Transaction data processing method and device

Abstract

The invention discloses a transaction data processing method and device, and relates to the technical field of transactions, and the method comprises the steps: receiving a plurality of transactions to be approved; the approval passing probability of each to-be-approved transaction is calculated; calculating a predicted accumulated value of each transaction opponent; determining the generated value of each transaction opponent according to the real generated library; receiving a first approval result of the first to-be-approved transaction; modifying the approval passing probability of the first to-be-approved transaction according to the first approval result, and updating the predicted accumulated value of the first transaction opponent; wherein when the first approval result is passed, deleting the to-be-approved transactions with the priority lower than that of the first to-be-approved transactions from the to-be-approved transactions of the first trade opponent, and updating the predicted accumulated value of the first trade opponent according to all the to-be-approved transactions of the deleted first trade opponent; and determining whether the first to-be-approved transaction triggers a major transaction or not according to the generated value and the updated predicted accumulated value. According to the invention, the recognition accuracy and efficiency of major transactions can be improved.

Description

Technical Field

[0001] The present invention relates to the field of transaction technology, and in particular to a transaction data processing method and device. Background Art

[0002] This section is intended to provide a background or context to the embodiments of the invention recited in the claims. No admission is made that the description herein is prior art by inclusion in this section.

[0003] There are many types of transactions that are closely related to enterprises, such as regular business transactions, supply chain transactions, financial services transactions, etc. The party that interacts with the other party in the transaction is called the counterparty. Among them, relatively significant transactions are those that have a significant impact on the company's financial status, operating results or business decisions. The management and disclosure of significant transactions is an important part of corporate transaction management.

[0004] In the prior art, the judgment of major transactions is generally made by monitoring the amount of a single transaction to determine whether it reaches the threshold for major transaction judgment. However, for more complex transactions, such as when there are multiple transactions with the same counterparty at the same time, the existing monitoring method cannot make an accurate judgment on all transaction statuses of the same counterparty. At this time, it is necessary to monitor and calculate the same counterparty manually, compare the counterparty's regulatory status, use the counterparty identifier to filter out all transactions of the counterparty, and then mark the transaction with the transaction amount reaching the regulatory threshold standard as a major transaction according to the regulatory rules. However, manual processing has the following shortcomings: First, manually matching each transaction according to the counterparty list requires huge manpower, but the matching work efficiency is low. Second, when calculating manually, due to the message lag in the transaction information to be approved, for example, the information on the counterparty's regulatory status, the number of transactions, and the transaction approval status is not fully grasped, so it is easy to misjudge whether the transaction meets the major transaction standard, resulting in missed identification and wrong identification of major transactions. Summary of the invention

[0005] The embodiment of the present invention provides a transaction data processing method for reducing labor costs and improving the recognition accuracy and efficiency of major transactions. The method includes:

[0006] Receiving multiple pending transactions; the pending transactions carry a counterparty identifier and a transaction amount;

[0007] The probability of approval of each pending transaction is calculated using a preset probability calculation model; the preset probability calculation model is obtained by training a logistic regression model using the approval results and counterparty credit information of multiple historical transactions in advance; the approval result includes whether the approval is passed and the approval time point;

[0008] The estimated cumulative value of each counterparty is calculated based on the transaction amount and approval probability corresponding to the same counterparty ID; the estimated cumulative value reflects the total transaction amount expected to occur by the counterparty;

[0009] Determine the incurred value of each transaction counterparty based on the transactions obtained from the real incurred database; the incurred value reflects the total amount of transactions that have occurred cumulatively with the transaction counterparty; the real incurred database stores the incurred transactions;

[0010] Receiving a first approval result of a first transaction to be approved; the first transaction to be approved carries an identifier of a first transaction counterparty;

[0011] Modify the approval probability of the first pending transaction according to the first approval result, and update the estimated cumulative value of the first transaction counterparty; wherein, when the first approval result is approval, updating the estimated cumulative value of the first transaction counterparty includes: deleting pending transactions with a lower priority than the first pending transaction from the pending transactions of the first transaction counterparty, and updating the estimated cumulative value of the first transaction counterparty according to the transaction amount and transaction occurrence probability of all pending transactions of the first transaction counterparty after deletion; the priority reflects the business importance and approval time of the pending transaction;

[0012] Determine whether the first transaction to be approved triggers a major transaction based on the updated estimated cumulative value of the first transaction counterparty, the incurred value of the first transaction counterparty and the preset cumulative threshold; the preset cumulative threshold is used to determine whether the total amount of all transactions of the transaction counterparty triggers a major transaction.

[0013] The embodiment of the present invention further provides a transaction data processing device for reducing labor costs and improving the recognition accuracy and efficiency of major transactions. The device includes:

[0014] A data receiving module, used to receive multiple pending transactions; the pending transactions carry a counterparty identifier and a transaction amount;

[0015] The approval probability prediction module is used to calculate the approval probability of each transaction to be approved by using a preset probability calculation model; the preset probability calculation model is obtained by training a logistic regression model in advance using the approval results and counterparty credit information of multiple historical transactions; the approval result includes whether the approval is passed and the approval time point;

[0016] The estimated cumulative value calculation module is used to calculate the estimated cumulative value of each transaction counterparty based on the transaction amount and approval probability corresponding to the same transaction counterparty identifier; the estimated cumulative value reflects the total transaction amount expected to occur by the transaction counterparty;

[0017] An already occurred value calculation module is used to determine the already occurred value of each transaction counterparty based on transactions obtained from a real already occurred database; the already occurred value reflects the total amount of transactions that have occurred cumulatively by the transaction counterparty; the real already occurred database stores already occurred transactions;

[0018] The estimated cumulative value updating module is used to receive a first approval result of a first pending transaction; the first pending transaction carries an identifier of a first transaction counterparty; the approval probability of the first pending transaction is modified according to the first approval result, and the estimated cumulative value of the first transaction counterparty is updated; wherein, when the first approval result is passed, updating the estimated cumulative value of the first transaction counterparty includes: deleting pending transactions with a lower priority than the first pending transaction from the pending transactions of the first transaction counterparty, and updating the estimated cumulative value of the first transaction counterparty according to the transaction amount and transaction occurrence probability of all pending transactions of the first transaction counterparty after deletion; the priority reflects the business importance and approval time of the pending transaction;

[0019] The trigger judgment module is used to determine whether the first pending transaction triggers a major transaction based on the updated estimated cumulative value of the first transaction counterparty, the incurred value of the first transaction counterparty and the preset cumulative threshold; the preset cumulative threshold is used to determine whether the total amount of all transactions of the transaction counterparty triggers a major transaction.

[0020] An embodiment of the present invention further provides a computer device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the above-mentioned transaction data processing method when executing the computer program.

[0021] An embodiment of the present invention further 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 above-mentioned transaction data processing method is implemented.

[0022] An embodiment of the present invention further provides a computer program product, which includes a computer program. When the computer program is executed by a processor, the above-mentioned transaction data processing method is implemented.

[0023] In the embodiment of the present invention, the cumulative value of the expected transactions and the cumulative value of the actual transactions of all pending transactions of the same counterparty are automatically counted, and the cumulative value of the statistics is automatically updated according to the real-time approval results, and whether the approved transaction triggers a major transaction is determined according to the preset cumulative threshold used to determine whether the major transaction is triggered. Compared with the technical solutions in the prior art, the embodiment of the present invention first avoids manual calculations, reduces a large amount of labor costs, and improves the efficiency of identifying major transactions. Secondly, in the case of multiple pending transactions at the same time, the expected amount of the transaction can be adjusted in real time according to the approval results of the transaction, thereby making the identification of major transactions more accurate and improving the accuracy of identifying major transactions. BRIEF DESCRIPTION OF THE DRAWINGS

[0024] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments or the prior art descriptions. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work. In the drawings:

[0025] Figure 1 Schematic diagram of a transaction data processing method according to an embodiment of the present invention;

[0026] Figure 2 A diagram showing a specific example of a transaction data processing method in an embodiment of the present invention;

[0027] Figure 3 FIG. 2 is another specific example diagram of the transaction data processing method in the embodiment of the present invention;

[0028] Figure 4 Schematic diagram of a transaction data processing device in an embodiment of the present invention;

[0029] Figure 5 Schematic diagram of a computer device in an embodiment of the present invention. DETAILED DESCRIPTION

[0030] To make the purpose, technical solution and advantages of the embodiments of the present invention more clear, the embodiments of the present invention are further described in detail below in conjunction with the accompanying drawings. Here, the exemplary embodiments of the present invention and their descriptions are used to explain the present invention, but are not intended to limit the present invention.

[0031] The acquisition, transmission, storage, use, and processing of data in the technical solution of this application comply with the relevant provisions of national laws and regulations.

[0032] It should be noted that in the embodiments of the present application, certain software, components, models and other existing solutions in the industry may be mentioned, and they should be regarded as exemplary. Their purpose is only to illustrate the feasibility of implementing the technical solution of the present application, but it does not mean that the applicant has or will necessarily use the solution.

[0033] Figure 1 FIG. 1 is a flow chart of a transaction data processing method according to an embodiment of the present invention. Figure 1 As shown, the method includes:

[0034] Step 101: receiving multiple pending transactions; the pending transactions carry counterparty identification and transaction amount;

[0035] Step 102: Calculate the approval probability of each transaction to be approved using a preset probability calculation model; the preset probability calculation model is obtained by training a logistic regression model using the approval results and counterparty credit information of multiple historical transactions in advance; the approval result includes whether the approval is passed and the approval time point;

[0036] Step 103: Calculate the estimated cumulative value of each transaction counterparty based on the transaction amount and approval probability corresponding to the same transaction counterparty identifier; the estimated cumulative value reflects the total transaction amount that is expected to occur with the transaction counterparty;

[0037] Step 104: Determine the value of each transaction counterparty based on the transactions obtained from the real transaction database; the value of the transaction reflects the total amount of transactions that have occurred cumulatively by the transaction counterparty; the real transaction database stores the transactions that have occurred;

[0038] Step 105: receiving a first approval result of a first transaction to be approved; the first transaction to be approved carries an identifier of a first transaction counterparty;

[0039] Step 106: modify the approval probability of the first pending transaction according to the first approval result, and update the estimated cumulative value of the first transaction counterparty; wherein, when the first approval result is approval, updating the estimated cumulative value of the first transaction counterparty includes: deleting pending transactions with a lower priority than the first pending transaction from the pending transactions of the first transaction counterparty, and updating the estimated cumulative value of the first transaction counterparty according to the transaction amount and transaction occurrence probability of all pending transactions of the first transaction counterparty after deletion; the priority reflects the business importance and approval time of the pending transaction;

[0040] Step 107: Determine whether the first transaction to be approved triggers a major transaction based on the updated estimated cumulative value of the first transaction counterparty, the incurred value of the first transaction counterparty, and a preset cumulative threshold; the preset cumulative threshold is used to determine whether the total amount of all transactions of the transaction counterparty triggers a major transaction.

[0041] from Figure 1 It can be seen from the process shown that the embodiment of the present invention automatically counts the cumulative value of the expected transactions and the cumulative value of the actual transactions that have occurred for all pending transactions of the same counterparty, and automatically updates the statistical cumulative value based on the real-time approval results, and judges whether the approved transaction triggers a major transaction based on the preset cumulative threshold used to judge whether the major transaction is triggered. Compared with the technical solutions in the prior art, the embodiment of the present invention first avoids manual calculations, reduces a large amount of labor costs, and improves the efficiency of identifying major transactions. Secondly, when there are multiple pending transactions at the same time, the expected amount of the transaction can be adjusted in real time according to the approval results of the transaction, thereby making the identification of major transactions more accurate and improving the accuracy of identifying major transactions.

[0042] In the embodiment of the present invention, a trading system is established, which realizes Figure 1 In the method shown, the transaction system interacts with the business system and receives multiple pending transactions pushed by the business system, including transaction information of each pending transaction. The transaction information includes but is not limited to transaction identifiers such as serial numbers, counterparty identifiers such as personal names of counterparties, corporate information of counterparties, transaction amounts, transaction types such as fixed assets, transportation costs, sales transactions, etc., descriptive information on transaction impacts, etc.

[0043] In one embodiment, the method described in the embodiment of the present invention is to process transactions of a specified type, such as establishing an expected change library and an actual occurred library for transactions of the same specified type, and performing corresponding processing. The transaction of the specified type can be a transaction with a special counterparty, where the enterprise and its subsidiaries transfer resources, services, obligations or other interests with the special counterparty, and there is a designated regulatory department to supervise the transaction of the specified type.

[0044] The transaction system can store the received multiple pending transactions into the expected change library. The expected change library is mainly used to store the transaction information of pending transactions.

[0045] In one embodiment, after receiving multiple transactions to be approved, the method may further include the following steps:

[0046] Step 1: Obtain the credit information and transaction data of the counterparty in real time through the streaming architecture; the credit information includes but is not limited to one or any combination of the following: enterprise registration information, enterprise operation information, enterprise compliance supervision information, enterprise public opinion and market evaluation information, natural person credit information, natural person public affairs payment information, natural person tax information, natural person public opinion information;

[0047] Step 2: Store multiple pending transactions and real-time acquired credit information and transaction data in the expected change library; the expected change library stores pending and approved transactions in the form of a graph structure, in which the pending transaction identifier and counterparty identifier are nodes, and the transaction amount, time information, credit information and transaction data are node attributes.

[0048] In this example, in order to realize probability prediction, priority prediction, etc. faster and more efficiently in the future, the latest counterparty credit information and transaction data are obtained in real time, and the counterparty credit information and transaction data are stored as node attributes together with the pending transaction information.

[0049] During implementation, in order to quickly obtain pending transactions, counterparty credit information and transaction data may also be stored separately, that is, not placed in the expected change library, but stored in other databases, which may be stored in the form of a structure tree.

[0050] Credit information about the counterparty:

[0051] If the counterparty is a legal person, the counterparty credit information may include: the industry to which the legal person belongs, the registered amount, the approval time of the approved historical transactions (if included), the operating status of the legal person, the frequency of information changes of the legal person, water and electricity fee records, and the internal credit rating of the institution;

[0052] If the counterparty is a natural person, the counterparty credit information may include: the natural person's age, the natural person's social security record, the natural person's length of service, the nature of the natural person's work industry, the approval time of historical transactions that have passed the approval (if included), and the institution's internal credit rating, etc.

[0053] Transaction data of counterparties: such as transaction frequency, transaction type, approval time information, etc., which are not listed here one by one.

[0054] Afterwards, the preset probability calculation model is used to calculate the approval probability of each transaction to be approved.

[0055] In one embodiment, a preset probability calculation model is used to calculate the approval probability of each pending transaction, including: obtaining node information and node attribute information of the pending transaction from the expected change library; using the node information, node attribute information and the preset probability calculation model to calculate the approval probability of each pending transaction.

[0056] In one embodiment, the node information, node attribute information and a preset probability calculation model are used to calculate the approval probability of each pending transaction, including:

[0057] Step 1: preprocess the node information and node attribute information to obtain preprocessed data; data preprocessing includes natural language processing, data cleaning, and data conversion;

[0058] Step 2: Input the preprocessed data into the following preset probability calculation model to output the approval probability of each transaction to be approved:

[0059]

[0060] In the formula, p represents the probability value of approval, for example, approval is 1, and approval is 0; ψ represents the probability calculation function, which is used to judge the p value result based on the probability result, b represents the sample deviation, and a n is the weight coefficient, x n is the nth preprocessed data, i.e., the independent variable of the constant term, and θ is the influencing factor, which is obtained based on the public opinion information related to the counterparty.

[0061] During implementation, the approval results and counterparty credit information of multiple historical transactions are used in advance to train, verify and test the logistic regression model to obtain the model parameters of the trained preset probability calculation model, such as weight coefficients and influencing factors.

[0062] Specifically, the p value can be obtained by constructing a probability function of the transaction occurring (i.e., approval being passed), and by using the characteristic parameters (e.g., a n and b), calculate the probability, compare the obtained probability result with the preset threshold (e.g. 50%), and then determine the value of p of the corresponding single transaction in formula (1). For example, if the probability function calculation result is 60%, which is greater than 50%, then p is 1. The preset moment can be understood as the preset probability calculation model needs to be updated in real time using historical data, and its characteristic parameters will change, so the characteristic parameters up to the preset moment are obtained for calculation.

[0063] The p-value can be obtained as follows:

[0064] Step 1: For a transaction to be approved, extract the node features corresponding to each node and the edge features of each directed edge from the historical transactions of the counterparty of the transaction to be approved;

[0065] Step 2: construct a probability function of basic features, and fit the basic probability function to obtain the basic features of the transaction occurrence probability function;

[0066] Step 3: Extract the basic features of the moment stage, construct a new function (logistic regression model) with the basic features of the transaction occurrence probability function, and obtain the matrix required by the target.

[0067] When using a logistic regression model to predict the probability of approval, the specific steps are as follows:

[0068] Step 1: Data Preparation

[0069] Data collection: Collect historical transaction data, including but not limited to approval, failure, approval time information, etc.

[0070] Data cleaning: remove duplicate records, handle missing values ​​(delete or fill), detect and handle outliers.

[0071] Data transformation: Convert categorical variables to numerical representations, such as one-hot encoding; standardize or normalize continuous variables (such as transaction amounts) to facilitate model training.

[0072] Step 2: Feature Selection

[0073] Correlation analysis: Calculate the correlation between each feature and the target variable (approval result), and use methods such as correlation coefficient matrix or chi-square test to screen out important features.

[0074] Step 3: Model training

[0075] Divide the dataset: Divide the dataset into training and testing sets (the ratio is 70% training and 30% testing).

[0076] Initialize the model: Use the logistic regression model and select the binary cross entropy loss function.

[0077] Parameter setting: Set regularization parameters (such as L1 or L2 regularization) to prevent overfitting; set the number of iterations and other hyperparameters.

[0078] Train Model: Fit a logistic regression model using the training set data.

[0079] Step 4: Model Evaluation

[0080] Evaluate model performance on the test set using metrics such as accuracy, precision, recall, and F1 score.

[0081] Step 5: Model maintenance and update

[0082] As more data accumulates, the model is retrained periodically to adapt to the new data distribution.

[0083] Formula (1) is obtained through the specific steps of the above logistic regression model.

[0084] The steps for determining the impact factor θ are as follows: (1) Based on the counterparty’s credit information, public opinion is collected and analyzed by calling the big prediction model; (2) The big language model analyzes the counterparty’s public opinion results and obtains the public opinion outlook level (such as severe negative impact and negative impact, maintain the status quo, positive impact and optimistic positive impact); (3) The big language model feeds back the public opinion outlook level to the trading system; (4) The trading system converts the public opinion outlook level of the big language model into an impact factor (numerical value) and adds it to formula (1) for calculation.

[0085] In step 103, the estimated cumulative value of each transaction counterparty is calculated based on the transaction amount and approval probability corresponding to the same transaction counterparty identifier; the estimated cumulative value reflects the total transaction amount expected to occur by the transaction counterparty.

[0086] During implementation, the estimated cumulative value of each counterparty is calculated separately, for example, the transaction amount and the probability of approval are multiplied together, and then the products are accumulated.

[0087] At this time, in the process of accumulating the product, the pending transactions with a p value of 0 can be placed at the end. For the pending transactions with a p value of 0, the pending transactions can be sorted in chronological order to participate in the calculation of the estimated cumulative value.

[0088] For example, there are four transactions, n, n+1, n+2 and n+3. The order of transactions by time is n, n+1, n+2 and n+3. After the probability of approval is calculated, the result of n+1 is 0, so the order is n, n+2, n+3 and n+1.

[0089] Taking into account that when some pending transactions are approved, the p-value needs to be changed later, in an embodiment of the present invention, pending transactions with different p-values ​​can be stored separately. For example, the data in the expected change library is stored in a structured manner in batches according to the p-value results, and two database tables are used to store transactions with unchanged p-value results and transactions with a p-value of 0, respectively.

[0090] In one embodiment, the estimated cumulative value of each transaction counterparty is calculated based on the transaction amount and approval probability corresponding to the same transaction counterparty identifier, which may include the following steps:

[0091] Step 1: Calculate the priority of pending transactions of the same counterparty using a preset linear regression model, sort the pending transactions of the same counterparty according to the priority, and obtain a sequence of pending transactions; the counterparty identifiers of any two pending transaction sequences are different; the preset linear regression model is pre-trained using the transaction approval information, credit information and transaction data of multiple historical counterparties; the transaction approval information includes the time point of initiating approval, whether the approval is passed, and the time point of approval;

[0092] Step 2: Based on the transaction amount, approval probability and priority order in the sequence of pending transactions, the estimated cumulative value of each counterparty is calculated using the following formula:

[0093] A 变 =P 1 ×A 1 +P 2 ×A 2 +P 3 ×A 3 +......+P n ×A n (2);

[0094] In the formula, A 变 Indicates the expected cumulative value, P n A represents the probability of approval of the nth pending transaction. n Indicates the transaction amount of the nth pending transaction. The priority of the n-1th pending transaction is higher than that of the nth pending transaction.

[0095] In this example, the priority of pending transactions is introduced. The priority can reflect the importance of the business and the length of the approval time, which is more in line with the needs of actual work.

[0096] The priority of pending transactions of the same counterparty is calculated by using a preset linear regression model, which may include:

[0097] Step 1: Obtain node information and node attribute information of pending transactions of the same counterparty from the expected change database;

[0098] Step 2: Calculate the priority of pending transactions of the same counterparty using the node information, node attribute information and preset linear regression model of pending transactions of the same counterparty according to the following formula:

[0099] y'=(X×β+∈)×θ (3);

[0100] Where y' represents the priority; X is the input data; β is the model parameter matrix, which is obtained by least squares fitting; ∈ is the random error term; θ is the influencing factor, which is obtained based on the public opinion information related to the counterparty.

[0101] Specifically, the preset linear regression model is:

[0102]

[0103] Among them, x nnis a constant term independent variable, that is, the independent variable of the constant term, which can be used as a vector feature for function construction for the relevant factors in historical transactions. The construction steps are as follows: take the nature of the counterparty (legal person and natural person) in the variable feature as the basic feature to select the influencing factors. If the nature of the counterparty is a legal person, the influencing factors are: the industry to which the legal person belongs, the historical transaction amount, the transaction frequency, the operating status of the legal person, the frequency of information changes of the legal person, the water and electricity fee records, the credit rating within the institution, the recent public opinion information of the legal person, the approval probability of historical transactions, etc.; if the nature of the counterparty is a natural person, the influencing factors are: the age of the natural person, the social security record of the natural person, the length of service of the natural person, the nature of the natural person's work industry, the credit rating within the institution, the recent public opinion information of the natural person, the historical transaction amount, the transaction frequency, the approval probability of historical transactions, etc.; β n is a variable parameter obtained by function fitting, and θ is an influencing factor, which is usually a constant. The process of obtaining the influencing factor in the preset probability calculation model can be referred to.

[0104] For example, for the four specified types of transactions whose sorting results are n, n+2, n+3 and n+1, they are re-sorted according to the priority sorting results. If transaction n+3 is more important and has a shorter approval time, n is second, and n+2 is the longest, then the sorting results are updated to: n+3, n, n+2 and n+1.

[0105] In step 104, the incurred value of each transaction counterparty is determined based on the transactions obtained from the real incurred database; the incurred value reflects the total amount of transactions that have occurred cumulatively by the transaction counterparty; the real incurred database stores the transactions that have occurred.

[0106] For example, the incurred value of each counterparty is determined based on the transactions obtained from the real incurred database according to the following formula:

[0107] B 历 =k 1 ×B 1 +k 2 ×B 2 +k 3 ×B 3 +......+k n ×B n (5);

[0108] In the formula, B 历 Indicates the value that has occurred, k n Indicates the preset weight coefficient of the nth approved transaction, which is generally a business experience value and a constant. n Indicates the transaction amount that has occurred. n , which can be understood as leaving room for variation in calculations. For example, even if a transaction is approved, something unexpected may happen.

[0109] In summary, the calculation of the expected cumulative value and the incurred value for each counterparty has been completed.

[0110] When the transaction system receives the approval result of a pending transaction from the business system, it starts to update and modify the estimated cumulative value. In step 105, a first approval result of a first pending transaction is received; the first pending transaction carries the identifier of the first transaction counterparty. The first pending transaction can be any pending transaction or any number of pending transactions.

[0111] In step 106, the approval probability of the first transaction to be approved is modified according to the first approval result, and the estimated accumulated value of the first transaction counterparty is updated.

[0112] When the first approval result is failure, the approval probability is modified according to the failure result, and formula (2) is modified at the same time, waiting for the approval processing of other subsequent transactions.

[0113] When the first approval result is passed, updating the estimated cumulative value of the first transaction counterparty includes: deleting pending transactions with a lower priority than the first pending transaction from the pending transactions of the first transaction counterparty, and updating the estimated cumulative value of the first transaction counterparty according to the transaction amount and transaction occurrence probability of all pending transactions of the first transaction counterparty after deletion.

[0114] For example, if the third transaction in formula (2) is approved, the estimated cumulative value of the first three transactions will be calculated.

[0115] Finally, in step 107, it is determined whether the first transaction to be approved triggers a major transaction based on the updated estimated cumulative value of the first transaction counterparty, the incurred value of the first transaction counterparty, and the preset cumulative threshold.

[0116] In the embodiment of the present invention, a significant transaction includes a transaction whose transaction amount exceeds a specified threshold, wherein the transaction amount includes a single transaction amount and a cumulative transaction amount of a single counterparty.

[0117] When implemented, it can be calculated according to the following formula:

[0118] Z 组合 =ψ(ω1×A 变 +ω2×B 历 ) (6);

[0119] In the formula, Z 组合 Indicates whether a major transaction is triggered. For example, if it is a major transaction, Z 组合 Yes, not for major transactions, Z 组合No; ψ() represents a judgment function. The data input into ψ() is subtracted from the preset cumulative threshold. If it is less than 0, it is not a major transaction. If it is greater than 0, it is a major transaction. ω1 and ω2 are variable parameters that can be obtained from business experience and regulatory requirements.

[0120] In one embodiment, a counterparty supervision flag is introduced, and the counterparty supervision flag is used to reflect whether the counterparty is still subject to supervision.

[0121] When the counterparty supervision mark expires, the expiration date of the counterparty is taken as the starting point, and the proposed amount in the estimated change database after this time point is fully deducted. The accumulated amount in the actual occurrence database up to the last day of the validity period of the counterparty supervision mark is used as the calculation amount of major transactions and stored in the historical transactions of the trading system.

[0122] When the validity period of the counterparty's regulatory mark increases, if the validity period of the regulatory mark is continuously increased on the original validity period, it will be calculated according to the existing rules. If the counterparty's regulatory mark becomes effective again within a period of time after expiration, the major transactions will continue to be calculated according to the existing rules based on the historical transactions in the actual database.

[0123] In specific implementation, after the business system pushes the approval results to the trading system, the trading system receives the approval results text data, converts the text data into signal information, and performs transaction information combination calculation according to formula (2). After the counterparty regulatory mark in the trading system is changed, the regulatory mark change information is converted into signal information in major transactions, and the transaction information combination calculation is performed. The basic characteristics of the converted signal information are fed back in the change of the P value. The vector eigenvalue changes of the P value at the preset time node are as follows.

[0124] The changes approved by the approval results are shown in Table 1:

[0125] Table 1

[0126]

[0127] The changes with disapproval results are shown in Table 2:

[0128] Table 2

[0129]

[0130]

[0131] The changes in counterparty supervision marks are shown in Table 3:

[0132] Table 3

[0133]

[0134] In one embodiment, determining whether the first transaction to be approved triggers a major transaction based on the updated estimated cumulative value of the first transaction counterparty, the incurred value of the first transaction counterparty, and a preset cumulative threshold may include:

[0135] comparing the sum of the updated estimated cumulative value of the first counterparty and the incurred value of the first counterparty to a preset cumulative threshold;

[0136] When the total is greater than the preset cumulative threshold, it is determined that the first transaction to be approved triggers a major transaction; when the total is not greater than the preset cumulative threshold, it is determined that the first transaction to be approved does not trigger a major transaction.

[0137] In one embodiment, after determining that the first pending transaction triggers a major transaction, the process may further include: pushing the first pending transaction to a high-level approval process for approval, where the high-level approval process is used to approve major transactions. For example, the senior management of the enterprise decides whether to approve the major transaction.

[0138] In one embodiment, after determining that the first pending transaction did not trigger a major transaction, it may also include: deleting the node corresponding to the first pending transaction from the expected change library, updating the expected cumulative value of the first transaction counterparty; storing the first pending transaction as a real transaction that has occurred in the real occurred library, and updating the occurred value of the first transaction counterparty.

[0139] Similarly, after the approval of the advanced approval process, if it is determined that no major transaction has been triggered, the node corresponding to the first pending transaction will be deleted from the expected change library, and the expected cumulative value of the first transaction counterparty will be updated; the first pending transaction will be stored in the real transaction library as a real transaction, and the first transaction counterparty's actual value will be updated. After the approval of the advanced approval process, if it is determined that a major transaction has been triggered, it will be executed in accordance with the corresponding regulations for major transactions, which will not be expanded here.

[0140] Figure 2 FIG. 1 is a specific example of a transaction data processing method according to an embodiment of the present invention. Figure 2 As shown, after determining the value of each counterparty's transaction based on the transactions obtained from the real transaction database, it can also include:

[0141] Step 201: using the trained time prediction model, predict the approval time point of each pending transaction; the time prediction model is obtained by training the autoregressive integrated moving average ARIMA model using historical transaction approval information, counterparty credit information and transaction data in advance;

[0142] Step 202: Set the approval probability of each pending transaction to approval;

[0143] Step 203: Calculate the estimated accumulated value and the incurred value for each counterparty at the same approval time point in the order of the approval time points of all pending transactions from earliest to latest;

[0144] Step 204: output the time point when each trading counterparty reaches the preset cumulative threshold value based on the preset cumulative threshold value and the estimated cumulative value and the incurred value at the same approval time point for each trading counterparty.

[0145] In this example, at a preset time point, the future transaction trend of the counterparty is predicted, and the predicted value is used to perform a combined calculation in formula (6). The combined calculation result is used to predict whether a major transaction is triggered. When implementing, first obtain historical data, then construct a time series (Time Series Analysis) function, use historical data to fit the basic probability function, and obtain the basic characteristics of the time series (Time Series Analysis) function. Finally, based on the basic characteristics of the function, introduce data to calculate the transaction value under the future time vector. The constructed functional relationship is:

[0146] Z 未来 ={ARIMA(v,d,q)}×θ (7);

[0147] In the formula, Z 未来 It represents the predicted cumulative value, parameter v is the parameter of AR autoregressive model, parameter d is the difference parameter, and parameter q is the parameter of MA moving average model. All three parameters are obtained by model fitting. The model needs to add historical transaction data as the constant term independent variable, that is, the independent variable of the constant term, which can be the relevant factors in historical transactions as vector features for function construction. The construction steps are as follows: Take the nature of the counterparty (legal person and natural person) in the variable special type as the basic feature to select the influencing factors. If the nature of the counterparty is a legal person, the influencing factors are selected as follows: the industry to which the legal person belongs, the historical transaction amount, the transaction frequency, the nature of the legal person, the operating status of the legal person, the frequency of changes in the supervisor of the legal person, the water and electricity fee records, and the recent public opinion information of the legal person; if the nature of the counterparty is a natural person, the influencing factors are selected as follows: the age of the natural person, the social security record of the natural person, the length of service of the natural person, the historical transaction amount, the transaction frequency, etc.; θ is the influencing factor (generally a constant).

[0148] The calculated value Z at the preset time node 未来 Substitute it into formula (6) for calculation, and send a risk warning message of upcoming major transactions to the business system based on the calculation result, so that the enterprise can prepare in advance and improve the user experience.

[0149] In one embodiment, after outputting the time point when each transaction counterparty reaches the preset cumulative threshold, it can also include: counting the actual approval result and actual approval time information of each transaction to be approved, and obtaining the latest credit information and transaction data of the transaction counterparty; based on the principal component analysis algorithm, using the actual approval result, actual approval time information, the latest credit information and transaction data of each transaction to be approved, updating the model parameters of the time prediction model, the preset linear regression model, and the preset probability calculation model.

[0150] During implementation, every time a transaction is approved, based on the approval result, the new approval time is used as a node, and the approval factors and approval results such as the credit data and transaction data of the counterparty at that time point are used as samples, and included in the training data of the new transaction time prediction model, the preset linear regression model, and the preset probability calculation model. According to the training results, the trading system automatically adjusts the parameters of the model. At the same time, according to the new model parameters, the probability of occurrence, ranking and future transaction amount of the approved transaction are adjusted in real time according to the model after the system adjusts the parameters, and the adjusted results are included in the formula for major transaction calculation.

[0151] Figure 3 FIG. 1 is another specific example of the transaction data processing method in the embodiment of the present invention. Figure 3 As shown in the figure, in this method, the business system pushes pending transactions to the trading system, specifically, pending transactions 1, 2, ..., N information. After receiving the pending transactions 1, 2, ..., N information, the trading system Figure 1 The method described in the above sequentially calculates the approval probability p, the estimated cumulative value, and the priority sorting, etc., determines the actual value of each counterparty according to the actual actual database, and receives the approval information pushed by the business system in real time, updates the approval probability and the estimated cumulative value according to the latest approval information, and then determines whether the currently approved transaction triggers a major transaction of the counterparty according to the real-time estimated cumulative value and the actual value. If a major transaction is triggered, the business system is prompted to further approve it. If no major transaction is triggered, the transaction is deleted from the estimated change database, and the transaction amount of the counterparty in the actual actual database is updated.

[0152] In summary, the transaction data processing method in the embodiment of the present invention is based on the real-time detection of the transaction system, counterparty supervision signs, transaction status and transaction cumulative amount, and can calculate major transactions more accurately, and has the following beneficial technical effects.

[0153] 1) The calculation is switched from manual matching to trading system, which reduces a lot of labor costs. At the same time, it can make real-time judgments based on the validity period of the counterparty supervision mark, making the identification and calculation of major transactions more accurate.

[0154] 2) When there are multiple transactions at the same time, the proposed transaction amount can be adjusted in real time based on the approval results of the transaction, making the identification of major transactions more accurate.

[0155] 3) In the case of multiple transactions, it can measure in real time whether it constitutes a major transaction based on the approval results of the transaction, provide advance reminders for transactions that meet the major transaction standards, conduct relevant approval processes, and reduce compliance risks.

[0156] 4) In the case of multiple transactions, the algorithm effectively excludes unreasonable transactions from the calculation in advance, improving the accuracy of major transactions.

[0157] 5) Predict trading behaviors with counterparties in advance and respond in a timely and proactive manner to transactions that may trigger major transactions, thereby improving business operation efficiency.

[0158] The present invention also provides a transaction data processing device, as described in the following embodiments. Since the principle of solving the problem by the device is similar to that of the transaction data processing method, the implementation of the device can refer to the implementation of the transaction data processing method, and the repeated parts will not be repeated.

[0159] Figure 4 Schematic diagram of a transaction data processing device according to an embodiment of the present invention. Figure 4 The transaction data processing device 400 includes:

[0160] The data receiving module 401 is used to receive multiple pending transactions; the pending transactions carry the counterparty identification and transaction amount;

[0161] The approval probability prediction module 402 is used to calculate the approval probability of each transaction to be approved by using a preset probability calculation model; the preset probability calculation model is obtained by training a logistic regression model in advance using the approval results and counterparty credit information of multiple historical transactions; the approval result includes whether the approval is passed and the approval time point;

[0162] The estimated cumulative value calculation module 403 is used to calculate the estimated cumulative value of each transaction counterparty according to the transaction amount and approval probability corresponding to the same transaction counterparty identifier; the estimated cumulative value reflects the total transaction amount expected to occur by the transaction counterparty;

[0163] The occurred value calculation module 404 is used to determine the occurred value of each transaction counterparty based on the transactions obtained from the real occurred database; the occurred value reflects the total amount of transactions that have occurred cumulatively by the transaction counterparty; the real occurred database stores the transactions that have occurred;

[0164] The estimated cumulative value updating module 405 is used to receive a first approval result of a first pending transaction; the first pending transaction carries an identifier of a first transaction counterparty; the approval probability of the first pending transaction is modified according to the first approval result, and the estimated cumulative value of the first transaction counterparty is updated; wherein, when the first approval result is passed, updating the estimated cumulative value of the first transaction counterparty includes: deleting pending transactions with a lower priority than the first pending transaction from the pending transactions of the first transaction counterparty, and updating the estimated cumulative value of the first transaction counterparty according to the transaction amount and transaction occurrence probability of all pending transactions of the first transaction counterparty after deletion; the priority reflects the business importance and approval time of the pending transaction;

[0165] The trigger judgment module 406 is used to determine whether the first pending transaction triggers a major transaction based on the updated estimated cumulative value of the first transaction counterparty, the incurred value of the first transaction counterparty and the preset cumulative threshold; the preset cumulative threshold is used to determine whether the total amount of all transactions of the transaction counterparty triggers a major transaction.

[0166] In one embodiment, the apparatus 400 further includes: an expected change library processing module; the expected change library processing module is used to:

[0167] After the data receiving module 401 receives multiple pending transactions, the credit information and transaction data of the counterparty are obtained in real time through the streaming architecture; the counterparty credit information includes one or any combination of the following: enterprise registration information, enterprise operation information, enterprise compliance supervision information, enterprise public opinion and market evaluation information, natural person credit information, natural person public affairs payment information, natural person tax information, natural person public opinion information;

[0168] Store multiple pending transactions and real-time counterparty credit information and transaction data in the expected change library; the expected change library stores pending and under-approval transactions in the form of a graph structure, in which the pending transaction identifier and counterparty identifier are nodes, and the transaction amount, time information, credit information and transaction data are node attributes.

[0169] In one embodiment, the approval probability prediction module 402 is specifically used to: obtain node information and node attribute information of the pending transaction from the expected change library; and calculate the approval probability of each pending transaction using the node information, node attribute information and a preset probability calculation model.

[0170] In one embodiment, the approval probability prediction module 402 is specifically used to:

[0171] Perform data preprocessing on node information and node attribute information to obtain preprocessed data; data preprocessing includes natural language processing, data cleaning, and data conversion;

[0172] The preprocessed data is input into the following preset probability calculation model to output the approval probability of each pending transaction: wherein the preset probability calculation model uses a binary cross entropy loss function;

[0173]

[0174] In the formula, p represents the probability of approval, ψ represents the probability calculation function, b represents the sample deviation, and a n is the weight coefficient, x n is the nth preprocessed data, and θ is the influencing factor, which is obtained based on the public opinion information related to the counterparty.

[0175] In one embodiment, the estimated accumulated value calculation module 403 is specifically used for:

[0176] The priority of pending transactions of the same counterparty is calculated by using a preset linear regression model, and the pending transactions of the same counterparty are sorted by priority to obtain a sequence of pending transactions; the counterparty identifiers of any two pending transaction sequences are different; the preset linear regression model is pre-trained using the transaction approval information, credit information and transaction data of multiple historical counterparties; the transaction approval information includes the time point of initiating approval, whether the approval is passed, and the time point of approval;

[0177] Based on the transaction amount, approval probability and priority order of the pending transactions, the estimated cumulative value of each counterparty is calculated using the following formula:

[0178] A 变 =P 1 ×A 1 +P 2 ×A 2 +P 3 ×A 3 +......+P n ×A n ;

[0179] In the formula, A 变 Indicates the expected cumulative value, P n A represents the probability of approval of the nth pending transaction. n Indicates the transaction amount of the nth pending transaction. The priority of the n-1th pending transaction is higher than that of the nth pending transaction.

[0180] In one embodiment, the estimated accumulated value calculation module 403 is specifically used for:

[0181] Obtain node information and node attribute information of pending transactions of the same counterparty from the expected change library;

[0182] The priority of pending transactions of the same counterparty is calculated using the node information, node attribute information and the preset linear regression model of pending transactions of the same counterparty according to the following formula:

[0183] y'=(X×β+∈)×θ;

[0184] Where y' represents the priority; X is the input data; β is the model parameter matrix, which is obtained by least squares fitting; ∈ is the random error term; θ is the influencing factor, which is obtained based on the public opinion information related to the counterparty.

[0185] In one embodiment, the occurred value calculation module 404 is specifically used for:

[0186] The incurred value of each counterparty is determined based on the transactions obtained from the real incurred database according to the following formula:

[0187] B 历 =k 1 ×B 1 +k 2 ×B 2 +k 3 ×B 3 +......+k n ×B n ;

[0188] In the formula, B 历 Indicates the value that has occurred, k n Indicates the preset weight coefficient of the nth approved transaction, B n Indicates the transaction amount that has occurred.

[0189] In one embodiment, the trigger determination module 406 is specifically used to:

[0190] comparing the sum of the updated estimated cumulative value of the first counterparty and the incurred value of the first counterparty to a preset cumulative threshold;

[0191] When the sum is greater than the preset cumulative threshold, the first pending transaction is determined to trigger a major transaction;

[0192] When the sum is not greater than the preset cumulative threshold, it is determined that the first transaction to be approved does not trigger a major transaction.

[0193] In one embodiment, the device 400 also includes: an advanced approval processing module, which is used to: after the trigger judgment module 406 determines that the first transaction to be approved triggers a major transaction, push the first transaction to be approved to the advanced approval process for approval, and the advanced approval process is used to approve major transactions.

[0194] In one embodiment, the device 400 also includes: a post-approval processing module, which is used to: after the trigger judgment module 406 determines that the first pending transaction has not triggered a major transaction, delete the node corresponding to the first pending transaction from the expected change library, and update the expected cumulative value of the first transaction counterparty; store the first pending transaction as a real transaction that has occurred in the real occurred library, and update the occurred value of the first transaction counterparty.

[0195] In one embodiment, the device 400 further includes: a major transaction advance prediction module, the major transaction advance prediction module is used to:

[0196] After the occurred value calculation module 404 determines the occurred value of each transaction counterparty according to the transactions obtained from the real occurred database, the time point of approval of each pending transaction is predicted using the trained time prediction model; the time prediction model is pre-trained using the historical transaction approval information, the credit information of the transaction counterparty and the transaction data to train the autoregressive integrated moving average ARIMA model;

[0197] Set the approval probability of each pending transaction to approved;

[0198] According to the order of approval time points of all pending transactions, the following are calculated for each counterparty at the same approval time point: the expected cumulative value and the incurred value;

[0199] Based on the preset cumulative threshold and the estimated cumulative value and the actual value at the same approval time point for each counterparty, the time point at which each counterparty reaches the preset cumulative threshold is output.

[0200] In one embodiment, the device 400 further includes: a model iteration updating module, the model iteration updating module is used to:

[0201] After the major transaction advance prediction module outputs the time point when each transaction counterparty reaches the preset cumulative threshold, the actual approval result and actual approval time information of each pending transaction are counted to obtain the latest credit information and transaction data of the counterparty; based on the principal component analysis algorithm, the actual approval result, actual approval time information, the latest credit information and transaction data of each pending transaction are used to update the model parameters of the time prediction model, the preset linear regression model, and the preset probability calculation model.

[0202] Figure 5 Schematic diagram of a computer device in an embodiment of the present invention. Figure 5As shown, an embodiment of the present invention further provides a computer device 500, including a processor 501, a memory 502, and a computer program 503 stored in the memory 502 and executable on the processor 501, and the processor 501 implements the above-mentioned transaction data processing method when executing the computer program 503.

[0203] An embodiment of the present invention further 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 above-mentioned transaction data processing method is implemented.

[0204] An embodiment of the present invention further provides a computer program product, which includes a computer program. When the computer program is executed by a processor, the above-mentioned transaction data processing method is implemented.

[0205] In the embodiment of the present invention, the cumulative value of the expected transactions and the cumulative value of the actual transactions of all pending transactions of the same counterparty are automatically counted, and the cumulative value of the statistics is automatically updated according to the real-time approval results, and whether the approved transaction triggers a major transaction is determined according to the preset cumulative threshold used to determine whether the major transaction is triggered. Compared with the technical solutions in the prior art, the embodiment of the present invention first avoids manual calculations, reduces a large amount of labor costs, and improves the efficiency of identifying major transactions. Secondly, in the case of multiple pending transactions at the same time, the expected amount of the transaction can be adjusted in real time according to the approval results of the transaction, thereby making the identification of major transactions more accurate and improving the accuracy of identifying major transactions.

[0206] Those skilled in the art will appreciate that embodiments of the present invention may be provided as methods, systems, or computer program products. Therefore, the present invention may take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware. Moreover, the present invention may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0207] The present invention is described with reference to flowcharts and / or block diagrams of methods, devices (systems), and computer program products according to embodiments of the present invention. It should be understood that each process and / or block in the flowchart and / or block diagram, as well as the combination of processes and / or blocks 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 produce 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 1A device that provides the functions specified in a block or multiple blocks.

[0208] 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.

[0209] 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 The steps for the functions specified in one or more boxes.

[0210] The specific embodiments described above further illustrate the objectives, technical solutions and beneficial effects of the present invention in detail. It should be understood that the above description is only a specific embodiment of the present invention and is not intended to limit the scope of protection of the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the scope of protection of the present invention.

Claims

1. A transaction data processing method, characterized in that: include: Receiving multiple pending transactions; the pending transactions carry a counterparty identifier and a transaction amount; The probability of approval of each pending transaction is calculated using a preset probability calculation model; the preset probability calculation model is obtained by training a logistic regression model using the approval results and counterparty credit information of multiple historical transactions in advance; the approval result includes whether the approval is passed and the approval time point; The estimated cumulative value of each counterparty is calculated based on the transaction amount and approval probability corresponding to the same counterparty ID; the estimated cumulative value reflects the total transaction amount expected to occur by the counterparty; Determine the incurred value of each counterparty based on the transactions obtained from the real incurred database; The value that has occurred reflects the total amount of transactions that have occurred cumulatively with the counterparty; the real transaction database stores transactions that have occurred; Receiving a first approval result of a first transaction to be approved; The first transaction to be approved carries an identifier of the first transaction counterparty; Modify the approval probability of the first pending transaction according to the first approval result, and update the estimated cumulative value of the first transaction counterparty; wherein, when the first approval result is approval, updating the estimated cumulative value of the first transaction counterparty includes: deleting pending transactions with a lower priority than the first pending transaction from the pending transactions of the first transaction counterparty, and updating the estimated cumulative value of the first transaction counterparty according to the transaction amount and transaction occurrence probability of all pending transactions of the first transaction counterparty after deletion; the priority reflects the business importance and approval time of the pending transaction; Determine whether the first transaction to be approved triggers a major transaction based on the updated estimated cumulative value of the first transaction counterparty, the incurred value of the first transaction counterparty and the preset cumulative threshold; the preset cumulative threshold is used to determine whether the total amount of all transactions of the transaction counterparty triggers a major transaction.

2. The method according to claim 1, characterized in that After receiving multiple pending transactions, it also includes: Through the streaming architecture, the credit information and transaction data of the counterparty are obtained in real time; the credit information includes one or any combination of the following: enterprise registration information, enterprise operation information, enterprise compliance supervision information, enterprise public opinion and market evaluation information, natural person credit information, natural person public affairs payment information, natural person tax information, natural person public opinion information; Store multiple pending transactions and real-time acquired credit information and transaction data in the expected change library; the pending and under-approval transactions are stored in the expected change library in the form of a graph structure, in which the pending transaction identifier and counterparty identifier are nodes, and the transaction amount, time information, credit information and transaction data are node attributes.

3. The method according to claim 2, characterized in that The preset probability calculation model is used to calculate the approval probability of each pending transaction, including: Obtain node information and node attribute information of pending transactions from the expected change library; The node information, node attribute information and the preset probability calculation model are used to calculate the approval probability of each transaction to be approved.

4. The method according to claim 3, characterized in that Using node information, node attribute information and the preset probability calculation model, the approval probability of each pending transaction is calculated, including: Perform data preprocessing on node information and node attribute information to obtain preprocessed data; data preprocessing includes natural language processing, data cleaning, and data conversion; The preprocessed data is input into the following preset probability calculation model to output the approval probability of each pending transaction: wherein the preset probability calculation model uses a binary cross entropy loss function; In the formula, p represents the probability of approval, ψ represents the probability calculation function, b represents the sample deviation, and a n is the weight coefficient, x n is the nth preprocessed data, and θ is the influencing factor, which is obtained based on the public opinion information related to the counterparty.

5. The method according to claim 2, characterized in that Based on the transaction amount and approval probability corresponding to the same counterparty ID, the estimated cumulative value of each counterparty is calculated, including: The priority of pending transactions of the same counterparty is calculated by using a preset linear regression model, and the pending transactions of the same counterparty are sorted by priority to obtain a sequence of pending transactions; the counterparty identifiers of any two pending transaction sequences are different; the preset linear regression model is pre-trained using the transaction approval information, credit information and transaction data of multiple historical counterparties; the transaction approval information includes the time point of initiating approval, whether the approval is passed, and the time point of approval; Based on the transaction amount, approval probability and priority order of the pending transactions, the estimated cumulative value of each counterparty is calculated using the following formula: A 变 =P1×A1+P2×A2+P3×A3+......+P n ×A n ; In the formula, A 变 Indicates the expected cumulative value, P n A represents the probability of approval of the nth pending transaction. n Indicates the transaction amount of the nth pending transaction. The priority of the n-1th pending transaction is higher than that of the nth pending transaction.

6. The method according to claim 5, characterized in that The preset linear regression model is used to calculate the priority of pending transactions with the same counterparty, including: Obtain node information and node attribute information of pending transactions of the same counterparty from the expected change library; The priority of pending transactions of the same counterparty is calculated using the node information, node attribute information and the preset linear regression model of pending transactions of the same counterparty according to the following formula: y'=(X×β+∈)×θ; Where y' represents the priority; X is the input data; β is the model parameter matrix, which is obtained by least squares fitting; ∈ is the random error term; θ is the influencing factor, which is obtained based on the public opinion information related to the counterparty.

7. The method according to claim 1, characterized in that Determine the incurred value of each counterparty based on the transactions obtained from the real incurred database, including: The incurred value of each counterparty is determined based on the transactions obtained from the real incurred database according to the following formula: B 历 =k1×B1+k2×B2+k3×B3+......+k n ×B n ; In the formula, B 历 Indicates the value that has occurred, k n Indicates the preset weight coefficient of the nth approved transaction, B n Indicates the transaction amount that has occurred.

8. The method according to claim 2, characterized in that Determine whether the first pending transaction triggers a major transaction based on the updated estimated cumulative value of the first transaction counterparty, the incurred value of the first transaction counterparty, and the preset cumulative threshold, including: comparing the sum of the updated estimated cumulative value of the first counterparty and the incurred value of the first counterparty to a preset cumulative threshold; When the sum is greater than the preset cumulative threshold, the first pending transaction is determined to trigger a major transaction; When the sum is not greater than the preset cumulative threshold, it is determined that the first transaction to be approved does not trigger a major transaction.

9. The method according to claim 8, characterized in that After determining that the first pending transaction triggers a major transaction, it also includes: The first transaction to be approved is pushed to the advanced approval process for approval. The advanced approval process is used to approve major transactions.

10. The method according to claim 8, characterized in that After determining that the first pending transaction does not trigger a major transaction, the following also applies: Deleting the node corresponding to the first pending transaction from the estimated change library, and updating the estimated cumulative value of the first transaction counterparty; The first pending transaction is stored as a real transaction in the real transaction database, and the transaction value of the first transaction counterparty is updated.

11. The method according to claim 5, characterized in that After determining the value of each counterparty's transactions based on the transactions obtained from the real transaction database, it also includes: The time point of approval of each pending transaction is predicted by using the trained time prediction model; the time prediction model is obtained by training the autoregressive integrated moving average ARIMA model using historical transaction approval information, counterparty credit information and transaction data in advance; Set the approval probability of each pending transaction to approved; According to the order of approval time points of all pending transactions, the following are calculated for each counterparty at the same approval time point: the expected cumulative value and the incurred value; Based on the preset cumulative threshold and the estimated cumulative value and the actual value at the same approval time point for each counterparty, the time point at which each counterparty reaches the preset cumulative threshold is output.

12. The method according to claim 11, characterized in that Outputs the time point at which each counterparty reaches the preset cumulative threshold, and also includes: Statistics on the actual approval results and actual approval time of each pending transaction, and obtain the latest counterparty credit information and transaction data; Based on the principal component analysis algorithm, the model parameters of the time prediction model, the preset linear regression model, and the preset probability calculation model are updated using the actual approval results of each pending transaction, the actual approval time information, the latest counterparty's credit information and transaction data.

13. A transaction data processing device, characterized in that: include: A data receiving module, used to receive multiple pending transactions; the pending transactions carry a counterparty identifier and a transaction amount; The approval probability prediction module is used to calculate the approval probability of each transaction to be approved by using a preset probability calculation model; the preset probability calculation model is obtained by training a logistic regression model in advance using the approval results and counterparty credit information of multiple historical transactions; the approval result includes whether the approval is passed and the approval time point; The estimated cumulative value calculation module is used to calculate the estimated cumulative value of each transaction counterparty based on the transaction amount and approval probability corresponding to the same transaction counterparty identifier; the estimated cumulative value reflects the total transaction amount expected to occur by the transaction counterparty; The incurred value calculation module is used to determine the incurred value of each counterparty based on the transactions obtained from the real incurred database; The value that has occurred reflects the total amount of transactions that have occurred cumulatively with the counterparty; the real transaction database stores transactions that have occurred; An estimated accumulated value updating module, configured to receive a first approval result of a first transaction to be approved; The first pending transaction carries the identifier of the first transaction counterparty; the approval probability of the first pending transaction is modified according to the first approval result, and the estimated cumulative value of the first transaction counterparty is updated; wherein, when the first approval result is approved, updating the estimated cumulative value of the first transaction counterparty includes: deleting pending transactions with a lower priority than the first pending transaction from the pending transactions of the first transaction counterparty, and updating the estimated cumulative value of the first transaction counterparty according to the transaction amount and transaction occurrence probability of all pending transactions of the first transaction counterparty after deletion; the priority reflects the business importance and approval time of the pending transaction; The trigger judgment module is used to determine whether the first pending transaction triggers a major transaction based on the updated estimated cumulative value of the first transaction counterparty, the incurred value of the first transaction counterparty and the preset cumulative threshold; the preset cumulative threshold is used to determine whether the total amount of all transactions of the transaction counterparty triggers a major transaction.

14. 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 12 is implemented.

15. 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 12 is implemented.

16. 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 12 is implemented.

Images