Transaction data processing method, device, equipment, medium and program product

Abstract

The invention provides a transaction data processing method and device, equipment, a storage medium and a program product, and can be applied to the technical field of financial science and technology and the technical field of big data. The transaction data processing method comprises the steps of obtaining current transaction behavior data corresponding to a transaction behavior of a target user in a current time interval; based on at least one predetermined target index, performing first calculation processing on the current transaction behavior data to obtain a first calculation result; based on the historical transaction behavior data of the target user in the first predetermined historical time interval, second calculation processing is carried out to obtain a second calculation result, and the second calculation result represents the probability of occurrence of predetermined behaviors of the target user in a predetermined future time interval; and performing fusion processing on the first calculation result and the second calculation result to obtain a transaction data processing result corresponding to the target user.

Description

Technical Field

[0001] This application relates to the fields of financial technology and big data technology, specifically to a transaction data processing method, apparatus, equipment, medium, and program product. Background Technology

[0002] In the field of financial risk management, continuous risk assessment of users is necessary to achieve accurate credit limit allocation and dynamic credit granting decisions. However, static credit models mainly rely on historical data to build a fixed system, which makes it difficult to dynamically reflect the user's current risk status and lacks sensitivity to real-time behavioral data. On the other hand, fixed threshold-based triggering mechanisms lack flexibility, have slow response speeds, and are difficult to adapt to complex and ever-changing risk management scenarios, thus limiting the accuracy and timeliness of risk assessment and credit limit adjustments. Summary of the Invention

[0003] In view of the above problems, this application provides a transaction data processing method, apparatus, device, medium and program product.

[0004] According to a first aspect of this application, a transaction data processing method is provided, comprising: acquiring current transaction behavior data corresponding to the transaction behavior of a target user in the current time interval; performing a first calculation on the current transaction behavior data based on at least one predetermined target indicator to obtain a first calculation result, wherein the first calculation result represents the probability of the target user performing a predetermined behavior in the current time interval, and the target indicator is determined from multiple initial indicators based on the importance of each of the multiple initial indicators; performing a second calculation on the target user's historical transaction behavior data in a first predetermined historical time interval to obtain a second calculation result, wherein the second calculation result represents the probability of the target user performing a predetermined behavior in a predetermined future time interval; and fusing the first calculation result and the second calculation result to obtain a transaction data processing result corresponding to the target user.

[0005] According to an embodiment of this application, the transaction data processing method further includes: acquiring a training sample set, the training sample set including: multiple sets of training transaction behavior data generated by multiple users in a second predetermined historical time interval, wherein each set of training transaction behavior data includes multiple historical indicator values ​​of multiple initial indicators, each set of training transaction behavior data has a predetermined label, the predetermined label being used to characterize the probability of the corresponding user performing a predetermined behavior in the second predetermined time interval; inputting the training sample set into a first calculation model, calling the first calculation model to determine the mapping relationship between multiple historical indicator values ​​and labels, and determining the importance of each initial indicator according to the mapping relationship, and outputting a target indicator, wherein the importance of the target indicator satisfies a preset numerical condition.

[0006] According to an embodiment of this application, based on the historical transaction behavior data of a target user within a first predetermined historical time interval, a second calculation process is performed to obtain a second calculation result, including: constructing a historical transaction behavior time series based on the historical transaction behavior data; calling a predetermined time series prediction model and using a sliding time window to extract multiple sets of target historical transaction behavior data corresponding to the sliding time window from the historical transaction behavior time series; determining multiple trend data corresponding to the sliding time window based on the multiple sets of target historical transaction behavior data, wherein the trend data is used to characterize the behavioral change trend corresponding to the target user's transaction behavior within the sliding time window; and performing a second calculation process based on the multiple trend data to obtain a second calculation result.

[0007] According to an embodiment of this application, the second calculation process based on multiple trend data to obtain the second calculation result includes: determining the correlation between each trend data and the predetermined transaction risk level based on the time interval between the sliding time window corresponding to each trend data and the current time interval, and the behavioral change trend represented by the trend data; determining the attention weight of each trend data based on the correlation; performing a weighted summation of multiple trend data based on the attention weight of each trend data; and determining the second calculation result based on the weighted summation result.

[0008] According to an embodiment of this application, the transaction data processing method includes: determining the window size of the sliding time window based on the transaction scenario corresponding to the transaction behavior and the collection frequency of historical transaction behavior data.

[0009] According to an embodiment of this application, the transaction data processing method further includes: determining a risk threshold corresponding to the current time interval based on transaction environment information and the user type of the target user; matching the transaction data processing result with the risk threshold; and determining a quota determination strategy for the target user based on the matching result.

[0010] According to an embodiment of this application, based on at least one predetermined target indicator, a first calculation process is performed on the current transaction behavior data to obtain a first calculation result, including: determining at least one indicator value of the current transaction behavior data relative to at least one target indicator; and performing a weighted summation of the at least one indicator value according to the respective weights of the at least one predetermined target indicator to obtain the first calculation result.

[0011] A second aspect of this application provides a transaction data processing apparatus, comprising: a first acquisition module for acquiring current transaction behavior data corresponding to the transaction behavior of a target user in the current time interval; a first processing module for performing a first calculation on the current transaction behavior data based on at least one predetermined target indicator to obtain a first calculation result, wherein the first calculation result represents the probability of the target user performing a predetermined behavior in the current time interval, and the target indicator is determined from multiple initial indicators based on the importance of each of the multiple initial indicators; a second processing module for performing a second calculation on the target user's historical transaction behavior data in a first predetermined historical time interval to obtain a second calculation result, wherein the second calculation result represents the probability of the target user performing a predetermined behavior in a predetermined future time interval; and a third processing module for fusing the first calculation result and the second calculation result to obtain a transaction data processing result corresponding to the target user.

[0012] A third aspect of this application provides an electronic device comprising: one or more processors; and a memory for storing one or more computer programs, wherein the one or more processors execute the one or more computer programs to implement the steps of the method described above.

[0013] A fourth aspect of this application also provides a computer-readable storage medium having a computer program or instructions stored thereon, which, when executed by a processor, implement the steps of the above-described method.

[0014] The fifth aspect of this application also provides a computer program product, including a computer program or instructions that, when executed by a processor, implement the steps of the above-described method.

[0015] According to embodiments of this application, target indicators are selected from multiple initial indicators based on importance, thereby identifying more representative evaluation indicators and improving the accuracy of transaction data processing results. By applying different calculation methods to transaction behavior data in the current time interval and risk data in historical time intervals, a first calculation result representing the probability of a predetermined behavior occurring immediately and a second calculation result representing future risk trends are generated, achieving real-time capture of short-term risk fluctuations and continuous assessment of long-term credit status. Finally, a weighted fusion algorithm replaces hard threshold rules for comprehensive calculation, resulting in transaction data processing results that simultaneously reflect both immediate risk status and long-term risk trends. This overcomes the data lag of static models and the untimely decision-making of rule engines, enabling dynamic risk decision-making without frequent manual intervention, thus providing a judgment basis with both high timeliness and high accuracy for applications such as credit limit adjustments. Attached Figure Description

[0016] The above-mentioned contents, other objects, features and advantages of this application will become clearer from the following description of embodiments with reference to the accompanying drawings, in which:

[0017] Figure 1 The illustration shows an application scenario diagram of a transaction data processing method, apparatus, device, medium, and program product according to embodiments of this application;

[0018] Figure 2 A flowchart illustrating a transaction data processing method according to an embodiment of this application is shown schematically.

[0019] Figure 3 A flowchart illustrating a transaction data processing method according to another embodiment of this application is shown schematically;

[0020] Figure 4 A flowchart illustrating a method for determining a second calculation result according to an embodiment of this application is shown schematically.

[0021] Figure 5 A flowchart illustrating a method for determining a second calculation result according to another embodiment of this application is shown schematically;

[0022] Figure 6 This schematically illustrates a structural block diagram of a transaction data processing apparatus according to an embodiment of the present application; and

[0023] Figure 7 A block diagram schematically illustrates an electronic device suitable for implementing a transaction data processing method according to an embodiment of this application. Detailed Implementation

[0024] The embodiments of this application will now be described with reference to the accompanying drawings. However, it should be understood that these descriptions are exemplary only and are not intended to limit the scope of this application. In the following detailed description, numerous specific details are set forth to provide a thorough understanding of the embodiments of this application for ease of explanation. However, it will be apparent that one or more embodiments may be implemented without these specific details. Furthermore, descriptions of well-known structures and technologies are omitted in the following description to avoid unnecessarily obscuring the concepts of this application.

[0025] The terminology used herein is for the purpose of describing particular embodiments only and is not intended to limit the scope of this application. The terms “comprising,” “including,” etc., as used herein indicate the presence of the stated features, steps, operations, and / or components, but do not exclude the presence or addition of one or more other features, steps, operations, or components.

[0026] All terms used herein (including technical and scientific terms) have the meanings commonly understood by those skilled in the art, unless otherwise defined. It should be noted that the terms used herein are to be interpreted in a manner consistent with the context of this specification, and not in an idealized or overly rigid way.

[0027] When using expressions such as "at least one of A, B and C", they should generally be interpreted in accordance with the meaning that is commonly understood by those skilled in the art (e.g., "a system having at least one of A, B and C" should include, but is not limited to, a system having A alone, a system having B alone, a system having C alone, a system having A and B, a system having A and C, a system having B and C, and / or a system having A, B and C, etc.).

[0028] It should be noted that the transaction data processing method and apparatus of this application can be used in the fields of fintech and big data technology, as well as in any field other than fintech and big data technology. The application fields of the transaction data processing method and apparatus of this application are not limited.

[0029] In the technical solution of this application, the user information (including but not limited to user personal information, user image information, user device information, such as location information) and data (including but not limited to data used for analysis, stored data, and displayed data) involved are all information and data authorized by the user or fully authorized by all parties. Furthermore, the collection, storage, use, processing, transmission, provision, disclosure, and application of related data all comply with relevant laws, regulations, and standards, take necessary confidentiality measures, do not violate public order and good morals, and provide corresponding operation entry points for users to choose to authorize or refuse.

[0030] In scenarios where personal information is used for automated decision-making, the methods, devices, and systems provided in this application all provide users with corresponding operation entry points for users to choose to agree to or reject the automated decision results; if the user chooses to reject, the process enters the expert decision-making process.

[0031] Financial institutions need to continuously assess the risk level of users and dynamically adjust their credit limits accordingly in order to achieve a balance between risk control and user experience.

[0032] On the one hand, the currently prevalent static credit models are entirely built upon users' historical data. Once established, the model's input features, weight allocation, and computational logic are fixed. This prevents the model from processing users' latest behavioral data, such as recent transaction sequences or changes in consumption patterns. Therefore, the model's output reflects a user's credit status over a specific period in the past, making it difficult to serve as a reliable basis for assessing their current risk status.

[0033] On the other hand, the decision-making logic of the rule engine used to ultimately trigger action relies on pre-set hard thresholds. This mechanism judges risk in a black-and-white manner, lacking the ability to quantitatively and comprehensively assess continuous, multi-dimensional risk signals. In addition, the addition, modification, or removal of rules usually requires a complete manual code development, testing, and review process, with a long update cycle, making it difficult to respond promptly to rapidly evolving risk patterns.

[0034] In view of this, embodiments of this application provide a transaction data processing method, comprising: acquiring current transaction behavior data corresponding to the transaction behavior of a target user in the current time interval; performing a first calculation on the current transaction behavior data based on at least one predetermined target indicator to obtain a first calculation result, wherein the first calculation result represents the probability of the target user performing a predetermined behavior in the current time interval, and the target indicator is determined from multiple initial indicators based on the importance of each of the multiple initial indicators; performing a second calculation on the target user's historical transaction behavior data in a first predetermined historical time interval to obtain a second calculation result, wherein the second calculation result represents the probability of the target user performing a predetermined behavior in a predetermined future time interval; and performing a fusion processing on the first calculation result and the second calculation result to obtain a transaction data processing result corresponding to the target user.

[0035] Figure 1 The diagram illustrates an application scenario of the account information reading method according to an embodiment of this application.

[0036] like Figure 1 As shown, application scenario 100 according to this embodiment may include a first terminal device 101, a second terminal device 102, a third terminal device 103, a network 104, and a server 105. The network 104 serves as a medium for providing a communication link between the first terminal device 101, the second terminal device 102, the third terminal device 103, and the server 105. The network 104 may include various connection types, such as wired or wireless communication links, or fiber optic cables, etc.

[0037] Users can use the first terminal device 101, the second terminal device 102, and the third terminal device 103 to interact with the server 105 via the network 104 to receive or send messages, etc. Various communication client applications can be installed on the first terminal device 101, the second terminal device 102, and the third terminal device 103, such as account management applications, shopping applications, web browser applications, search applications, instant messaging tools, email clients, social media platform software, etc. (for example only).

[0038] The first terminal device 101, the second terminal device 102, and the third terminal device 103 can be various electronic devices with displays and supporting web browsing, including but not limited to smartphones, tablets, laptops, and desktop computers. In the application scenario of this application embodiment, the first terminal device 101, the second terminal device 102, and the third terminal device 103 can be terminal devices capable of inserting and recognizing bank cards or target components. Users insert bank cards through the terminal devices of the first terminal device 101, the second terminal device 102, and the third terminal device 103 and operate on various accounts.

[0039] Server 105 can be a server that provides various services, such as a backend management server that supports websites browsed by users using the first terminal device 101, the second terminal device 102, and the third terminal device 103 (this is just an example). The backend management server can analyze and process data such as received user requests, and feed back the processing results (such as web pages, information, or data obtained or generated according to user requests) to the terminal devices.

[0040] It should be noted that the data processing method provided in the embodiments of this application can generally be executed by server 105. Correspondingly, the data processing device provided in the embodiments of this application can generally be located in server 105. The data processing method provided in the embodiments of this application can also be executed by a server or server cluster that is different from server 105 and capable of communicating with the first terminal device 101, the second terminal device 102, the third terminal device 103, and / or server 105. Correspondingly, the data processing device provided in the embodiments of this application can also be located in a server or server cluster that is different from server 105 and capable of communicating with the first terminal device 101, the second terminal device 102, the third terminal device 103, and / or server 105.

[0041] It should be understood that Figure 1 The number of terminal devices, networks, and servers shown is merely illustrative. Depending on implementation needs, any number of terminal devices, networks, and servers can be included.

[0042] The following will be based on Figure 1 The described scene, through Figures 2-6 The method for reading account information according to embodiments of this application will be described in detail.

[0043] Figure 2 A flowchart illustrating a transaction data processing method according to an embodiment of this application is shown.

[0044] like Figure 2 As shown, the account information reading method 200 of this embodiment includes operations S210 to S240.

[0045] In operation S210, obtain the current transaction behavior data corresponding to the target user's transaction behavior in the current time interval.

[0046] According to embodiments of this application, the target user refers to a user for whom risk prediction is required. The current time interval refers to the period most recent to the present. For example, the past 7 days, the past 30 days, or the period from this month to the present are significantly more timely than historical time intervals. Transaction behavior refers to all fund transfers by the user in financial transactions. Examples include: credit card spending, online payments, installment bill payments, cash withdrawals, and money transfers. Current transaction behavior data refers to the original set of records of all transactions that occurred within the aforementioned current time interval. In this embodiment, transaction behavior can also broadly refer to operations without actual transactions, such as access activity and login frequency.

[0047] For example, upon detecting that user A (the target user) has submitted a credit limit adjustment request, a transaction data processing flow is triggered. The transaction database is retrieved to show all of user A's spending, cash withdrawals, and installment payment records (transaction behaviors) over the past 30 days (the current time period), and these records are used as the current transaction behavior data. Alternatively, transaction data processing and monitoring of users in the system can be performed periodically.

[0048] In embodiments of this application, user consent or authorization can be obtained before acquiring user information. For example, after a user initiates a request, user consent or authorization can be obtained through a pop-up window. By clicking on a confirmation or agreement button, the user indicates that they have agreed or authorized the acquisition of user information, including user transaction behavior.

[0049] In operation S220, based on at least one predetermined target indicator, the current transaction behavior data is subjected to a first calculation process to obtain a first calculation result, wherein the first calculation result represents the probability that the target user will perform a predetermined behavior in the current time interval, and the target indicator is determined from multiple initial indicators according to the importance of each of the multiple initial indicators.

[0050] According to embodiments of this application, the target indicator refers to a specific metric deemed effective in measuring the currently occurring predetermined behavior after screening, and is calculated directly from current transaction behavior data. The predetermined behavior is related to transaction risk. Initial indicators refer to all potential risk assessment dimensions that may be considered in the early stages of model construction. Examples include: basic user information, including age, gender, occupation, income, and education level; credit history information, including historical overdue payments and credit card usage; user behavior data, including APP activity, login frequency, and repayment behavior; external data, including credit reports and third-party data; and transaction data, including repayment amount, debt-to-income ratio, total transaction amount, proportion of nighttime transactions, number of cross-border transactions, frequency of purchases at luxury merchants, and number of failed transactions. Importance is used to characterize the degree or weight of each initial indicator's contribution to the final risk. It can be calculated using algorithms such as logistic regression and decision tree feature importance. The first calculation result reflects whether the user's transaction behavior in the current short term is abnormal or risky; a higher value generally indicates a higher risk.

[0051] Before determining at least one target indicator, indicators with a strong correlation to risk are selected from multiple initial indicators based on a large amount of sample data, and a target indicator is determined according to its importance. Once current trading behavior data is acquired, the values ​​of these target indicators are calculated in real time, and then the first calculation result of these indicator values ​​is calculated.

[0052] For example, initial metrics included: total transaction amount over the past 7 days, percentage of nighttime transactions over the past 7 days, maximum single transaction amount over the past 7 days, and number of merchants participating in transactions over the past 7 days. Importance analysis revealed that the percentage of nighttime transactions and the maximum single transaction amount were the most important for identifying the probability of pre-planned behavior. Therefore, these two were ultimately determined as the target metrics. During calculation, it was found that user A's nighttime transaction amount accounted for as much as 60% of the past 7 days (current time interval), and the maximum single transaction amount reached 90% of their limit. Based on these two metric values, the model calculated a very high first result, indicating that the user's recent transaction behavior was abnormal and posed a potential risk.

[0053] In operation S230, based on the target user's historical transaction behavior data in the first predetermined historical time interval, a second calculation process is performed to obtain a second calculation result, wherein the second calculation result represents the probability that the target user will perform a predetermined behavior in the predetermined future time interval.

[0054] According to embodiments of this application, a first predetermined historical time interval is used to characterize a relatively long past period for observing a user's long-term risk trends. For example, the past 24 months or the past year. Historical transaction behavior data refers to the historical record of a user's risk in each past calculation period (e.g., monthly). This may include external credit scores, etc. Transaction data processing is based on statistics of time-series data to discover patterns and trends in historical data and predict future trends accordingly. The second calculation result is used to characterize the predicted future credit risk trend of the user; the higher the result, the greater the probability of an increase in future risk. The predetermined future time interval refers to the future period in which transaction data processing is desired. For example, the next 3 months or the next billing cycle.

[0055] For example, we obtain user A's monthly risk (historical transaction behavior data) for each month over the past two years (the first predetermined historical time interval). Analysis reveals that although the user's recent first calculation result is high, their credit risk over the past two years has been very stable and excellent. Another possibility is that the user's historical risk has shown a continuous, slow downward trend over the past six months. In this case, the time series forecasting model will process the transaction data based on this downward trend to obtain a second calculation result, which indicates that user A's credit risk is likely to gradually increase over the next three months (the predetermined future time interval). The time series forecasting model could, for example, be a Long Short-Term Memory Network (LSTM) model.

[0056] In operation S240, the first calculation result and the second calculation result are merged to obtain the transaction data processing result corresponding to the target user.

[0057] According to embodiments of this application, the fusion processing can be, for example, a weighted summation process, where the transaction data processing result = (first calculation result × weight A) + (second calculation result × weight B). The weight represents its importance in the final decision. Therefore, the final, comprehensive transaction data processing result considers both the user's recent instantaneous risk and long-term risk trends, providing the most comprehensive basis for the final credit limit adjustment decision. The weights of the first or second calculation result can be dynamically adjusted based on user behavior and transaction data.

[0058] For example, the first calculation result (recent risk) is very high (e.g., 85 points, out of 100 representing the highest risk). The second calculation result (future risk trend) is moderate (e.g., 60 points) due to good long-term credit but a slight decline in recent trend. If the risk control strategy focuses more on immediate risk, the first calculation result can be weighted at 0.7, and the second calculation result at 0.3. The transaction data processing result would then be 85 * 0.7 + 60 * 0.3 = 77.5. This score is high and may lead to a rejection of the credit limit increase application. If the strategy values ​​the user's long-term credit more, the first calculation result can be weighted at 0.4, and the second calculation result at 0.6. The transaction data processing result would then be 85 * 0.4 + 60 * 0.6 = 70. This score is relatively moderate and may lead to approval of a moderate credit limit increase.

[0059] For example, a user who has lived in China for a long time suddenly makes multiple large purchases overseas within a short period of time; or engages in continuous high-frequency transactions during inactive periods (such as late at night). In this case, the user's long-term credit history is no longer very important, and immediate action must be taken (such as freezing the account or refusing transactions) to prevent further losses. In this case, the weight of the first calculation result needs to be increased (e.g., 0.8-0.9).

[0060] According to embodiments of this application, target indicators are selected from multiple initial indicators based on importance, thereby identifying more representative evaluation indicators and improving the accuracy of transaction data processing results. By applying different calculation methods to transaction behavior data in the current time interval and risk data in historical time intervals, a first calculation result representing the probability of a predetermined behavior occurring immediately and a second calculation result representing future risk trends are generated, achieving real-time capture of short-term risk fluctuations and continuous assessment of long-term credit status. Finally, a weighted fusion algorithm replaces hard threshold rules for comprehensive calculation, resulting in transaction data processing results that simultaneously reflect both immediate risk status and long-term risk trends. This overcomes the data lag of static models and the untimely decision-making of rule engines, enabling dynamic risk decision-making without frequent manual intervention, thus providing a judgment basis with both high timeliness and high accuracy for applications such as credit limit adjustments.

[0061] According to an embodiment of this application, based on at least one predetermined target indicator, a first calculation process is performed on the current transaction behavior data to obtain a first calculation result, including: determining at least one indicator value of the current transaction behavior data relative to at least one target indicator; and performing a weighted summation of the at least one indicator value according to the respective weights of the at least one predetermined target indicator to obtain the first calculation result.

[0062] Target metrics are specific dimensions selected to measure immediate risk. Metric values ​​refer to the numerical results calculated based on current transaction behavior data for a specific target metric. For example, from the current transaction behavior data, all transactions that occurred between 10 PM and 6 AM the following day within the past 7 days are selected, their amounts are summed, and then divided by the total transaction amount for those 7 days. Let's assume the result is 35%. This 35% is the metric value (transactions occurring between 10 PM and 6 AM the following day within the past 7 days) in this round of calculation. Similarly, other metrics can be calculated, such as a metric value of 20,000 yuan for the largest single transaction amount in the past 30 days.

[0063] The first calculation result = (Indicator 1 value × Indicator 1 weight) + (Indicator 2 value × Indicator 2 weight) + ... + (Indicator N value × Indicator N weight). The weight of each indicator can be predefined, for example, regression coefficients obtained by training a logistic regression model on a historical dataset. During the model development phase, logistic regression analysis is performed on multiple initial indicators using historical user samples with risk labels (such as "overdue"). The regression coefficient corresponding to each indicator is calculated, quantifying the indicator's contribution to the risk outcome. These regression coefficients are then normalized to determine the weight of each indicator in the final result. For example, if the normalized regression coefficient of the percentage of nighttime transaction amount in the past 7 days is 0.6, then the normalized coefficient of the largest single transaction amount in the past 30 days is 0.4. This method of determining weights based on historical data, compared to relying on subjective assignments based on expert experience, can more objectively reflect the true statistical relationship between each indicator and risk, thus making the first calculation result obtained after weighted summation more accurate in prediction.

[0064] Furthermore, if indicator A represents the percentage of nighttime transaction amount in the past 7 days, with a weight of 0.6; and indicator B represents the largest single transaction amount in the past 30 days, with a weight of 0.4, the calculated indicator values ​​are: A = 35%, B = 20,000 yuan. Therefore, the first calculation result = 80 (points) × 0.6 + 60 (points) × 0.4 = 48 + 24 = 72 points. Here, 72 points is the final first calculation result, which comprehensively reflects the user's recent risk performance in both nighttime and large-scale transactions. It is important to note that before performing weighted summation, the original indicator values ​​usually need to be standardized or mapped to points.

[0065] Figure 3 A flowchart illustrating a transaction data processing method according to another embodiment of this application is shown.

[0066] like Figure 3 As shown, the transaction data processing method 300 further includes operations S310 to S320 in addition to operations S210 to S240.

[0067] In operation S310, a training sample set is obtained. The training sample set includes multiple sets of training transaction behavior data generated by multiple users in a second predetermined historical time interval. Each set of training transaction behavior data includes multiple historical index values ​​of multiple initial indicators. Each set of training transaction behavior data has a predetermined label, which is used to characterize the probability of the corresponding user performing a predetermined behavior in the second predetermined time interval.

[0068] In operation S320, the training sample set is input into the first calculation model, the first calculation model is called to determine the mapping relationship between multiple historical indicator values ​​and labels, and the importance of each initial indicator is determined according to the mapping relationship, and the target indicator is output. The importance of the target indicator meets the preset numerical conditions.

[0069] According to an embodiment of this application, the training sample set is a dataset used to train the model. Each user's complete set of training transaction behavior data within a second predetermined historical time interval, along with their corresponding risk labels, constitutes a training sample. The second predetermined historical time interval represents a specific past period used for model training. It is typically longer than the current time interval used in real-time calculations, such as the past 24 months, to ensure sufficient data to learn stable patterns. The training transaction behavior data consists of all transaction records for each user within the second predetermined historical time interval. Historical indicator values ​​are specific values ​​calculated for each set of training transaction behavior data based on multiple initial indicators. The data used to calculate these indicator values ​​is historical, with the aim of identifying historical patterns. A label is a known user risk level, such as 0 representing normal, 1 representing overdue; or more granular categories like low risk, medium risk, and high risk. For example, data from 100,000 users over the past two years (the second predetermined historical time interval) is selected. For user A, all their consumption and repayment records during this period are extracted (training transaction behavior data), and historical indicator values ​​for dozens of initial indicators, such as the percentage of nighttime transactions in the past 30 days and the average bill amount in the past six months, are calculated. Meanwhile, based on whether user A experienced a delinquency exceeding 90 days after those two years, they were labeled as either "good" (label=0) or "bad" (label=1). Data from 100,000 users collectively constituted the training sample set.

[0070] According to an embodiment of this application, the first computational model is, for example, an extreme gradient boosting (XGBoost) model. Multiple sets of training transaction behavior data are used as input features and fed into the XGBoost model. The label of the probability level of a predetermined behavior is used as the target to be predicted by the model. The XGBoost model first performs interpolation processing (such as mean imputation) on missing values, truncates outliers, performs One-Hot encoding on categorical variables, and standardizes continuous variables. The XGBoost model parameters are initialized as follows: learning rate (0.1), maximum tree depth (5), and minimum number of sample splits (100). By employing gradient boosting and constructing multiple decision trees, and continuously correcting errors, a complex function is ultimately formed to capture the intrinsic patterns between historical indicator values ​​and risk labels. In other words, the trained XGBoost model represents the mapping relationship between multiple historical indicator values ​​and labels. Through cross-validation by splitting the data into training and validation sets, the performance of different hyperparameter combinations is comprehensively tested, avoiding overfitting or underfitting misjudgments caused by a single validation set, ultimately finding the optimal hyperparameter combination. By processing high-dimensional data to avoid learning ineffective features, and by optimizing hyperparameters through cross-validation to avoid overfitting to the accidental patterns in the training data, the model can consistently assess risk in scenarios such as bank risk control, even when facing new customers and new transaction data.

[0071] In the financial sector, initial risk indicators are high-dimensional data, such as customer transaction history, credit scores, and behavioral tags in bank risk control. Direct processing can easily lead to data sparsity and a surge in computational costs. Therefore, using models for feature selection and dimensionality reduction to control computational costs preserves key feature information and avoids interference from invalid features. Missing data often results from customers not filling in non-mandatory information or system malfunctions causing missing segments of transaction history. Directly deleting samples with missing values ​​would result in the loss of a large amount of crucial data. Models, by interpolating missing values ​​and truncating outliers, reduce data preprocessing costs by eliminating excessive reliance on manual completion.

[0072] According to embodiments of this application, importance is determined based on mapping relationships. After the XGBoost model is trained, the more times a feature is used as a split point across all trees, the more important it is generally. The greater the reduction in loss function a feature brings when used as a split point, the more important it is. Thus, we obtain an importance score for each initial metric. Importance is used to measure the contribution and influence of each initial metric in predicting labels.

[0073] According to embodiments of this application, the preset condition may be, for example, Top-K selection, that is, only retaining the top K indicators in terms of importance. The preset condition may be, for example, threshold selection, retaining only indicators whose importance exceeds a preset threshold (e.g., 0.05). The preset condition may be, for example, cumulative contribution selection, accumulating from the most important indicators until the cumulative importance reaches a percentage of the total importance (e.g., 95%), and then selecting these indicators.

[0074] For example, if the preset condition is to select the top 5 indicators by importance, then the 5 most important indicators are selected from dozens of initial indicators, such as: the proportion of nighttime transactions, the number of recent credit inquiries, the current debt-to-income ratio, the recent transaction failure rate, and the largest single transaction amount. These 5 indicators constitute the final target indicator set used for online real-time calculation.

[0075] At the data processing level, the automatic feature importance-based selection mechanism during model training can identify and retain the most predictive feature subset from a large number of initial risk assessment indicators, effectively reducing the data dimensionality required for subsequent real-time risk value calculation. In terms of computational efficiency, the inherent computational efficiency of the XGBoost model, combined with the reduced computational load due to lower feature dimensionality, allows the system to complete model inference with fewer computational resources when generating the first calculation result. This makes this application suitable for real-time risk control scenarios requiring rapid response, improving the overall operating efficiency of the system while ensuring assessment accuracy.

[0076] In the above embodiments, training samples containing user behavior data and risk labels are obtained. A transaction data processing model is trained using the XGBoost algorithm. Based on the feature importance provided by XGBoost, target indicators are selected. Based on the selected key indicators and the trained XGBoost model, a risk assessment system is determined. Furthermore, the model can be differentially analyzed. Specifically, 50 sets of parameter combinations are randomly generated; the impact of these parameter combinations on risk indicators is simulated on a historical dataset, and the loss function is calculated; the population is updated through mutation, crossover, and selection operations, retaining the top 10% of parameter combinations after each iteration; the search stops when the loss function shows no significant decrease for five consecutive iterations, and the optimal parameter combination is output.

[0077] Furthermore, reinforcement learning algorithms are used to input a normalized, encoded real-time state vector, including multi-dimensional features such as the user's current risk, industry risk index, and historical default rate. The output layer generates continuous-value actions through activation functions, corresponding to adjustments in risk model parameters, such as threshold offsets and feature weight coefficients. During training, parameter adjustment actions are generated based on the current policy network. After execution, changes in indicators such as user default rate and capital turnover rate are observed, and corresponding reward values ​​are calculated. These state-action-reward tuples are stored in the experience replay pool. During the learning phase, the temporal difference error is calculated through the target network, and the policy network parameters are updated using gradient descent.

[0078] Figure 4 A flowchart illustrating a method for determining a second calculation result according to an embodiment of this application is shown.

[0079] like Figure 4 As shown, the method 400 for performing a second calculation process based on the historical transaction behavior data of the target user in a first predetermined historical time interval to obtain a second calculation result includes operations S410 to S440.

[0080] Operate S410 to construct a time-series sequence of historical transaction behavior based on historical transaction behavior data.

[0081] Operation S420 calls the predefined time series prediction model and uses a sliding time window to extract multiple sets of target historical transaction behavior data corresponding to the sliding time window from the historical transaction behavior time series.

[0082] Operation S430 determines multiple trend data corresponding to the sliding time window based on multiple sets of target historical transaction behavior data. The trend data is used to characterize the behavioral risk change trend corresponding to the target user's transaction behavior within the sliding time window.

[0083] Operation S440 performs a second calculation based on multiple trend data to obtain a second calculation result.

[0084] According to embodiments of this application, historical transaction behavior data refers to the risk values ​​recorded for a user at a series of consecutive points in time in the past. A historical transaction behavior time series is an ordered data sequence formed by arranging the aforementioned historical transaction behavior data in chronological order. For example, obtaining the historical transaction behavior data of user B at the end of each month over the past 24 months. The specific risk values ​​can be calculated from this historical transaction behavior data using a predetermined algorithm. Arranging these risk values ​​in chronological order constitutes a time series of length 24, for example: [72,72.5,71.8,71.5,71,70.5,70,69.5,...]. This sequence clearly shows the overall trend of user B's transaction data over two years.

[0085] According to an embodiment of this application, a sliding time window is a fixed-length data extraction frame that slides across a time series. For example, a window with a length of 6 months. The target historical transaction behavior data refers to the small segment of historical data selected at each specific position within the sliding time window. Using a sliding window of 6 months, the data is slid across the 24-month series constructed in S410. The first slide selects the data from January to June: [72.0, 72.5, 71.8, 71.5, 71, 70.5], which is the first group of target historical transaction behavior data. The window slides forward one month, selecting the data from February to July: [72.5, 71.8, 71.5, 71.0, 70.5, 70.0], which is the second group of target historical transaction behavior data. And so on, ultimately multiple groups of such continuous data can be extracted from the entire series.

[0086] According to an embodiment of this application, the transaction data processing method further includes: determining the window size of the sliding time window based on the transaction scenario corresponding to the transaction behavior and the collection frequency of historical transaction behavior data.

[0087] The sliding window size should not be fixed, but dynamically adjusted according to the speed of business operations and the frequency of data updates. The goal is to use historical data sequences for analysis that capture meaningful risk patterns without including too much outdated noise.

[0088] First, determine the size of the sliding time window based on the transaction scenario corresponding to the transaction behavior.

[0089] A transaction scenario refers to the specific business environment or product type in which a user engages in a transaction. In different scenarios, user behavior patterns and the speed of risk evolution vary. The window size refers to the length of time covered by the sliding time window, such as 30 days, 90 days, or 180 days.

[0090] For high-frequency scenarios with rapidly evolving risks, use shorter windows. This is because short-term behavioral data already contains sufficient risk signals; using excessively long windows introduces outdated information irrelevant to the current situation, diluting the importance of recent risk signals. For low-frequency scenarios with slow-evolving risks, use longer windows. This is because sufficient time is needed to accumulate meaningful behavioral patterns; excessively short windows cannot capture the complete risk cycle or trend.

[0091] For example, in high-frequency payment scenarios such as third-party payments and QR code payments, the sheer volume of transactions and behavioral data means that risks can erupt within hours or days. A short-term window, such as 7, 15, or 30 days, can be set. Analyzing a user's transaction trends over the past 30 days is sufficient to effectively identify abnormal changes in spending frequency, amount, or merchant type, without needing to obtain payment habits from a year ago.

[0092] Secondly, the size of the sliding time window is determined based on the collection frequency of historical transaction behavior data.

[0093] Here, collection frequency refers to the time interval between updates to user risk data, such as daily, weekly, or monthly updates. A high collection frequency allows for a shorter window, as high-frequency data provides enough data points in a short time to fit the trend. A low collection frequency requires a longer window, as a sufficiently long time span is needed to collect a sufficient number of data points to ensure the statistical significance of the trend analysis.

[0094] For example, a bank calculates risk values ​​based on real-time behavior derived from daily transaction records. The window size can be set to 30 data points (i.e., 30 days). One risk value per day, 30 points is sufficient to fit a short-term trend line well. Another example is calculating the monthly credit score in an individual's credit report. The window size is set to 24 data points (i.e., 24 months). One risk value per month; if only 6 points (six months) are used, it is difficult to reliably determine whether the credit score is in a cyclical fluctuation or a long-term downward trend. 24 points provide a more robust trend analysis.

[0095] In summary, the window size of the sliding time window can be determined by considering both the transaction scenario and the collection frequency.

[0096] Specifically, one implementation method for determining the window size is a lookup table method. This involves pre-defining a configuration table that specifies the transaction scenario, the base window size, and the correspondence between different collection frequencies. For example, in the case of online transfer approval, the base window size is 90 days. If the collection frequency is high, the final window size is determined to be 90 days; if the collection frequency is medium, the final window size is determined to be 180 days; and if the collection frequency is low, the final window size is determined to be 270 days or 360 days, etc.

[0097] Specifically, another embodiment of determining the window size can be a weighted method based on data points. An ideal number of data points is set, for example, N=30, meaning 30 historical risk value points are used to predict the trend. A base time period (days) is defined for each scenario. For example, high-frequency scenario (S1): 30 days; medium-frequency scenario (S2): 90 days; low-frequency scenario (S3): 720 days. The final window size is calculated as min(S, N×F), where F is the conversion value for the collection frequency (daily update = 1, monthly update = 30, quarterly update = 90).

[0098] Therefore, by dynamically determining the window size, the transaction data processing model can use the most relevant and effective historical information according to the rhythm of different businesses and the data foundation, thereby improving the accuracy of transaction data processing.

[0099] Based on multiple sets of historical transaction behavior data of the target, several risk change data corresponding to the sliding time window are determined. Among them, the risk change data is used to characterize the risk change trend of the probability of the target user performing a predetermined behavior within the sliding time window.

[0100] According to embodiments of this application, risk change data refers to statistical characteristics calculated from each target historical transaction behavior data segment that can quantify the trend of risk change. For window data [720,725,718,715,710,705], various risk change data can be calculated. For example, the slope of the sequence can be calculated using linear regression, resulting in -2.8. This indicates that the user's credit score has been decreasing at an average rate of 2.8 points per month over the past 6 months. Alternatively, the standard deviation of the sequence can be calculated, resulting in 7.2. This indicates that the user's credit score fluctuated significantly during this period. Alternatively, the coefficient of variation can be calculated, i.e., the standard deviation divided by the mean, eliminating dimensions and facilitating comparisons between different users. These calculated slopes, standard deviations, etc., are the risk change data used for prediction, which can quantitatively characterize the trend of risk change in the probability of a user engaging in a predetermined behavior within a sliding time window.

[0101] According to an embodiment of this application, a predetermined time-series prediction model is used to process transaction data based on multiple risk change data to obtain a second calculation result. The predetermined time-series prediction model can be a model capable of processing time-series data, such as an LSTM model. The training objective of the model is to predict the risk value at the next time point, or to predict the change in risk value over a future period (such as the next quarter), based on risk change data (such as slope) within a time window. For each set of target historical transaction behavior data, the corresponding risk change data is the input feature (X), and the actual risk value (or the amount of risk value change) at the first time point after that window is the label (Y). The model is trained using all these samples. When a prediction is needed for user B, the risk value data for their most recent time window (such as the last 6 months) is taken, their risk change data is calculated, and then input into the trained model. The model outputs a predicted value, for example, that their credit score is expected to decrease by 5 points in the next 3 months. This prediction result is determined as the second calculation result, which characterizes the probability that the target user will exhibit a predetermined behavior within a predetermined future time interval.

[0102] By introducing the LSTM model, the system can automatically learn and memorize long-term dependencies and temporal evolution patterns in users' historical transaction behavior data, effectively solving the problem of traditional cycle models' insensitivity to short-term risk fluctuations. When processing risk value time series, the LSTM model can dynamically identify key time nodes and change patterns in risk trends, thereby generating a second calculation result with higher predictive accuracy. This enhances the ability to perceive and predict dynamic changes in users' risk status, providing a more reliable trend-based judgment basis for credit limit adjustment decisions.

[0103] Figure 5 A flowchart illustrating a method for determining a second calculation result according to another embodiment of this application is shown schematically.

[0104] like Figure 5 As shown, a second calculation is performed based on multiple trend data to obtain the second calculation result, including operations S510 to S540.

[0105] Operation S510 determines the correlation between each trend data and the predetermined transaction risk level based on the time interval between the sliding time window corresponding to each trend data and the current time interval, and the behavioral trend represented by the trend data.

[0106] According to embodiments of this application, the time interval is the time distance between the historical time window represented by the risk change data (e.g., January-June 2023) and the current time interval (e.g., January 2024). The closer the distance, the stronger the correlation is generally. The probability level of the predetermined behavior, also known as the risk inflection point, refers to the critical point where a risk state undergoes a drastic change, such as the sudden moment when it jumps from low risk to medium risk. The time interval can be used to represent historical trends that are closer to the present, and generally have a stronger correlation with the future. For example, a slow downward trend a year ago may not be as important as an accelerated downward trend a month ago. The risk change trend itself can determine whether an inflection point is imminent. For example, a trend with a sharply negative slope and greatly increased volatility is more correlated with a risk inflection point than a slightly negative and stable trend. For example, trend A (8 months ago): slope -0.5 (slightly slow decline), time interval is relatively far. Trend B (3 months ago): slope -3.0 (rapid decline), and large standard deviation (violent volatility), time interval is relatively recent. The model calculates that the correlation between trend B and the impending surge in risk level is much stronger than that between trend B and trend A. This is because trend B is more recent in time, indicating that the risk change data corresponding to trend B is more closely related to the probability level of the predetermined behavior.

[0107] Operate S520 to determine the attention weight of each trend data based on the correlation.

[0108] According to an embodiment of this application, a normalization function is used to transform the correlation scores between all historical risk change data and risk inflection points into an attention weight distribution with a sum of 1. Assume the calculated correlation scores are: Trend A = 5, Trend B = 95. After normalization, the attention weight of Trend A may be only 0.04, while the attention weight of Trend B is as high as 0.96. This means that when the model makes predictions, 96% of its attention is focused on the information contained in Trend B.

[0109] Operate S530 to perform a weighted summation of multiple trend data based on the attention weights of each trend data point.

[0110] According to an embodiment of this application, the weighted summation result = (overall value of trend A × weight 0.04) + (overall value of trend B × weight 0.96). This result represents the essence of trends most valuable for predicting the future, extracted from the entire history.

[0111] Operate S540 to determine the second calculation result based on the weighted summation result.

[0112] According to an embodiment of the present application, the weighted summation result after attention weighting is input into the prediction layer of the model and mapped to the final second calculation result. Based on this, the prediction layer outputs a very high second calculation result, such as 85 points (out of 100), indicating a very high risk that the user will default in a predetermined future time interval (such as the next quarter).

[0113] According to an embodiment of the present application, by introducing an attention mechanism, the model can automatically judge the importance of data in different historical periods. For example, it may find that the trend pattern before a similar risk inflection point a year ago is highly valuable for the current prediction.

[0114] According to an embodiment of the present application, the transaction data processing method further includes: determining a risk threshold corresponding to the current time interval according to the transaction environment information corresponding to the current time interval and the user type of the target user; matching the transaction data processing result with the risk threshold, and determining a quota determination strategy for the target user according to the matching result.

[0115] The transaction environment information may be external risk factors at the macro level, such as the current overall economic situation, industry fluctuations, and specific sales cycles (such as the credit risk may increase short-term during the Double Eleven period). When the external environment deteriorates, the system will automatically lower the risk threshold to make the risk control strategy tend to be conservative; conversely, during a stable and prosperous period, the threshold can be appropriately relaxed.

[0116] The user type refers to the individual attributes of the user. The user type can be divided from multiple dimensions. For example, from the value dimension, it includes high-net-worth customers, ordinary customers, and new customers; from the behavior dimension, it includes high-frequency consumption users, installment preference users, and dormant users; from the risk dimension, it includes customers with good history, customers with minor overdue records, etc. For high-net-worth customers or customers with good history, the system will set a more relaxed threshold to provide a better experience and maintain customer relationships; while for new customers or customers with overdue records, a more stringent threshold will be adopted to prudently control risks.

[0117] The risk threshold is one or more critical scores for decision-making. For example, two thresholds can be set to divide the risks into three levels: a low-risk threshold (T1) and a high-risk threshold (T2). When the risk value < T1, it is low risk; T1 ≤ risk value < T2, it is medium risk; risk value ≥ T2, it is high risk. Compare the calculated single transaction data processing result with the dynamically set risk threshold and classify it into a preset risk level.

[0118] The credit limit determination strategy consists of pre-defined business operation instructions linked to each risk level. When a target customer's risk level changes, the credit limit is redefined based on the mapping relationship between the risk level and the credit limit adjustment. For example, for low risk (transaction data processing result < 70), the credit limit determination strategy is to approve a 50% increase; for medium risk (70 ≤ transaction data processing result < 85), the credit limit determination strategy is to maintain the original credit limit; for high risk (transaction data processing result ≥ 85), the credit limit determination strategy is to reduce the credit limit by 20% and send a risk warning SMS.

[0119] By using transaction environment information and user type as input parameters for dynamic risk thresholds, the risk control strategy achieves context-adaptive adjustment. Specifically, it can adjust risk assessment criteria in real time based on environmental factors such as macroeconomic indicators and industry risk data, ensuring that credit granting decisions are synchronized with the external risk environment. Simultaneously, by segmenting user groups with different characteristics and configuring differentiated threshold parameters, the credit limit management strategy can take into account both users' historical behavioral characteristics and differences in customer value. This effectively improves the adaptability of risk assessment to business scenarios, providing a more targeted decision-making basis for credit limit adjustments.

[0120] A dynamic risk threshold is calculated based on the transaction environment information and user type within the current time interval. The risk value of the target user is then matched against this threshold, and a corresponding credit limit determination strategy is generated based on the matching result. Furthermore, the generated credit limit determination strategy is encapsulated as an instruction and distributed in real-time to relevant business systems via a message queue. The rules engine, acting as a producer, publishes the instructions to designated topics. Each business system (such as the transaction system and the alert system), acting as a consumer, subscribes to these topics, receives the instructions, and executes specific credit limit changes, process triggers, or notification push operations.

[0121] Based on the above-described transaction data processing method, this application also provides a transaction data processing apparatus. The following will be combined with... Figure 6 The device is described in detail.

[0122] Figure 6 A schematic block diagram of a transaction data processing apparatus according to an embodiment of this application is shown.

[0123] like Figure 6 As shown, the account information transaction data processing device 600 of this embodiment includes a first acquisition module 610, a first processing module 620, a second processing module 630 and a third processing module 640.

[0124] The first acquisition module 610 is used to acquire current transaction behavior data corresponding to the target user's transaction behavior in the current time interval. In one embodiment, the first acquisition module 610 can be used to perform the operation S210 described above, which will not be repeated here.

[0125] The first processing module 620 is used to perform a first calculation on the current transaction behavior data based on at least one predetermined target indicator to obtain a first calculation result. The first calculation result represents the probability that a target user will perform a predetermined behavior within the current time interval. The target indicator is determined from multiple initial indicators based on the importance of each initial indicator. In one embodiment, the first processing module 620 can be used to execute the operation S220 described above, which will not be repeated here.

[0126] The second processing module 630 is used to perform a second calculation based on the target user's historical transaction behavior data within a first predetermined historical time interval, to obtain a second calculation result, wherein the second calculation result represents the probability that the target user will perform a predetermined behavior within a predetermined future time interval. In one embodiment, the second processing module 630 may be used to execute the operation S230 described above, which will not be repeated here.

[0127] The third processing module 640 is used to fuse the first calculation result and the second calculation result to obtain the transaction data processing result corresponding to the target user. In one embodiment, the third processing module 640 can be used to perform the operation S240 described above, which will not be repeated here.

[0128] According to embodiments of this application, the transaction data processing apparatus further includes a second acquisition module, an input module, a calling module, and an output module.

[0129] The second acquisition module is used to acquire a training sample set, which includes: multiple sets of training transaction behavior data generated by multiple users within a second predetermined historical time interval. Each set of training transaction behavior data includes multiple historical indicator values ​​of multiple initial indicators. Each set of training transaction behavior data has a predetermined label, which is used to characterize the probability of the corresponding user performing a predetermined behavior within the second predetermined time interval. The input module is used to input the training sample set into the first calculation model. The calling module is used to call the first calculation model to determine the mapping relationship between multiple historical indicator values ​​and labels, and to determine the importance of each initial indicator based on the mapping relationship. The output module is used to output the target indicator, wherein the importance of the target indicator meets a preset numerical condition.

[0130] According to an embodiment of this application, the second processing module 630 includes a construction submodule, an extraction submodule, a first determination submodule, and a fourth processing submodule.

[0131] The system comprises four modules: a construction submodule for constructing a historical transaction behavior time series based on historical transaction behavior data; an extraction submodule for calling a predefined time series prediction model and using a sliding time window to extract multiple sets of target historical transaction behavior data corresponding to the sliding time window from the historical transaction behavior time series; a first determination submodule for determining multiple trend data corresponding to the sliding time window based on the multiple sets of target historical transaction behavior data, wherein the trend data is used to characterize the trend of behavioral risk changes corresponding to the target user's transaction behavior within the sliding time window; and a fourth processing submodule for performing a second calculation based on the multiple trend data to obtain a second calculation result.

[0132] According to an embodiment of this application, the fourth processing submodule includes a first determining unit, a second determining unit, a calculation unit, and a third determining unit.

[0133] The first determining unit is used to determine the correlation between each trend data and the predetermined transaction risk level based on the time interval between the sliding time window corresponding to each trend data and the current time interval, and the behavioral trend represented by the trend data; the second determining unit is used to determine the attention weight of each trend data based on the correlation; the calculation unit is used to perform a weighted summation of multiple trend data based on the attention weight of each trend data; and the third determining unit is used to determine the second calculation result based on the weighted summation result.

[0134] According to an embodiment of this application, the transaction data processing apparatus further includes a fourth determining module.

[0135] The fourth determination module is used to determine the window size of the sliding time window based on the transaction scenario corresponding to the transaction behavior and the collection frequency of historical transaction behavior data.

[0136] According to an embodiment of this application, the transaction data processing apparatus further includes a fifth determining module and a matching module.

[0137] The fifth determination module is used to determine the risk threshold corresponding to the current time interval based on the transaction environment information and the user type of the target user; the matching module is used to match the transaction data processing results with the risk threshold and determine the quota determination strategy for the target user based on the matching results.

[0138] According to an embodiment of this application, the first processing module 620 includes a second determining submodule and a calculation submodule.

[0139] The second determining submodule is used to determine at least one indicator value of the current transaction behavior data for at least one target indicator; the calculation submodule is used to perform a weighted summation of at least one indicator value according to the weights of each of the at least one target indicator determined in advance, to obtain a first calculation result.

[0140] According to embodiments of this application, any plurality of modules among the first acquisition module 610, the first processing module 620, the second processing module 630, and the third processing module 640 may be combined into one module, or any one of these modules may be split into multiple modules. Alternatively, at least a portion of the functionality of one or more of these modules may be combined with at least a portion of the functionality of other modules and implemented in one module. According to embodiments of this application, at least one of the first acquisition module 610, the first processing module 620, the second processing module 630, and the third processing module 640 may be at least partially implemented as hardware circuitry, such as a field-programmable gate array (FPGA), a programmable logic array (PLA), a system-on-a-chip, a system-on-a-substrate, a system-on-package, an application-specific integrated circuit (ASIC), or implemented in hardware or firmware by any other reasonable means of integrating or packaging the circuitry, or implemented in any one of the three implementation methods of software, hardware, and firmware, or in a suitable combination of any of these. Alternatively, at least one of the first acquisition module 610, the first processing module 620, the second processing module 630, and the third processing module 640 may be implemented at least partially as a computer program module, which can perform corresponding functions when the computer program module is run.

[0141] Figure 7 A block diagram schematically illustrates an electronic device suitable for implementing a transaction data processing method according to an embodiment of this application.

[0142] like Figure 7 As shown, an electronic device 700 according to an embodiment of this application includes a processor 701, which can perform various appropriate actions and processes according to a program stored in a read-only memory (ROM) 702 or a program loaded from a storage portion 708 into a random access memory (RAM) 703. The processor 701 may include, for example, a general-purpose microprocessor (e.g., a CPU), an instruction set processor and / or an associated chipset and / or a special-purpose microprocessor (e.g., an application-specific integrated circuit (ASIC)), etc. The processor 701 may also include onboard memory for caching purposes. The processor 701 may include a single processing unit or multiple processing units for performing different actions of the method flow according to an embodiment of this application.

[0143] RAM 703 stores various programs and data required for the operation of electronic device 700. Processor 701, ROM 702, and RAM 703 are interconnected via bus 704. Processor 701 executes various operations of the method flow according to embodiments of this application by executing programs in ROM 702 and / or RAM 703. It should be noted that programs may also be stored in one or more memories other than ROM 702 and RAM 703. Processor 701 may also execute various operations of the method flow according to embodiments of this application by executing programs stored in one or more memories.

[0144] According to embodiments of this application, the electronic device 700 may further include an input / output (I / O) interface 705, which is also connected to a bus 704. The electronic device 700 may also include one or more of the following components connected to the input / output (I / O) interface 705: an input section 706 including a keyboard, mouse, etc.; an output section 707 including a cathode ray tube (CRT), liquid crystal display (LCD), etc., and a speaker, etc.; a storage section 708 including a hard disk, etc.; and a communication section 709 including a network interface card such as a LAN card, modem, etc. The communication section 709 performs communication processing via a network such as the Internet. A drive 710 is also connected to the input / output (I / O) interface 705 as needed. A removable medium 711, such as a disk, optical disk, magneto-optical disk, semiconductor memory, etc., is installed on the drive 710 as needed so that computer programs read from it can be installed into the storage section 708 as needed.

[0145] This application also provides a computer-readable storage medium, which may be included in the device / apparatus / system described in the above embodiments; or it may exist independently and not assembled into the device / apparatus / system. The computer-readable storage medium carries one or more programs, which, when executed, implement the method according to the embodiments of this application.

[0146] According to embodiments of this application, the computer-readable storage medium can be a non-volatile computer-readable storage medium, such as including but not limited to: portable computer disks, hard disks, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), portable compact disk read-only memory (CD-ROM), optical storage devices, magnetic storage devices, or any suitable combination thereof. In this application, the computer-readable storage medium can be any tangible medium containing or storing a program that can be used by or in conjunction with an instruction execution system, apparatus, or device. For example, according to embodiments of this application, the computer-readable storage medium may include ROM 702 and / or RAM 703 and / or one or more memories other than ROM 702 and RAM 703 described above.

[0147] Embodiments of this application also include a computer program product comprising a computer program containing program code for performing the methods shown in the flowchart. When the computer program product is run on a computer system, the program code enables the computer system to implement the account information reading method provided in the embodiments of this application.

[0148] When the computer program is executed by the processor 701, it performs the functions defined in the system / apparatus of this application embodiment. According to the embodiments of this application, the systems, apparatuses, modules, units, etc., described above can be implemented by computer program modules.

[0149] In one embodiment, the computer program may rely on a tangible storage medium such as an optical storage device or a magnetic storage device. In another embodiment, the computer program may also be transmitted and distributed in the form of signals over a network medium, and may be downloaded and installed via the communication section 709, and / or installed from a removable medium 711. The program code contained in the computer program can be transmitted using any suitable network medium, including but not limited to: wireless, wired, etc., or any suitable combination thereof.

[0150] In such an embodiment, the computer program can be downloaded and installed from a network via the communication section 709, and / or installed from the removable medium 711. When the computer program is executed by the processor 701, it performs the functions defined in the system of this application embodiment. According to the embodiments of this application, the systems, devices, apparatuses, modules, units, etc., described above can be implemented by computer program modules.

[0151] According to embodiments of this application, program code for executing the computer programs provided in the embodiments of this application can be written in any combination of one or more programming languages. Specifically, these computational programs can be implemented using high-level procedural and / or object-oriented programming languages, and / or assembly / machine languages. Programming languages ​​include, but are not limited to, languages ​​such as Java, C++, Python, "C", or similar programming languages. The program code can be executed entirely on the user's computing device, partially on the user's device, partially on a remote computing device, or entirely on a remote computing device or server. In cases involving remote computing devices, the remote computing device can be connected to the user's computing device via any type of network, including a local area network (LAN) or a wide area network (WAN), or it can be connected to an external computing device (e.g., via the Internet using an Internet service provider).

[0152] The flowcharts and block diagrams in the accompanying drawings illustrate the architecture, functionality, and operation of possible implementations of systems, methods, and computer program products according to various embodiments of this application. In this regard, each block in a flowchart or block diagram may represent a module, segment, or portion of code containing one or more executable instructions for implementing a specified logical function. It should also be noted that in some alternative implementations, the functions indicated in the blocks may occur in a different order than those indicated in the drawings. For example, two consecutively indicated blocks may actually be executed substantially in parallel, and they may sometimes be executed in reverse order, depending on the functions involved. It should also be noted that each block in a block diagram or flowchart, and combinations of blocks in a block diagram or flowchart, may be implemented using a dedicated hardware-based system that performs the specified function or operation, or using a combination of dedicated hardware and computer instructions.

[0153] Those skilled in the art will understand that the features described in the various embodiments of this application can be combined and / or combined in various ways, even if such combinations or combinations are not explicitly described in this application. In particular, the features described in the various embodiments of this application can be combined and / or combined in various ways without departing from the spirit and teachings of this application. All such combinations and / or combinations fall within the scope of this application.

Claims

1. A method for processing transaction data, characterized in that, The method includes: Obtain current transaction behavior data corresponding to the target user's transaction behavior in the current time interval; Based on at least one predetermined target indicator, the current transaction behavior data is subjected to a first calculation process to obtain a first calculation result, wherein the first calculation result represents the probability that the target user will perform a predetermined behavior in the current time interval, and the target indicator is determined from the plurality of initial indicators according to the importance of each of the plurality of initial indicators. Based on the target user's historical transaction behavior data within a first predetermined historical time interval, a second calculation process is performed to obtain a second calculation result, wherein the second calculation result represents the probability that the target user will perform a predetermined behavior within a predetermined future time interval. The first calculation result and the second calculation result are fused together to obtain the transaction data processing result corresponding to the target user.

2. The method according to claim 1, characterized in that, The method further includes: Obtain a training sample set, which includes: multiple sets of training transaction behavior data generated by multiple users in a second predetermined historical time interval, wherein each set of training transaction behavior data includes multiple historical index values ​​of the multiple initial indicators, and each set of training transaction behavior data has a predetermined label, which is used to characterize the probability of the corresponding user performing a predetermined behavior in the second predetermined time interval. The training sample set is input into the first calculation model, the first calculation model is called to determine the mapping relationship between the multiple historical indicator values ​​and the label, and the importance of each initial indicator is determined according to the mapping relationship, and the target indicator is output, wherein the importance of the target indicator satisfies a preset numerical condition.

3. The method according to claim 1, characterized in that, The second calculation process based on the target user's historical transaction behavior data within a first predetermined historical time interval, to obtain the second calculation result, includes: Based on the aforementioned historical transaction data, a time-series sequence of historical transaction behavior is constructed; A predetermined time series prediction model is invoked, and a sliding time window is used to extract multiple sets of target historical transaction behavior data corresponding to the sliding time window from the historical transaction behavior time series. Based on the multiple sets of target historical transaction behavior data, multiple trend data corresponding to the sliding time window are determined, wherein the trend data is used to characterize the behavioral risk change trend corresponding to the target user's transaction behavior within the sliding time window; The second calculation result is obtained by performing a second calculation based on the multiple trend data.

4. The method according to claim 3, characterized in that, The second calculation process based on the multiple trend data to obtain the second calculation result includes: Based on the time interval between the sliding time window corresponding to each of the trend data and the current time interval, and the behavioral trend represented by the trend data, the correlation between each of the trend data and the predetermined transaction risk level is determined. Based on the aforementioned correlation, determine the attention weight for each of the aforementioned trend data; The multiple trend data are weighted and summed according to the attention weight of each trend data; The second calculation result is determined based on the weighted summation result.

5. The method according to claim 3, characterized in that, The method includes: The window size of the sliding time window is determined based on the transaction scenario corresponding to the transaction behavior and the collection frequency of the historical transaction behavior data.

6. The method according to claim 1, characterized in that, The method further includes: The risk threshold corresponding to the current time interval is determined based on the transaction environment information corresponding to the current time interval and the user type of the target user. The transaction data processing results are matched with risk thresholds, and a credit limit determination strategy for the target user is determined based on the matching results.

7. The method according to claim 1, characterized in that, The first calculation process, based on at least one predetermined target indicator, on the current transaction behavior data to obtain a first calculation result includes: Determine at least one indicator value of the current transaction behavior data for the at least one target indicator; The first calculation result is obtained by weighting and summing the values ​​of the at least one target indicator according to their respective weights as predetermined.

8. A transaction data processing apparatus, characterized in that, The device includes: The first acquisition module is used to acquire current transaction behavior data corresponding to the target user's transaction behavior in the current time interval; A first processing module is configured to perform a first calculation on the current transaction behavior data based on at least one predetermined target indicator to obtain a first calculation result, wherein the first calculation result represents the probability that the target user will perform a predetermined behavior in the current time interval, and the target indicator is determined from the plurality of initial indicators according to the importance of each of the plurality of initial indicators. The second processing module is used to perform a second calculation based on the target user's historical transaction behavior data in a first predetermined historical time interval to obtain a second calculation result, wherein the second calculation result represents the probability that the target user will perform the predetermined behavior in a predetermined future time interval. The third processing module is used to fuse the first calculation result and the second calculation result to obtain the transaction data processing result corresponding to the target user.

9. An electronic device, comprising: One or more processors; Memory, used to store one or more computer programs. The characteristic feature is that the one or more processors execute the one or more computer programs to implement the steps of the method according to any one of claims 1 to 7.

10. A computer-readable storage medium having a computer program or instructions stored thereon, characterized in that, When the computer program or instructions are executed by a processor, they implement the steps of the method according to any one of claims 1 to 7.

11. A computer program product, comprising a computer program or instructions, characterized in that, When the computer program or instructions are executed by a processor, they implement the steps of the method according to any one of claims 1 to 7.

Images