Transaction data processing method and device, storage medium and electronic equipment

Abstract

The invention discloses a transaction data processing method and device, a storage medium and electronic equipment, and relates to the field of artificial intelligence. The method comprises the steps of obtaining transaction data of a target merchant and attribute information of the target merchant; screening at least one feature value from the transaction data of the target merchant and the attribute information of the target merchant according to the target feature sequence; inputting the screened feature value into a target prediction model, converting the feature value into a target feature vector according to priori knowledge learned by the target prediction model in a model training stage, and determining a target proportionality coefficient for commission extraction based on transaction data and attribute information of the target merchant according to the target feature vector; and calculating a difference value between a preset proportionality coefficient set by the user and the target proportionality coefficient to obtain a processing result of the transaction data of the target merchant. The technical problem of low analysis accuracy caused by manual analysis of merchant transaction data in the prior art is solved.

Description

Technical Field

[0001] The present application relates to the field of artificial intelligence, and more specifically, to a method, device, storage medium and electronic device for processing transaction data. Background Art

[0002] With the rapid development of e-commerce, massive amounts of merchant transaction data and multi-dimensional market information have provided unprecedented opportunities for business analysis and intelligent decision-making. However, in the face of this data torrent, traditional methods for processing merchant transaction data have exposed serious limitations, becoming a bottleneck restricting business precision and intelligent upgrades.

[0003] Taking the merchant rebate rate prediction as an example, traditional methods often rely on manual experience analysis and manual data processing. This process is not only cumbersome and inefficient, but also very easy to introduce human errors in the data processing process, resulting in significant deviations between the analysis results and the actual rebate rate. More importantly, the traditional prediction model lacks sufficient flexibility and intelligence, and it is difficult to capture the dynamic changes and complex correlations of merchant market behavior. Therefore, there are significant deficiencies in analysis accuracy and it is unable to meet the rapid response and refined management needs of merchants.

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

[0005] The main purpose of the present application is to provide a transaction data processing method, device, storage medium and electronic device to at least solve the technical problem in the prior art of low analysis accuracy based on manual analysis of merchant transaction data.

[0006] In order to achieve the above-mentioned purpose, according to one aspect of the present application, a method for processing transaction data is provided, including: obtaining the transaction data of a target merchant and the attribute information of the target merchant; screening at least one feature value from the transaction data of the target merchant and the attribute information of the target merchant according to a target feature sequence, wherein the target feature sequence is a feature parameter determined according to pre-training; inputting the screened feature value into a target prediction model, converting the feature value into a target feature vector according to the prior knowledge learned by the target prediction model in the model training stage, and determining a target proportional coefficient for commission extraction based on the transaction data and attribute information of the target merchant according to the target feature vector, wherein the target prediction model is a model obtained by updating the initial prediction model through a target parameter combination, wherein the target parameter combination is a model parameter combination that meets a preset condition, wherein the preset condition is used to constrain the difference between the prediction result and the actual result of the initial prediction model to be less than a first preset threshold; calculating the difference between the preset proportional coefficient set by the user and the target proportional coefficient to obtain the processing result of the transaction data of the target merchant.

[0007] Optionally, the target prediction model is obtained through the following steps: obtaining the merchant's historical target information from a database, wherein the historical target information includes the merchant's historical transaction data and the merchant's historical attribute information; performing feature screening on the merchant's historical transaction data and the merchant's historical attribute information to obtain a target feature sequence consisting of X historical transaction features and Y historical attribute features, wherein X and Y are both integers greater than or equal to 1, and the feature screening is used to digitize the merchant's historical transaction data and historical attribute information, and continuously iteratively screen the digitized data to select features that meet the screening conditions, wherein the screening conditions are used to select features whose model prediction errors are less than a second preset threshold; iteratively training the first prediction model according to the target feature sequence and the historical target information to obtain the target prediction model.

[0008] Optionally, feature screening is performed on the merchant's historical transaction data and the merchant's historical attribute information to obtain a target feature sequence consisting of X historical transaction features and Y historical attribute features, including: when the historical transaction data and historical attribute information of N merchants are included in the historical target information, the historical transaction data and historical attribute information belonging to the same merchant are divided into a data set to obtain N data sets, wherein N is an integer greater than or equal to 1; a feature sequence is extracted from each data set in the N data sets to obtain N feature sequences; the N feature sequences are numerically converted to obtain N first feature sequences, wherein each first feature sequence includes an initial value and an initial weight, wherein the initial value is used to characterize whether each feature in the first feature sequence is selected; multiple target operations are performed on the initial prediction model according to the N first feature sequences until the number of target operations is greater than or equal to the first iteration number, thereby obtaining a target feature sequence consisting of X historical transaction data and Y historical attribute information, wherein each target operation is used to train the initial prediction model according to the prediction result of the initial prediction model for each first feature sequence.

[0009] Optionally, the target operation includes the following steps: setting a first function, wherein the first function is used to calculate the difference value between the model prediction result and the actual result obtained according to each feature sequence; obtaining a second function and a third function, wherein the second function is used to calculate the feature selection state in each of the N first feature sequences, and the third function is used to calculate the weight of each of the N first feature sequences; inputting the N first feature sequences into the initial prediction model for training, and calculating the prediction error corresponding to each of the N first feature sequences according to the first function; updating the feature selection state and weight in each first feature sequence according to the prediction error corresponding to each of the N first feature sequences, the second function and the third function, to obtain N second feature sequences; obtaining a third feature sequence from the N second feature sequences, and using the third feature sequence as the target feature sequence, wherein the third feature sequence is a second feature sequence among the N second feature sequences whose prediction error is less than the corresponding preset threshold.

[0010] Optionally, the feature selection state and weight in each first feature sequence are updated according to the prediction error corresponding to each first feature sequence in the N first feature sequences, the second function and the third function, including: in the case of the first target operation, the prediction error corresponding to each first feature sequence in the N first feature sequences is used as the corresponding initial prediction error, and the weight of each first feature sequence is updated according to the third function; in the case of the i-th target operation, if the prediction error of the j-th first feature sequence is less than the prediction error of the j-th first feature sequence under the i-1-th target operation, the feature selection state in the j-th first feature sequence is updated according to the second function, and the weight of the j-th first feature sequence is updated according to the third function, wherein i is an integer greater than 1, and j is a positive integer less than or equal to N.

[0011] Optionally, before iteratively training the first prediction model according to the target feature sequence and historical target information to obtain the target prediction model, the method for processing transaction data also includes: setting a parameter set for the initial prediction model, wherein the parameter set includes S groups of parameter combinations, each group of parameter combinations includes L parameters, wherein S is an integer greater than or equal to 1, and L is an integer greater than or equal to 1; obtaining a target algorithm, wherein the target algorithm is used to search for a parameter combination that meets preset conditions in the parameter space; determining the target parameter combination from the parameter set according to the target algorithm; and determining the first prediction model according to the target parameter combination and the initial prediction model.

[0012] Optionally, the first prediction model is iteratively trained according to the target feature sequence and the historical target information to obtain the target prediction model, including: extracting X historical transaction features and Y historical attribute features from each feature sequence of N feature sequences to obtain N model training features; obtaining N commission ratio coefficients of N merchants corresponding to the N feature sequences; normalizing the N model training features to obtain the target features; and iteratively training the first prediction model according to the target features and the N commission ratio coefficients to obtain the target prediction model.

[0013] Optionally, the difference between a preset proportional coefficient and a target proportional coefficient set by a user is calculated to obtain a processing result of the transaction data of the target merchant. The method for processing the transaction data includes: when the difference between the preset proportional coefficient and the target proportional coefficient is less than a set threshold, calculating the target commission value corresponding to the target merchant according to the target proportional coefficient; setting a first interface, wherein the first interface is used to issue the commission value; and sending the target commission value to a target platform through the first interface, wherein the target platform is a platform associated with the target merchant.

[0014] Optionally, after sending the target commission value to the target platform through the first interface, the transaction data processing method also includes: setting a target detection indicator, and setting a corresponding alarm threshold for each detection indicator in the target detection indicator; detecting the transaction data and target commission value of the target merchant to obtain a detection result set, wherein the detection result set includes abnormal transaction data between the target merchant and the user and / or the target platform; when the detection indicator corresponding to the tth detection result in the detection result set is greater than or equal to the alarm threshold corresponding to the detection indicator, issuing an early warning for the tth detection result, wherein t is an integer greater than or equal to 1.

[0015] In order to achieve the above-mentioned purpose, according to another aspect of the present application, a transaction data processing device is also provided, including: an acquisition unit, which acquires the transaction data of a target merchant and the attribute information of the target merchant; a screening unit, which screens at least one feature value from the transaction data of the target merchant and the attribute information of the target merchant according to a target feature sequence, wherein the target feature sequence is a feature parameter determined according to pre-training; a determination unit, which inputs the screened feature value into a target prediction model, converts the feature value into a target feature vector according to the prior knowledge learned by the target prediction model in the model training stage, and determines a target proportional coefficient for commission extraction based on the transaction data and attribute information of the target merchant according to the target feature vector, wherein the target prediction model is a model obtained by updating the initial prediction model through a target parameter combination, wherein the target parameter combination is a model parameter combination that meets a preset condition, wherein the preset condition is used to constrain the difference between the prediction result and the actual result of the initial prediction model to be less than a first preset threshold value; a processing unit, which calculates the difference between the preset proportional coefficient set by the user and the target proportional coefficient, and obtains the processing result of the transaction data of the target merchant.

[0016] According to another aspect of an embodiment of the present application, a computer-readable storage medium is also provided, the computer-readable storage medium including a stored executable program, wherein when the executable program runs, the device where the computer-readable storage medium is located is controlled to execute the above-mentioned transaction data processing method.

[0017] According to another aspect of an embodiment of the present application, there is also provided an electronic device, comprising: a memory storing an executable program; and a processor for running the program, wherein the above-mentioned transaction data processing method is executed when the program is running.

[0018] According to another aspect of an embodiment of the present application, a computer program product is also provided, including computer instructions, which implement the steps of the above-mentioned transaction data processing method when executed by a processor.

[0019] In the present application, the transaction data and attribute information of the target merchant are first obtained, and then at least one feature value is screened from the transaction data and attribute information of the target merchant according to a target feature sequence, wherein the target feature sequence is a feature parameter determined according to pre-training, and then the screened feature value is input into the target prediction model, and the feature value is converted into a target feature vector according to the prior knowledge learned by the target prediction model in the model training stage, and the target proportional coefficient for commission extraction based on the transaction data and attribute information of the target merchant is determined according to the target feature vector, wherein the target prediction model is a model obtained by updating the initial prediction model through a target parameter combination, wherein the target parameter combination is a model parameter combination that meets a preset condition, wherein the preset condition is used to constrain the difference between the prediction result and the actual result of the initial prediction model to be less than a first preset threshold value, and finally the difference between the preset proportional coefficient set by the user and the target proportional coefficient is calculated to obtain the processing result of the transaction data of the target merchant. That is, through intelligent feature selection and model prediction, the target proportional coefficient is obtained, and the preset proportional coefficient set by the user is compared with the target proportional coefficient, so as to achieve the purpose of improving the matching degree between the prediction model and the actual situation of the merchant, thereby achieving the technical effect of improving the analysis accuracy, and further solving the technical problem of low analysis accuracy based on manual analysis of merchant transaction data in the existing technology. BRIEF DESCRIPTION OF THE DRAWINGS

[0020] The drawings constituting a part of the present application are used to provide a further understanding of the present application. The illustrative embodiments and descriptions of the present application are used to explain the present application and do not constitute an improper limitation on the present application. In the drawings:

[0021] Figure 1 A hardware structure block diagram of a computer terminal for implementing a method for processing transaction data is shown;

[0022] Figure 2 is a flowchart of an optional transaction data processing method according to an embodiment of the present application;

[0023] Figure 3 is a schematic diagram of an optional optimization prediction model according to an embodiment of the present application;

[0024] Figure 4 is a schematic diagram of an optional prediction model application according to an embodiment of the present application;

[0025] Figure 5 is a flowchart of an optional prediction model application according to an embodiment of the present application;

[0026] Figure 6 is a block diagram of an optional prediction model application according to an embodiment of the present application;

[0027] Figure 7 It is a schematic diagram of an optional transaction data processing device according to an embodiment of the present application. DETAILED DESCRIPTION

[0028] In order to enable those skilled in the art to better understand the solution of the present application, the technical solution in the embodiments of the present application will be clearly and completely described below in conjunction with the drawings in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without creative work should fall within the scope of protection of the present application.

[0029] It should be noted that the terms "first", "second", etc. in the specification and claims of the present application and the above-mentioned drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence. It should be understood that the data used in this way can be interchangeable where appropriate, so that the embodiments of the present application described herein can be implemented in an order other than those illustrated or described herein. In addition, the terms "including" and "having" and any of their variations are intended to cover non-exclusive inclusions, for example, a process, method, system, product or device comprising a series of steps or units is not necessarily limited to those steps or units clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or devices.

[0030] It should be noted that the collected information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data for display, data for analysis, etc.) involved in this application are information and data authorized by the user or fully authorized by all parties, and the collection, storage, use, processing, transmission, provision, disclosure and application of relevant data are in compliance with relevant laws, regulations and standards, necessary confidentiality measures are taken, and public order and good customs are not violated, and corresponding operation entrances are provided for users to choose to authorize or refuse. If an interface is set up between this system and relevant users or institutions, the corresponding operation entrance is provided for users to choose to agree or refuse the automated decision-making results; if the user chooses to refuse, the expert decision-making process will be entered.

[0031] According to an embodiment of the present application, an embodiment of a method for processing transaction data is provided. It should be noted that the steps shown in the flowchart of the accompanying drawings can be executed in a computer system such as a set of computer executable instructions, and although a logical order is shown in the flowchart, in some cases, the steps shown or described can be executed in an order different from that shown here.

[0032] It should be noted that a merchant data processing system can be used as the execution subject of the transaction data processing method of the embodiment of the present application. It is understandable that the transaction data processing method provided in the embodiment of the present application can also be executed by other systems or devices, and the embodiment of the present application does not specifically limit this.

[0033] The method embodiments provided in the present application may be executed in a mobile terminal, a computer terminal or a similar computing device. Figure 1 The hardware structure block diagram of a computer terminal (or mobile device) for implementing a method for processing transaction data is shown. Figure 1 As shown, the computer terminal 10 (or mobile device) may include one or more (102a, 102b, ..., 102n are used to illustrate) processors 102 (the processor 102 may include but is not limited to a processing device such as a microprocessor MCU or a programmable logic device FPGA), a memory 104 for storing data, and a transmission device 106 for communication functions. In addition, it may also include: a display, an input / output interface (I / O interface), a universal serial bus (USB) port (which may be included as one of the ports of the BUS bus), a network interface, a power supply and / or a camera. It can be understood by those skilled in the art that Figure 1 The structure shown is only for illustration and does not limit the structure of the above electronic device. Figure 1 More or fewer components as shown, or with Figure 1 Different configurations are shown.

[0034] It should be noted that the one or more processors 102 and / or other data processing circuits described above may generally be referred to herein as "data processing circuits". The data processing circuit may be embodied in whole or in part as software, hardware, firmware, or any other combination thereof. In addition, the data processing circuit may be a single independent processing module, or may be fully or partially integrated into any of the other components in the computer terminal 10 (or mobile device). As in the method for processing transaction data involved in the embodiments of the present application, the data processing circuit serves as a processor control (e.g., selection of a variable resistor terminal path connected to an interface).

[0035] The memory 104 can be used to store software programs and modules of application software, such as program instructions / data storage devices corresponding to the method for processing transaction data in the embodiment of the present application. The processor 102 executes various functional applications and data processing by running the software programs and modules stored in the memory 104, that is, realizing the above-mentioned method for processing transaction data. The memory 104 may include a high-speed random access memory, and may also include a non-volatile memory, such as one or more magnetic storage devices, flash memory, or other non-volatile solid-state memory. In some examples, the memory 104 may further include a memory remotely arranged relative to the processor 102, and these remote memories may be connected to the computer terminal 10 via a network. Examples of the above-mentioned network include, but are not limited to, the Internet, an intranet, a local area network, a mobile communication network, and combinations thereof.

[0036] The transmission device 106 is used to receive or send data via a network. The specific example of the above network may include a wireless network provided by a communication provider of the computer terminal 10. In one example, the transmission device 106 includes a network adapter (Network Interface Controller, NIC), which can be connected to other network devices through a base station so as to communicate with the Internet. In one example, the transmission device 106 can be a radio frequency (RF) module, which is used to communicate with the Internet wirelessly.

[0037] The display may be, for example, a touch screen liquid crystal display (LCD) that enables a user to interact with a user interface of the computer terminal 10 (or mobile device).

[0038] Under the above operating environment, this application provides Figure 2 The method of processing transaction data shown. Figure 2 FIG. 1 is a flowchart of an optional transaction data processing method according to an embodiment of the present application. Figure 2 As shown, the method comprises the following steps:

[0039] Step S201, acquiring the transaction data and attribute information of the target merchant.

[0040] Optionally, the merchant data processing system first obtains the transaction data of the target merchant and the attribute information of the target merchant.

[0041] Optionally, the transaction data includes but is not limited to transaction records, transaction amount, transaction time, transaction type, and rebate rate data.

[0042] Optionally, the attribute information includes but is not limited to industry category, geographical location, number of people, asset size, and annual revenue size.

[0043] Step S202: screening at least one feature value from the transaction data of the target merchant and the attribute information of the target merchant according to the target feature sequence.

[0044] In step S202, the target feature sequence is feature parameters determined according to pre-training.

[0045] Optionally, the target feature sequence refers to a set of features that are determined based on previous training and are of great value to the predicted proportional coefficient. The merchant data processing system selects feature values ​​that match the target feature sequence by analyzing the data of the target merchant for subsequent model input.

[0046] Step S203, input the filtered feature values ​​into the target prediction model, convert the feature values ​​into target feature vectors based on the prior knowledge learned by the target prediction model in the model training phase, and determine the target proportional coefficient for commission extraction based on the target merchant's transaction data and attribute information based on the target feature vector.

[0047] In step S203, the target prediction model is a model obtained by updating the initial prediction model through the target parameter combination.

[0048] In step S203, the target parameter combination is a model parameter combination that meets preset conditions.

[0049] In step S203, the preset condition is used to constrain the difference between the prediction result of the initial prediction model and the actual result to be less than a first preset threshold.

[0050] Optionally, the initial prediction model is an SVR model (Support Vector Regression, support vector regression model). The SVR model is a machine learning method based on support vector machine theory, which is mainly used to solve regression analysis problems.

[0051] Optionally, the merchant data processing system inputs the screened feature values ​​into a trained target prediction model, converts the feature values ​​into target feature vectors based on the prior knowledge learned by the target prediction model during the model training phase, and obtains the target proportional coefficient corresponding to the target merchant based on the target feature vector.

[0052] Step S204, calculating the difference between the preset proportional coefficient set by the user and the target proportional coefficient, and obtaining the processing result of the transaction data of the target merchant.

[0053] Optionally, the merchant data processing system can obtain a processing result, that is, the commission extraction ratio to be adjusted, by comparing the preset ratio coefficient with the target ratio coefficient.

[0054] It can be seen from the contents of steps S201 to S204 that in the present application, the transaction data and attribute information of the target merchant are first obtained, and then at least one feature value is screened from the transaction data and attribute information of the target merchant according to the target feature sequence, wherein the target feature sequence is a feature parameter determined according to pre-training, and then the screened feature value is input into the target prediction model, and the feature value is converted into a target feature vector according to the prior knowledge learned by the target prediction model in the model training stage, and the target proportional coefficient for commission extraction based on the transaction data and attribute information of the target merchant is determined according to the target feature vector, wherein the target prediction model is a model obtained by updating the initial prediction model through a target parameter combination, wherein the target parameter combination is a model parameter combination that meets a preset condition, wherein the preset condition is used to constrain the difference between the prediction result and the actual result of the initial prediction model to be less than a first preset threshold value, and finally the difference between the preset proportional coefficient set by the user and the target proportional coefficient is calculated to obtain the processing result of the transaction data of the target merchant. That is, through intelligent feature selection and model prediction, the target proportional coefficient is obtained, and the preset proportional coefficient set by the user is compared with the target proportional coefficient, so as to achieve the purpose of improving the matching degree between the prediction model and the actual situation of the merchant, thereby achieving the technical effect of improving the analysis accuracy, and further solving the technical problem of low analysis accuracy based on manual analysis of merchant transaction data in the existing technology.

[0055] In an optional embodiment, the merchant data processing system obtains the merchant's historical target information from a database, wherein the historical target information includes the merchant's historical transaction data and the merchant's historical attribute information, and then performs feature screening on the merchant's historical transaction data and the merchant's historical attribute information to obtain a target feature sequence consisting of X historical transaction features and Y historical attribute features, wherein X and Y are integers greater than or equal to 1, and the feature screening is used to digitize the merchant's historical transaction data and historical attribute information, and the digitized data is continuously iteratively screened to select features that meet the screening conditions, wherein the screening conditions are used to select features whose model prediction errors are less than a second preset threshold, and then iteratively train the first prediction model according to the target feature sequence and the historical target information to obtain a target prediction model. The target feature sequence consists of historical transaction data and Y historical attribute information, wherein each target operation is used to train the initial prediction model according to the prediction results of the initial prediction model for each first feature sequence.

[0056] Optionally, the merchant data processing system will obtain the basic information of the merchant and establish a comprehensive merchant profile to provide basic information support for the entire system. The merchant data processing system will include the merchant's contact information, such as address, contact number, email, etc., into the database to ensure that the system can communicate with the merchant quickly and accurately, and at the same time obtain the merchant's transaction records, covering key data such as transaction amount, sales volume, and commodity types. This initiative aims to gain an in-depth understanding of the merchant's sales status, thereby providing sufficient historical transaction data support for the prediction and processing of the rebate rate. At the same time, the timestamp of the transaction record is recorded to conduct time series analysis to discover sales trends and seasonal changes. The system also collects customer evaluations, comments, and feedback information. Based on this information, the system can objectively evaluate the merchant's reputation and user satisfaction in the market. These data will provide the system with a more comprehensive merchant portrait, which will help intelligently analyze the rebate rate and provide merchants with more accurate service recommendations. Finally, these data are integrated and mined to form a database.

[0057] Optionally, the database may include multiple tables, such as a customer information table, a customer history data table, a rebate rate table, etc., for storing and managing various types of data.

[0058] Optionally, the merchant data processing system obtains the merchant's historical target information from the database, and the historical target information includes but is not limited to industry category, geographic location, number of people, asset size, annual revenue size, transaction records, transaction amount, transaction time, transaction type, and commission rate data.

[0059] Optionally, the merchant data processing system performs feature screening on the acquired merchant's historical target information, aiming to identify features that have a significant impact on the prediction model from the historical target information, convert non-numerical features into numerical form that can be processed by the model through digitization, and then screen out features whose prediction errors meet threshold requirements through iterative training.

[0060] Optionally, feature screening can ensure a prediction result with a high accuracy by finding a subset as small as possible among multiple features and then learning through the features in this subset.

[0061] Optionally, the merchant data processing system obtains a target feature sequence consisting of X historical transaction features and Y historical attribute features through feature screening.

[0062] Optionally, the merchant data processing system iteratively trains the SVR model after parameter optimization according to the target feature sequence and historical target information, and finally obtains the target prediction model.

[0063] From the above content, it can be seen that the merchant data processing system digitizes the merchant's historical transaction data and attribute information, and through iterative screening, the system can select the features that have the greatest impact on the rebate rate prediction and form a target feature sequence. This process ensures the quality of input data in the model training phase, thereby improving the prediction accuracy of the first prediction model. In each target operation, the system iteratively trains the prediction results of each first feature sequence based on the initial prediction model, which not only optimizes the model parameters, but also enhances the model's ability to process unseen data. In addition, through automated feature screening and iterative training, it reduces reliance on manual analysis and decision-making, and improves the system's degree of automation. This not only reduces the errors and inconsistencies that may be caused by manual operations, but also significantly improves the efficiency and speed of data processing.

[0064] In an optional embodiment, when the merchant data processing system includes historical transaction data and historical attribute information of N merchants in the historical target information, the historical transaction data and historical attribute information belonging to the same merchant are divided into a data set to obtain N data sets, wherein N is an integer greater than or equal to 1, a feature sequence is extracted from each of the N data sets to obtain N feature sequences, and the N feature sequences are numerically converted to obtain N first feature sequences, wherein each first feature sequence includes an initial value and an initial weight, wherein the initial value is used to characterize whether each feature in the first feature sequence is selected, and finally, multiple target operations are performed on the initial prediction model according to the N first feature sequences until the number of target operations is greater than or equal to the first iteration number, thereby obtaining a target feature sequence consisting of X historical transaction data and Y historical attribute information, wherein each target operation is used to train the initial prediction model according to the prediction result of the initial prediction model for each first feature sequence.

[0065] Optionally, the merchant data processing system processes the data of N merchants separately, and integrates the transaction data and attribute information of each merchant into an independent data set, thereby forming N data sets. This approach ensures that the characteristics of each merchant can be considered independently, and enhances the personalized prediction ability of the model. Then, for each data set, the system performs a feature extraction operation to generate a feature sequence. The feature sequence here includes transaction data features (such as transaction amount, transaction frequency, etc.) and attribute information features (such as the industry, geographical location, scale, etc. of the merchant) related to the merchant rebate rate prediction. Then, the extracted N feature sequences are converted into numerical values ​​to obtain N first feature sequences, and each first feature sequence is The feature sequence contains not only the initial value of the feature (used to characterize whether the feature is selected), but also the initial weight of the feature. These two parameters are the basis for the subsequent model to extract features. Then, the initial SVR model is iteratively trained using the N first feature sequences, i.e., the "target operation". In each target operation, the model makes predictions based on the current feature sequence, and then adjusts the model based on the difference between the prediction result and the actual rebate rate until the number of iterations reaches the preset first number of iterations. Finally, after multiple iterations of training, the system determines a set of target feature sequences consisting of X historical transaction data features and Y historical attribute information features, which represent the optimal feature set involved in the prediction in the final model.

[0066] Optionally, the merchant data processing system may perform numerical processing on the features through a Sigmoid function (Sigmoid Function, S-type function), and the Sigmoid function may convert any input value in the real number domain into an output value between (0, 1).

[0067] Optionally, Table 1 is an example of a feature combination in an optional target feature sequence according to an embodiment of the present application. As shown in Table 1, the features included in the target feature sequence include the merchant's industry, geographical location, number of people, asset scale (10,000), and annual revenue scale (10,000). The input feature vector obtained according to the target feature sequence is: [(0,001,500,1000,500), (1,002,600,800,300)...].

[0068] Table 1

[0069]

[0070] From the above content, it can be seen that the merchant data processing system not only improves the model's prediction ability on a single merchant data through multiple target operations, but also enhances the model's generalization ability by processing N data sets, enabling it to more accurately predict merchant rebate rates with different attributes and historical records, providing more reliable data support for merchants and platform decisions, while realizing dynamic screening and optimization of features to ensure that the final model only contains the features that contribute most to the prediction results, thereby improving the accuracy of the prediction and the efficiency of the model. The target feature sequence finally obtained can significantly improve the accuracy of the prediction model and reduce the prediction error.

[0071] In an optional embodiment, the merchant data processing system first sets a first function, wherein the first function is used to calculate the difference value between the model prediction result and the actual result obtained according to each feature sequence, and then obtains a second function and a third function, wherein the second function is used to calculate the feature selection state in each of the N first feature sequences, and the third function is used to calculate the weight of each of the N first feature sequences, and then the N first feature sequences are input into the initial prediction model for training, and the prediction error corresponding to each of the N first feature sequences is calculated according to the first function, and then the feature selection state and weight in each first feature sequence are updated according to the prediction error corresponding to each of the N first feature sequences, the second function and the third function to obtain N second feature sequences, and finally the third feature sequence is obtained from the N second feature sequences, and the third feature sequence is used as the target feature sequence, wherein the third feature sequence is a second feature sequence in the N second feature sequences whose prediction error is less than the corresponding preset threshold.

[0072] Optionally, the first function may be mean square error, mean absolute error, root mean square error, etc., and the goal is to minimize these error indicators.

[0073] Optionally, the merchant data processing system extracts and numerically processes N first feature sequences from N merchant data and inputs them into the initial prediction model for training. During the model training process, the prediction error corresponding to each feature sequence is calculated according to the first function. This is a key indicator for evaluating the model training effect and making the next adjustment. Based on the calculated prediction error, the second function (feature selection status) and the third function (feature weight), the system updates the feature selection status and weight in each first feature sequence to generate N second feature sequences. Finally, from the N second feature sequences, those feature sequences whose prediction errors are less than a preset threshold corresponding to the target number of operations are screened out to form a third feature sequence. This third feature sequence is the optimal set of feature sequences under the current target operation.

[0074] From the above content, it can be seen that the merchant data processing system can obtain the optimal feature sequence combination corresponding to the current target operation through each target operation, and finally the final target feature sequence can be screened out through multiple groups of optimal feature sequence combinations. The determined target feature sequence can significantly improve the accuracy of the prediction model.

[0075] In an optional embodiment, in the case of the first target operation, the merchant data processing system takes the prediction error corresponding to each first feature sequence in N first feature sequences as the corresponding initial prediction error, and updates the weight of each first feature sequence according to a third function. In the case of the i-th target operation, if the prediction error of the j-th first feature sequence is less than the prediction error of the j-th first feature sequence under the i-1-th target operation, the feature selection state in the j-th first feature sequence is updated according to the second function, and the weight of the j-th first feature sequence is updated according to the third function, wherein i is an integer greater than 1, and j is a positive integer less than or equal to N.

[0076] Optionally, in the first target operation (i.e., the initial stage of model training), the merchant data processing system uses the prediction error corresponding to each sequence in the N first feature sequences (i.e., the feature sequences after digitization and weight initialization) as the initial prediction error. This step provides a benchmark for subsequent weight and feature state updates. Similarly, under the first target operation, the merchant data processing system will update the weight of each first feature sequence according to the third function. The weight update reflects the relative contribution of the feature to the model prediction results, which helps the model pay more attention to important features in subsequent iterations.

[0077] Optionally, in the i-th target operation (where i is an integer greater than 1), the system checks the prediction error of the j-th first feature sequence (j is a positive integer less than or equal to N). If the prediction error of the j-th feature sequence is smaller in the i-th time than in the i-1-th time, the feature selection status in the sequence is updated according to the second function, which usually means increasing or adjusting the numerical representation of the currently selected feature to further optimize the prediction performance of the model. At the same time, the weight of the j-th feature sequence is updated again according to the third function to continuously optimize the contribution of the feature to the model prediction.

[0078] Optionally, the essence of the feature screening problem is to select a suitable 0 / 1 string, the length of the string is the number of features N in the original data set, 0 is an unselected attribute, and 1 is a selected attribute. The merchant data processing system can use the PSO algorithm (Particle Swarm Optimization) to complete the feature screening operation. The idea of ​​the PSO algorithm is: assume that there are some particles in the space that are randomly distributed in different spatial positions, and each particle has two attributes: position and speed. The initial values ​​of these two attributes are random. Each particle maintains its own optimal point pbest (optimal refers to the position value when the fitness function is optimal), and the entire particle group maintains a global optimal point gbest (that is, the one with the largest fitness function corresponding to the pbest of all particles), and then iterates. The iterative process can be described as the particle changing its speed in a certain proportion in the direction of its own optimal pbest and the direction of the global optimal gbest, and then changing the position reached by the particle in this iteration. The global optimal solution gbest (the position value when the fitness function is optimal) at the end of the iteration is the solution obtained by the algorithm, where the fitness function can be defined as the size of the prediction error. When the prediction error takes the minimum value, the optimal solution is obtained. In the PSO algorithm, the calculation methods for updating the position (that is, the selection state of the above-mentioned feature) and updating the speed (that is, the above-mentioned weight) are shown in formula (1) and formula (2) respectively:

[0079]

[0080]

[0081] Among them, x is the position of the particle in the PSO algorithm, i represents the number or sequence of the particle in the particle swarm, t represents the number of iterations, v is the velocity of the particle in the PSO algorithm, w, c1, c2 are all weight parameters, rand is a random number [0, 1], pbest is the optimal position of each particle itself, and gbest is the position of the best particle among all particles.

[0082] It should be noted that the speed and position of each particle here are N-dimensional vectors, and since the PSO algorithm is an algorithm for solving continuous space problems, in order to find the corresponding optimal 0 / 1 string, it is necessary to convert the N-dimensional position vector of the continuous space into a 0 / 1 string of the discrete space (the distribution of 0 and 1 in the high-dimensional space). Because the Sigmoid function can be used for conversion, the obtained x is a 0 / 1 string, and the feature that is 1 in gbest is the feature to be finally selected. At this time, the formula for updating the particle position is shown in formula (3), and the calculation method of the Sigmoid function is shown in formula (4):

[0083]

[0084] Among them, i represents the number or sequence of the particle in the particle group, j represents the dimension number in the particle position and velocity vector, and otherwise represents other situations, namely Sigmoid (v ij ) is greater than or equal to rand.

[0085] From the above content, it can be seen that the merchant data processing system dynamically adjusts the feature selection status and weights, so that the model can find the target feature sequence, improve the accuracy of the prediction, and provide a basis for the subsequent construction of the target prediction model.

[0086] In an optional embodiment, the merchant data processing system first sets a parameter set of an initial prediction model, wherein the parameter set includes S groups of parameter combinations, each parameter combination includes L parameters, wherein S is an integer greater than or equal to 1, and L is an integer greater than or equal to 1, and then obtains a target algorithm, wherein the target algorithm is used to search for a parameter combination that meets preset conditions in the parameter space, and then determines the target parameter combination from the parameter set according to the target algorithm, and finally determines the first prediction model according to the target parameter combination and the initial prediction model.

[0087] Optionally, the merchant data processing system sets a set of initial parameter sets, which includes S groups of different parameter combinations, each group of combinations consists of L parameters, which may include learning rate, regularization term coefficient, SVR epsilon (insensitive bandwidth of loss function, defining the width of an error tolerance or insensitive zone), C (penalty coefficient, used to control the trade-off between the generalization ability of the model and the training error) and gamma (kernel function coefficient, controlling the nonlinear degree of data mapping to high-dimensional space, thereby affecting the complexity of the model), etc., forming the hyperparameter space of the prediction model, and then obtaining the target algorithm, i.e., the PSO algorithm, to search for those parameters that can meet the preset requirements in the parameter space of the model. Parameter combination of performance conditions. In the embodiment of the present application, the PSO algorithm simulates the search behavior of particles in multidimensional space to find the optimal parameter combination and improve the accuracy of model prediction. According to the PSO algorithm, the system iteratively searches for the best parameter combination from the parameter set. In each iteration, the algorithm evaluates the performance of each parameter combination according to the fitness function (such as minimizing prediction error, maximizing classification accuracy, etc.), and adjusts the position of the particles (i.e., parameter combination) to find a better solution. Finally, the PSO algorithm will determine a set of target parameter combinations that perform best under preset performance indicators. The system uses the determined target parameter combination to update or construct the initial prediction model, thereby obtaining a first prediction model.

[0088] Optionally, Figure 3 is a schematic diagram of an optional optimization prediction model according to an embodiment of the present application, such as Figure 3As shown, the merchant data processing system first obtains relevant data of the customer, including but not limited to transaction records, merchant attributes and market feedback, then extracts the characteristics of the customer from the obtained data, performs feature selection through the PSO algorithm, identifies the key features that have the greatest influence on the prediction of the commission ratio coefficient, and constructs a data matrix based on these selected features. Then, it is determined whether the preset termination conditions are met, such as the number of iterations reaching an upper limit or the prediction accuracy being stable at a certain level. If so, the optimal feature sequence is output, and the corresponding feature matrix is ​​constructed. At the same time, the parameters of the initial prediction model are optimized. Finally, the constructed feature matrix is ​​input into the initial prediction model after parameter optimization for prediction, and the prediction result is output. If the preset termination conditions are not met, the optimal feature selection state in each feature sequence (the optimal individual particle) and the optimal feature sequence in all feature combinations (the global optimal of the entire group) are obtained according to the principle of the PSO algorithm, and the feature selection state in each feature sequence (particle position update) and the speed of each feature sequence (particle speed update, i.e., the above-mentioned weight) are updated, and iteration is performed again until the termination conditions are met.

[0089] From the above content, it can be seen that the merchant data processing system optimizes the parameters of the initial prediction model through the target algorithm, ensuring that the prediction model can run with the most ideal parameter configuration, making the prediction of the merchant rebate ratio coefficient more accurate, and effectively improving the accuracy and efficiency of the prediction model.

[0090] In an optional embodiment, the merchant data processing system extracts X historical transaction features and Y historical attribute features from each of the N feature sequences to obtain N model training features, and then obtains N commission ratio coefficients of the N merchants corresponding to the N feature sequences, and then normalizes the N model training features to obtain target features, and then iteratively trains the first prediction model based on the target features and the N commission ratio coefficients to obtain a target prediction model.

[0091] Optionally, the merchant data processing system extracts X historical transaction features and Y historical attribute features in each sequence from the N feature sequences to form N model training features, and then further obtains the commission ratio coefficients of the N merchants corresponding to the N feature sequences. In order to eliminate the dimensional differences and numerical range influences between the features, the system normalizes the N model training features, converts all feature values ​​to the same numerical range (such as between 0 and 1), and obtains the target features. Finally, using the processed target features and the merchant's commission ratio coefficient, the system iteratively trains the first prediction model to obtain the target prediction model.

[0092] Optionally, Figure 4 is a schematic diagram of an optional prediction model application according to an embodiment of the present application, such as Figure 4 As shown, after obtaining the target prediction model, when predicting the rebate rate for the merchant, the relevant information of the merchant (industry, geographical location, number of people, asset scale, annual revenue scale, etc.) is obtained, and then the obtained data is directly input into the target prediction model to obtain the corresponding rebate rate of the merchant.

[0093] From the above content, it can be seen that the merchant data processing system can generate high-quality model training features by accurately extracting historical transaction features and historical attribute features, and combining them with merchant-specific commission ratio coefficients, which helps to improve the model's prediction accuracy and generalization ability.

[0094] In an optional embodiment, when the difference between the preset proportional coefficient and the target proportional coefficient is less than a set threshold, the merchant data processing system calculates the target commission value corresponding to the target merchant according to the target proportional coefficient, then sets a first interface, wherein the first interface is used to issue the commission value, and finally sends the target commission value to the target platform through the first interface, wherein the target platform is a platform associated with the target merchant.

[0095] Optionally, after obtaining the target proportional coefficient, the merchant data processing will calculate the difference between the preset proportional coefficient and the target proportional coefficient. When the difference is less than a set threshold, the specific commission value that the target merchant should pay is determined based on the target proportional coefficient, and the commission is automatically paid to the platform that has a business cooperation relationship with the target merchant through the first interface set by the system.

[0096] Optionally, merchant data processing monitors merchant transaction activities in real time. After each merchant transaction occurs, the prediction model is used to calculate the corresponding commission extraction ratio coefficient to ensure that the calculation process is efficient and accurate. The target prediction model can generate a commission data set containing merchant ID, transaction ID, commission rate and other information. This data set will be used for integration with the payment system or financial system. Merchant data processing can be achieved through interface calls, file transfers, message queues, etc. to communicate with the payment system or financial system, while ensuring that the interface is secure, reliable and efficient. After the target prediction model outputs the result, the commission data is passed to the payment system or financial system, and the automatic issuance mechanism is triggered. After the payment system or financial system receives the commission data, it performs corresponding payment processing through the first interface based on the calculated commission rate information. This step may include generating payment orders, transfer operations or other payment methods, depending on the agreement and settlement method between the merchant and the platform.

[0097] From the above content, it can be seen that the merchant data processing system has achieved automation and precision in the determination of the commission extraction ratio and the calculation of the commission value by introducing the target prediction model, which greatly reduces the need for manual intervention, reduces the error rate, and improves the satisfaction of the cooperation between merchants and the platform. In addition, through the setting of the first interface and the implementation of the automated commission payment process, the efficiency of commission management has been significantly improved, the delays and inconveniences of traditional manual operations have been avoided, and the trust and cooperation between merchants and the platform have been enhanced. At the same time, the calculation process of the commission extraction ratio and the specific commission value based on model prediction is transparent and data-based, ensuring the fairness and transparency of commission management.

[0098] In an optional embodiment, the merchant data processing system sets a target detection indicator, and sets a corresponding alarm threshold for each of the target detection indicators, detects the transaction data and target commission value of the target merchant, and obtains a detection result set, wherein the detection result set includes abnormal transaction data between the target merchant and the user and / or the target platform, and when the detection indicator corresponding to the tth detection result in the detection result set is greater than or equal to the alarm threshold corresponding to the detection indicator, an early warning is issued for the tth detection result, wherein t is an integer greater than or equal to 1.

[0099] Optionally, the merchant data processing system uses big data analysis tools to collect merchants' rebate rate data, including transaction amount, transaction time, transaction type, rebate rate, etc., and then performs data cleaning, processes missing values ​​and outliers, ensures data quality, and then extracts key features, such as transaction frequency, distribution of transaction amounts, changes in rebate rates in different time periods, etc., and combines other relevant data, such as the merchant's industry category, city, geographical location, etc., based on these data using time series analysis and other methods to explore the trend and periodicity of the rebate rate, identify peaks and troughs, grasp the seasonal fluctuations in the rebate rate, and provide merchants with more effective business strategies. At the same time, merchants are grouped according to different characteristics, such as transaction size, geographical location, etc., and the rebate rate performance of different groups is analyzed to find the commonalities and differences between high rebate rate groups and low rebate rate groups, and use the SVR model to provide merchants with future rebate rate forecasts. The merchant data processing system provides merchants with specific optimization suggestions based on the analysis results, including but not limited to recommending promotional activities during specific time periods to increase transaction volume and commission rates, providing more accurate target customer groups to optimize promotion effects, and providing suggestions for improving services and products for merchants with low commission rates, and optimizing merchants' pricing strategies to increase overall commission rates.

[0100] Optionally, the merchant data processing system provides a visualization function to display the above analysis results in the form of charts, reports, etc., which helps merchants to understand the data more intuitively and establish a mechanism for regularly updating data to ensure the continued accuracy of analysis and predictions, and optimize the effectiveness of recommendations by monitoring changes in rebate rates after merchants implement recommendations.

[0101] Optionally, the merchant data processing system obtains key system performance indicators and business indicators, including but not limited to system load, response time, transaction volume, rebate rate calculation time, etc. These indicators can reflect the health of the system, set appropriate thresholds for each key indicator, and use anomaly detection algorithms to analyze monitoring data, identify potential abnormalities, and trigger alarms when indicators exceed or fall below the set thresholds, which helps capture system anomalies or potential problems, while distinguishing the urgency of the alarm and setting different emergency levels for different abnormal situations. This helps administrators take timely measures based on the urgency of the situation.

[0102] Optionally, the merchant data processing system uses monitoring tools to monitor the system status in real time, and sets up a dashboard to display key indicators to help administrators quickly understand system performance and operating conditions. A detailed logging mechanism is added to the system to record key events, error messages, and important operations. The appropriate log level and format help track and analyze abnormal situations, and implement automated response mechanisms for some known abnormal situations. For example, when the server load is detected to be too high, more server instances can be automatically started to respond.

[0103] Optionally, Figure 5 is a flowchart of an optional prediction model application according to an embodiment of the present application, such as Figure 5 As shown, the merchant data processing system first obtains the current transaction data of the current merchant, and then directly inputs the current transaction data into a pre-created model to obtain the current merchant's commission rate, obtains the commission value based on the commission rate, and automatically distributes the commission value to the relevant platform. After distribution, the commission rate is analyzed to identify abnormal commission rates, and an alarm is issued after an abnormal commission rate is detected.

[0104] Optionally, Figure 6 is a block diagram of an optional prediction model application according to an embodiment of the present application, such as Figure 6As shown, it includes a bank transaction system, a data collection module, a database, a rebate rate prediction module, an automatic rebate rate issuance module, an intelligent analysis module, and a monitoring and alarm module. The data collection module provides an interface. By calling this interface, merchant data can be actively obtained from multiple channels (such as databases), ensuring that the latest basic information, transaction records, customer evaluations and other data are collected in a timely and accurate manner. Through real-time data collection, the system can obtain the latest merchant dynamics in a timely manner and provide real-time data support for other modules; the database is used to save merchant information tables (merchant industry, merchant identity, geographic location, number of people, asset size), merchant historical data tables (merchant historical transaction records for various products), rebate rate tables (actual rebate rate, predicted rebate rate, difference value (the difference between the actual rebate rate and the predicted rebate rate)); the rebate rate prediction module is the SVR prediction model. After predicting the customer's rebate rate data, the rebate rate data will be sent to the automated rebate rate issuance module and the intelligent analysis module; the automated rebate rate issuance module will automatically execute the rebate value issuance operation based on the rebate rate prediction data, reduce manual intervention, and ensure timely and accurate issuance of rebate values; the intelligent analysis module uses artificial intelligence technology to conduct in-depth analysis of the rebate rate data, explore business opportunities and trends, generate intelligent analysis reports and send them to the bank transaction system and monitoring alarm module at the same time; the monitoring alarm module receives and monitors the data in the database, the rebate rate data, and the data analyzed by the intelligent analysis module in real time, discovers potential problems or abnormal situations, and promptly notifies the bank transaction system through a flexible alarm mechanism.

[0105] From the above content, it can be seen that the merchant data processing system can quickly identify any possible abnormal situations by real-time monitoring of merchant-related data, thereby improving the timeliness of risk warnings. By setting dynamic alarm thresholds for different detection indicators, the system can automatically adjust the warning standards according to the characteristics of merchant transaction behavior, thereby improving the accuracy and flexibility of the warning.

[0106] The embodiment of the present application also provides a transaction data processing device. It should be noted that the transaction data processing device of the embodiment of the present application can be used to execute the transaction data processing method provided in the embodiment of the present application. The transaction data processing device provided in the embodiment of the present application is introduced below.

[0107] According to an embodiment of the present application, a device for implementing the above transaction data processing method is also provided. Figure 7 is a schematic diagram of an optional transaction data processing device according to an embodiment of the present application, such as Figure 7 As shown, the device includes: an acquisition unit 701, a screening unit 702, a determination unit 703, and a processing unit 704.

[0108] Optionally, an acquisition unit 701 is used to acquire the transaction data and attribute information of a target merchant; a screening unit 702 is used to screen at least one feature value from the transaction data and attribute information of a target merchant according to a target feature sequence, wherein the target feature sequence is a feature parameter determined according to pre-training; a determination unit 703 is used to input the screened feature value into a target prediction model, convert the feature value into a target feature vector according to prior knowledge learned by the target prediction model in the model training phase, and determine a target proportional coefficient for commission extraction based on the transaction data and attribute information of the target merchant according to the target feature vector, wherein the target prediction model is a model obtained by updating the initial prediction model through a target parameter combination, wherein the target parameter combination is a model parameter combination that meets a preset condition, wherein the preset condition is used to constrain the difference between the prediction result and the actual result of the initial prediction model to be less than a first preset threshold value; a processing unit 704 is used to calculate the difference between a preset proportional coefficient set by a user and a target proportional coefficient to obtain a processing result of the transaction data of the target merchant.

[0109] Optionally, the determination unit 703 includes: a first acquisition subunit, a first screening subunit, and a first training subunit. The first acquisition subunit is used to acquire the merchant's historical target information from the database, wherein the historical target information includes the merchant's historical transaction data and the merchant's historical attribute information; the first screening subunit is used to perform feature screening on the merchant's historical transaction data and the merchant's historical attribute information to obtain a target feature sequence consisting of X historical transaction features and Y historical attribute features, wherein X and Y are both integers greater than or equal to 1, and the feature screening is used to digitize the merchant's historical transaction data and historical attribute information, and continuously iterate and screen the digitized data to select features that meet the screening conditions, wherein the screening conditions are used to select features whose model prediction errors are less than a second preset threshold; the first training subunit is used to iteratively train the first prediction model according to the target feature sequence and the historical target information to obtain a target prediction model.

[0110] Optionally, the first screening subunit includes: a first partitioning module, a first extraction module, a first processing module, and a second processing module. The first partitioning module is used to divide the historical transaction data and historical attribute information of the same merchant into a data set when the historical target information includes the historical transaction data and historical attribute information of N merchants, so as to obtain N data sets, wherein N is an integer greater than or equal to 1; the first extraction module is used to extract a feature sequence from each of the N data sets to obtain N feature sequences; the first processing module is used to perform numerical conversion processing on the N feature sequences to obtain N first feature sequences, wherein each first feature sequence includes an initial value and an initial weight, wherein the initial value is used to characterize whether each feature in the first feature sequence is selected; the second processing module is used to perform multiple target operations on the initial prediction model according to the N first feature sequences, until the number of target operations is greater than or equal to the first iteration number, and a target feature sequence consisting of X historical transaction data and Y historical attribute information is obtained, wherein each target operation is used to train the initial prediction model according to the prediction result of the initial prediction model for each first feature sequence.

[0111] Optionally, the second processing module includes: a first setting submodule, a first acquiring submodule, a first calculating submodule, a first updating submodule, and a first processing submodule. Among them, the first setting submodule is used to set the first function, wherein the first function is used to calculate the difference value between the model prediction result obtained according to each feature sequence and the actual result; the first acquisition submodule is used to obtain the second function and the third function, wherein the second function is used to calculate the feature selection state in each of the N first feature sequences, and the third function is used to calculate the weight of each of the N first feature sequences; the first calculation submodule is used to input the N first feature sequences into the initial prediction model for training, and calculate the prediction error corresponding to each of the N first feature sequences according to the first function; the first updating submodule is used to update the feature selection state and weight in each first feature sequence according to the prediction error corresponding to each of the N first feature sequences, the second function and the third function, to obtain N second feature sequences; the first processing submodule is used to obtain the third feature sequence from the N second feature sequences, and use the third feature sequence as the target feature sequence, wherein the third feature sequence is the second feature sequence in the N second feature sequences whose prediction error is less than the corresponding preset threshold.

[0112] Optionally, the first updating submodule includes: a first processing component and a first updating component. The first processing component is used to, in the case of the first target operation, use the prediction error corresponding to each first feature sequence in the N first feature sequences as the corresponding initial prediction error, and update the weight of each first feature sequence according to a third function; the first updating component is used to, in the case of the i-th target operation, if the prediction error of the j-th first feature sequence is less than the prediction error of the j-th first feature sequence under the i-1 target operation, update the feature selection state in the j-th first feature sequence according to the second function, and update the weight of the j-th first feature sequence according to the third function, wherein i is an integer greater than 1, and j is a positive integer less than or equal to N.

[0113] Optionally, the first training subunit includes: a first setting module, a first acquisition module, a first determination module, and a second determination module. The first setting module is used to set the parameter set of the initial prediction model, wherein the parameter set includes S groups of parameter combinations, each group of parameter combinations includes L parameters, wherein S is an integer greater than or equal to 1, and L is an integer greater than or equal to 1; the first acquisition module is used to obtain a target algorithm, wherein the target algorithm is used to search for a parameter combination that meets a preset condition in the parameter space; the first determination module is used to determine a target parameter combination from the parameter set according to the target algorithm; the second determination module is used to determine the first prediction model according to the target parameter combination and the initial prediction model.

[0114] Optionally, the first training subunit includes: a second extraction module, a second acquisition module, a third processing module, and a first training module. The second extraction module is used to extract X historical transaction features and Y historical attribute features from each feature sequence of N feature sequences to obtain N model training features; the second acquisition module is used to obtain N commission ratio coefficients of N merchants corresponding to the N feature sequences; the third processing module is used to normalize the N model training features to obtain target features; the first training module is used to iteratively train the first prediction model according to the target features and the N commission ratio coefficients to obtain the target prediction model.

[0115] Optionally, the processing unit 704 includes: a first calculation subunit, a first setting subunit, and a first sending subunit. The first calculation subunit is used to calculate the target commission value corresponding to the target merchant according to the target proportion coefficient when the difference between the preset proportion coefficient and the target proportion coefficient is less than a set threshold; the first setting subunit is used to set a first interface, wherein the first interface is used to issue commission values; the first sending subunit is used to send the target commission value to the target platform through the first interface, wherein the target platform is a platform associated with the target merchant.

[0116] Optionally, the transaction data processing device further includes: a setting unit, a detection unit, and an early warning unit. Among them, the setting unit is used to set the target detection index, and set the corresponding alarm threshold for each detection index in the target detection index; the detection unit is used to detect the transaction data and target commission value of the target merchant to obtain a detection result set, wherein the detection result set includes abnormal transaction data between the target merchant and the user and / or the target platform; the early warning unit is used to issue an early warning for the tth detection result in the detection result set when the detection index corresponding to the tth detection result in the detection result set is greater than or equal to the alarm threshold corresponding to the detection index, wherein t is an integer greater than or equal to 1.

[0117] According to another aspect of the present application, a computer-readable storage medium is provided, the computer-readable storage medium including a stored executable program, wherein when the executable program runs, the device where the computer-readable storage medium is located is controlled to execute the above-mentioned transaction data processing method.

[0118] According to another aspect of the present application, an electronic device is provided, comprising: a memory storing an executable program; and a processor for running the program, wherein the above-mentioned transaction data processing method is executed when the program is running.

[0119] According to another aspect of the present application, a computer program product is also provided, comprising computer instructions, which implement the steps of the above-mentioned transaction data processing method when executed by a processor.

[0120] The serial numbers of the above-mentioned embodiments of the present application are for description only and do not represent the advantages or disadvantages of the embodiments.

[0121] In the above embodiments of the present application, the description of each embodiment has its own emphasis. For parts that are not described in detail in a certain embodiment, please refer to the relevant description of other embodiments.

[0122] In the several embodiments provided in this application, it should be understood that the disclosed technical content can be implemented in other ways. Among them, the device embodiments described above are only schematic. For example, the division of the units can be a logical function division. There may be other division methods in actual implementation. For example, multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some interfaces, indirect coupling or communication connection of units or modules, which can be electrical or other forms.

[0123] The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed on multiple units. Some or all of the units may be selected according to actual needs to achieve the purpose of the present embodiment.

[0124] In addition, each functional unit in each embodiment of the present application may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit. The above-mentioned integrated unit may be implemented in the form of hardware or in the form of software functional units.

[0125] If the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present application, in essence, or the part that contributes to the prior art or all or part of the technical solution can be embodied in the form of a software product, and the computer software product is stored in a storage medium, including a number of 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: U disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), mobile hard disk, disk or optical disk and other media that can store program codes.

[0126] The above is only a preferred embodiment of the present application. It should be pointed out that for ordinary technicians in this technical field, several improvements and modifications can be made without departing from the principles of the present application. These improvements and modifications should also be regarded as the scope of protection of the present application.

Claims

1. A method for processing transaction data, characterized in that: include: Acquire transaction data of a target merchant and attribute information of the target merchant; Filtering at least one feature value from the transaction data of the target merchant and the attribute information of the target merchant according to a target feature sequence, wherein the target feature sequence is a feature parameter determined according to pre-training; The eigenvalues ​​obtained by screening are input into the target prediction model, and according to the prior knowledge learned by the target prediction model in the model training stage, the eigenvalues ​​are converted into target eigenvectors, and according to the target eigenvectors, the target proportional coefficient for commission extraction based on the transaction data and attribute information of the target merchant is determined, wherein the target prediction model is a model obtained by updating the initial prediction model through the target parameter combination, wherein the target parameter combination is a model parameter combination that meets the preset conditions, wherein the preset conditions are used to constrain the difference between the prediction result of the initial prediction model and the actual result to be less than a first preset threshold value; The difference between the preset proportional coefficient set by the user and the target proportional coefficient is calculated to obtain the processing result of the transaction data of the target merchant.

2. The method for processing transaction data according to claim 1, characterized in that: The target prediction model is obtained by the following steps: Acquire historical target information of the merchant from a database, wherein the historical target information includes historical transaction data of the merchant and historical attribute information of the merchant; Performing feature screening on the merchant's historical transaction data and the merchant's historical attribute information to obtain the target feature sequence consisting of X historical transaction features and Y historical attribute features, wherein X and Y are both integers greater than or equal to 1, and the feature screening is used to digitize the merchant's historical transaction data and historical attribute information, and continuously iteratively screen the digitized data to select features that meet the screening conditions, wherein the screening conditions are used to select features whose model prediction error is less than a second preset threshold; The first prediction model is iteratively trained according to the target feature sequence and the historical target information to obtain a target prediction model.

3. The method for processing transaction data according to claim 2, characterized in that: Performing feature screening on the merchant's historical transaction data and the merchant's historical attribute information to obtain the target feature sequence consisting of X historical transaction features and Y historical attribute features, including: When the historical target information includes historical transaction data and historical attribute information of N merchants, the historical transaction data and historical attribute information belonging to the same merchant are divided into one data set to obtain N data sets, where N is an integer greater than or equal to 1; Extracting a feature sequence from each of the N data sets to obtain N feature sequences; Performing numerical conversion processing on the N feature sequences to obtain N first feature sequences, wherein each of the first feature sequences includes an initial numerical value and an initial weight, wherein the initial numerical value is used to indicate whether each feature in the first feature sequence is selected; The initial prediction model is subjected to multiple target operations according to the N first feature sequences until the number of the target operations is greater than or equal to the first iteration number, thereby obtaining the target feature sequence consisting of the X historical transaction data and the Y historical attribute information, wherein each target operation is used to train the initial prediction model according to the prediction result of the initial prediction model for each of the first feature sequences.

4. The method for processing transaction data according to claim 3, characterized in that: The target operation includes the following steps: Setting a first function, wherein the first function is used to calculate the difference between the model prediction result and the actual result obtained according to each feature sequence; Obtaining a second function and a third function, wherein the second function is used to calculate the feature selection state in each of the N first feature sequences, and the third function is used to calculate the weight of each of the N first feature sequences; Inputting the N first feature sequences into the initial prediction model for training, and calculating the prediction error corresponding to each first feature sequence in the N first feature sequences according to the first function; According to the prediction error corresponding to each first feature sequence in the N first feature sequences, the second function and the third function, the feature selection state and weight in each first feature sequence are updated to obtain N second feature sequences; A third feature sequence is obtained from the N second feature sequences, and the third feature sequence is used as the target feature sequence, wherein the third feature sequence is a second feature sequence whose prediction error is less than a corresponding preset threshold among the N second feature sequences.

5. The method for processing transaction data according to claim 4, characterized in that: Updating the feature selection state and weight in each of the N first feature sequences according to the prediction error corresponding to each of the N first feature sequences, the second function, and the third function, including: In the case of the first target operation, taking the prediction error corresponding to each first feature sequence in the N first feature sequences as the corresponding initial prediction error, and updating the weight of each first feature sequence according to the third function; In the case of the i-th target operation, if the prediction error of the j-th first feature sequence is less than the prediction error of the j-th first feature sequence under the i-1-th target operation, the feature selection state in the j-th first feature sequence is updated according to the second function, and the weight of the j-th first feature sequence is updated according to the third function, where i is an integer greater than 1 and j is a positive integer less than or equal to N.

6. The method for processing transaction data according to claim 2, characterized in that: Before iteratively training the first prediction model according to the target feature sequence and the historical target information to obtain the target prediction model, the transaction data processing method further includes: Setting a parameter set of the initial prediction model, wherein the parameter set includes S groups of parameter combinations, each group of parameter combinations includes L parameters, wherein S is an integer greater than or equal to 1, and L is an integer greater than or equal to 1; Acquire a target algorithm, wherein the target algorithm is used to search for a parameter combination that meets the preset condition in a parameter space; Determining the target parameter combination from the parameter set according to the target algorithm; The first prediction model is determined according to the target parameter combination and the initial prediction model.

7. The method for processing transaction data according to claim 3, characterized in that: Iteratively training the first prediction model according to the target feature sequence and the historical target information to obtain a target prediction model includes: Extracting X historical transaction features and Y historical attribute features from each of the N feature sequences to obtain N model training features; Obtaining N commission ratio coefficients of N merchants corresponding to the N feature sequences; Normalizing the N model training features to obtain target features; The first prediction model is iteratively trained according to the target feature and the N commission ratio coefficients to obtain the target prediction model.

8. The method for processing transaction data according to claim 1, characterized in that: The difference between the preset proportional coefficient set by the user and the target proportional coefficient is calculated to obtain the processing result of the transaction data of the target merchant. The method for processing the transaction data includes: When the difference between the preset proportional coefficient and the target proportional coefficient is less than a set threshold, the target commission value corresponding to the target merchant is calculated according to the target proportional coefficient; Setting a first interface, wherein the first interface is used to issue a commission value; The target commission value is sent to a target platform through the first interface, wherein the target platform is a platform associated with the target merchant.

9. The method for processing transaction data according to claim 8, characterized in that: After sending the target commission value to the target platform through the first interface, the transaction data processing method further includes: Setting target detection indicators, and setting corresponding alarm thresholds for each detection indicator in the target detection indicators; Detecting the transaction data of the target merchant and the target commission value to obtain a detection result set, wherein the detection result set includes abnormal transaction data between the target merchant and the user and / or the target platform; When the detection index corresponding to the t-th detection result in the detection result set is greater than or equal to the alarm threshold corresponding to the detection index, an early warning is issued for the t-th detection result, where t is an integer greater than or equal to 1.

10. A transaction data processing device, characterized in that: include: An acquisition unit, which acquires transaction data of a target merchant and attribute information of the target merchant; a screening unit, screening at least one feature value from the transaction data of the target merchant and the attribute information of the target merchant according to a target feature sequence, wherein the target feature sequence is a feature parameter determined according to pre-training; A determination unit inputs the eigenvalues ​​obtained by screening into a target prediction model, converts the eigenvalues ​​into target eigenvectors according to the prior knowledge learned by the target prediction model in the model training phase, and determines the target proportional coefficient for commission extraction based on the transaction data and attribute information of the target merchant according to the target eigenvectors, wherein the target prediction model is a model obtained by updating the initial prediction model through a target parameter combination, wherein the target parameter combination is a model parameter combination that satisfies a preset condition, wherein the preset condition is used to constrain the difference between the prediction result of the initial prediction model and the actual result to be less than a first preset threshold value; The processing unit calculates the difference between the preset proportional coefficient set by the user and the target proportional coefficient to obtain a processing result of the transaction data of the target merchant.

11. A computer-readable storage medium, characterized in that: The computer-readable storage medium includes a stored executable program, wherein when the executable program is run, the device where the computer-readable storage medium is located is controlled to execute the transaction data processing method according to any one of claims 1 to 9.

12. An electronic device, characterized in that: include: A memory storing an executable program; A processor is used to run the program, wherein the program, when running, executes the transaction data processing method described in any one of claims 1 to 9.

13. A computer program product comprising computer instructions, characterized in that: When the computer instructions are executed by a processor, the steps of the transaction data processing method described in any one of claims 1 to 9 are implemented.

Images