Fund scheduling method and device for distributed account

Abstract

The invention provides a fund scheduling method and device for a distributed account, which can be used in the financial field or other fields, and the method comprises the steps: obtaining the transaction data of the distributed account; generating a queuing result according to the transaction data and a pre-trained fund scheduling model; and if the queuing result is that transaction queuing may occur, performing fund scheduling processing of a distributed account by using the fund scheduling model. According to the fund scheduling method and device for the distributed account provided by the invention, whether the transaction queuing possibly occurs in the next time window is predicted, and the account adjustment action is implemented in advance through the fund scheduling decision to avoid the transaction queuing, so that the occurrence frequency of the transaction queuing is effectively reduced, and the stability and the service throughput of a transaction system are improved.

Description

Technical Field

[0001] The present application relates to the field of financial technology, and in particular to a method and device for allocating funds for a distributed account. Background Art

[0002] The distributed account multi-active trading system is a system designed in recent years in the field of financial trading to improve the throughput and stability of the trading system. The system splits the participant accounts in the financial trading system into two sub-accounts of "one master and one slave", and then splits the daily initial transaction funds of the participant accounts into two sub-accounts according to a certain ratio. All sub-accounts can participate in the transaction, and the multi-active method of "one master and one slave" is used to achieve the goal of improving the throughput and stability of the trading system.

[0003] Due to the different initial balances and transaction volumes among sub-accounts, it is easy for a single sub-account to have insufficient balance but sufficient total balance among all sub-accounts during a trading day, resulting in the inability of the participant's sub-account transaction to proceed smoothly, thus causing transaction queuing in the distributed account trading system.

[0004] In the existing technology, the distributed account multi-active transaction system is prone to transaction queues. The main technologies currently used are divided into two categories:

[0005] (1) Queue rescue method: The queue rescue method mainly solves the transaction queue problem of the distributed account multi-active transaction system through post-rescue. When the transaction queue occurs, the account multi-active state is degraded to the account single-active state, and all transactions are sent to the main account until manual intervention, and then the account will be split into two sub-accounts to continue to participate in the transaction. The queue rescue method ensures the continuous progress of transactions, but because it is a post-rescue, only a single account is retained during the account degradation period, so the queue rescue cannot keep the system in the state of multi-active accounts.

[0006] (2) Transaction queue prediction method based on supervised learning: This method uses transaction detail data to obtain a series of statistical features and trains a supervised model to predict whether false queues will occur in the next time period. This method can predict problems in advance, but after predicting the problem, a single supervised model cannot avoid the problem. It is still necessary to manually intervene to adjust the account amount after predicting the problem. Therefore, the transaction queue prediction method based on supervised learning is difficult to solve the transaction queue problem with a single model.

[0007] The transaction queue problem of the distributed account multi-active trading system is the main factor affecting the stable operation of the trading system. It is very necessary to schedule funds between sub-accounts in advance. However, none of the methods proposed so far can meet this requirement at the same time. Summary of the invention

[0008] In response to the problems in the prior art, the embodiments of the present application provide a method and device for fund scheduling of a distributed account, which can at least partially solve the problems in the prior art.

[0009] In a first aspect, the present application provides a distributed account fund scheduling method, comprising:

[0010] Get transaction data of distributed accounts;

[0011] Generate a queuing result according to the transaction data and a pre-trained funds scheduling model;

[0012] If the queuing result indicates that a transaction queue may occur, the fund scheduling model is used to perform fund scheduling processing on the distributed account.

[0013] Furthermore, generating a queuing result according to the transaction data and a pre-trained funds scheduling model includes:

[0014] Acquire data features of the transaction data from the transaction data;

[0015] generating statistical features of the transaction data according to the data features of the transaction data;

[0016] The statistical characteristics of the transaction data are input into the fund scheduling model for analysis to obtain the queuing result.

[0017] Furthermore, the step of training the fund scheduling model includes:

[0018] Acquiring training data and preprocessing the training data;

[0019] The preprocessed training data is input into the initialized reinforcement learning model for training to obtain the fund scheduling model.

[0020] Furthermore, the preprocessing of the training data includes:

[0021] Acquire data features of the training data from the training data;

[0022] The statistical features of the training data are generated according to the data features of the training data.

[0023] Furthermore, the step of initializing the reinforcement learning model includes:

[0024] Setting an observation environment of the reinforcement learning model and an action space of the reinforcement learning model;

[0025] Construct a reward function for the reinforcement learning model.

[0026] Furthermore, the pre-processed training data is input into the initialized reinforcement learning model for training to obtain the fund scheduling model, including:

[0027] The following steps are iterated until the number of iterations reaches a preset value to obtain the fund scheduling model:

[0028] Iteratively update the training data according to the action space and the reward function and generate environment-action-reward data;

[0029] Updating the parameters of the Critic-NN network in the reinforcement learning model using a back-propagation method according to the environment-action-reward data;

[0030] Updating the parameters of the actor-new network in the reinforcement learning model using a back-propagation method according to the environment-action-reward data;

[0031] The parameters of the actor-old network in the reinforcement learning model are updated using the parameters of the actor-new network.

[0032] Furthermore, the use of the fund scheduling model to perform fund scheduling processing of distributed accounts includes:

[0033] Generate a fund scheduling strategy according to the queuing results and the statistical characteristics of the transaction data;

[0034] The distributed account adjustment action is implemented according to the fund scheduling strategy.

[0035] In a second aspect, the present application provides a fund scheduling device for a distributed account, comprising:

[0036] A transaction data acquisition unit, used to acquire transaction data of distributed accounts;

[0037] A queuing result generating unit, used to generate a queuing result according to the transaction data and a pre-trained funds scheduling model;

[0038] The fund scheduling processing unit is used to use the fund scheduling model to perform fund scheduling processing on the distributed account if the queuing result is that transaction queuing may occur.

[0039] Furthermore, the queuing result generating unit includes:

[0040] A first data feature acquisition module, used to acquire data features of transaction data from the transaction data;

[0041] A first statistical feature acquisition module, configured to generate statistical features of the transaction data according to data features of the transaction data;

[0042] The queuing result generating module is used to perform internal cross-validation analysis using the statistical characteristics of the transaction data to obtain the queuing result.

[0043] Furthermore, it also includes:

[0044] A data preprocessing unit, used for acquiring training data and preprocessing the training data;

[0045] The model training unit is used to input the preprocessed training data into the initialized reinforcement learning model for training to obtain the fund scheduling model.

[0046] Furthermore, the data preprocessing unit includes:

[0047] A second data feature acquisition module, used to acquire data features of training data from the training data;

[0048] The second statistical feature acquisition module is used to generate statistical features of the training data according to the data features of the training data.

[0049] Furthermore, it also includes:

[0050] An observation environment and action space setting unit, used to set the observation environment of the reinforcement learning model and the action space of the reinforcement learning model;

[0051] A reward function construction unit is used to construct a reward function of the reinforcement learning model.

[0052] Furthermore, the model training unit includes:

[0053] Iterative operation module, used to perform iterative operations during the training of the fund scheduling model:

[0054] A training data updating module, used to iteratively update the training data according to the action space and the reward function and generate environment-action-reward data and generate environment-action-reward data after each round of iteration;

[0055] A Critic-NN network update module, used to iteratively update the parameters of the Critic-NN network in the reinforcement learning model using a back propagation method according to the training data after the last update;

[0056] An actor-new network update module is used to iteratively update the parameters of the actor-new network in the reinforcement learning model using a back propagation method according to the training data and the set of the training data after each round of update;

[0057] The actor-old network update module is used to update the parameters of the actor-old network in the reinforcement learning model using the parameters of the actor-new network.

[0058] Furthermore, the fund scheduling processing unit includes:

[0059] A fund scheduling strategy generating module, used for generating a fund scheduling strategy according to the queuing result and the statistical characteristics of the transaction data;

[0060] The account adjustment action implementation module is used to implement the distributed account adjustment action according to the fund scheduling strategy.

[0061] In a third aspect, the present application provides a computer device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the method described in any one of the above embodiments when executing the computer program.

[0062] In a fourth aspect, the present application provides a computer-readable storage medium, wherein the computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the method described in any of the above embodiments is implemented.

[0063] In a fifth aspect, the present application provides a computer program product, wherein the computer program product includes a computer program, and when the computer program is executed by a processor, the method described in any of the above embodiments is implemented.

[0064] The distributed account fund scheduling method and device provided by the present application obtains the transaction data of the distributed account; generates a queuing result according to the transaction data and the pre-trained fund scheduling model; if the queuing result is that a transaction queue may occur, the fund scheduling model is used to perform fund scheduling processing on the distributed account, so as to predict whether a transaction queue may occur in the next time window and implement the account adjustment action in advance through the fund scheduling decision to avoid the transaction queue, effectively reducing the number of transaction queues and improving the stability and business throughput of the transaction system. Among them, the generation of the fund scheduling model is realized by training the reinforcement learning model to implement reasonable account adjustment actions, which effectively improves the transaction efficiency of the distributed account. BRIEF DESCRIPTION OF THE DRAWINGS

[0065] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0066] Figure 1 It is a flow chart of a method for distributing funds of a distributed account provided in an embodiment of the present application;

[0067] Figure 2 It is a flow chart of a method for distributing funds of a distributed account provided in an embodiment of the present application;

[0068] Figure 3 It is a flow chart of a method for distributing funds of a distributed account provided in an embodiment of the present application;

[0069] Figure 4 It is a flow chart of a method for distributing funds of a distributed account provided in an embodiment of the present application;

[0070] Figure 5 It is a flow chart of a method for distributing funds of a distributed account provided in an embodiment of the present application;

[0071] Figure 6 It is a flow chart of a method for distributing funds of a distributed account provided in an embodiment of the present application;

[0072] Figure 7 It is a flow chart of a method for distributing funds of a distributed account provided in an embodiment of the present application;

[0073] Figure 8 It is a structural diagram of a fund scheduling device for a distributed account provided in an embodiment of the present application;

[0074] Fig. 9 It is a structural diagram of a fund scheduling device for a distributed account provided in an embodiment of the present application;

[0075] Fig.10 It is a structural diagram of a fund scheduling device for a distributed account provided in an embodiment of the present application;

[0076] Fig.11 It is a structural diagram of a fund scheduling device for a distributed account provided in an embodiment of the present application;

[0077] Fig.12 It is a structural diagram of a fund scheduling device for a distributed account provided in an embodiment of the present application;

[0078] Fig.13 It is a structural diagram of a fund scheduling device for a distributed account provided in an embodiment of the present application;

[0079] Fig.14 It is a structural diagram of a fund scheduling device for a distributed account provided in an embodiment of the present application;

[0080] Fig.15 is a schematic diagram of the physical structure of an electronic device provided in one embodiment of the present application;

[0081] Fig.16 It is a schematic diagram of a fund scheduling scenario of a distributed account provided by an embodiment of the present application;

[0082] Fig.17 It is a flowchart of training a fund scheduling model provided in an embodiment of the present application. DETAILED DESCRIPTION

[0083] In order to make the purpose, technical scheme and advantages of the embodiments of the present application clearer, the embodiments of the present application are further described in detail below in conjunction with the accompanying drawings. Here, the illustrative embodiments of the present application and their descriptions are used to explain the present application, but are not intended to limit the present application. It should be noted that, in the absence of conflict, the embodiments in the present application and the features in the embodiments can be combined arbitrarily with each other.

[0084] It should be noted that the distributed account fund scheduling method and device proposed in this application can be used in the field of financial technology, and can also be used in any field outside the field of financial technology. The application field of the distributed account fund scheduling method and device proposed in this application is not limited.

[0085] The following examples illustrate the scenarios of the embodiments of the present application, but the present application is not limited thereto.

[0086] Fig.16 is a schematic diagram of a fund scheduling scenario of a distributed account provided by an embodiment of the present application, such as Fig.16 As shown, the fund scheduling scenario of the distributed account may include a server 1601 and a trading system 1602. For simplicity, Fig.16 Only one server 1601 and one transaction system 1602 are taken as an example for description, but the present application is not limited thereto.

[0087] In one embodiment, information exchange may be performed between the server 1601 and the trading system 1602 , for example, the server 1601 obtains trading data from the trading system 1602 , and the server 1601 feeds back a scheduling strategy to the trading system 1602 , but the present application is not limited thereto.

[0088] The following uses the server as an example of the execution subject to illustrate the specific implementation process of the distributed account fund scheduling method provided in the embodiment of the present application.

[0089] Figure 1 : is a flow chart of a distributed account fund scheduling method provided by an embodiment of the present application, such as Figure 1 As shown, the distributed account fund scheduling method provided by this application includes:

[0090] S101: Obtain transaction data of distributed accounts;

[0091] S102: generating a queuing result according to the transaction data and a pre-trained funds scheduling model;

[0092] S103: If the queuing result indicates that a transaction queue may occur, the fund scheduling model is used to perform fund scheduling processing for the distributed account.

[0093] from Figure 1 As can be seen from the illustrated process, the distributed account fund scheduling method provided in the present application obtains the transaction data of the distributed account; generates a queuing result based on the transaction data and a pre-trained fund scheduling model; if the queuing result is that a transaction queue may occur, the fund scheduling model is used to perform fund scheduling processing on the distributed account, thereby predicting whether a transaction queue may occur in the next time window and avoiding transaction queues by implementing account adjustment actions in advance through fund scheduling decisions, effectively reducing the number of transaction queues and improving the stability of the transaction system and business throughput.

[0094] It should be noted that the acquisition, storage, use, and processing of data in the technical solution of this application are in compliance with the relevant provisions of national laws and regulations.

[0095] Each step is explained in detail below.

[0096] S101: Obtain transaction data of distributed accounts;

[0097] Specifically, the transaction data of the distributed account includes transaction details and balance data of the participant account, and the balance data of the participant account will change accordingly with the occurrence of transactions. Both the transaction details and the balance data of the participant account contain multiple data features, and data features can be extracted from the transaction details and the balance data of the participant account, and the distributed account can be analyzed and scheduled based on the data features.

[0098] S102: generating a queuing result according to the transaction data and a pre-trained funds scheduling model;

[0099] Specifically, the funds scheduling model is obtained by training the reinforcement learning model, and the reinforcement learning model needs to be initialized before training. The server inputs the transaction data into the pre-trained funds scheduling model, processes the transaction data and uses the processed data for analysis to determine whether a transaction queue will occur in the next time window and obtain the queue result.

[0100] Figure 2 : is a flow chart of a distributed account fund scheduling method provided by an embodiment of the present application, such as Figure 2 As shown, S102 includes:

[0101] S201: Acquire data features of transaction data from the transaction data;

[0102] Specifically, the transaction data of the distributed account includes transaction details and balance data of the participant accounts. The data features obtained from the transaction details include: debit participants, credit participants, transaction time, transaction amount and transaction instance (transaction sub-account); the data features obtained from the balance data of the participant accounts include: the total balance of each participant, the balance of the main sub-account and the balance of the subordinate sub-account.

[0103] S202: generating statistical features of the transaction data according to the data features of the transaction data;

[0104] Specifically, let the time window size be k minutes. The funds scheduling model reads the transaction details data and the balance data of the participant accounts in the current time window every k minutes. For each participant who has split the active-active account, multiple statistical features are generated for data analysis. The statistical features of the transaction data include: the maximum value of debit transactions, the average of the top 5 debit transactions, the sum of debit transaction amounts, the number of debit transactions, the maximum value of credit transactions, the sum of credit transaction amounts, the number of credit transactions and the current transaction time.

[0105] S203: Inputting the statistical features of the transaction data into the fund scheduling model for analysis to obtain a queuing result.

[0106] Specifically, after generating statistical features based on transaction data, the server automatically analyzes the features using a funds scheduling model to obtain the queuing results for the next time window. If the model determines that no transaction queuing will occur in the next time window, it will output a scheduling strategy that does not schedule funds.

[0107] Figure 3 : is a flow chart of a distributed account fund scheduling method provided by an embodiment of the present application, such as Figure 3 As shown, the steps of training the fund scheduling model include:

[0108] S301: Acquire training data and preprocess the training data;

[0109] Specifically, before using the training data to train the reinforcement learning model, the training data needs to be preprocessed accordingly to obtain the statistical characteristics of the training data and perform model training based on the statistical characteristics.

[0110] Figure 4 : is a flow chart of a distributed account fund scheduling method provided by an embodiment of the present application, such as Figure 4 As shown, S301 includes:

[0111] S401: Acquire data features of training data from the training data;

[0112] Specifically, the training data includes transaction details training data and participant account balance training data. The data features of the training data obtained from the transaction details training data include: debit participants, credit participants, transaction time, transaction amount and transaction instance (transaction sub-account); the data features of the training data obtained from the participant account balance training data include: the total balance of each participant, the balance of the main sub-account and the balance of the subordinate sub-account.

[0113] S402: Generate statistical features of the training data according to the data features of the training data.

[0114] Specifically, let the time window size be k minutes, and the reinforcement learning model read the transaction details training data and the balance training data of the participant accounts in the current time window every k minutes. For each participant who has split the active-active account, multiple statistical features are generated and input into the reinforcement learning model for training. The statistical features of the training data include: the maximum value of debit transactions, the average of the top 5 debit transactions, the sum of debit transaction amounts, the number of debit transactions, the maximum value of credit transactions, the sum of credit transaction amounts, the number of credit transactions and the current transaction time.

[0115] S302: Input the preprocessed training data into the initialized reinforcement learning model for training to obtain the fund scheduling model.

[0116] Specifically, Fig.17 The figure shows the process of training a fund scheduling model. After the server generates statistical features from the training data and initializes the reinforcement learning model, the statistical features are input into the initialized reinforcement learning model for training to obtain a fund scheduling model to realize fund scheduling of distributed accounts.

[0117] Figure 5 : is a flow chart of a distributed account fund scheduling method provided by an embodiment of the present application, such as Figure 5 As shown, the steps of initializing the reinforcement learning model include:

[0118] S501: Setting the observation environment of the reinforcement learning model and the action space of the reinforcement learning model;

[0119] Specifically, the observation environment is set as the statistical characteristics of the training data every k minutes, and the action space is the transfer ratio between accounts. Let the main sub-account be numbered 0, the subordinate sub-account be numbered 1, and the action space be a, the value range of a is (-100%, +100%), -α% means that the main sub-account 0 transfers α% of its own balance to the subordinate sub-account 1, and +α% means that the subordinate sub-account 1 transfers α% of its own balance to the main sub-account 0.

[0120] S502: Constructing a reward function of the reinforcement learning model.

[0121] Specifically, when the corresponding action is formed based on the statistical features to dispatch funds to the participant's sub-account, the reinforcement learning model will evaluate the action, and the expression of the reward function is as follows:

[0122]

[0123] Among them, a represents the percentage of funds allocated this time.

[0124] Figure 6 : is a flow chart of a distributed account fund scheduling method provided by an embodiment of the present application, such as Figure 6 As shown, S302 includes:

[0125] S601: Iteratively update the training data according to the action space and the reward function and generate environment-action-reward data;

[0126] Specifically, the statistical feature s (the transaction details data of the current time window and the balance data of the participant's account) is input into the actor-new network of the reinforcement learning model to obtain the normal distribution Normal, and the action a for the statistical feature s of the current time window is sampled according to the normal distribution Normal, and then the reward r and the statistical feature s_ of the next time window are calculated according to the action a, and the environment-action-reward data (s, a, r) ​​is generated. Among them, the reward r is calculated by the reward function, and the statistical feature s_ of the next time window is the transaction details data of the next time window and the balance data of the participant's account. The environment-action-reward data (s, a, r) ​​of the current time window and the statistical feature s_ of the next time window are input into the actor-new network of the reinforcement learning model, and the above steps are iterated until the preset threshold is reached to generate multiple sets of environment-action-reward data.

[0127] S602: updating the parameters of the Critic-NN network in the reinforcement learning model using a back propagation method according to the environment-action-reward data;

[0128] Specifically, the following steps are iterated until the preset threshold is reached: the statistical feature s_ obtained in the last round of iteration is input into the Critic-NN network, which is used to evaluate the stability, smoothness and rationality of the action, and the evaluation v_ corresponding to the statistical feature s_ is obtained. Then, the weighted sum R[t] of the reward in the t-th time window and the vector R of the weighted sum of the rewards in each time window are calculated:

[0129] R[t]=r[t]+γ1*r[t+1]+γ2*r[t+2]+…+γ T-t+1 r[T-1]+γ T-t *v_ (2)

[0130] R=[R[0],…,R[T]] (3)

[0131] Among them, T represents the last time window, γ1~γ T-t represents the weight of each reward, and v_ represents the evaluation corresponding to the statistical feature s_.

[0132] Next, the statistical features s extracted from each set of environment-action-reward data (s, a, r) ​​are combined and input into the Critic-NN network to obtain the evaluation v_ of each time window and calculate the advantage A of taking action a in state s t :

[0133] A t =R-v_ (4)

[0134] A t =mean(square(A t )) (5)

[0135] Among them, R represents the vector of the weighted sum of the rewards in each time window, and v_ represents the evaluation corresponding to the statistical feature s_.

[0136] Finally, the back propagation method (BP) is used to update the Critic-NN network.

[0137] S603: updating the parameters of the actor-new network in the reinforcement learning model using a back propagation method according to the environment-action-reward data;

[0138] Specifically, the following steps are iterated until a preset threshold is reached: the statistical feature s extracted from each set of environment-action-reward data (s, a, r) ​​is combined and input into the actor-old network and the actor-new network of the reinforcement learning model, wherein the actor-old network and the actor-new network have the same structure but different parameters, and normal distributions Normal1 and Normal2 about action a are obtained respectively, and the probability prob1 of each action a of the actor-old network and the probability prob2 of each action a of the actor-new network are obtained by sampling according to the normal distribution, and the weight ratio is calculated according to the probability prob1 of each action a of the actor-old network and the probability prob2 of each action a of the actor-new network, and then the weight ratio is calculated according to the weight ratio and the advantage A. t Calculate the loss aloss:

[0139] aloss=mean(min((ratio*A t ),clip(ratio,1-ξ,1+ξ)*A t ))) (6)

[0140] Among them, A t represents the advantage of taking action a in state s, ratio represents the weight calculated according to the probability prob1 of the actor-old network for each action a and the probability prob2 of the actor-new network for each action a, and ξ represents a hyperparameter.

[0141] Finally, the back propagation method (BP) is used to update the actor-new network.

[0142] S604: Using the parameters of the actor-new network, update the parameters of the actor-old network in the reinforcement learning model.

[0143] Specifically, after the critic-NN network and the actor-new network have been updated for a certain number of iterations, the parameters of the actor-new network are used to update the parameters of the actor-old network in the reinforcement learning model.

[0144] S605: Determine whether the number of iterations reaches a preset value.

[0145] Specifically, S601 to S604 are executed in a loop to determine whether the number of executions reaches a preset value. If the number of executions does not reach the preset value, S601 to S604 are continued to be executed; if the number of executions reaches the preset value, the loop is exited to allow the reinforcement learning model to converge and obtain a fund scheduling model.

[0146] S103: If the queuing result indicates that a transaction queue may occur, the fund scheduling model is used to perform fund scheduling processing for the distributed account.

[0147] Specifically, if the queuing result is that transaction queuing may occur, the fund scheduling model is used to perform fund scheduling processing for the distributed account; if the queuing result is that transaction queuing will not occur, no fund scheduling processing is performed.

[0148] Figure 7 : is a flow chart of a distributed account fund scheduling method provided by an embodiment of the present application, such as Figure 7 As shown, S103 includes:

[0149] S701: Generate a fund scheduling strategy according to the queuing result and the statistical characteristics of the transaction data;

[0150] Specifically, when the queuing result is that transaction queuing may occur, the server generates a corresponding fund scheduling strategy using the fund scheduling model according to the statistical characteristics of the transaction data; when the queuing result is that transaction queuing will not occur, the server does not perform fund scheduling. For example, when the fund scheduling strategy (action) a output by the model is not 0, it means that the server determines that transaction queuing will occur, and when the fund scheduling strategy a is 0, it means that the server determines that transaction queuing will not occur.

[0151] S702: Implementing distributed account adjustment actions according to the fund scheduling strategy.

[0152] Specifically, the distributed account adjustment action is implemented according to the fund scheduling strategy, and the funds of the sub-accounts with sufficient balances in the participant accounts are scheduled in advance to the sub-accounts that may have insufficient balances, thereby avoiding the occurrence of false queuing problems. For example, when the fund scheduling strategy a output by the model is 0.1, it means that the server will schedule 10% of the funds in account 0 to account 1. When the fund scheduling strategy a is 0, it means that the server will not make any adjustments to the account.

[0153] The distributed account fund scheduling method provided by the present application obtains the transaction data of the distributed account; generates a queuing result according to the transaction data and the pre-trained fund scheduling model; if the queuing result is that a transaction queue may occur, the fund scheduling model is used to perform fund scheduling processing on the distributed account, so as to predict whether a transaction queue may occur in the next time window and implement the account adjustment action in advance through the fund scheduling decision to avoid the transaction queue, effectively reducing the number of transaction queues and improving the stability and business throughput of the transaction system. Among them, the generation of the fund scheduling model is realized by training the reinforcement learning model to implement reasonable account adjustment actions, which effectively improves the transaction efficiency of the distributed account.

[0154] Based on the same inventive concept, the embodiments of the present application also provide a fund scheduling device for a distributed account, which can be used to implement the method described in the above embodiments, as described in the following embodiments. Since the principle of solving the problem by the fund scheduling device for a distributed account is similar to that of the fund scheduling method for a distributed account, the implementation of the fund scheduling device for a distributed account can refer to the implementation of the method based on software performance benchmark determination, and the repeated parts will not be repeated. As used below, the term "unit" or "module" can be a combination of software and / or hardware that implements a predetermined function. Although the system described in the following embodiments is preferably implemented in software, the implementation of hardware, or a combination of software and hardware, is also possible and conceived.

[0155] Figure 8 : is a schematic diagram of the structure of a fund scheduling device for a distributed account provided in an embodiment of the present application, such as Figure 8 As shown, the device comprises:

[0156] The transaction data acquisition unit 801 is used to acquire the transaction data of the distributed account;

[0157] A queuing result generating unit 802, configured to generate a queuing result according to the transaction data and a pre-trained funds scheduling model;

[0158] The fund scheduling processing unit 803 is used to perform fund scheduling processing of the distributed account using the fund scheduling model if the queuing result is that transaction queuing may occur.

[0159] Fig. 9 : is a schematic diagram of the structure of a fund scheduling device for a distributed account provided by an embodiment of the present application. Figure 8 Based on the embodiment, further, Fig. 9 As shown, the distributed account fund scheduling device provided by the present application also includes:

[0160] A first data feature acquisition module 901, configured to acquire data features of transaction data from the transaction data;

[0161] A first statistical feature acquisition module 902, configured to generate statistical features of the transaction data according to the data features of the transaction data;

[0162] The queuing result generating module 903 is used to perform internal cross-validation analysis using the statistical features of the transaction data to obtain a queuing result.

[0163] Fig.10 : is a schematic diagram of the structure of a fund scheduling device for a distributed account provided by an embodiment of the present application. Figure 8 Based on the embodiment, further, Fig.10 As shown, the distributed account fund scheduling device provided by the present application also includes:

[0164] The data preprocessing unit 1001 is used to obtain training data and preprocess the training data;

[0165] The model training unit 1002 is used to input the pre-processed training data into the initialized reinforcement learning model for training to obtain the fund scheduling model.

[0166] Fig.11 : is a schematic diagram of the structure of a fund scheduling device for a distributed account provided by an embodiment of the present application. Fig.10 Based on the embodiment, further, Fig.11 As shown, the distributed account fund scheduling device provided by the present application also includes:

[0167] A second data feature acquisition module 1101 is used to acquire data features of training data from the training data;

[0168] The second statistical feature acquisition module 1102 is used to generate statistical features of the training data according to the data features of the training data.

[0169] Fig.12 : is a schematic diagram of the structure of a fund scheduling device for a distributed account provided by an embodiment of the present application. Fig.10 Based on the embodiment, further, Fig.12 As shown, the distributed account fund scheduling device provided by the present application also includes:

[0170] An observation environment and action space setting unit 1201, used to set the observation environment of the reinforcement learning model and the action space of the reinforcement learning model;

[0171] The reward function construction unit 1202 is used to construct the reward function of the reinforcement learning model.

[0172] Fig.13 : is a schematic diagram of the structure of a fund scheduling device for a distributed account provided by an embodiment of the present application. Fig.12 Based on the embodiment, further, Fig.13 As shown, the distributed account fund scheduling device provided by the present application also includes:

[0173] The iterative operation module 1301 is used to perform iterative operations during the training of the fund scheduling model:

[0174] A training data updating module 1302, configured to iteratively update the training data according to the action space and the reward function and generate environment-action-reward data and generate environment-action-reward data after each round of iteration;

[0175] The Critic-NN network updating module 1303 is used to iteratively update the parameters of the Critic-NN network in the reinforcement learning model using a back propagation method according to the training data after the last update;

[0176] An actor-new network updating module 1304 is used to iteratively update the parameters of the actor-new network in the reinforcement learning model using a back propagation method according to the training data and the set of the training data after each round of update;

[0177] The actor-old network update module 1305 is used to update the parameters of the actor-old network in the reinforcement learning model using the parameters of the actor-new network.

[0178] Fig.14 : is a schematic diagram of the structure of a fund scheduling device for a distributed account provided by an embodiment of the present application. Fig. 9 Based on the embodiment, further, Fig.14 As shown, the distributed account fund scheduling device provided by the present application also includes:

[0179] A fund scheduling strategy generating module 1401 is used to generate a fund scheduling strategy according to the queuing result and the statistical characteristics of the transaction data;

[0180] The account adjustment action implementation module 1402 is used to implement the distributed account adjustment action according to the fund scheduling strategy.

[0181] The distributed account fund scheduling method and device provided by the present application obtains the transaction data of the distributed account; generates a queuing result according to the transaction data and the pre-trained fund scheduling model; if the queuing result is that a transaction queue may occur, the fund scheduling model is used to perform fund scheduling processing on the distributed account, so as to predict whether a transaction queue may occur in the next time window and implement the account adjustment action in advance through the fund scheduling decision to avoid the transaction queue, effectively reducing the number of transaction queues and improving the stability and business throughput of the transaction system. Among them, the generation of the fund scheduling model is realized by training the reinforcement learning model to implement reasonable account adjustment actions, which effectively improves the transaction efficiency of the distributed account.

[0182] Fig.15 is a schematic diagram of the physical structure of an electronic device provided in an embodiment of the present application, such as Fig.15 As shown, the electronic device may include: a processor 1501, a communication interface 1502, a memory 1503 and a communication bus 1504, wherein the processor 1501, the communication interface 1502 and the memory 1503 communicate with each other through the communication bus 1504. The processor 1501 may call the logic instructions in the memory 1503 to execute the following method: obtaining the transaction data of the distributed account; generating a queuing result according to the transaction data and the pre-trained fund scheduling model; if the queuing result indicates that a transaction queue may occur, using the fund scheduling model to perform fund scheduling processing of the distributed account.

[0183] In addition, the logic instructions in the above-mentioned memory 1503 can be implemented in the form of software functional units and can be stored in a computer-readable storage medium when sold or used as an independent product. Based on such an understanding, the technical solution of the present application can be essentially or partly embodied in the form of a software product that contributes to the prior art, and the computer software product is stored in a storage medium, including several instructions to enable a computer device (which can be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the method described in each embodiment of the present application. The aforementioned storage medium includes: various media that can store program codes, such as a USB flash drive, a mobile hard disk, a read-only memory (ROM), a random access memory (RAM), a disk or an optical disk.

[0184] This embodiment discloses a computer program product, which includes a computer program stored on a computer-readable storage medium, and the computer program includes program instructions. When the program instructions are executed by a computer, the computer can execute the methods provided by the above-mentioned method embodiments, for example, including: obtaining transaction data of a distributed account; generating a queuing result based on the transaction data and a pre-trained funds scheduling model; if the queuing result is that a transaction queue may occur, using the funds scheduling model to perform funds scheduling processing on the distributed account.

[0185] The present embodiment provides a computer-readable storage medium, which stores a computer program. The computer program enables the computer to execute the methods provided by the above-mentioned method embodiments, for example, including: obtaining transaction data of distributed accounts; generating queuing results based on the transaction data and a pre-trained funds scheduling model; if the queuing result is that transaction queuing may occur, using the funds scheduling model to perform funds scheduling processing on distributed accounts.

[0186] Those skilled in the art will appreciate that the embodiments of the present application may be provided as methods, systems, or computer program products. Therefore, the present application may adopt the form of a complete hardware embodiment, a complete software embodiment, or an embodiment in combination with software and hardware. Moreover, the present application may adopt the form of a computer program product implemented in one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) that include computer-usable program code.

[0187] The present application is described with reference to the flowcharts and / or block diagrams of the methods, devices (systems), and computer program products according to the embodiments of the present application. It should be understood that each process and / or box in the flowchart and / or block diagram, as well as the combination of the processes and / or boxes in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to generate a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0188] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0189] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process. Figure 1 A process or multiple processes and / or boxes Figure 1 A step that specifies a function in one or more boxes.

[0190] In the description of this specification, the description with reference to the terms "one embodiment", "a specific embodiment", "some embodiments", "for example", "example", "specific example", or "some examples" means that the specific features, structures, materials or characteristics described in conjunction with the embodiment or example are included in at least one embodiment or example of the present application. In this specification, the schematic representation of the above terms does not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials or characteristics described may be combined in any one or more embodiments or examples in a suitable manner.

[0191] The specific embodiments described above further illustrate the purpose, technical solutions and beneficial effects of the present application. It should be understood that the above description is only a specific embodiment of the present application and is not intended to limit the scope of protection of the present application. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present application should be included in the scope of protection of the present application.

Claims

1. A method for distributing funds of a distributed account, characterized in that: include: Get transaction data of distributed accounts; Generate a queuing result according to the transaction data and a pre-trained funds scheduling model; If the queuing result indicates that a transaction queue may occur, the fund scheduling model is used to perform fund scheduling processing on the distributed account.

2. The distributed account fund scheduling method according to claim 1, characterized in that: The generating of the queuing result according to the transaction data and the pre-trained funds scheduling model includes: Acquire data features of the transaction data from the transaction data; generating statistical features of the transaction data according to the data features of the transaction data; The statistical characteristics of the transaction data are input into the fund scheduling model for analysis to obtain the queuing result.

3. The distributed account fund scheduling method according to claim 1, characterized in that: The steps of training the fund scheduling model include: Acquiring training data and preprocessing the training data; The preprocessed training data is input into the initialized reinforcement learning model for training to obtain the fund scheduling model.

4. The distributed account fund scheduling method according to claim 3, characterized in that: The preprocessing of the training data comprises: Acquire data features of the training data from the training data; The statistical features of the training data are generated according to the data features of the training data.

5. The distributed account fund scheduling method according to claim 3, characterized in that: The step of initializing the reinforcement learning model comprises: Setting an observation environment of the reinforcement learning model and an action space of the reinforcement learning model; Construct a reward function for the reinforcement learning model.

6. The distributed account fund scheduling method according to claim 5, characterized in that: The pre-processed training data is input into the initialized reinforcement learning model for training to obtain the fund scheduling model, including: The following steps are iterated until the number of iterations reaches a preset value to obtain the fund scheduling model: Iteratively update the training data according to the action space and the reward function and generate environment-action-reward data; Updating the parameters of the Critic-NN network in the reinforcement learning model using a back-propagation method according to the environment-action-reward data; Updating the parameters of the actor-new network in the reinforcement learning model using a back-propagation method according to the environment-action-reward data; The parameters of the actor-old network in the reinforcement learning model are updated using the parameters of the actor-new network.

7. The distributed account fund scheduling method according to claim 2, characterized in that: The method of using the fund scheduling model to perform fund scheduling processing of distributed accounts includes: Generate a fund scheduling strategy according to the queuing results and the statistical characteristics of the transaction data; The distributed account adjustment action is implemented according to the fund scheduling strategy.

8. A fund scheduling device for a distributed account, characterized in that: include: A transaction data acquisition unit, used to acquire transaction data of distributed accounts; A queuing result generating unit, used to generate a queuing result according to the transaction data and a pre-trained funds scheduling model; The fund scheduling processing unit is used to use the fund scheduling model to perform fund scheduling processing on the distributed account if the queuing result is that transaction queuing may occur.

9. The distributed account fund scheduling device according to claim 8, characterized in that: The queuing result generating unit comprises: A first data feature acquisition module, used to acquire data features of transaction data from the transaction data; A first statistical feature acquisition module, configured to generate statistical features of the transaction data according to data features of the transaction data; The queuing result generating module is used to perform internal cross-validation analysis using the statistical characteristics of the transaction data to obtain the queuing result.

10. The distributed account fund scheduling device according to claim 8, characterized in that: Also includes: A data preprocessing unit, used for acquiring training data and preprocessing the training data; The model training unit is used to input the preprocessed training data into the initialized reinforcement learning model for training to obtain the fund scheduling model.

11. The distributed account fund scheduling device according to claim 10, characterized in that: The data preprocessing unit comprises: A second data feature acquisition module, used to acquire data features of training data from the training data; The second statistical feature acquisition module is used to generate statistical features of the training data according to the data features of the training data.

12. The distributed account fund scheduling device according to claim 10, characterized in that: Also includes: An observation environment and action space setting unit, used to set the observation environment of the reinforcement learning model and the action space of the reinforcement learning model; A reward function construction unit is used to construct a reward function of the reinforcement learning model.

13. The distributed account fund scheduling device according to claim 12, characterized in that: The model training unit comprises: Iterative operation module, used to perform iterative operations during the training of the fund scheduling model: A training data updating module, used to iteratively update the training data according to the action space and the reward function and generate environment-action-reward data and generate environment-action-reward data after each round of iteration; A Critic-NN network update module, used to iteratively update the parameters of the Critic-NN network in the reinforcement learning model using a back propagation method according to the training data after the last update; An actor-new network update module is used to iteratively update the parameters of the actor-new network in the reinforcement learning model using a back propagation method according to the training data and the set of the training data after each round of update; The actor-old network update module is used to update the parameters of the actor-old network in the reinforcement learning model using the parameters of the actor-new network.

14. The distributed account fund scheduling device according to claim 9, characterized in that: The fund scheduling processing unit includes: A fund scheduling strategy generating module, used for generating a fund scheduling strategy according to the queuing result and the statistical characteristics of the transaction data; The account adjustment action implementation module is used to implement the distributed account adjustment action according to the fund scheduling strategy.

15. A computer device comprising a memory, a processor and a computer program stored in the memory and executable on the processor, characterized in that: When the processor executes the computer program, the method according to any one of claims 1 to 7 is implemented.

16. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the method according to any one of claims 1 to 7 is implemented.

17. A computer program product, characterized in that The computer program product comprises a computer program, and when the computer program is executed by a processor, the method according to any one of claims 1 to 7 is implemented.

Images