantic recognition result according to the associated feature template corresponding to the associated data.

[0099] In specific implementation, the above transaction features and associated features can be mapped into feature codes of unified dimensions, namely, transaction feature codes and associated feature codes, respectively, according to preset splicing rules; and then the above transaction feature codes and associated feature codes are connected once in sequence to obtain joint features that can be recognized and processed by the preset account management strategy prediction model.

[0100] In some embodiments, see Figure 3 As shown, when the method is implemented specifically, it may also include the following contents:

[0101] S1: Initiate a model training request for a preset account control policy prediction model in the alliance chain according to preset protocol rules; wherein the first server responds to the model training request, generates an initial prediction model based on an extreme gradient boosting tree structure; and uploads the initial prediction model to the alliance chain;

[0102] S2: Download the initial prediction model from the alliance chain; and train the initial prediction model locally using the sample data of the first customer to obtain a first intermediate model; wherein the second customer terminal downloads the initial prediction model; and trains the initial prediction model locally using the sample data of the associated customer to obtain a second intermediate model;

[0103] S3: uploading the model parameters of the first intermediate model to the alliance chain; wherein the second client terminal uploads the model parameters of the second intermediate model to the alliance chain; the first server combines the model parameters of the first intermediate model and the model parameters of the second intermediate model to perform model aggregation to obtain the corresponding intermediate global model;

[0104] S4: Downloading the intermediate global model from the alliance chain; and detecting whether the intermediate global model meets the preset model requirements;

[0105] S5: When it is determined that the intermediate global model meets the preset model requirements, the intermediate global model is determined as the preset account control policy prediction model; and an end instruction for the end of the training of the preset account control policy prediction model is initiated in the alliance chain.

[0106] Based on the above embodiment, the first client terminal can cooperate with the first server deployed on the side of the financial service institution and the second client terminal deployed on the side of the second client that has a cooperative relationship with the first client according to the preset protocol rules, and use the sample data held by each of them locally, and through federated learning based on the alliance chain, without leaking the data privacy of the participants, jointly train to obtain a more comprehensive and effective preset account management strategy prediction model based on the extreme gradient boosting tree structure, which can effectively protect the data security of the participants and prevent the data of the participants from being leaked during the training process.

[0107] The sample data of the first customer mentioned above can be specifically understood as the historical transaction data of the sample transaction account of the first customer and the target sample transaction account held by the first customer, as well as the relevant sample association data. The sample data of the associated customer mentioned above can be specifically understood as the historical transaction data of the sample transaction account of the associated customer and the target sample transaction account held by the associated customer, as well as the relevant sample association data.

[0108] Specifically, the above initial prediction model can be constructed in the following form:

[0109]

[0110] in, is the predicted value of the t-th decision tree based on the sample data numbered i, is the predicted value of the t-1th decision tree, f t (x i ) represents the tth decision tree.

[0111] Based on the above initial prediction model, the number of decision trees corresponds to the number of iterations. In specific implementation, the results of all decision trees in the model are accumulated to obtain a value as the prediction value of the model based on the sample data.

[0112] When conducting local training specifically, taking the first customer terminal as an example, the above-mentioned locally using the sample data of the first customer to train the initial prediction model to obtain the first intermediate model may include: locally using the initial prediction model to perform multiple iterative operations based on the sample data of the first customer; and after each completion of the current iterative operation, using a preset first objective function to detect whether the prediction value obtained by the current iterative operation based on the sample data of the first customer meets the iteration stop condition.

[0113] Specifically, when it is determined that the iteration stop condition is met, the iteration is stopped and the current model is used as the first intermediate model. On the contrary, when it is determined that the iteration stop condition is not met, a new decision tree is added to the current model to obtain an updated model as the next model; and the next iteration operation is performed using the next model until the iteration stop condition is met.

[0114] The above preset first objective function can be expressed as follows:

[0115]

[0116]

[0117] Among them, obj t is the function value of the preset first objective function, t represents the number of decision trees in the model, and cons is a constant term. is the error term, Ω(f t ) is the regular term sum expression, Ω(f k ) is a regular term used to reduce the complexity of the model and prevent the model from overfitting. k , ω decibels represent the number and score of leaf nodes of the kth decision tree, and γ and λ are the parameters used to control the number and score scale factor of leaf nodes, respectively.

[0118] In specific implementation, when it is detected that the difference between the function value obtained by using the preset first objective function based on the predicted value of the sample data of the first customer obtained based on the current iterative operation and the reference value of the preset first objective function value is less than or equal to the preset first difference threshold, it is determined that the iteration stop condition is met.

[0119] During specific implementation, depending on the specific situation and preset protocol rules, the first server may not participate in local training and may only be responsible for model aggregation; the first server may also participate in local training like the first client terminal and the second client terminal and be responsible for model aggregation at the same time.

[0120] Specifically, for example, the first server can also synchronously train the initial prediction model locally using the sample data of the financial service institution to obtain the third intermediate model; and upload the model parameters of the third intermediate model to the alliance chain. Correspondingly, the first server can combine the model parameters of the first intermediate model, the model parameters of the second intermediate model, and the model parameters of the third model to perform model aggregation and obtain the corresponding intermediate global model.

[0121] In some embodiments, after detecting whether the intermediate global model meets the preset model requirements, the method may further include the following when implemented:

[0122] S1: When it is determined that the intermediate global model does not meet the preset model requirements, locally training is performed using the sample training data of the first customer and the intermediate global model to obtain an updated first intermediate model; wherein the second customer terminal locally trains using the sample data of the associated customer and the intermediate global model to obtain an updated second intermediate model;

[0123] S2: Upload the updated model parameters of the first intermediate model to the alliance chain; wherein the second client terminal uploads the updated model parameters of the second intermediate model to the alliance chain; the first server obtains the corresponding updated intermediate global model through model aggregation based on the updated model parameters of the first intermediate model and the updated model parameters of the second intermediate model.

[0124] Based on the above embodiment, when it is determined that the intermediate global model obtained by the current round of training does not meet the preset model requirements, the first client terminal, the first server, and the second client terminal participating in the training can repeat the above joint training process to continue one or more rounds of joint training until the updated intermediate global model meets the preset model requirements. Thus, a preset account control strategy prediction model that meets the requirements can be obtained through multiple rounds of joint training.

[0125] In specific implementation, taking the first client terminal in the current joint training in multiple rounds of joint training as an example, the first client terminal downloads the intermediate global model of the current round from the alliance chain; and when it is determined that the intermediate global model of the current round does not meet the preset model requirements, the first client terminal can perform local training locally using the sample training data of the first client and the intermediate global model of the current round to obtain an updated first intermediate model as the intermediate model for the next round: update the updated local model according to the intermediate global model of the current round and the local first intermediate model of the current round; then locally use the sample training data of the first client, perform multiple iterative operations on the updated local model until the iteration stop condition is met, and determine the final local model as the first intermediate model for the next round.

[0126] Specifically, the updated local model can be obtained by updating the intermediate global model of the current round and the first intermediate model of the local current round according to the following formula:

[0127]

[0128] Wherein, b is the serial number of the first client terminal, is the model parameter of the updated local model, w p is the model parameter of the intermediate global model of the current round, η is the model update learning rate of the intermediate global model of the current round, represents the gradient function of the first client terminal, w p Represents the model parameters of the first intermediate model of the current round.

[0129] In some embodiments, when implementing Figure 4 As shown, the server and related client terminals can jointly build a consortium chain according to preset protocol rules; and deploy corresponding smart contracts on the consortium chain.

[0130] The above-mentioned server may specifically include multiple servers deployed in multiple different financial service institutions, for example, server 1 deployed in XX Bank, server 2 deployed in YY Bank, ... server N deployed in a transaction regulatory agency, etc.

[0131] The above-mentioned client terminal may specifically include a plurality of client terminals deployed in a plurality of different companies and institutions, for example, client terminal 1 deployed in AA company, client terminal 2 deployed in BB merchant, ... client terminal M deployed in WW enterprise, and so on.

[0132] The above-mentioned alliance chain can be specifically understood as a blockchain suitable for a specific group composed of different companies and institutions. Based on the above-mentioned alliance chain, it only targets members within a specific group and limited third-party members, and it specifies multiple pre-selected nodes as account keepers. The generation of each block is jointly determined by all pre-selected nodes.

[0133] The above-mentioned smart contract can be specifically understood as a computer protocol designed to disseminate, verify or execute contract rules in an information-based manner. Based on the above-mentioned smart contract, it supports reliable data interaction without a third party and ensures that the above-mentioned data interaction is traceable and irreversible.

[0134] The above-mentioned preset protocol rules may specifically be a rule set formed by integrating alliance chain, federated learning, and related encryption algorithms and other technologies.

[0135] The above-mentioned federated learning (FL) can specifically refer to a distributed machine learning framework. Participants holding data can conduct joint training of models while keeping the data locally, thereby breaking down data barriers and achieving data confidentiality.

[0136] In this embodiment, the server may specifically include a background server applied to a financial service institution and capable of realizing functions such as data transmission and data processing. Specifically, the server may be, for example, an electronic device having data calculation, storage and network interaction functions. Alternatively, the server may also be a software program running in the electronic device to provide support for data processing, storage and network interaction. In this embodiment, the number of the servers is not specifically limited. The server may specifically be one server, or several servers, or a server cluster formed by several servers.

[0137] In this embodiment, the client terminal may specifically include a front end applied to the client side and capable of realizing functions such as data collection and data transmission. Specifically, the client terminal may be, for example, an electronic device such as a desktop computer, a tablet computer, a laptop computer, etc. Alternatively, the client terminal may also be a software application that can be run in the above electronic devices.

[0138] In specific implementation, whenever a new member (for example, a new client terminal or a new server) wants to join the alliance chain, it can initiate a joining application to the alliance chain; wherein the joining application at least carries the identity of the new member. The smart contract receives and responds to the joining application, obtains the basic information of the new member according to the preset protocol rules, and initiates a proposal in the alliance chain on whether to approve the new member to join; wherein the proposal at least carries the identity of the new member and the basic information of the new member.

[0139] The old members in the alliance chain (including the existing servers and client terminals in the alliance chain) can reach a consensus on the proposal based on the identity of the new member and the basic information of the new member. The smart contract can determine whether to agree to the new member to join the alliance chain based on the consensus results of the old members in the alliance chain.

[0140] When it is determined that the new member is included in the joint chain, the smart contract can generate a key pair for the new member based on the identity of the new member and a random number generator; wherein the key pair contains the private key and public key corresponding to the new member. Furthermore, the smart contract broadcasts the public key of the new member in the alliance chain; at the same time, the private key is sent to the new member. The private key is not disclosed to the public, but is held and kept by the server or client terminal deployed on the new member's side.

[0141] Currently, the first customer wants to generate a preset account control strategy prediction model that can assist in automatically generating a control setting strategy for the first trading account. Accordingly, the first customer can initiate a model training request for the preset account control strategy prediction model in the alliance chain through his own client terminal (referred to as the first client terminal) according to the preset protocol rules; wherein the above training request carries at least the account identifier of the first trading account.

[0142] The smart contract receives the above-mentioned model training request and, based on the account identifier of the first trading account, usually determines the opening bank of the first trading account through data query, and further determines the server of the opening bank as the first server; and forwards the above-mentioned model training request to the first server.

[0143] At the same time, the smart contract can also further interact with the first client terminal to determine the associated client who is suitable and agrees to participate in the model training, and transfer the above model training to the second client terminal of the associated client.

[0144] The first server receives and responds to the model training request according to the preset protocol rules, generates an initial prediction model based on the extreme gradient boosting tree structure, and uploads the initial prediction model to the alliance chain.

[0145] Specifically, the extreme gradient boosting tree (XGBoost) can be understood as a tree-based gradient boosting algorithm. Based on the extreme gradient boosting tree, the corresponding prediction function can be achieved by combining multiple decision tree models constructed by iterative training, which has higher prediction accuracy, faster operation efficiency, and can effectively prevent model overfitting.

[0146] The first client terminal and the second client terminal can download the initial prediction model from the alliance chain according to the preset protocol rules, and train the initial prediction model locally with the sample data they hold (perform local training) to obtain the corresponding first intermediate model and second intermediate model. Since the sample data held by the first client terminal and the second client terminal respectively do not leave the local client terminal during the above-mentioned local training process, the relevant sample data can be effectively prevented from being leaked during the joint training process.

[0147] Furthermore, the first client terminal and the second client terminal may upload the model parameters of the first intermediate model and the second intermediate model obtained through training to the alliance chain respectively according to preset protocol rules.

[0148] Next, the first server may perform model aggregation by combining the model parameters of the first intermediate model and the model parameters of the second intermediate model according to preset protocol rules to obtain a corresponding intermediate global model; and upload the intermediate global model to the alliance chain.

[0149] Correspondingly, the first client terminal can download the intermediate global model from the alliance chain according to the preset protocol rules; and detect whether the intermediate global model meets the preset model requirements. If it is determined that the intermediate global model meets the preset model requirements, the intermediate global model can be determined as the preset account control strategy prediction model; and an end instruction for the end of the preset account control strategy prediction model training is initiated in the alliance chain to end the model training.

[0150] In addition, after detecting the above-mentioned end instruction, the smart contract will mark the current latest intermediate global model as the preset account control strategy prediction model that has been trained; and store the preset account control strategy prediction model on the alliance chain for subsequent query use.

[0151] On the contrary, when the first client terminal detects that the intermediate global model does not meet the preset model requirements, it can repeat the above joint training process with the first server and the second client terminal through a smart contract to continue one or more rounds of joint training until the obtained model meets the preset model requirements.

[0152] In specific implementation, taking the current round of joint training in multiple rounds of joint training as an example, the first server can combine the model parameters of the first intermediate model of the current round and the model parameters of the second intermediate model of the current round according to the following formula, and combine the intermediate global model of the previous round to perform model aggregation to obtain the intermediate global model of the current round:

[0153]

[0154] Among them, W p is the model parameter of the intermediate global model of the current round, W p-1 is the intermediate global model of the previous round, p is the number of the current round, W0 is the initial prediction model, η is the update learning rate of the intermediate global model of the current round, n represents the total amount of sample data participating in the model training of the current round, K represents the number of local training participants participating in the model training of the current round, is the model parameter of the intermediate model of the current round uploaded by the participant numbered k (eg, the first client terminal or the second client terminal) in the current round, Represents the training gradient of the participant numbered k during local training.

[0155] In specific implementation, the first client terminal can use the preset second objective function to detect whether the intermediate global model of the current round meets the preset model requirements. The preset second objective function can be expressed in the following form:

[0156]

[0157] Where n represents the total amount of sample data participating in the current round of model training, K represents the number of local training participants participating in the current round of model training, Represents the loss value corresponding to the participant numbered i.

[0158] Specifically, the function value of the second objective function preset in the current round can be calculated based on the intermediate global model of the current round, and by detecting whether the function value of the second objective function preset in the current round is less than a preset function value threshold, it is determined whether the preset model requirement is met. Specifically, when it is determined that the function value of the second objective function preset in the current round is less than or equal to the preset function value threshold, it is determined that the preset model requirement is met. Otherwise, it is determined that the preset model requirement is not met.

[0159] During specific implementation, the learning rate may also be determined and updated according to the difference between the function value of the preset second objective function of the current round and the preset function value threshold.

[0160] In some embodiments, see Figure 5 As shown, the above uploading of the model parameters of the first intermediate model to the alliance chain may include the following contents during specific implementation:

[0161] S1: Encrypting the model parameters of the first intermediate model using a preset encryption algorithm to obtain ciphertext data of the model parameters of the first intermediate model; wherein the preset encryption algorithm includes: Paillier semi-homomorphic encryption algorithm;

[0162] S2: Upload the ciphertext data of the model parameters of the first intermediate model to the alliance chain;

[0163] Correspondingly, when the first server is implemented, it can combine the model parameters of the first intermediate model and the model parameters of the second intermediate model to perform model aggregation according to a preset encryption algorithm to obtain the ciphertext data of the corresponding intermediate global model.

[0164] Based on the above embodiment, by introducing and using the Paillier semi-homomorphic encryption algorithm as the preset encryption algorithm to encrypt and protect the model parameters, the data security of the participants can be more effectively protected and the relevant data of the participants can be prevented from being leaked.

[0165] In some embodiments, the model parameters of the first intermediate model uploaded to the alliance chain may also carry the identity certificate of the first client terminal; wherein the identity certificate of the first client terminal may be generated according to the first private key of the first client terminal;

[0166] Correspondingly, before the first server combines the model parameters of the first intermediate model and the model parameters of the second intermediate model for model aggregation, it can also authenticate the first client terminal based on the identity credentials of the first client terminal carried by the model parameters of the first intermediate model; and generate a first type of error prompt when it is determined that the identity authentication of the first client terminal has failed.

[0167] Based on the above embodiments, by introducing identity credentials for identity authentication, data security in the joint training process can be protected more comprehensively and effectively.

[0168] Specifically, the model parameters of the first intermediate model may also carry the identity identifier of the first client terminal. Accordingly, after receiving the model parameters of the first intermediate model, the first server may first select the first public key corresponding to the identity identifier from the previously received public keys according to the identity identifier of the first client terminal; and then use the first public key to verify the authenticity of the identity certificate of the first client terminal. When it is determined that the identity certificate is authentic, it can be determined that the identity authentication of the first client terminal has been passed, and subsequent data processing can be performed normally.

[0169] On the contrary, if it is determined that the identity credential is false, it can be determined that the identity authentication of the first client terminal has not been passed, and then it can be judged that the model parameters of the first intermediate model may be uploaded by other data parties impersonating the first client terminal, and the model parameters of the first intermediate model may have security risks. At this time, the first server will suspend subsequent data processing and generate a first-class error prompt; at the same time, the first-class error prompt will be sent to the first client terminal. In addition, the first server will also store the above-mentioned first-class error prompt in the alliance chain for evidence storage.

[0170] In some embodiments, the first server uses the model parameters of the first intermediate model and the model parameters of the second intermediate model to perform model aggregation to obtain the corresponding intermediate global model. When specifically implemented, the following may be included:

[0171] The model is aggregated according to the following formula to obtain the corresponding intermediate global model (i.e., the intermediate global model of the first round):

[0172]

[0173] Wherein, W1 is the model parameter of the intermediate global model, W0 is the model parameter of the initial prediction model, η is the update learning rate of the global model (i.e., the learning rate of the first round), n is the total amount of sample data used in this round of joint training, K is the number of participants participating in this round of joint training of the preset account control strategy prediction model, k is the number of the participant, The model parameters of the intermediate model uploaded by participant numbered k.

[0174] Based on the above embodiment, the first server can effectively combine and use the model parameters of the first intermediate model and the model parameters of the second intermediate model to perform model aggregation, and obtain an intermediate global model that meets the requirements.

[0175] In some embodiments, when the model parameters of the first intermediate model and the model parameters of the second intermediate model uploaded to the alliance chain are ciphertext data encrypted based on a preset encryption algorithm. The first server can use the relevant semi-homomorphic encryption characteristics according to the preset encryption algorithm to combine the model parameters of the first intermediate model and the model parameters of the second intermediate model in the ciphertext state to perform model aggregation to obtain the ciphertext data of the corresponding intermediate global model. Correspondingly, after the first client terminal obtains the ciphertext data of the intermediate global model through the alliance chain, it can locally decrypt the ciphertext data of the above intermediate global model according to the preset encryption algorithm, obtain the intermediate global model in plain text form, and then perform subsequent data processing.

[0176] In some embodiments, the first server uses the model parameters of the first intermediate model and the model parameters of the second intermediate model to perform model aggregation to obtain the corresponding intermediate global model. When specifically implemented, the following contents may be included:

[0177] The first server obtains noise data that obeys the Laplace distribution; and combines the model parameters of the first intermediate model, the model parameters of the second intermediate model, and the noise data to perform model aggregation to obtain the corresponding intermediate global model.

[0178] Based on the above embodiments, by introducing and using noise data that obeys the Laplace distribution in the process of model aggregation, on the one hand, it will not affect the subsequent use of the intermediate global model; on the other hand, the introduced noise data can confuse the model parameters of the first intermediate model and the model parameters of the second intermediate model, so that even if the intermediate global model is leaked, it is impossible to reversely infer the accurate model parameters of the first intermediate model and the model parameters of the second intermediate model, and it is even more impossible to reversely infer the sample data of the first customer used in training the first intermediate model and the sample data of the associated customers used in training the second intermediate model, thereby being able to more effectively protect the data security of the participants.

[0179] In specific implementation, the noise data that obeys the Laplace distribution can be determined according to the following formula:

[0180]

[0181] Wherein, Lap(μ,λ) represents Laplace distribution, μ and λ are location parameter and scale parameter respectively, and h is the noise data that obeys Laplace distribution.

[0182] The use of the above-mentioned noise data that obeys the Laplace distribution can effectively prevent malicious data parties from inferring whether a certain sample data belongs to the sample data held by a certain participant by accessing and inferring the global model during the training process or the final model output after the training is completed, thereby better protecting the data privacy of the participants and avoiding the leakage of the relevant data information of the participants.

[0183] In some embodiments, see Figure 6 As shown, the above downloading of the intermediate global model from the alliance chain may include the following contents during specific implementation:

[0184] S1: Generate a download request for the intermediate global model; wherein the download request carries at least the identity certificate of the first client terminal;

[0185] S2: Initiate the download request in the alliance chain; wherein the alliance chain also deploys a smart contract; the smart contract responds to the download request and performs permission verification on the first client terminal according to the identity certificate of the first client terminal; when it is determined that the permission verification passes, the smart contract sends the download certificate to the first client terminal;

[0186] S3: Download the intermediate global model from the consortium chain based on the download credentials.

[0187] Based on the above embodiment, it is possible to effectively prevent a third party who does not have relevant authority and is not a participant in the joint training from illegally downloading and obtaining the above intermediate global model, thereby more effectively protecting the data security of the model joint training process.

[0188] When performing specific permission verification, the smart contract can use the first public key of the first client terminal to verify the authenticity of the identity credentials of the first client terminal; after determining that it is true, further query the preset permission list based on the identity identification of the first client terminal to determine whether the first client terminal currently has the download permission for the intermediate global model of the second client terminal.

[0189] In some embodiments, after the intermediate global model is determined as the preset account management and control strategy prediction model, the method may further include the following contents when implemented:

[0190] S1: Initiate a model update request for the preset account control strategy prediction model in the alliance chain at every preset time period;

[0191] S2: Respond to the model update request, and according to the preset protocol rules, cooperate with the first server and the second client terminal respectively using the updated sample data of the preset time period to update the preset account management and control strategy prediction model based on the alliance chain.

[0192] Based on the above embodiment, the first client terminal can, at intervals of a preset time period, and according to preset protocol rules, work with the first server and the second client terminal to continuously train and update the preset account management strategy prediction model using the latest sample data obtained in the preset time period, so as to continuously improve the model accuracy.

[0193] As can be seen from the above, based on the control setting processing method of the transaction account provided in the embodiment of this specification, before the specific implementation, the first client terminal deployed on the first client side can cooperate with the first server deployed on the financial service institution side and the second client terminal deployed on the second client side that has a cooperative relationship with the first client according to the preset protocol rules, respectively using the sample data held by each locally, based on the alliance chain through federated learning, without leaking the data privacy of the participants, and jointly training to obtain a relatively comprehensive and effective preset account control strategy prediction model based on the extreme gradient boosting tree structure. In the specific implementation, the first client terminal obtains the historical transaction data of the first transaction account and the target transaction account, as well as the associated data related to the first transaction account and the target transaction account; and uses the preset account control strategy prediction model to process the transaction data and the associated data to obtain the corresponding target processing result; according to the target processing result, determine the target control setting strategy of the first transaction account for the target transaction account; and according to the target control setting strategy, perform corresponding account control settings on the first transaction account. Thereby, the workload of the staff of the first customer can be effectively simplified, and the account control settings of the first transaction account can be automatically and accurately completed.

[0194] This specification also provides a client terminal, see Figure 7 As shown. The client terminal includes a network communication port 701, a processor 702 and a memory 703, and the above structures are connected by internal cables so that each structure can perform specific data interaction. The client terminal can be connected to the alliance chain according to the preset protocol rules.

[0195] The network communication port 701 may be specifically used to obtain historical transaction data of the first transaction account and the target transaction account, as well as associated data related to the first transaction account and the target transaction account; wherein the first transaction account is a transaction account held by the first customer, and the target transaction account is a transaction account held by the target customer.

[0196] The processor 702 can be specifically used to process the historical transaction data and related data using a preset account control strategy prediction model to obtain a corresponding target processing result; wherein the preset account control strategy prediction model is a model based on an extreme gradient boosting tree structure obtained by the first client terminal in advance according to preset protocol rules in cooperation with the first server and the second client terminal based on the alliance chain through federated learning training; the first server is deployed on the side of a financial service institution; the second client terminal is deployed on the side of an associated client that has a cooperative relationship with the first client; according to the target processing result, the target control setting strategy of the first trading account for the target trading account is determined; and according to the target control setting strategy, the corresponding account control setting is performed on the first trading account.

[0197] The memory 703 may be specifically used to store corresponding instruction programs.

[0198] In this embodiment, the network communication port 701 can be a virtual port that is bound to different communication protocols so that different data can be sent or received. For example, the network communication port can be a port responsible for web data communication, a port responsible for FTP data communication, or a port responsible for email data communication. In addition, the network communication port can also be a physical communication interface or communication chip. For example, it can be a wireless mobile network communication chip, such as GSM, CDMA, etc.; it can also be a Wifi chip; it can also be a Bluetooth chip.

[0199] In this embodiment, the processor 702 may be implemented in any appropriate manner. For example, the processor may take the form of a microprocessor or processor and a computer-readable medium storing a computer-readable program code (such as software or firmware) executable by the (micro)processor, a logic gate, a switch, an application specific integrated circuit (ASIC), a programmable logic controller, and an embedded microcontroller, etc. This specification does not limit this.

[0200] In this embodiment, the memory 703 may include multiple levels. In a digital system, anything that can store binary data can be a memory; in an integrated circuit, a circuit with a storage function that has no physical form is also called a memory, such as RAM, FIFO, etc.; in a system, a storage device with a physical form is also called a memory, such as a memory stick, TF card, etc.

[0201] The embodiment of the present specification also provides a computer-readable storage medium based on the above-mentioned transaction account control setting processing method, wherein the computer-readable storage medium stores computer program instructions, and when the computer program instructions are executed, the following are implemented: obtaining historical transaction data of a first transaction account and a target transaction account, and associated data related to the first transaction account and the target transaction account; wherein the first transaction account is a transaction account held by a first customer, and the target transaction account is a transaction account held by a target customer; processing the historical transaction data and the associated data using a preset account control strategy prediction model to obtain a corresponding target processing result; wherein the preset account control strategy prediction model is a model based on an extreme gradient boosting tree structure obtained by the first customer terminal in advance according to preset protocol rules in cooperation with the first server and the second customer terminal based on the alliance chain through federated learning training; the first server is deployed on the side of a financial service institution; the second customer terminal is deployed on the side of an associated customer that has a cooperative relationship with the first customer; according to the target processing result, determining the target control setting strategy of the first transaction account for the target transaction account; and performing corresponding account control settings on the first transaction account according to the target control setting strategy.

[0202] In this embodiment, the storage medium includes, but is not limited to, a random access memory (RAM), a read-only memory (ROM), a cache, a hard disk (HDD), or a memory card. The memory may be used to store computer program instructions. The network communication unit may be an interface for network connection communication set in accordance with the standard specified by the communication protocol.

[0203] In this embodiment, the functions and effects specifically implemented by the program instructions stored in the computer-readable storage medium can be explained in comparison with other implementations and will not be described in detail here.

[0204] See also Figure 8 As shown, the embodiment of this specification also provides a transaction account management and control setting processing device, which may specifically include the following structural modules:

[0205] The acquisition module 801 may be specifically used to acquire historical transaction data of a first transaction account and a target transaction account, and associated data related to the first transaction account and the target transaction account; wherein the first transaction account is a transaction account held by a first customer, and the target transaction account is a transaction account held by a target customer;

[0206] The prediction module 802 can be specifically used to process the historical transaction data and the associated data using a preset account control strategy prediction model to obtain a corresponding target processing result; wherein the preset account control strategy prediction model is a model based on an extreme gradient boosting tree structure obtained by the first client terminal in advance in cooperation with the first server and the second client terminal based on the alliance chain through federated learning training according to preset protocol rules; the first server is deployed on the side of the financial service institution; the second client terminal is deployed on the side of an associated client that has a cooperative relationship with the first client;

[0207] The processing module 803 may be specifically configured to determine a target control setting strategy for the first trading account with respect to the target trading account according to the target processing result; and perform corresponding account control settings on the first trading account according to the target control setting strategy.

[0208] In some embodiments, the associated data may specifically include at least one of the following: business transaction records between the first customer and the target customer, attribute parameters of the first transaction account, attribute parameters of the target transaction account, business transaction records between the target customer and associated customers, associated files related to the first customer, etc.

[0209] In some embodiments, the target transaction account may specifically include multiple target transaction accounts;

[0210] Accordingly, when the above-mentioned processing module 803 is specifically implemented, the corresponding account control settings can be performed on the first trading account according to the target control setting strategy in the following manner: according to the target control setting strategy, the first trading account is set to support the interaction of trading data with the target trading account; and / or, according to the target control setting strategy, the first trading account is set to prohibit the interaction of trading data with the target trading account; and / or, according to the target control setting strategy, the first trading account is set to support receiving the trading data initiated by the target trading account, but prohibit initiating trading data to the target trading account; and / or, according to the target control setting strategy, the first trading account is set to support initiating trading data to the target trading account, but prohibit receiving the trading data initiated by the target trading account.

[0211] In some embodiments, when the above-mentioned prediction module 802 is implemented, the historical transaction data and related data can be processed by using a preset account control strategy prediction model in the following manner to obtain a corresponding target processing result: extracting corresponding transaction features from the historical transaction data; extracting corresponding related features from the related data; splicing the transaction features and related features according to preset splicing rules to obtain joint features; inputting the joint features into the preset account control strategy prediction model for processing to obtain the corresponding target processing result.

[0212] In some embodiments, when the device is implemented, it can also be used to: initiate a model training request for a preset account control policy prediction model in the alliance chain according to preset protocol rules; wherein the first server responds to the model training request and generates an initial prediction model based on the extreme gradient boosting tree structure; and uploads the initial prediction model to the alliance chain; downloads the initial prediction model from the alliance chain; and locally trains the initial prediction model using the sample data of the first customer to obtain a first intermediate model; wherein the second customer terminal downloads the initial prediction model; and locally trains the initial prediction model using the sample data of the associated customer to obtain a second intermediate model; uploads the model parameters of the first intermediate model to the alliance chain; wherein the second customer terminal uploads the model parameters of the second intermediate model to the alliance chain; the first server combines the model parameters of the first intermediate model and the model parameters of the second intermediate model for model aggregation to obtain a corresponding intermediate global model; downloads the intermediate global model from the alliance chain; and detects whether the intermediate global model meets the preset model requirements; when it is determined that the intermediate global model meets the preset model requirements, the intermediate global model is determined as the preset account control policy prediction model; and initiates an end instruction in the alliance chain regarding the end of the training of the preset account control policy prediction model.

[0213] In some embodiments, after detecting whether the intermediate global model meets the preset model requirements, the device can also be used when it is implemented: when it is determined that the intermediate global model does not meet the preset model requirements, locally perform local training using the sample training data of the first customer and the intermediate global model to obtain an updated first intermediate model; wherein the second client terminal locally performs local training using the sample data of the associated customer and the intermediate global model to obtain an updated second intermediate model; upload the model parameters of the updated first intermediate model to the alliance chain; wherein the second client terminal uploads the model parameters of the updated second intermediate model to the alliance chain; the first server obtains the corresponding updated intermediate global model through model aggregation based on the model parameters of the updated first intermediate model and the model parameters of the updated second intermediate model.

[0214] In some embodiments, when the device is implemented, the model parameters of the first intermediate model can be uploaded to the alliance chain in the following manner: the model parameters of the first intermediate model are encrypted using a preset encryption algorithm to obtain ciphertext data of the model parameters of the first intermediate model; the ciphertext data of the model parameters of the first intermediate model is uploaded to the alliance chain; accordingly, the first server can perform model aggregation by combining the model parameters of the first intermediate model and the model parameters of the second intermediate model according to the preset encryption algorithm to obtain the ciphertext data of the corresponding intermediate global model.

[0215] In some embodiments, the model parameters of the first intermediate model uploaded to the alliance chain also carry the identity credentials of the first client terminal; wherein, the identity credentials of the first client terminal are generated based on the first private key of the first client terminal; accordingly, before the first server combines the model parameters of the first intermediate model and the model parameters of the second intermediate model for model aggregation, it also authenticates the first client terminal based on the identity credentials of the first client terminal carried by the model parameters of the first intermediate model; when it is determined that the identity authentication of the first client terminal has failed, a first type of error prompt is generated.

[0216] In some embodiments, during specific implementation, the model aggregation can be performed according to the following formula to obtain the corresponding intermediate global model:

[0217]

[0218] Wherein, W1 is the model parameter of the intermediate global model, W0 is the model parameter of the initial prediction model, η is the update learning rate of the global model, n is the total amount of sample data used in this round of joint training, K is the number of participants participating in this round of joint training of the preset account control strategy prediction model, k is the number of participants, The model parameters of the intermediate model uploaded by participant numbered k.

[0219] In some embodiments, the first server combines the model parameters of the first intermediate model and the model parameters of the second intermediate model to perform model aggregation to obtain a corresponding intermediate global model, which may specifically include: the first server obtains noise data that obeys the Laplace distribution; and combines the model parameters of the first intermediate model, the model parameters of the second intermediate model, and the noise data to perform model aggregation to obtain a corresponding intermediate global model.

[0220] In some embodiments, when the device is implemented, the intermediate global model can be downloaded from the alliance chain in the following manner: generate a download request for the intermediate global model; wherein the download request carries at least the identity credentials of the first client terminal; initiate the download request in the alliance chain; wherein the alliance chain also deploys a smart contract; the smart contract responds to the download request and verifies the authority of the first client terminal based on the identity credentials of the first client terminal; when it is determined that the authority verification is passed, the smart contract sends the download credential to the first client terminal; and downloads the intermediate global model from the alliance chain based on the download credential.

[0221] In some embodiments, after the intermediate global model is determined as the preset account control policy prediction model, the device can also be used when it is implemented to: initiate a model update request for the preset account control policy prediction model in the alliance chain at every preset time period; respond to the model update request, and according to the preset protocol rules, cooperate with the first server and the second client terminal to use the updated sample data of the preset time period respectively, and update the preset account control policy prediction model based on the alliance chain.

[0222] It should be noted that the units, devices or modules described in the above embodiments can be implemented by computer chips or entities, or by products with certain functions. For the convenience of description, the above devices are described separately by functions divided into various modules. Of course, when implementing this specification, the functions of each module can be implemented in the same or more software and / or hardware, or the modules that implement the same function can be implemented by a combination of multiple sub-modules or sub-units. The device embodiments described above are only schematic. For example, the division of the units is only a logical function division. There may be other division methods in actual implementation, such as multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some interfaces, indirect coupling or communication connection of devices or units, which can be electrical, mechanical or other forms.

[0223] It can be seen from the above that the trading account management and control setting processing device provided in the embodiment of this specification can effectively simplify the workload of the first customer's staff and automatically and accurately complete the account management and control settings of the first trading account.

[0224] In a specific scenario example, the transaction account control setting processing method provided in this specification can be applied to implement account income and expenditure control account counterparty setting recommendation. The specific implementation process can be found in the following content.

[0225] 1. XX Bank (corresponding to the first server) publishes the demand for training the XGBoost model recommended by the counterparty of the account control account, and uploads the initialized model (e.g., the initial prediction model) with randomly selected parameters to the consortium chain;

[0226] 2. Each participant (including the first client terminal, the second client terminal, the first server, etc.) downloads the global model from the alliance chain, collects data locally and iterates the training of the local model, updates the model parameters (obtains the intermediate model), and encrypts it using Paillier before uploading it to the alliance chain.

[0227] 3. XX Bank aggregates all local model updates recorded on the alliance chain, uses the federation algorithm to implement global model updates, and submits the global model updates to the alliance chain. The federation algorithm formula is as follows:

[0228]

[0229] Where W t+1 , W t represents the parameters after the t+1th, tth global update, η is the update learning rate of the global model, n is the total amount of data of customers participating in this round of training, K is the number of customers participating in this round of training, is the local model of the kth participant selected in the t+1th iteration after training with local data, is the updated gradient of the local data training for the kth participant selected in the t+1th iteration.

[0230] In addition, in order to prevent malicious users from inferring whether a sample belongs to the user's local training set by accessing the global model during training or the output model after training, thereby endangering the client data privacy, protection is provided by adding noise that follows the Laplace distribution during aggregation. The Laplace distribution is as follows:

[0231]

[0232] Where μ and λ are location parameters and scale parameters, and h is the noise variable that obeys the Laplace distribution.

[0233] 4. Each participant downloads the updated global model from the alliance chain and updates the local model, and trains again. After the training is completed, it is encrypted using Paillier, and the local model update will be uploaded to the alliance chain. The formula for updating the local model is as follows:

[0234]

[0235] In the formula, w t+1 represents the parameters after the t+1th global update, represents the parameters updated by the Kth participant after the t+2th time, K is the number of customers participating in this round of training, η represents the learning rate, Represents the gradient function of the Kth participant.

[0236] Repeat steps 3 and 4 above until the federated learning objective function reaches the optimum or the number of iterations reaches the preset maximum number of iterations. At this point, the training is completed and XX Bank obtains a high-performance account control account counterparty recommended F-XGBoost model (e.g., a preset account control strategy prediction model). The formula of the objective function (e.g., the preset second objective function) is specifically as follows:

[0237]

[0238] In the formula, K is the number of customers participating in this round of training, and n represents the total amount of data participating in this round of training. is the predicted loss for the ith customer.

[0239] Through the above scenario examples, it is verified that the control setting and processing method of the transaction account provided in this specification can overcome the shortcomings of high subjectivity and incomplete generalization factors in the existing methods. By combining XGBoost technology with federated learning technology, under given hardware constraints, multiple participants (customers involving financial institutions, customers' related partners, and ICBC) are allowed to participate in model training to obtain a comprehensive, high-performance, and multi-party recognized F-XGBoost recommendation model, which avoids the risk of transaction failure and complaint caused by improper settings under the existing account counterparty setting method. In addition, considering that the central server of federated learning may become a single point of failure for the security and performance of the entire system, there are problems such as privacy leakage and model poisoning; and blockchain technology is a distributed, tamper-proof, traceable, decentralized distributed database that can be used to replace the central server of federated learning and provide a secure data exchange environment for federated learning. Therefore, the alliance chain is selected to replace the central server of federated learning to meet the needs of participants for review and relatively stable settings. Therefore, under the premise of protecting corporate data privacy and obtaining recognition from multiple parties, we can provide customers with reasonable account control counterparties (realizing account control settings for trading accounts) quickly, accurately, efficiently and comprehensively, effectively reducing the time and labor costs of subsequent frequent changes.

[0240] Although the present specification provides the method operation steps as described in the embodiment or flow chart, more or less operation steps may be included based on conventional or non-creative means. The order of steps listed in the embodiment is only one way of executing the order of many steps, and does not represent the only execution order. When the device or client terminal product in practice is executed, it can be executed in sequence or in parallel according to the method shown in the embodiment or the accompanying drawings (for example, a parallel processor or a multi-threaded processing environment, or even a distributed data processing environment). The term "include", "include" or any other variant thereof is intended to cover non-exclusive inclusion, so that the process, method, product or equipment including a series of elements includes not only those elements, but also includes other elements that are not explicitly listed, or also includes elements inherent to such process, method, product or equipment. In the absence of more restrictions, it is not excluded that there are other identical or equivalent elements in the process, method, product or equipment including the elements. The first, second, etc. words are used to represent the name, and do not represent any particular order.

[0241] Those skilled in the art also know that, in addition to implementing the controller in a purely computer-readable program code, the controller can be made to implement the same function in the form of logic gates, switches, application-specific integrated circuits, programmable logic controllers, and embedded microcontrollers by logically programming the method steps. Therefore, such a controller can be considered as a hardware component, and the devices for implementing various functions included therein can also be considered as structures within the hardware component. Or even, the devices for implementing various functions can be considered as both software modules for implementing the method and structures within the hardware component.

[0242] This specification may be described in the general context of computer-executable instructions executed by a computer, such as program modules. Generally, program modules include routines, programs, objects, components, data structures, classes, etc. that perform specific tasks or implement specific abstract data types. This specification may also be practiced in distributed computing environments where tasks are performed by remote processing devices connected through a communication network. In a distributed computing environment, program modules may be located in local and remote computer-readable storage media including storage devices.

[0243] Through the description of the above embodiments, it can be known that those skilled in the art can clearly understand that the present specification can be implemented by means of software plus a necessary general hardware platform. Based on such an understanding, the technical solution of the present specification can essentially be embodied in the form of a software product, which can be stored in a storage medium such as ROM / RAM, a magnetic disk, an optical disk, etc., and includes a number of instructions for enabling a computer device (which can be a personal computer, a mobile terminal, a client terminal, or a network device, etc.) to execute the methods described in each embodiment of the present specification or some parts of the embodiments.

[0244] The various embodiments in this specification are described in a progressive manner. The same or similar parts between the various embodiments can be referred to each other. Each embodiment focuses on the differences from other embodiments. This specification can be used in many general or special computer system environments or configurations. For example: personal computers, client terminal computers, handheld devices or portable devices, tablet devices, multi-processor systems, microprocessor-based systems, set-top boxes, programmable electronic devices, network PCs, minicomputers, mainframe computers, distributed computing environments including any of the above systems or devices, etc.

[0245] Although the present specification is described through embodiments, those skilled in the art will appreciate that there are many modifications and changes to the present specification without departing from the spirit of the present specification, and it is intended that the appended claims include these modifications and changes without departing from the spirit of the present specification.

Claims

1. A method for controlling and setting a transaction account, characterized in that: Applied to a first client terminal, the first client terminal is deployed at the first client side, the method includes: Acquire historical transaction data of a first transaction account and a target transaction account, and associated data related to the first transaction account and the target transaction account; wherein the first transaction account is a transaction account held by a first customer, and the target transaction account is a transaction account held by a target customer; The historical transaction data and associated data are processed using a preset account control strategy prediction model to obtain a corresponding target processing result; wherein the preset account control strategy prediction model is a model based on an extreme gradient boosting tree structure obtained by the first client terminal in advance in cooperation with the first server and the second client terminal based on the alliance chain through federated learning training according to preset protocol rules; the first server is deployed on the side of the financial service institution; the second client terminal is deployed on the side of an associated client that has a cooperative relationship with the first client; According to the target processing result, a target control setting strategy of the first trading account for the target trading account is determined; and according to the target control setting strategy, corresponding account control settings are performed on the first trading account.

2. The method according to claim 1, characterized in that The associated data includes at least one of the following: business transaction records between the first customer and the target customer, attribute parameters of the first transaction account, attribute parameters of the target transaction account, business transaction records between the target customer and associated customers, and associated files related to the first customer.

3. The method according to claim 1, characterized in that The target transaction account includes multiple target transaction accounts; Accordingly, according to the target control setting strategy, corresponding account control settings are performed on the first trading account, including at least one of the following: According to the target control setting strategy, the first trading account is set to support the transaction data interaction with the target trading account; According to the target control setting strategy, the first trading account is prohibited from interacting with the target trading account in terms of trading data; According to the target control setting strategy, the first trading account is set to support receiving transaction data initiated by the target trading account, but is prohibited from initiating transaction data to the target trading account; According to the target control setting strategy, the first trading account is set to support initiating transaction data to the target trading account, but is prohibited from receiving transaction data initiated by the target trading account.

4. The method according to claim 1, characterized in that: The historical transaction data and associated data are processed using a preset account control strategy prediction model to obtain corresponding target processing results, including: Extracting corresponding transaction features from the historical transaction data; extracting corresponding association features from the association data; According to a preset splicing rule, the transaction feature and the associated feature are spliced ​​to obtain a joint feature; The joint features are input into a preset account management and control strategy prediction model for processing to obtain the corresponding target processing results.

5. The method according to claim 1, characterized in that The method further comprises: According to the preset protocol rules, a model training request for a preset account control policy prediction model is initiated in the alliance chain; wherein the first server responds to the model training request, generates an initial prediction model based on an extreme gradient boosting tree structure; and uploads the initial prediction model to the alliance chain; Download the initial prediction model from the alliance chain; and train the initial prediction model locally using the sample data of the first customer to obtain a first intermediate model; wherein the second customer terminal downloads the initial prediction model; and trains the initial prediction model locally using the sample data of the associated customer to obtain a second intermediate model; The model parameters of the first intermediate model are uploaded to the alliance chain; wherein the second client terminal uploads the model parameters of the second intermediate model to the alliance chain; the first server uses the model parameters of the first intermediate model and the model parameters of the second intermediate model in combination to perform model aggregation to obtain a corresponding intermediate global model; Downloading the intermediate global model from the alliance chain; and detecting whether the intermediate global model meets the preset model requirements; When it is determined that the intermediate global model meets the preset model requirements, the intermediate global model is determined as the preset account control policy prediction model; and an end instruction for the end of the training of the preset account control policy prediction model is initiated in the alliance chain.

6. The method according to claim 5, characterized in that After detecting whether the intermediate global model meets the preset model requirements, the method further includes: When it is determined that the intermediate global model does not meet the preset model requirements, local training is performed locally using the sample training data of the first customer and the intermediate global model to obtain an updated first intermediate model; wherein the second customer terminal locally performs local training using the sample data of the associated customer and the intermediate global model to obtain an updated second intermediate model; The model parameters of the updated first intermediate model are uploaded to the alliance chain; wherein the second client terminal uploads the model parameters of the updated second intermediate model to the alliance chain; the first server obtains the corresponding updated intermediate global model through model aggregation based on the model parameters of the updated first intermediate model and the model parameters of the updated second intermediate model.

7. The method according to claim 5, characterized in that Uploading the model parameters of the first intermediate model to the alliance chain includes: Encrypting the model parameters of the first intermediate model using a preset encryption algorithm to obtain ciphertext data of the model parameters of the first intermediate model; Upload the ciphertext data of the model parameters of the first intermediate model to the alliance chain; Correspondingly, the first server uses a combination of model parameters of the first intermediate model and model parameters of the second intermediate model to perform model aggregation according to a preset encryption algorithm to obtain ciphertext data of the corresponding intermediate global model.

8. The method according to claim 5, characterized in that The model parameters of the first intermediate model uploaded to the alliance chain also carry the identity certificate of the first client terminal; wherein the identity certificate of the first client terminal is generated according to the first private key of the first client terminal; Correspondingly, before the first server combines the model parameters of the first intermediate model and the model parameters of the second intermediate model for model aggregation, the first server also authenticates the first client terminal based on the identity credentials of the first client terminal carried by the model parameters of the first intermediate model; and generates a first type of error prompt when it is determined that the identity authentication of the first client terminal has failed.

9. The method according to claim 5, characterized in that The first server uses the model parameters of the first intermediate model and the model parameters of the second intermediate model to perform model aggregation to obtain a corresponding intermediate global model, including: Perform model aggregation according to the following formula to obtain the corresponding intermediate global model: Wherein, W1 is the model parameter of the intermediate global model, W0 is the model parameter of the initial prediction model, η is the update learning rate of the global model, n is the total amount of sample data used in this round of joint training, K is the number of participants participating in this round of joint training of the preset account control strategy prediction model, k is the number of participants, The model parameters of the intermediate model uploaded by participant numbered k.

10. The method according to claim 5, characterized in that The first server uses the model parameters of the first intermediate model and the model parameters of the second intermediate model to perform model aggregation to obtain a corresponding intermediate global model, including: The first server obtains noise data that obeys the Laplace distribution; and combines the model parameters of the first intermediate model, the model parameters of the second intermediate model, and the noise data to perform model aggregation to obtain the corresponding intermediate global model.

11. The method according to claim 5, characterized in that Download the intermediate global model from the consortium chain, including: Generate a download request for the intermediate global model; wherein the download request carries at least the identity certificate of the first client terminal; Initiate the download request in the alliance chain; wherein the alliance chain also deploys a smart contract; the smart contract responds to the download request and verifies the authority of the first client terminal according to the identity certificate of the first client terminal; when it is determined that the authority verification passes, the smart contract sends the download certificate to the first client terminal; Download the intermediate global model from the consortium chain based on the download credentials.

12. The method according to claim 5, characterized in that After determining the intermediate global model as the preset account management and control strategy prediction model, the method further includes: At every preset time period, a model update request for the preset account control strategy prediction model is initiated in the consortium chain; In response to the model update request, according to the preset protocol rules, cooperate with the first server and the second client terminal respectively using the updated sample data of the preset time period to update the preset account management and control strategy prediction model based on the alliance chain.

13. A transaction account management and control setting processing device, characterized in that: Applied to a first client terminal, the first client terminal is deployed at the first client side, and includes: an acquisition module, configured to acquire historical transaction data of a first transaction account and a target transaction account, and associated data related to the first transaction account and the target transaction account; wherein the first transaction account is a transaction account held by a first customer, and the target transaction account is a transaction account held by a target customer; A prediction module, used to process the historical transaction data and associated data using a preset account control strategy prediction model to obtain a corresponding target processing result; wherein the preset account control strategy prediction model is a model based on an extreme gradient boosting tree structure obtained by the first client terminal in advance in cooperation with the first server and the second client terminal based on the alliance chain through federated learning training according to preset protocol rules; the first server is deployed on the side of the financial service institution; the second client terminal is deployed on the side of an associated client that has a cooperative relationship with the first client; The processing module is used to determine the target control setting strategy of the first trading account for the target trading account according to the target processing result; and perform corresponding account control settings on the first trading account according to the target control setting strategy.

14. A client terminal, characterized in that: The method comprises a processor and a memory for storing processor-executable instructions, wherein the processor implements the steps of the method according to any one of claims 1 to 12 when executing the instructions.

15. A computer-readable storage medium, characterized in that: Computer instructions are stored thereon, and when the instructions are executed by a processor, the steps of the method according to any one of claims 1 to 12 are implemented.

Images