Payment interface market dynamic matching and optimizing method

Abstract

The invention relates to the technical field of intelligent payment, and discloses a payment interface market dynamic matching and optimization method, which comprises the steps of obtaining multi-dimensional data, and constructing a user portrait based on the multi-dimensional data; based on a stream processing engine, aggregating the user behavior data, and updating a user portrait; identifying the user portrait based on the decision tree model, determining a transaction scene, constructing a differentiation strategy, and generating a payment mode recommendation list based on the differentiation strategy; inputting the user portrait into a pre-trained probability model for comparison, and updating a payment mode recommendation list based on a multi-target algorithm; determining a payment channel state based on the multi-dimensional data, obtaining a comprehensive score of a payment channel based on the payment channel state, and generating a candidate channel based on the comprehensive score; and obtaining an equipment health degree state, and determining an optimal payment channel based on the equipment health degree state. According to the invention, the operation fluency of a user in a complex use scene is improved.

Description

Technical Field

[0001] The present invention relates to the technical field of intelligent payment, and more specifically, to a method for dynamically matching and optimizing the payment interface market. Background Art

[0002] Under the background of the rapid development of the current digital economy, the payment interface market presents the characteristics of diversification, high-frequency updates, and high competition. The payment system not only needs to support traditional bank card payments but also needs to be compatible with various third-party payment methods such as WeChat Pay, Alipay, and UnionPay QuickPass, as well as meet the emerging payment needs that continue to emerge.

[0003] Currently, there are various payment solutions in the market, but most solutions still focus on the integration and management of a single payment channel and lack the ability to dynamically adapt to and optimize different payment scenarios and hardware devices. Moreover, existing payment solutions are often designed for specific payment channels or hardware devices and it is difficult to achieve seamless docking across channels and devices, resulting in technical barriers for merchants when accessing new payment channels or updating hardware devices. In complex and changing payment scenarios, the existing technology is difficult to achieve dynamic matching of payment interfaces, leading to cumbersome payment processes and slow response speeds, affecting the user experience. Therefore, there are still many deficiencies in existing payment solutions.

[0004] Therefore, it is necessary to design a method for dynamically matching and optimizing the payment interface market to solve the problems existing in the current technology. Summary of the Invention

[0005] In view of this, the present invention proposes a method for dynamically matching and optimizing the payment interface market, aiming to solve the problem of low efficiency of the current payment interface.

[0006] The present invention proposes a method for dynamically matching and optimizing the payment interface market, including: Obtain multi-dimensional data and construct a user profile based on the multi-dimensional data; Aggregate user behavior data based on a stream processing engine and update the user profile; Identify the transaction scenario based on a decision tree model for the user profile, construct a differentiation strategy, and generate a payment method recommendation list based on the differentiation strategy; Input the user profile into a pre-trained probability model and update the payment method recommendation list based on a multi-objective algorithm; Determine the payment channel status based on the multi-dimensional data, obtain the comprehensive score of the payment channel based on the payment channel status, and generate candidate channels based on the comprehensive score; Obtain the device health status and determine the optimal payment channel based on the device health status.

[0007] Further, when obtaining multi-dimensional data and constructing a user portrait based on the multi-dimensional data, it includes: Collect the user's historical transaction records through the payment gateway, determine the payment methods based on the historical transaction records, and count the usage frequency and success rate of each payment method to generate user payment habit tags; Collect device operation data through device fingerprints, construct a device compatibility matrix, and determine the current device characteristics based on the device compatibility matrix; Determine the payment area through the IP address and GPS positioning data, and determine the regional payment method white list based on the payment area; Construct a user portrait based on the regional payment method white list, the device characteristics, and the user payment habit tags.

[0008] Further, when real-time monitoring the payment channel and triggering the fusing mechanism when the health value of the payment channel is lower than the threshold, it includes: Construct a computing pipeline based on the Flink streaming computing engine and obtain the transaction data of the payment channel; Obtain the health metrics of the payment channel based on a sliding time window; The health metrics include the transaction success rate, average latency, and number of transactions; Perform weighted calculation based on the transaction success rate, the average latency, and the number of transactions to obtain the health value data; When the health value data is lower than the threshold, trigger weight degradation and update the weight parameters through the configuration library; When the health value data is lower than the threshold for 3 consecutive times, trigger the fusing mechanism and suspend the traffic allocation of the payment channel.

[0009] Further, when determining the payment area through the IP address and GPS positioning data, it includes: Convert the IP address into a country / city code based on the MaxMind GeoIP service; Convert the GPS positioning data into a standardized geographical location, fuse the country / city code with the standardized geographical location to obtain geographical fusion data; Establish a geographical database based on map data, determine the boundary information through the geographical database, and determine the shape and radius of the boundary information through the longitude and latitude coordinates; Emit horizontal rays based on the geographical fusion data, count the number of intersections with the boundary information, and when the number of intersections is odd, determine that the geographical fusion data is inside the boundary information, and when the number of intersections is even, determine that the geographical fusion data is outside the boundary information; Calculate the spherical distance between the geographical fusion data and the center point of the boundary information. When the distance is less than the radius, determine that the geographical fusion data is within the boundary information; when the distance is greater than the radius, determine that the geographical fusion data is outside the boundary information. Determine the payment area based on the spherical distance information and the horizontal ray information.

[0010] Further, when determining the regional payment method whitelist based on the payment area, it includes: Based on the determination result of the payment area, obtain the recommended payment method list of the payment area, and sort the payment methods in the recommended payment method list in combination with the user portrait. When both the IP address and the GPS positioning data are unavailable, use the user portrait data to determine the preferred payment method.

[0011] Further, when aggregating user behavior data and updating the user portrait, it includes: Obtain payment success / failure data, page view data, and device usage data, and combine the payment success / failure data, the page view data, and the device usage data to obtain the user behavior data. Subscribe to the user behavior data stream based on the stream processing engine, and remove duplicate data based on the device fingerprint. Aggregate the user behavior data based on a sliding window, group it based on the user ID, and store the cumulative metrics. Statistically analyze the cumulative metrics based on the Flink window, and update the user portrait weight based on the cumulative metrics.

[0012] Further, when constructing a differentiation strategy based on the decision tree model for the user portrait and identifying its transaction scenario, it includes: Obtain historical transaction scenarios and historical user portraits to generate a training set. Use information gain to screen key features based on the training set, train a decision tree based on Scikit-learn, and obtain the decision tree model. Input the transaction data and the user portrait into the decision tree model, obtain the scenario classification result, and generate a recommended payment method list based on the scenario classification result.

[0013] Further, when inputting the user portrait into a pre-trained probability model, it includes: Determine the global success rate of each payment method based on the Bayesian probability model for the user portrait, and sort the payment methods based on the global success rate. The global success rate includes the user habit matching probability, the device matching probability, and the geographical matching probability.

[0014] Further, when updating the payment method recommendation list based on the multi-objective algorithm, it includes: Calculating the priority based on the global success rate, transaction time consumption, and handling fee cost, and sorting the payment methods in the payment method recommendation list according to the priority; The priority is obtained through the following formula: ; where H is the user portrait, D is the device feature, L is the geographical location, is the priority, is the global success rate, is the user habit matching probability, is the device matching probability, is the geographical matching probability, is the joint probability; Sorting the payment methods in the payment method recommendation list according to the priority: ; where, is the priority of the payment method, is the historical average time consumption, is the handling fee rate of the payment method, i is the index of the payment method, , , is the weight, and + + = 1, n is the total number of payment methods.

[0015] Further, when obtaining the comprehensive score of the payment channel based on the payment channel status and generating candidate channels based on the comprehensive score, it includes: The comprehensive score of the payment channel is obtained through the following formula: ; where, is the comprehensive score, , , and are the weight coefficients, and + + + = 1, is the delay, is the device compatibility, is the handling fee rate; Sorting in descending order based on the comprehensive score, and selecting the top three channels as candidate channels.

[0016] Further, when obtaining the device health status and determining the optimal payment channel based on the device health status, it includes: Obtain the device health status and screen the payment channels. The device health status includes the camera status, NFC module status, network latency, and processor occupancy rate. Based on the device health status, eliminate the unavailable payment channels, sort the remaining payment channels by score, and determine the optimal payment channel.

[0017] Compared with the prior art, the beneficial effects of the present invention are as follows: By integrating multi-dimensional data such as user basic attributes, historical transaction records, behavior preferences, and device characteristics, it is possible to comprehensively depict the user portrait, break through the limitations of single-dimensional analysis, and more accurately identify their consumption ability, risk preference, and usage scenario characteristics. This helps to reduce the probability of misjudging user needs, provides a reliable basis for subsequent differential strategies, and at the same time reduces resource waste caused by data one-sidedness. Based on the stream processing engine to perform real-time aggregation and update of user behavior data, it can capture the instantaneous changes of user behavior, instantly adjust the evaluation of their payment willingness and decision-making anxiety level, improve the timeliness of the portrait, avoid the information lag problem in the traditional batch update mode, make the recommendation strategy closer to the user's current state, and reduce the decision-making deviation caused by stale information. Using the decision tree model to intelligently identify the transaction scenario, it can match different recommendation logics according to the scenario characteristics of the user (such as emergency recharge, cross-border shopping, large amount transfer). Through the hierarchical decision-making mechanism, it breaks through the rigidity of a single recommendation rule, improves the user operation efficiency, and reduces the payment interruption risk caused by strategy mismatch through scenario adaptation. Through the multi-objective algorithm to dynamically optimize the payment method recommendation list, it can coordinate the balance of indicators such as success rate, cost, and user experience, enhance the ability to handle complex business constraints, and reduce the strategic short-sightedness problem caused by goal fragmentation. Calculating the comprehensive score based on the real-time channel status data and generating a candidate channel set can avoid the risk of sudden failures or network congestion, improve the fault tolerance of the payment system, reduce the probability of transaction failure caused by single-point failures, and at the same time avoid local overload of channel resources through intelligent load balancing. By monitoring the device operation status (such as battery power, network stability, security environment), it can predict potential payment interruption risks and optimize the channel selection in advance, reduce the payment process interruption caused by terminal abnormalities, and at the same time improve the user's operation fluency in complex usage scenarios through environment adaptation. Description of the Drawings

[0018] By reading the following detailed description of the preferred embodiments, various other advantages and benefits will become clear to those of ordinary skill in the art. The drawings are only for the purpose of showing the preferred embodiments and are not considered to be a limitation of the present invention. Moreover, throughout the drawings, the same reference numerals are used to represent the same components. In the drawings: Figure 1Flowchart of the payment interface market dynamic matching and optimization method provided by the embodiments of the present invention. Detailed implementation manners

[0019] The exemplary embodiments of the present disclosure will be described in more detail below with reference to the accompanying drawings. Although the exemplary embodiments of the present disclosure are shown in the drawings, it should be understood that the present disclosure can be implemented in various forms and should not be limited by the embodiments set forth herein. On the contrary, these embodiments are provided so that the present disclosure can be more thoroughly understood and the scope of the present disclosure can be completely conveyed to those skilled in the art. It should be noted that, without conflict, the embodiments in the present invention and the features in the embodiments can be combined with each other. The present invention will be described in detail below with reference to the drawings and in conjunction with the embodiments.

[0020] In some embodiments of the present application, referring to Figure 1 as shown, a payment interface market dynamic matching and optimization method includes: S100: Obtain multi-dimensional data and construct a user profile based on the multi-dimensional data.

[0021] S200: Aggregate user behavior data based on a stream processing engine and update the user profile.

[0022] S300: Identify the user profile based on a decision tree model, determine the transaction scenario, construct a differentiation strategy, and generate a payment method recommendation list based on the differentiation strategy.

[0023] S400: Input the user profile into a pre-trained probability model and update the payment method recommendation list based on a multi-objective algorithm.

[0024] S500: Determine the payment channel status based on the multi-dimensional data, obtain the comprehensive score of the payment channel based on the payment channel status, and generate candidate channels based on the comprehensive score.

[0025] S600: Obtain the device health status and determine the optimal payment channel based on the device health status.

[0026] Specifically, the historical transaction records of users are captured through the payment gateway, and the usage frequency and success rate of each payment method (face recognition payment, QR code payment, NFC payment) are statistically analyzed to generate user payment habit tags. For example, if a user's face recognition payment accounts for 85% in the past 30 days, they are marked as "face recognition preference users". When constructing the user portrait, device fingerprint technology is also integrated to collect data such as device model, operating system version, and sensor types (such as NFC module, 3D structured light camera) to construct a device compatibility matrix. For example, when it is detected that the user's device is equipped with an NFC chip, the NFC payment function is automatically unlocked. Furthermore, the current location of the user portrait is determined through the geographical location, and based on the IP address and GPS positioning data, a regional payment method whitelist is defined in combination with the geofencing algorithm. For example, in a cross-border scenario, when the user's IP address is located overseas, payment interfaces that support cross-border settlement (such as Visa, MasterCard) are preferentially matched. The multi-dimensional data includes user payment preferences, device characteristics, and geographical location information. Through these data, a user portrait is constructed. Then, through the stream processing engine, window aggregation calculations are performed on user behavior data (such as single transaction duration, number of payment failure retries). Based on the real-time feature calculation built on Apache Flink, a stream-batch integrated processing mode is adopted, and the accuracy of out-of-order data processing is ensured through the EventTime semantics and Watermark mechanism. It supports multi-type window calculations of sliding windows (5-second granularity), session windows (30-second timeout), and global windows, and realizes aggregation analysis with millisecond-level latency for high-frequency user behaviors (such as payment requests). The fault tolerance mechanism of Flink ensures the Exactly-Once semantics through state snapshots (Checkpoint), and can guarantee the integrity and consistency of feature calculations even in case of node failures. It includes quantile values (P50 / P90 / P99) of single transaction duration and exponentially decaying counts of payment failure times, and realizes real-time statistics at the user granularity through rolling windows (1-minute granularity) and KeyedState. Abnormal behavior identifiers are generated by combining standard deviation and threshold rules (such as 3 consecutive payment failures within 1 minute), and user portraits are constructed by integrating activity metrics of multiple time windows (Z-Score normalization of daily / weekly / monthly operation frequencies). A dynamic weight adjustment mechanism is adopted during the calculation process, and business metric changes (such as fluctuations in payment success rate) are responded to through side output streams to achieve hot updates of feature weights. Then, based on transaction attributes (such as transaction amount, merchant type) and context information (such as network environment, timestamp), a decision tree model is used for scenario classification. For example, when a transaction occurs between 22:00 and 06:00 and the merchant type is "convenience store", it is determined as a "nighttime offline small-amount payment scenario". Through differential strategies, for example: when the device supports NFC → preferentially match NFC payment. Otherwise, when the geographical location is cross-border → match international credit cards.Otherwise, recommend in descending order according to the historical success rate. This rule base supports dynamic loading, and the policy weights can be adjusted in real time through the management background. Based on the above method, a preliminary payment method recommendation list is generated (the payment solutions in the payment method recommendation list are sorted according to historical preferences). Then, the priority of the payment methods in the payment method recommendation list is calculated through the Bayesian probability model, and a Top-3 payment method candidate list is generated for each transaction request. Furthermore, during the matching process, the success rate, transaction speed, and handling fee cost are balanced, and the linear programming model is used to solve the optimal solution. Finally, the optimal payment method is obtained. After the payment method is determined, the comprehensive quality score of the payment channel is determined by obtaining the status of the payment channel, and candidate channels for the payment channel are generated through the score to ensure that when there is a problem with the current channel, the candidate channels can be switched in real time without affecting the user's payment process. Since there are multiple candidate channels, the health of the payment device is monitored, and unusable payment channels are excluded. For example, if mobile payment is used but the mobile phone does not have NFC, the NFC payment channel is excluded, and the remaining channels are sorted according to the priority, and then the optimal payment channel is obtained.

[0027] It is understandable that by integrating multi-dimensional data such as user payment habits (e.g., payment method preferences, usage frequency), device characteristics (e.g., hardware compatibility, sensor types), and real-time geographical location, a dynamically updated user profile is constructed to achieve in-depth insight into user needs. For example, by combining device fingerprint technology with geofencing algorithms, the payment methods supported by the device (e.g., unlocking via NFC function) and regional compliance interfaces (e.g., cross-border settlement channels) are automatically adapted to avoid payment failure problems caused by hardware limitations or regional policies, and the compatibility and scenario coverage capabilities of payment interfaces are enhanced. Based on a decision tree model, transaction scenarios (e.g., amount range, merchant type, time window) are intelligently classified, and payment recommendation strategies highly matched with the scenarios are dynamically generated. For example, in the case of small-value night-time payments, quick payment methods are preferentially recommended, while for cross-border large-value transactions, high-security international channels are automatically matched, thus reducing the complexity of manual selection by users and shortening the transaction decision-making path. Using a stream processing engine to aggregate and calculate user payment behaviors (e.g., transaction duration, number of failed retries), the dynamic tags in the user profile (e.g., payment success rate trend, abnormal behavior identification) are updated in real time. Through the flexible combination of sliding windows and session windows, short-term changes in user behaviors (e.g., migration of payment habits) are quickly captured to provide immediate feedback for subsequent recommendation strategies. By introducing probability models and multi-objective algorithms, key indicators such as success rate, transaction speed, and handling fee cost are balanced in payment method recommendations. For example, the payment methods preferred by users in history are preferentially recommended to improve the success rate, and at the same time, the sorting is dynamically adjusted in combination with the handling fee cost of channels to avoid business losses caused by single-dimensional optimization and achieve the maximization of overall benefits. Based on the comprehensive scores of payment channels (e.g., response latency, success rate, load status), a list of candidate channels is generated to support fault switching and load balancing. When an abnormality occurs in the primary channel, the system can seamlessly switch to the backup channel to avoid transaction interruption caused by a single-point failure and ensure the continuity of the payment process. By real-time monitoring the device status (e.g., sensor availability, network stability), incompatible payment methods are automatically excluded (e.g., hiding the NFC payment option when there is no NFC module), and the optimal channel is recommended according to the device performance (e.g., selecting a payment interface with low bandwidth dependence in a weak network environment) to reduce the risk of payment failure caused by device problems.

[0028] In some embodiments of the present application, when obtaining multi-dimensional data and constructing a user profile based on the multi-dimensional data, it includes: Collecting user historical transaction records through a payment gateway, determining payment methods based on the historical transaction records, and counting the usage frequency and success rate of each payment method to generate user payment habit tags.

[0029] Collecting device operation data through device fingerprints, constructing a device compatibility matrix, and determining the current device characteristics based on the device compatibility matrix.

[0030] Determine the payment area through the IP address and GPS positioning data, and determine the regional payment method whitelist based on the payment area.

[0031] Construct a user profile based on the regional payment method whitelist, device characteristics, and user payment habit tags.

[0032] It can be understood that the original transaction data recorded by the payment gateway is obtained, including user ID, transaction time, payment method (such as face recognition payment / scan code payment / NFC payment, etc.), transaction status (success / failure), and duplicate transaction records are removed. For each user, the proportion of the usage times of each payment method in the total usage times is statistically calculated, and then the success transaction ratio of each payment method is determined. When a certain payment method meets one of the following conditions, it is marked as a preference tag: usage frequency ≥ 60%, success rate ≥ 95%. For the characteristics of the device fingerprint collection software and hardware, including device model, operating system version, sensor list (such as NFC module, camera module), network type, and then the device characteristics are converted into a numerical feature vector, and then the compatibility is detected. The payment area is determined through the IP address and GPS positioning to determine the payment area, and it is judged whether the user is located in a specific area (such as a cross-border area). The payment whitelist is that when the user is within the country, the payment whitelist includes but is not limited to scan code payment, face recognition payment, and UnionPay cards. When the user is outside the country, the payment whitelist includes but is not limited to Visa, MasterCard, and international scan code payment.

[0033] In some embodiments of the present application, when determining the payment area through the IP address and GPS positioning data, it includes: Convert the IP address into a country / city code based on the MaxMind GeoIP service.

[0034] Convert the GPS positioning data into a standardized geographical location, fuse the country / city code with the standardized geographical location, and obtain the geographical fusion data.

[0035] Establish a geographical database based on the map data, and determine the boundary information through the geographical database. Determine the shape and radius of the boundary information through the longitude and latitude coordinates.

[0036] Emit horizontal rays based on the geographical fusion data, and count the number of intersections with the boundary information. When the number of intersections is odd, it is determined that the geographical fusion data is inside the boundary information. When the number of intersections is even, it is determined that the geographical fusion data is outside the boundary information.

[0037] Calculate the spherical distance between the geographical fusion data and the center point of the boundary information. When the distance is less than the radius, it is determined that the geographical fusion data is inside the boundary information. When the distance is greater than the radius, it is determined that the geographical fusion data is outside the boundary information.

[0038] Determine the payment area based on spherical distance information and horizontal ray information.

[0039] It can be understood that the IP address of the user device is obtained and converted into geographical information. For example, the MaxMind GeoIP service API is called. The geographical information includes country code, city name, and longitude and latitude coordinates. Then, the original GPS coordinates of the device are converted into the WGS84 standard format. By judging the spherical distance between the IP address and the GPS coordinates, it is determined whether the gap is greater than 1 kilometer. If the gap is greater than 1 kilometer, the GPS coordinates are used as the standard. At the same time, a geographical database is established to store coordinate sequences (such as national boundaries) and store the center coordinates and radius (such as the coverage range of a business district). By judging the vertex coordinates to be judged and the vertex sequence of the coordinate sequence, the longitude and latitude are converted into a plane rectangular coordinate system. By emitting a horizontal ray from the point of the vertex coordinates to be judged to the right, and by counting the number of intersections between the ray and the coordinate sequence, when the number of intersections is odd, the vertex coordinates to be judged are inside the coordinate sequence, and when the number of intersections is even, the vertex coordinates to be judged are outside the coordinate sequence. Then, by calculating the vertex coordinates to be judged and the center coordinates and radius: 。

[0040] Among them, , , R is the radius of the earth, is the latitude difference (in radians), is the longitude difference (in radians), D is the spherical distance, is the dimension of the point to be judged (the current position of the user), is the latitude of the reference point (center coordinates and radius), is the longitude of the point to be judged, is the longitude of the reference point (center coordinates and radius). It is necessary to convert the longitude and latitude into radians. Radian = longitude and latitude * π / 180. If the spherical distance between the user coordinate point and the center point of the circular area ≤ radius, the determination result is passed. If the distance > radius, the determination result is not passed. For example, for a local preferential activity, the coordinate sequence is the administrative boundary of Shanghai, and the center coordinates and radius are centered on the center of Shanghai with a radius of 50 kilometers. Then, the user needs to be located inside the administrative boundary of Shanghai (horizontal ray method) and the distance from the city center ≤ 50 kilometers (spherical distance method) at the same time to enjoy the local preferential payment channel.

[0041] In some embodiments of the present application, when determining the regional payment method whitelist based on the payment area, it includes: Based on the determination result of the payment area, obtain the recommended payment method list of the payment area, and sort the payment methods in the recommended payment method list in combination with the user portrait.

[0042] When both the IP address and GPS positioning data are unavailable, user portrait data is used to determine the preferred payment method.

[0043] It can be understood that generating a recommendation list based on the payment area determination result can match the local mainstream payment tools with user habits. For example, for the Southeast Asian market, e-wallet payment methods are preferentially recommended, while in the European region, credit cards and local bank transfer channels are emphasized. This regional screening mechanism enhances the usability of payment options, reduces the risk of transaction interruption caused by the payment method not being supported by local merchants, and at the same time avoids compliance risks by following regional financial regulatory requirements. Introducing user portrait sorting on the basis of the regional whitelist achieves a balance between "regional commonalities" and "individual characteristics". For example, although a user is in a region with a high credit card penetration rate, historical data shows that they use mobile payment more frequently. In this case, mobile payment is recommended at the top. Through the dual screening logic, both the basic guarantee of regional adaptation is retained, and the operation inertia formed by users over a long time is respected, reducing the sense of experience fragmentation caused by overly mechanical recommendation strategies and avoiding decision-making fatigue caused by users having to repeatedly switch payment methods. When the IP and GPS positioning data fails, the preferred payment method is deduced through user portrait data, constructing a backup decision-making path that does not rely on geographical information. For example, if a user has frequently used a certain cross-border payment tool recently, or the device language is set to a specific region, their potential payment preferences can be inferred even if real-time location cannot be obtained. The fault tolerance setting breaks through the absolute dependence on a single data source, reduces the probability of service interruption caused by technical failures or permission restrictions, and ensures the completability of the payment process under extreme conditions. Deducing the payment method based on portrait data such as historical behavior and device characteristics continues the user's past operation habits. For example, if a user has successfully used the offline payment function in a no-network environment, when the positioning data is missing again, this method is preferentially recommended instead of forcing a jump to an unfamiliar channel, reducing the user's cognitive burden and operation complexity in abnormal scenarios and avoiding a decrease in trust or abandonment of the payment process due to sudden environmental changes.

[0044] In some embodiments of the present application, when aggregating user behavior data and updating the user portrait, it includes: Obtain payment success / failure data, page view data, and device usage data, and merge the above data with the device fingerprint to obtain user behavior data.

[0045] Subscribe to the user behavior data stream based on the stream processing engine, and remove duplicate data based on the device fingerprint.

[0046] Aggregate user behavior data based on a sliding window, group it based on the user ID, and store cumulative metrics.

[0047] Statistical cumulative metrics based on Flink windows, and update the user portrait weights based on the cumulative metrics.

[0048] It is understandable that by merging multi-dimensional data such as payment results, page views, and device operations, a three-dimensional record of user behavior is formed. For example, by associating the payment failure event with the duration data of the user viewing the payment instructions page before the failure, it is possible to distinguish between payment interruptions caused by operational errors and active abandonment due to unclear page information. This cross-link data association reduces the one-sidedness of single-event analysis and improves the accuracy of behavior attribution. Using device fingerprints (such as device models, operating system characteristics, network environment) as the basis for data merging solves the problem of user identification in cross-terminal and cross-session scenarios. For example, when a user logs in to the same account using a different mobile phone, the new device login behavior can be identified through the difference in device fingerprints, avoiding misclassifying the operation errors of different devices into the same user profile. This mechanism reduces the analysis deviation caused by identity confusion and enhances the correspondence between data and real users. Subscribing to data streams based on a stream processing engine enables real-time calculation. For example, when a user continuously tries multiple payment methods, the system can immediately capture their behavior sequence (such as first trying a credit card and failing, then switching to an e-wallet), rather than waiting for batch data synchronization to complete before processing. This low-latency processing mechanism reduces the risk of losing key behavior signals and ensures the synchronization of the user profile update with the user's current state. Removing duplicates from the data stream through device fingerprints identifies and filters redundant data that is reported repeatedly or collected abnormally. For example, when network fluctuations cause the same payment success event to be reported multiple times, duplicate records are automatically removed. This cleaning mechanism reduces the resource occupancy of invalid data for storage and calculation, and at the same time avoids statistical indicator distortion caused by data duplication. Aggregating data using a sliding window can identify sudden changes in behavior patterns within a short time window. For example, by counting cumulative metrics such as the duration of staying on the payment page and the number of times of returning to modify within the last 5 minutes of the user, it is possible to timely perceive whether there is confusion or a fluctuation in payment willingness in their current operation. This time-sensitive analysis breaks through the coarseness of traditional daily granularity statistics and improves the response sensitivity to immediate behavior changes. Storing cumulative metrics grouped by user ID retains the complete context of the behavior trajectory. For example, by correlating and analyzing metrics such as the frequency of device portrait and landscape switching and the number of page refreshes after the first payment failure of a certain user within a single day, it is possible to determine whether their subsequent operations are active retries or passive abandonments. This reduces the misjudgment probability of interpreting isolated events and enhances the depth of understanding of the continuity of user intentions.

[0049] In some embodiments of the present application, when constructing a differential strategy based on a decision tree model for the user profile and identifying its transaction scenario, it includes: Obtain historical transaction scenarios and historical user profiles to generate a training set.

[0050] Based on the training set, use information gain to screen key features, and train a decision tree based on Scikit-learn to obtain a decision tree model.

[0051] Input the transaction data and user profile into the decision tree model to obtain the scene classification result, and generate a recommended payment method list based on the scene classification result.

[0052] It can be understood that by inputting the scenes of historical transactions and user profiles into the decision tree model, the historical transaction scene data includes transaction time, amount, merchant type, payment method, and transaction result, and the historical user profile data includes payment habit tags, device compatibility levels, and geographical location whitelists. Extract all the above data for standardization processing, and then perform feature screening and decision tree training. By measuring the degree of chaos of the dataset: 。

[0053] Among them, is the entropy of the dataset S, is the proportion of the i-th type of sample in the dataset (such as the proportion of "night transactions"), c is the total number of categories (transaction scene types), and i is the i-th type of sample. Then, define the standard for selecting split features: 。

[0054] Among them, A is the candidate feature (such as the transaction amount range), is the subset where the value of feature A is v (such as the amount range = 100 - 500 yuan), is the total number of samples in the dataset, is the number of samples in the subset , and then train the decision tree to obtain a trained decision tree model.

[0055] In some embodiments of the present application, when inputting the user profile into the pre-trained probability model for comparison, it includes: Determine the global success rate of each payment method based on the user profile using the Bayesian probability model, and sort the payment methods based on the global success rate. The global success rate includes the user habit matching probability, device matching probability, and geographical matching probability.

[0056] In some embodiments of the present application, when updating the payment method recommendation list based on the multi-objective algorithm, it includes: Calculate the priority based on the global success rate, transaction time consumption, and handling fee cost, and sort the payment methods in the payment method recommendation list according to the priority.

[0057] The priority is obtained through the following formula: 。

[0058] Among them, H is the user profile, D is the device feature, L is the geographical location, is the priority, is the global success rate, is the matching probability of user habits, is the matching probability of the device, is the matching probability of geography, is the joint probability.

[0059] Sort the payment methods in the payment method recommendation list based on the priority: .

[0060] Among them, is the priority of the payment method, is the historical average time consumption, is the handling fee rate of the payment method, i is the index of the payment method, , , is the weight, and + + = 1, n is the total number of payment methods.

[0061] It is understandable that by inputting three types of data: user portraits, device characteristics, and geographical locations, the Bayesian model synthesizes the matching probabilities of user habits, device characteristics, and geographical locations to form a multi-dimensional global success rate assessment. For example, although a certain user often uses an e-wallet (high probability of user habits), but the current device network environment is unstable (low probability of device matching), or in a region outside the country where the wallet is not enabled (low probability of geographical matching), at this time the model automatically reduces the priority of this payment method, breaks through the recommendation inertia that simply relies on historical preferences through a three-dimensional assessment mechanism, and reduces the risk of payment failure caused by environmental incompatibility. The probability update characteristic of the Bayesian model can continuously absorb new data and adjust the weights of each dimension. For example, when the service stability of a certain payment channel decreases in a specific geographical area, the model automatically reduces the contribution value of the geographical matching probability in this area, rather than relying on a fixed rule library for manual adjustment. It reduces the lag of manual maintenance strategies and improves the response sensitivity to changes in the market environment. Optimize resource consumption based on handling fee costs while ensuring the success rate. For example, for small-value payment scenarios, preferentially recommend channels with similar success rates but lower handling fees instead of blindly choosing expensive channels with the highest absolute success rate, which reduces the comprehensive costs of payment service providers and at the same time avoids users abandoning payments due to excessive handling fees. Transaction time solves the potential conflict between payment speed and success guarantee. For example, in scenarios where users urgently need to complete transactions (such as flash sales), appropriately increase the weight of the time-consuming factor and preferentially recommend channels with faster processing speeds, even if their historical success rates are slightly lower. It reduces the experience rigidity caused by excessive pursuit of stability and improves the user's operation fluency in emergency scenarios. The weight parameters allow adjusting the target focus according to different business scenarios. For example, for transactions of high-net-worth users, reduce the weight of handling fees to improve the guarantee of payment success. For the promotion period, increase the weight of time-consuming to accelerate the conversion of the transaction funnel, enhancing the carrying capacity of the recommendation mechanism for diverse commercial demands.

[0062] In some embodiments of the present application, when obtaining the comprehensive score of the payment channel based on the payment channel state and generating candidate channels based on the comprehensive score, it includes: The comprehensive score of the payment channel is obtained through the following formula.

[0063] 。

[0064] Wherein, is the comprehensive score, , , and are weight coefficients, and + + + = 1, is the delay, For device compatibility, is the handling fee rate.

[0065] Based on the descending order of the comprehensive score, select the top three channels as candidate channels.

[0066] It can be understood that by comprehensively considering the three core indicators of latency, device compatibility, and handling fee rate, it is possible to avoid resource misallocation caused by a single indicator dominating the decision-making. For example, although a certain channel has the lowest latency but high handling fees, and another channel has low handling fees but poor device compatibility, the comprehensive scoring mechanism can identify intermediate options with moderate latency, good compatibility, and reasonable rates, reducing the overall efficiency loss caused by excessive pursuit of local optimality. For example, it can avoid a sharp increase in costs due to blindly choosing a low-latency channel, or sacrificing the payment success rate by overly suppressing the handling fees. The configurability of the weight coefficients allows for dynamically adjusting the evaluation focus according to business needs. For example: during the peak promotion period, focus on the latency indicator to prioritize the payment response speed. In response to the fragmented device situation in emerging markets, increase the weight of device compatibility to reduce payment interruptions caused by device incompatibility. For business lines sensitive to profit margins, focus on optimizing the handling fee rate to reduce the proportion of payment costs. By flexibly adjusting the weight ratio, the rigid constraints of strategy adjustment are reduced, and the ability to carry different business goals is enhanced. Selecting the top three channels with the highest comprehensive scores as candidates not only ensures the overall quality of the recommended options but also guards against sudden risks by retaining alternative channels. For example, when the latency of the preferred channel surges due to network fluctuations, it can be seamlessly switched to the second-best channel instead of triggering a full-channel evaluation again. This reduces the probability of transaction failures caused by single-point failures and also reduces the computational load of real-time decision-making through pre-screening.

[0067] In some embodiments of the present application, when obtaining the device health status and determining the optimal payment channel based on the device health status, it includes: Obtain the device health status, screen the payment channels. The device health status includes the camera status, NFC module status, network latency, and processor occupancy rate. Based on the device health status, eliminate the unavailable payment channels, sort the remaining payment channels by score, and determine the optimal payment channel.

[0068] It is understandable that by obtaining the operating status of key hardware such as cameras and NFC modules in real time, an accurate match between payment methods and device functions is established. For example, when the abnormal focus of the camera is detected, the QR code scanning payment channels are automatically excluded to avoid transaction interruption caused by the inability to recognize the QR code after the user selects it. When it is recognized that the power supply of the NFC module is unstable, the near-field payment option is actively blocked to prevent the risk of amount freezing caused by transaction interruption. Through hardware-level compatibility verification, the availability detection that only relies on software interfaces in the traditional sense is broken through, and the probability of implicit payment failure caused by physical device failures is reduced. A dynamic evaluation mechanism is established for device performance differences. For example, on devices with a processor occupancy rate exceeding the threshold, payment channels with a simple interaction process (such as one-click payment) are recommended first to avoid the complex verification process from exacerbating the system load and causing interface lags. The sense of experience fragmentation of low-end device users is reduced, and the universality of service coverage is improved. The optimal communication protocol is selected according to the real-time network latency characteristics. For example, in a network environment with high latency and high jitter, the payment channel using the UDP protocol is preferred, sacrificing some data integrity in exchange for the reachability of transaction requests. The TCP protocol channel is selected in a low-latency and stable network to ensure the complete verification of transaction data.

[0069] In summary, the beneficial effects of the present invention are as follows: By integrating multi-dimensional data such as user basic attributes, historical transaction records, behavior preferences, and device characteristics, it is possible to comprehensively depict the user portrait, break through the limitations of single-dimensional analysis, more accurately identify their consumption ability, risk preference, and usage scenario characteristics, help reduce the probability of misjudging user needs, provide a reliable basis for subsequent differentiation strategies, and at the same time reduce resource waste caused by data one-sidedness. Based on the stream processing engine, real-time aggregation and update of user behavior data can capture the instantaneous changes in user behavior, instantly adjust the assessment of their payment willingness and decision-making anxiety level, improve the timeliness of the portrait, avoid the information lag problem in the traditional batch update mode, make the recommendation strategy closer to the user's current state, and reduce decision-making biases caused by stale information. Using the decision tree model to intelligently identify transaction scenarios, it is possible to match different recommendation logics according to the characteristics of the user's current scenario (such as emergency recharge, cross-border shopping, large amount transfer). Through the hierarchical decision-making mechanism, the rigidity of a single recommendation rule is broken through, which not only improves the user operation efficiency but also reduces the payment interruption risk caused by strategy mismatch through scenario adaptation. By dynamically optimizing the payment method recommendation list through a multi-objective algorithm, it is possible to coordinate the balance of indicators such as success rate, cost, and user experience, enhance the ability to handle complex business constraints, and reduce strategy myopia problems caused by goal fragmentation. Calculating a comprehensive score based on real-time channel status data and generating a candidate channel set can avoid the risk of sudden failures or network congestion, improve the fault tolerance of the payment system, reduce the probability of transaction failures caused by single-point failures, and at the same time avoid local overload of channel resources through intelligent load balancing. By monitoring the device operating status (such as battery power, network stability, security environment), it is possible to predict potential payment interruption risks and optimize channel selection in advance, reduce payment process interruptions caused by terminal abnormalities, and at the same time improve the operation fluency of users in complex usage scenarios through environment adaptation.

[0070] Those skilled in the art should understand that the embodiments of the present application can be provided as a method, a system, or a computer program product. Therefore, the present application can adopt the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware aspects. Moreover, the present application can adopt the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0071] This application is described with reference to the flowcharts and / or block diagrams of methods, apparatus (systems) and computer program products according to embodiments of the present application. It should be understood that each flow and / or block in the flowchart and / or block diagram can be implemented by computer program instructions, as well as the combination of flows and / or blocks in the flowchart and / or block diagram. These computer program instructions can be provided to the processor of a general-purpose computer, a special-purpose computer, an embedded processor or other programmable data processing device to generate a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate means for implementing the functions specified in one or more of the flows Figure 1 one or more of the flows and / or blocks Figure 1 or means for implementing the functions specified in one or more of the blocks.

[0072] These computer program instructions can also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to work in a specific manner, so that the instructions stored in the computer-readable storage medium generate a manufactured article including instruction means that implement the functions specified in one or more of the flows Figure 1 one or more of the flows and / or blocks Figure 1 or means for implementing the functions specified in one or more of the blocks.

[0073] These computer program instructions can also be loaded onto a computer or other programmable data processing device, so that a series of operation steps are executed on the computer or other programmable device to generate a computer-implemented process, and thus the instructions executed on the computer or other programmable device provide steps for implementing the functions specified in one or more of the flows Figure 1 one or more of the flows and / or blocks Figure 1 or means for implementing the functions specified in one or more of the blocks.

[0074] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit them. Although the present invention has been described in detail with reference to the above embodiments, those of ordinary skill in the art should understand that: it is still possible to modify the specific implementation manners of the present invention or make equivalent replacements, and any modification or equivalent replacement that does not depart from the spirit and scope of the present invention should be covered by the protection scope of the claims of the present invention.

Claims

1. A method for dynamically matching and optimizing a payment interface market, characterized in that, Including: Obtain multi-dimensional data and construct a user profile based on the multi-dimensional data; Aggregate user behavior data based on a stream processing engine and update the user profile; Identify the user profile based on a decision tree model, determine a transaction scenario, construct a differentiation strategy, and generate a recommended payment method list based on the differentiation strategy; Input the user profile into a pre-trained probability model and update the recommended payment method list based on a multi-objective algorithm; Determine the payment channel status based on the multi-dimensional data, obtain the comprehensive score of the payment channel based on the payment channel status, and generate candidate channels based on the comprehensive score; Obtain the device health status and determine the optimal payment channel based on the device health status.

2. The payment interface market dynamic matching and optimization method according to claim 1, wherein When obtaining multi-dimensional data and constructing a user profile based on the multi-dimensional data, it includes: Collect the user's historical transaction records through a payment gateway, determine the payment methods based on the historical transaction records, and count the usage frequency and success rate of each payment method to generate user payment habit tags; Collect device operation data through a device fingerprint, construct a device compatibility matrix, and determine the current device characteristics based on the device compatibility matrix; Determine the payment area through the IP address and GPS positioning data, and determine the regional payment method whitelist based on the payment area; Construct a user profile based on the regional payment method whitelist, the device characteristics, and the user payment habit tags.

3. The payment interface market dynamic matching and optimization method according to claim 2, wherein When determining the payment area through the IP address and GPS positioning data, it includes: Convert the IP address into a country / city code based on the MaxMind GeoIP service; Convert the GPS positioning data into a standardized geographical location, fuse the country / city code with the standardized geographical location to obtain geographical fusion data; Establish a geographical database based on map data, determine the boundary information through the geographical database, and determine the shape and radius of the boundary information through the longitude and latitude coordinates; Emit a horizontal ray based on the geographical fusion data and count the number of intersection points with the boundary information. When the number of intersection points is odd, determine that the geographical fusion data is inside the boundary information. When the number of intersection points is even, determine that the geographical fusion data is outside the boundary information; Calculate the spherical distance between the geographical fusion data and the center point of the boundary information. When the distance is less than the radius, determine that the geographical fusion data is inside the boundary information. When the distance is greater than the radius, determine that the geographical fusion data is outside the boundary information; Determine the payment area based on the spherical distance information and the horizontal ray information.

4. The payment interface market dynamic matching and optimization method according to claim 3, characterized in that When determining the regional payment method whitelist based on the payment area, it includes: Based on the determination result of the payment area, obtain the recommended payment method list of the payment area, and sort the payment methods in the recommended payment method list in combination with the user profile; When both the IP address and GPS positioning data are unavailable, use the user profile data to determine the preferred payment method.

5. The payment interface market dynamic matching and optimization method according to claim 4, wherein When aggregating user behavior data and updating the user profile, it includes: Obtain payment success / failure data, page view data, and device usage data, and obtain the user behavior data from the payment success / failure data, the page view data, and the device usage data; Subscribe to the user behavior data stream based on the stream processing engine, and remove duplicate data based on the device fingerprint; Aggregate the user behavior data based on a sliding window, group it based on the user ID, and store the cumulative metrics; Statistically analyze the cumulative metrics based on Flink windows, and update the user profile weights based on the cumulative metrics.

6. The payment interface market dynamic matching and optimization method according to claim 5, wherein When constructing a differentiation strategy based on the decision tree model for the user profile and identifying its transaction scenarios, it includes: Obtain historical transaction scenarios and historical user profiles to generate a training set; Use information gain to screen key features based on the training set, train a decision tree based on Scikit-learn, and obtain the decision tree model; Input the transaction data and the user profile into the decision tree model, obtain the scenario classification result, and generate a recommended payment method list based on the scenario classification result.

7. The payment interface market dynamic matching and optimization method according to claim 6, characterized in that When inputting the user profile into a pre-trained probability model, it includes: Determine the global success rate of each payment method based on the Bayesian probability model for the user profile, and sort the payment methods based on the global success rate. The global success rate includes the user habit matching probability, the device matching probability, and the geographical matching probability.

8. The payment interface market dynamic matching and optimization method according to claim 7, characterized in that When updating the payment method recommendation list based on a multi-objective algorithm, it includes: Calculate the priority based on the global success rate, transaction time consumption, and handling fee cost, and sort the payment methods in the payment method recommendation list according to the priority; The priority is obtained through the following formula: ; Among them, H is the user profile, D is the device feature, L is the geographical location, is the priority, is the global success rate, is the probability of user habit matching, is the device matching probability, is the geographical matching probability, is the joint probability; Sort the payment methods in the payment method recommendation list according to the priority: ； wherein, is the priority of the payment method, is the historical average time consumption, is the handling fee rate of the payment method, and i is the index of the payment method, , , is the weight, and + + = 1, where n is the total number of payment methods.

9. The payment interface market dynamic matching and optimization method according to claim 8, characterized in that Obtain the comprehensive score of the payment channel based on the payment channel status. When generating candidate channels based on the comprehensive score, it includes: The comprehensive score of the payment channel is obtained through the following formula: ； Among them, is the comprehensive score, , , and are the weight coefficients, and + + + = 1, is the delay, is the device compatibility, is the handling fee rate; Arrange in descending order based on the comprehensive score, and select the top three channels as candidate channels.

10. The method for dynamically matching and optimizing the payment interface market according to claim 9, characterized in that, Obtain the device health status. When determining the optimal payment channel based on the device health status, it includes: Obtain the device health status and screen the payment channels. The device health status includes the camera status, NFC module status, network latency, and processor occupancy rate. Eliminate unavailable payment channels based on the device health status, sort the remaining payment channels by score, and determine the optimal payment channel.

Images