Payment tool intelligent recommendation method and system, storage medium and program product

Abstract

The application relates to the technical field of financial technology, and more particularly to a payment tool intelligent recommendation method, a payment tool intelligent recommendation system for implementing the method, a computer readable storage medium and a computer program product. The method comprises the following steps: in response to a user initiating a payment request, acquiring real-time scene data of a current payment scene; based on the real-time scene data and user historical behavior data, generating a user dynamic portrait feature; acquiring a benefit knowledge base, the benefit knowledge base containing benefit marketing activity information corresponding to a plurality of payment tools; based on the user dynamic portrait feature and the benefit knowledge base, generating a payment recommendation decision through a pre-trained intelligent recommendation model; and based on the payment recommendation decision, outputting a recommendation result containing a recommended payment tool and a recommendation reason thereof.

Description

Technical Field

[0001] This application relates to the field of financial technology, and more specifically to a payment instrument intelligent recommendation method, a payment instrument intelligent recommendation system implementing the method, a computer-readable storage medium, and a computer program product. Background Technology

[0002] With the widespread adoption of electronic payments and the rapid development of fintech, individual users typically possess multiple bank cards and other digital payment tools. Meanwhile, commercial banks, card organizations, and various payment platforms have launched diverse and complex promotional activities and offers to attract users and increase activity. Faced with increasingly abundant payment options and ever-changing promotional information, how users can conveniently and efficiently choose the payment tool that best suits their preferences and maximizes their immediate benefits during transactions has become crucial for improving the payment experience. Currently, some payment applications offer basic bank card lists or payment order settings based on simple rules (such as recent use), but further exploration is needed in dynamically, in real-time, and personalizedly matching users' current scenarios with massive amounts of promotional information.

[0003] It should be noted that the information disclosed in the background section above is only used to enhance the understanding of the background of this application, and therefore may include information that does not constitute prior art known to those skilled in the art. Summary of the Invention

[0004] To address or at least mitigate one or more of the above problems, embodiments of this application provide a payment tool intelligent recommendation method, a payment tool intelligent recommendation system implementing the method, a computer-readable storage medium, and a computer program product. This method uses a pre-trained intelligent recommendation model to match and decide on real-time generated user dynamic profiles with a real-time rights knowledge base, ultimately providing users with recommended payment tools with clear reasons during payment, thereby improving the efficiency of users' payment decisions and their rights usage experience.

[0005] According to a first aspect of this application, a method for intelligently recommending payment tools is provided. The method includes: in response to a user initiating a payment request, acquiring real-time scenario data of the current payment scenario; generating a dynamic user profile feature based on the real-time scenario data and the user's historical behavior data; acquiring a rights and benefits knowledge base, the rights and benefits knowledge base containing information on rights and benefits marketing activities corresponding to multiple payment tools; generating a payment recommendation decision based on the user dynamic profile feature and the rights and benefits knowledge base through a pre-trained intelligent recommendation model; and outputting a recommendation result containing the recommended payment tool and its recommendation reason based on the payment recommendation decision.

[0006] As an alternative or supplement to the above solutions, in a payment tool intelligent recommendation method according to an embodiment of this application, the real-time scene data includes at least one of the following: current transaction amount, merchant type, geographical location, payment time, and terminal device information that triggered the payment request.

[0007] As an alternative or supplement to the above solutions, in a payment tool intelligent recommendation method according to an embodiment of this application, the user's historical behavior data includes at least one of the following: a list of payment tools bound to the user, historical transaction records, historical participation records of benefit marketing activities, and geographical location access history.

[0008] As an alternative or supplement to the above solutions, in the intelligent recommendation method for payment tools according to an embodiment of this application, generating the user dynamic profile features includes: determining the user's basic preference weights on multiple preset dimensions based on the user's historical behavior data, wherein the multiple preset dimensions include at least two of consumption category, payment time period, geographical location, merchant type and benefit marketing activity category; and adjusting the basic preference weights based on the real-time scene data to generate a comprehensive feature set reflecting the user's current state as the user dynamic profile features.

[0009] As an alternative or supplement to the above solutions, in the payment tool intelligent recommendation method according to an embodiment of this application, the basic preference weight of the user in a specific consumption category is determined by statistically analyzing the ratio of the user's historical consumption frequency in that specific consumption category to the user's total historical consumption frequency.

[0010] As an alternative or supplement to the above solutions, in the payment tool intelligent recommendation method according to an embodiment of this application, the rights knowledge base is connected to the marketing systems of bank card organizations and banks through a real-time data interface, and the rights marketing activity information stored in the rights knowledge base includes at least one of the following: activity rules, the current stage of the activity, the remaining participation quota, the list of merchants associated with the activity, the type of rights, the conditions for achieving the rights, the value of the rights, and the validity period of the activity.

[0011] As an alternative or supplement to the above solutions, in the payment tool intelligent recommendation method according to an embodiment of this application, the types of benefits include at least one of the following: instant discounts, dynamic achievement benefits, task-based marketing activities, and points rewards.

[0012] As an alternative or supplement to the above solutions, in a payment tool intelligent recommendation method according to an embodiment of this application, generating payment recommendation decisions through a pre-trained intelligent recommendation model includes: performing multi-label dynamic matching of the user dynamic profile features with the benefit marketing activity information; and predicting the feasibility of the user completing the qualification conditions of each benefit marketing activity based on the matching results.

[0013] As an alternative or supplement to the above solutions, in a payment tool intelligent recommendation method according to an embodiment of this application, generating payment recommendation decisions through a pre-trained intelligent recommendation model further includes: determining the user's preference for various benefit marketing activities based on the user's dynamic profile features; comprehensively evaluating the preference and the feasibility to obtain a comprehensive evaluation result; and determining at least one recommended payment tool based on the comprehensive evaluation result.

[0014] As an alternative or supplement to the above solutions, in a payment tool intelligent recommendation method according to an embodiment of this application, the comprehensive evaluation is achieved by the following weighting method: multiplying the preference degree and the feasibility by preset weight coefficients respectively and summing them to obtain a comprehensive score; and determining at least one recommended payment tool based on the comprehensive score and according to preset selection rules.

[0015] As an alternative or supplement to the above solutions, in a payment tool intelligent recommendation method according to an embodiment of this application, the preset selection rules include at least one of the following: comparing the comprehensive score with a preset score threshold and selecting payment tools whose comprehensive scores reach the score threshold; selecting a predetermined number of payment tools ranked first according to the comprehensive scores from high to low; when multiple payment tools have the same comprehensive score, prioritizing the recommendation of payment tools with higher benefit value.

[0016] As an alternative or supplement to the above solutions, in a payment tool intelligent recommendation method according to an embodiment of this application, the recommendation reason is generated by large model language generation technology, which generates text information including the following content by parsing the matching tags and recommendation logic on which the payment recommendation decision is based: the name of the recommended payment tool, the name of the associated rights and benefits marketing activity, and the relevant benefits that the user can obtain from it.

[0017] As an alternative or supplement to the above solutions, a payment tool intelligent recommendation method according to an embodiment of this application further includes: receiving feedback data from the user on the recommendation result, wherein the feedback data includes: whether the user adopts the recommendation, the user's satisfaction score with the recommendation result, and whether the user successfully obtains the relevant rights after payment.

[0018] As an alternative or supplement to the above solutions, the payment tool intelligent recommendation method according to an embodiment of this application further includes: dynamically optimizing the parameters of the pre-trained intelligent recommendation model using a reinforcement learning algorithm based on the feedback data; and / or optimizing the logic for generating the user dynamic profile features using the feedback data.

[0019] As an alternative or supplement to the above solutions, in the intelligent recommendation method for payment tools according to an embodiment of this application, the feasibility prediction is also based on risk control and business rationality assessment, including: determining whether the current transaction amount complies with the risk control rules of the payment tool; and determining whether the current recommendation complies with the business rules of the payment tool itself.

[0020] As an alternative or supplement to the above solutions, in the payment tool intelligent recommendation method according to an embodiment of this application, the intelligent recommendation model adopts a collaborative filtering strategy, which analyzes the user's historical behavior and preferences, matches the current user with similar user groups, and generates payment recommendation decisions based on the usage of payment tools by similar user groups.

[0021] According to a second aspect of this application, a payment tool intelligent recommendation system is provided, comprising: a memory; a processor; and a computer program stored on the memory and executable on the processor, wherein execution of the computer program causes any one of the payment tool intelligent recommendation methods according to the first aspect of this application to be performed.

[0022] According to a third aspect of this application, a computer-readable storage medium is provided, the computer-readable storage medium including instructions that, when executed, perform any one of the payment tool intelligent recommendation methods according to the first aspect of this application.

[0023] According to a fourth aspect of this application, a computer program product is provided, comprising a computer program that, when executed by a processor, implements any one of the payment tool intelligent recommendation methods described in the first aspect of this application.

[0024] The intelligent payment tool recommendation scheme according to one or more embodiments of this application acquires scenario data in real time in response to payment requests, dynamically generates profile features by integrating user historical behavior, connects to a rights and benefits knowledge base, and uses a pre-trained intelligent recommendation model for personalized matching and decision-making, ultimately outputting payment tool recommendations with clear reasons to the user. This scheme automates and automates the payment decision-making process, enabling real-time matching of the most suitable payment tool and its associated benefits based on the user's current scenario and historical preferences. This significantly reduces the time and effort required for users to manually compare and select, improving the efficiency and smoothness of the payment process. Simultaneously, the output recommendation reasons enhance the transparency and explainability of the decision, helping users understand the basis for the recommendation, thereby improving user trust and overall experience. Furthermore, by accurately connecting user needs with the supply of benefits, this scheme effectively increases the exposure and potential utilization rate of the additional benefits of payment tools such as bank cards, providing a technical foundation for precise reach in benefits marketing. Attached Figure Description

[0025] The above and / or other aspects and advantages of this application will become clearer and more readily understood from the following description taken in conjunction with the accompanying drawings, in which the same or similar elements are denoted by the same reference numerals. In the drawings: Figure 1 A schematic flowchart illustrating a payment tool intelligent recommendation method 10 according to one or more embodiments of this application; and Figure 2 This is a schematic block diagram of a payment tool intelligent recommendation system 20 according to one or more embodiments of this application. Detailed Implementation

[0026] The following detailed description is merely exemplary in nature and is not intended to limit the disclosed technology or its application and use. Furthermore, it is not intended to be bound by any express or implied theory presented in the foregoing technical fields, background art, or the following detailed description.

[0027] In the following detailed description of the embodiments, numerous specific details are set forth in order to provide a more thorough understanding of the disclosed technology. However, it will be apparent to those skilled in the art that the disclosed technology can be practiced without these specific details. In other instances, well-known features have not been described in detail to avoid unnecessarily complicating the description.

[0028] Terms such as "comprising" and "including" indicate that, in addition to the units and steps that are directly and explicitly described in the specification, the technical solution of this application does not exclude the presence of other units and steps that are not directly or explicitly described. Terms such as "first" and "second" do not indicate the order of the units in terms of time, space, size, etc., but are merely used to distinguish the units.

[0029] Furthermore, it should be noted that the user information and data involved in one or more embodiments of this application (including but not limited to data used for analysis, stored data, and displayed data) are all information and data authorized by the user or fully authorized by all parties. The collection, use, and processing of such data must comply with the relevant laws, regulations, and standards of the relevant countries and regions, and corresponding operation entry points are provided for users to choose to authorize or refuse. This application attaches great importance to the security of user personal information and has adopted industry-standard and reasonable security protection measures to protect user information and prevent unauthorized access, public disclosure, use, modification, damage, or loss of personal information.

[0030] In the following, exemplary embodiments according to the present disclosure will be described in detail with reference to the accompanying drawings.

[0031] Please refer to the attached diagram below. Figure 1This is a schematic flowchart of a payment tool intelligent recommendation method 10 according to one or more embodiments of this application.

[0032] Method 10 can be implemented in various client applications, payment platform systems, or banking service systems that support payment functions. For example, when a user makes an online or offline payment using a mobile wallet app, bank app, or third-party payment platform, the system can automatically trigger this method. In an exemplary application scenario, a user uses a payment app to check out at a chain supermarket. When the user enters the payment interface and triggers a payment request, the payment system backend can automatically call this method, using an intelligent recommendation model to recommend the most suitable bank card for the payment, and display the recommended card type, the reason for the recommendation, and related benefits information on the payment interface, thereby assisting the user in making a quick payment decision.

[0033] like Figure 1 As shown, in step 101, in response to the user initiating a payment request, real-time scene data of the current payment scenario is obtained.

[0034] Real-time scenario data refers to dynamic information directly related to a user's current payment behavior. This information reflects the specific environment and conditions under which the payment occurs, providing accurate input for subsequent intelligent recommendations. By acquiring real-time scenario data, the system can promptly capture changes in the payment scenario, thereby enhancing the timeliness and scenario adaptability of the recommendation results.

[0035] In one or more embodiments, real-time scenario data may include at least one of the following: current transaction amount, merchant type, geographical location, payment time, and terminal device information that triggered the payment request. For example, the current transaction amount refers to the amount of funds involved in this payment operation, which can be directly parsed from the payment order; the merchant type refers to the category of the merchant conducting the transaction, such as a supermarket, restaurant, e-commerce platform, or offline specialty store, which can be obtained from the merchant code or partner interface; the geographical location refers to the physical or network location of the user when initiating the payment, which can be obtained through GPS, cell tower triangulation, or Internet Protocol addressing (IPA) technology; the payment time includes the specific date, day of the week, time, or time period, provided by the system clock or Network Time Protocol (NTP); and the terminal device information may include the user's device type, operating system version, device identifier, and network connection status, which can be collected through the device interface or user agent string. The comprehensive acquisition of this data helps the system characterize the real-time features of the payment scenario from multiple perspectives.

[0036] In one or more embodiments, real-time scene data is acquired through a data acquisition module built into the payment platform, which is tightly integrated with the payment request processing flow. When a user initiates a payment through the client application, the data acquisition module automatically intercepts and parses the payment request. Relevant scene parameters are extracted from the data. For example, transaction amount and merchant identifier can be directly read from transaction messages; geolocation information can be obtained in real time through device sensors or network positioning services; payment timestamps are automatically generated by the system; and terminal device information is obtained by calling the device's application programming interface (API). Furthermore, real-time scene data collection can be combined with external data sources, such as obtaining merchant details through near-field communication tags or QR code content, or enhancing indoor positioning accuracy through Bluetooth beacons. This multi-channel data integration ensures the integrity and real-time nature of scene information.

[0037] In one or more embodiments, the real-time scenario data can be further expanded to include other related information, thereby providing a more comprehensive description of the payment scenario. For example, it may include the transaction channel (such as online payment, offline QR code payment, near-field payment, etc.), the user's network environment (such as Wi-Fi, mobile data), and even contextual information such as weather conditions or the popularity of surrounding businesses.

[0038] The applicant reiterates that user privacy will be strictly protected during the implementation of this plan. For example, data collection will adhere to the principle of minimum necessity, collecting only information directly related to the recommendation function, and sensitive data will be anonymized or encrypted. Furthermore, the data acquisition process will strictly comply with relevant laws and regulations to ensure informed consent and data security for users.

[0039] In step 103, dynamic user profile features are generated based on real-time scene data and user historical behavior data.

[0040] User dynamic profile features refer to a structured feature representation that can comprehensively reflect a user's historical behavior patterns and current real-time status. Its core function is to combine static user preferences with dynamic payment scenarios, thereby providing accurate and timely user feature input for subsequent intelligent recommendations.

[0041] The generation of dynamic user profile features relies on the collection and analysis of users' historical behavior data. In one or more embodiments, user historical behavior data includes at least one of the following: a list of payment tools linked to the user, historical transaction records, records of participation in historical benefit marketing activities, and geolocation access history. Specifically, the list of payment tools linked to the user refers to the collection of payment methods such as bank cards and third-party payment accounts that the user has linked and can use for payments on payment platforms or applications; historical transaction records cover information such as the user's past transaction time, transaction amount, merchant, payment tool used, and transaction result; records of participation in historical benefit marketing activities detail the user's past behavior in registering for, achieving, or redeeming various benefit marketing activities; and geolocation access history records the locations the user has visited and their frequency through location services. This historical data collectively forms the basis for understanding users' long-term preferences and behavioral habits.

[0042] In one or more embodiments, the process of generating user dynamic profile features may include two stages.

[0043] The first phase involves determining the user's basic preference weights across multiple preset dimensions based on historical user behavior data. For example, these preset dimensions may include at least two of the following: consumption category, payment period, geographical location, merchant type, and benefit / marketing activity category. The basic preference weights aim to quantify the user's long-term tendencies across different dimensions. In one specific implementation, the user's basic preference weight in a specific consumption category (such as "dining," "shopping," or "transportation") can be calculated by comparing the user's historical purchases in that specific category with their total historical purchases. For instance, if a user has a total of 100 historical transactions in the past month, with 30 belonging to the "dining" category, their basic preference weight in the "dining" category can be calculated as 0.3. Similarly, the preference weight for payment time period can be quantified by analyzing the transaction frequency of users at different times, such as weekends and weekday nights; the preference weight for geographical location can be determined based on the frequency of users' visits or consumption in different areas (such as business districts, office buildings, and residential areas); the preference weight for merchant type can be calculated based on the proportion of users' consumption at different types of merchants, such as supermarkets, gas stations, and online platforms; and the preference weight for benefit marketing activity category can be evaluated based on the degree of users' historical participation in various benefit activities (such as instant discounts, points, and membership privileges). In one or more embodiments, the method for determining the basic preference weight is not limited to frequency statistics, but can also combine transaction amount, time decay factor, or use machine learning models (such as clustering and classification models) for more complex feature extraction and weight allocation, thereby more deeply mining user behavior patterns.

[0044] The second stage involves adjusting the weights of the aforementioned basic preferences based on real-time scenario data to generate a comprehensive feature set reflecting the user's current state as the user's dynamic profile features. Real-time scenario data (such as current transaction amount, merchant type, geographical location, and payment time) provides immediate context for weight adjustment. In one or more embodiments, the adjustment strategy may include: if the current transaction occurs at a merchant type that the user has historically preferred and during their active payment period, the weight of the corresponding dimension will be appropriately strengthened; if the current geographical location is a new area that the user rarely visits, the weight of the geographical location dimension can be temporarily reduced. In addition, if a specific benefit marketing activity is detected in the real-time scenario (such as a merchant offering an exclusive discount), the system can also dynamically increase the user's attention weight for that type of benefit. Through this dynamic adjustment, the final generated user dynamic profile features are no longer just an average reflection of historical behavior, but a more timely and personalized feature vector that integrates the characteristics of the current scenario, such as a multi-dimensional feature set represented as [consumption category weight, payment time weight, geographical location weight].

[0045] From a technical perspective, the user dynamic profile features generated in step 103 organically integrate users' long-term historical preferences with the current real-time scenario. This allows the system to not only understand what users "usually like" but also to perceive what users "might need at this moment," thereby significantly improving the personalization and scenario adaptability of recommendation results. For example, a user who usually prefers restaurant discounts, if paying at a gas station, the system can, through real-time scenario recognition and weight adjustment, appropriately increase their attention to gas station benefits and recommend bank cards that offer gas station cashback, achieving precise scenario-based outreach.

[0046] In one or more embodiments, the process of generating dynamic user profile features can also integrate richer data sources and more complex modeling techniques. For example, in addition to the dimensions mentioned above, static or semi-static labels such as user demographic attributes (e.g., age, occupation), device usage habits, and social network influence can also be considered. In terms of modeling techniques, deep learning models (e.g., recurrent neural networks or attention mechanisms) can be used to model user behavior sequences to capture the dynamic patterns of their preferences evolving over time. Furthermore, the generated feature set can also include cross-features or higher-order features constructed through feature engineering to reveal the correlations between different dimensions.

[0047] In step 105, the rights and interests knowledge base is obtained, which contains information on rights and interests marketing activities corresponding to multiple payment tools.

[0048] A rights and benefits knowledge base is a structured database or knowledge graph that centrally stores, organizes, and manages information on rights and benefits associated with various payment tools (such as credit cards, debit cards, and third-party payment accounts from different banks). Its core function is to provide intelligent recommendation systems with accurate, comprehensive, and timely data sources of rights and benefits. By maintaining such a knowledge base, the system can keep abreast of various available offers, rewards, and marketing rules in the market, thus providing crucial information for subsequent matching and recommendations.

[0049] In one or more embodiments, the rights and benefits knowledge base connects to the marketing systems of bank card organizations and individual banks via real-time data interfaces to ensure timely synchronization and updates of rights and benefits information. These interfaces can be implemented using data exchange technologies such as application programming interfaces (APIs) or message queues, supporting the periodic or trigger-based retrieval of the latest activity information from authoritative data sources. For example, when a bank launches a new "weekend spending cashback" promotion, its marketing system can push data such as the promotion rules, time, and scope of application to the rights and benefits knowledge base via the interface; similarly, the knowledge base can also query the card organization for global platform activities via the interface. This connection method ensures that the data in the rights and benefits knowledge base remains highly consistent with external business systems, avoiding the problem of recommendation failure due to information lag.

[0050] In one or more embodiments, the rights and benefits marketing activity information stored in the rights and benefits knowledge base may include at least one of the following: activity rules, current stage of the activity, remaining participation slots, list of merchants associated with the activity, rights type, rights qualification conditions, rights value, and activity validity period. Specifically, the activity rules detail the conditions required to participate in the activity, the benefits obtainable, and the specific operational procedures; the current stage of the activity indicates whether the activity is in the pre-heating period, in progress, about to end, or has ended; the remaining participation slots reflect the remaining capacity or slot limit of the activity; the list of merchants associated with the activity specifies which specific merchants or scenarios the rights can be used in; the rights type categorizes the nature of the benefits offered by the activity; the rights qualification conditions clarify what actions users need to complete (such as cumulative spending amount, number of transactions) to obtain the rights; the rights value quantifies the actual benefits that the rights can bring to users (such as discount amount, number of points, value of gifts); and the activity validity period specifies the start and end time when the rights can be obtained or used. This information collectively constitutes a complete "rights and benefits map," enabling the system to accurately assess the applicability and value of each rights and benefits.

[0051] In one or more embodiments, the types of benefits may include at least one of the following, but are not limited to: instant discounts, dynamic achievement benefits, task-based marketing activities, and points rewards. Instant discounts refer to benefits that directly reduce a certain amount at the time of payment, such as "10 off for every 100 spent"; dynamic achievement benefits typically refer to benefits that require users to accumulate spending or complete specified behaviors within a certain period to unlock, such as "a free monthly video card for spending 5 times this month"; task-based marketing activities require users to complete specific tasks (such as check-in, sharing, or browsing) to earn rewards; points rewards refer to benefits that accumulate points through transactions and can be redeemed for gifts or discounts. In addition, benefit types may also include discount coupons, cash back, interest-free installments, VIP services, physical gifts, membership privileges, and other forms. By clearly defining the types of benefits, the system can better understand the essence of the benefits and perform differentiated matching based on user preferences during recommendations.

[0052] From a technical perspective, the acquisition and maintenance of the rights and benefits knowledge base in step 105 provides a reliable and dynamically updated data foundation for the entire intelligent recommendation process. It not only ensures the high accuracy and timeliness of the rights and benefits information upon which recommendations are based, avoiding misleading users due to outdated information, but also enables the system to efficiently perform multi-dimensional matching and calculations through structured information storage. For example, when the system detects that a user is currently making a payment at a partner supermarket, it can quickly retrieve all rights and benefits activities associated with that supermarket and still valid from the knowledge base, and then combine this with user profiles for precise filtering. Furthermore, the real-time updating feature allows the system to promptly capture newly launched popular activities or scarce rights and benefits that are about to run out, thus prioritizing recommendations and increasing users' chances of obtaining high-quality rights and benefits. It also helps banks and merchants improve the reach and engagement of their marketing activities.

[0053] In one or more embodiments, the implementation of the rights knowledge base can also employ more advanced data management techniques to improve its performance and scalability. For example, distributed databases or caching mechanisms can be used to handle high-concurrency queries; natural language processing techniques can be used to parse and structure unstructured activity description text; or a graph database can be introduced to characterize the complex relationship network between rights, payment tools, merchants, and users to support more complex association reasoning and recommendations. Simultaneously, the management of the knowledge base can also include monitoring data quality, such as detecting and removing expired activities, identifying conflicting rules, and verifying data integrity to ensure its reliability as a basis for recommendations.

[0054] In step 107, a payment recommendation decision is generated based on the user's dynamic profile features and rights knowledge base through a pre-trained intelligent recommendation model.

[0055] This step is the core of the intelligent recommendation method. Its goal is to intelligently analyze, match, and make decisions based on the features reflecting the user's current state and the rich information on benefits generated in the previous steps, thereby outputting one or more recommended payment tools that best suit the user's current interests and needs. The intelligent recommendation model refers to a computational model that is pre-trained using artificial intelligence technologies such as machine learning and deep learning, capable of processing complex feature inputs and outputting decision suggestions. This model learns from massive amounts of historical transaction, user behavior, and benefit redemption data to grasp the complex correlations between user preferences, benefit value, and scenario conditions.

[0056] In one or more embodiments, the steps for generating payment recommendation decisions may include: First, performing multi-tag dynamic matching of user dynamic profile features with information on benefit marketing activities in the benefit knowledge base. Multi-tag dynamic matching refers to the process of comparing and calculating the correlation between multiple dimensional features (such as consumption category preference, geographical location preference, time preference, benefit type preference, etc.) contained in the user profile and multiple attribute tags (such as applicable merchants, applicable time periods, benefit type, target customer group, etc.) associated with each benefit activity in real time. For example, for the tags "dining consumption preference" and "weekend activity" in a user dynamic profile, the system will quickly filter out benefit activities with the tags "applicable to dining merchants" and "activity validity period includes weekends" from the benefit knowledge base. This matching is not a simple binary judgment, but rather forms a matching score by calculating the similarity or correlation strength between feature vectors, thereby dynamically and refinedly evaluating the degree of fit between each benefit and the current user.

[0057] Based on the matching results above, the model will further predict the feasibility of users meeting the eligibility criteria for various benefit marketing activities. Feasibility prediction refers to assessing how likely a user is to meet the eligibility criteria (such as cumulative spending amount, number of transactions, completion of specific tasks, etc.) in the present and foreseeable future. In one or more embodiments, feasibility prediction is not only based on the current single transaction amount, but also comprehensively considers information such as the user's historical spending capacity, spending frequency, recent behavioral trends, and the time window required to meet the benefit criteria. For example, for an activity that "accumulates spending of 5,000 yuan in the current month to obtain airport VIP lounge access," the model will predict the probability of meeting the criteria based on the user's spending amount this month, historical average monthly spending level, and the number of days remaining in the current month, using regression or classification algorithms. In one or more embodiments, feasibility prediction can also be based on risk control and business rationality assessment, including determining whether the current transaction amount complies with the risk control rules of the payment instrument (e.g., whether a single transaction exceeds the card's transaction limit or triggers an abnormal transaction risk control model), and determining whether the current recommendation complies with the payment instrument's own business rules (e.g., some benefits may be limited to new users or users with specific card levels, and the system needs to verify user eligibility). This comprehensive assessment ensures that the recommendations are not only attractive to users, but also robust and actionable from a business and risk perspective.

[0058] In one or more embodiments, the payment recommendation decision generation step may further include: determining the user's preference for various benefit marketing activities based on the user's dynamic profile characteristics. Preference is a quantification of the user's subjective liking for different types of benefits, which may stem from the analysis of data such as the user's historical participation in similar activities, clicks, or attention behavior regarding specific benefit types. Next, the calculated preference and predicted feasibility can be comprehensively evaluated to obtain a comprehensive evaluation result. In one specific implementation, this comprehensive evaluation can be achieved through a weighted approach: multiplying preference and feasibility by preset weight coefficients and then summing the results to obtain a comprehensive score. The weight coefficients can be dynamically configured according to business objectives; for example, when focusing on improving user satisfaction, the weight of preference can be increased; when focusing on improving benefit redemption rates, the weight of feasibility can be increased. Finally, based on the comprehensive evaluation result, at least one recommended payment tool is determined. In one or more embodiments, the determination of the recommended payment tool can be based on preset selection rules. The preset selection rules may include at least one of the following: comparing the overall score with a preset scoring threshold and selecting payment tools whose overall scores meet the threshold; selecting a predetermined number (e.g., the top 3) of payment tools ranked from highest to lowest overall score; and prioritizing the payment tools with higher benefit value or lower thresholds when multiple payment tools have the same overall score. These rules ensure the optimal selection and practicality of the recommendation results.

[0059] The implementation strategies for intelligent recommendation models can be diverse. In one or more embodiments, the intelligent recommendation model employs a collaborative filtering strategy. By analyzing massive amounts of users' historical behavior and preference data, it matches the current user with "similar user groups" whose behavioral patterns are similar. Based on the frequency of use, satisfaction, or success rate of acquiring benefits for different payment tools within these similar user groups, it infers and generates payment recommendation decisions for the current user. Besides collaborative filtering, the model can also employ or combine other strategies, such as content-based recommendations (directly matching user characteristics and benefit attributes), decision tree or random forest-based models (handling structured rules and features), or neural network-based deep learning models (handling complex nonlinear relationships and high-dimensional features). The model is typically continuously optimized through offline training and online updates. Training data includes historical transaction records, user profiles, benefit information, and user feedback data on historical recommendations.

[0060] From a technical perspective, step 107, through a pre-trained intelligent recommendation model, achieves a crucial transformation from data to intelligent decision-making. It integrates seemingly discrete user characteristics and benefit information into a clearly targeted recommendation decision through a complex computational model. This process not only significantly improves the accuracy of recommendations—ensuring that the recommended payment tools are indeed those that users like and can genuinely enjoy their benefits—but also greatly enhances personalization, achieving precise marketing tailored to each individual. Simultaneously, by incorporating feasibility prediction and risk control, the recommendation results are both attractive and business-reasonable, avoiding ineffective recommendations or inducing users to perform impossible tasks. This, in turn, improves user experience and benefit utilization while ensuring the security and stability of the payment business.

[0061] In step 109, based on the payment recommendation decision, a recommendation result containing the recommended payment tool and the reasons for the recommendation is output.

[0062] This step is the output stage of the intelligent recommendation process, aiming to transform machine decisions into interactive information that users can intuitively understand and easily operate, thereby directly guiding users' payment behavior and improving their decision-making efficiency and experience. The output recommendation results are usually presented in the payment application interface in the form of graphical user interface elements, text messages, or voice prompts. Its core components include the recommended payment tool (such as the specific bank card name or icon) and an explanation of why the tool is recommended.

[0063] In one or more embodiments, the recommendation reason is generated using large-model language generation technology. Large-model language generation technology refers to the natural language generation capability based on a pre-trained large language model, which can automatically generate fluent, accurate, and easily understandable text descriptions based on input information. In such embodiments, the key matching tags and inherent recommendation logic upon which the payment recommendation decision is based can be analyzed first, such as matched user preference tags (e.g., "weekend dining enthusiast"), applicable benefit activity tags (e.g., "Saturday discount activity"), and feasibility prediction results (e.g., "high probability of meeting the target"). Subsequently, the large model can use these structured logics as input prompts and, through its learned language rules and domain knowledge, generate a complete and personalized recommendation reason text. In one specific implementation, the generated text information includes the name of the recommended payment tool, the name of the associated benefit marketing activity, and the relevant benefits that the user can obtain from it. For example, the reason given might be: "We recommend using 'XX Bank Credit Card' for payment. Your purchase qualifies you for the 'Weekend Food Carnival' event, where you can enjoy a 30 RMB discount on purchases over 200 RMB and earn double points." In this way, the recommendation not only tells users "what to use" but also explains "why," enhancing the transparency and persuasiveness of the recommendation. This helps users quickly understand the value of the recommendation, thereby increasing their willingness to adopt it.

[0064] In one or more embodiments, the output of the recommendation result is not the end point of the process; a closed-loop optimization mechanism can also be added. Specifically, method 10 may further include: receiving user feedback data on the recommendation result. Feedback data is the user's direct or indirect response to the system's recommendation, and its collection can be done explicitly or implicitly. For example, explicit feedback includes providing a satisfaction rating button or an option to accept or reject the recommendation on the payment interface; implicit feedback can be inferred indirectly by monitoring user behavior, such as whether the user completed the payment according to the recommendation, whether they viewed the details of the benefits after payment, and whether any subsequent redemption behavior related to the benefits occurred. Specifically, feedback data may include whether the user accepted the recommendation, their satisfaction rating of the recommendation result, and whether they successfully obtained the relevant benefits after payment. This data reflects the recommendation effect and the user's true feelings in real time.

[0065] Furthermore, in one or more embodiments, method 10 may further include: continuously optimizing the system based on the collected feedback data. Optimization can be reflected in two aspects: First, using the feedback data, the parameters of the pre-trained intelligent recommendation model are dynamically tuned through a reinforcement learning algorithm. The reinforcement learning algorithm treats positive feedback such as user adoption of recommendations and acquisition of benefits as reward signals, and negative feedback such as ignoring recommendations or failing to meet targets as penalty signals, continuously adjusting the policy function and value function in the model, thereby making the model more inclined to produce decisions that yield higher cumulative rewards (i.e., higher user satisfaction and higher benefit utilization) in subsequent recommendations. Second, the logic for generating dynamic user profile features can be optimized using the feedback data. For example, if the system finds that a user frequently adopts recommendations for "coffee benefits," the weight of their "coffee consumption preference" in the user profile can be strengthened accordingly; conversely, if a user repeatedly ignores a certain type of benefit recommendation, their preference estimate for that type of benefit can be lowered accordingly. This optimization enables the user profile to more sensitively and accurately reflect their true and potentially evolving interests.

[0066] From a technical perspective, step 109 and its associated feedback and optimization mechanisms together form a complete "recommendation-feedback-learning" closed loop. Providing clear justifications enhances the system's explainability and user trust, improving the interactive experience. The collection and utilization of feedback data continuously injects practical application effects into the system, driving the self-evolution and improvement of the intelligent recommendation model and user profile generation logic. This ensures that the entire recommendation system can continuously adapt and optimize over time, adapt to changes in user preferences, and evolve market activities, thereby maintaining its recommendation accuracy, personalization level, and user satisfaction in the long term, achieving a leap from static intelligent recommendation to dynamic, self-learning intelligent recommendation services.

[0067] Figure 2This is a schematic block diagram of a payment tool intelligent recommendation system 20 according to one or more embodiments of this application.

[0068] For example, the payment tool intelligent recommendation system 20 can be deployed in various computing environments, such as integrated into the back-end server cluster of a payment platform, running as an auxiliary decision-making module of a bank's core system, or built into the intelligent POS system of a large merchant.

[0069] like Figure 2 As shown, the payment tool intelligent recommendation system 20 includes a memory 210, a processor 220, and a computer program 230 stored on the memory 210 and executable on the processor 220. The memory 210 stores data and program instructions, and may be volatile memory (such as random access memory), non-volatile memory (such as flash memory, hard disk), or a combination of both. The processor 220 may be a central processing unit, a graphics processing unit, a dedicated artificial intelligence chip, or a combination thereof, and is responsible for executing the instructions in the computer program 230 to perform logical and arithmetic operations. The computer program 230 contains a series of executable instructions, which, when loaded and run by the processor 220, cause the processor 220 to perform operations such as... Figure 1 The payment tool intelligent recommendation method 10 shown includes all or part of the steps to achieve the intelligent recommendation function. In one or more embodiments, the payment tool intelligent recommendation system 20 may also include necessary communication interfaces (not shown in the figure), such as a network interface card, for exchanging data with user terminals, bank card organization systems, bank marketing systems, and merchant systems to receive payment requests, obtain real-time scenario data and rights information, and return recommendation results.

[0070] The operating logic of computer program 230 and Figure 1 The steps of method 10 shown are closely related. Specifically, the execution of program 230 enables processor 220 to perform the following operations: in response to a user payment request received through a communication interface, obtain real-time scenario data of the current payment scenario; retrieve the user's historical behavior data from memory 210 or from an external user database through an interface, and execute the logic for generating user dynamic profile features based on the real-time scenario data and the user's historical behavior data; obtain a rights knowledge base containing information on rights marketing activities corresponding to multiple payment tools from a connected rights knowledge base or external marketing system through a communication interface; call a pre-trained intelligent recommendation model, and perform model inference based on the generated user dynamic profile features and the obtained rights knowledge base to generate a payment recommendation decision; finally, based on the decision, generate a recommendation result containing recommended payment tools and their recommendation reasons, and output it to the user terminal that initiated the payment request through a communication interface.

[0071] In one or more embodiments, the software functional modules of the payment tool intelligent recommendation system 20 logically correspond to the steps of method 10 and can be implemented through different code segments or sub-modules of program 230. For example, program 230 may include a data acquisition and processing module, specifically responsible for acquiring and preliminarily processing real-time scene data in step 101; a user profile construction module, used to generate user dynamic profile features in step 103; a rights management module, used to connect, query, and maintain the rights knowledge base in step 105; an intelligent recommendation engine module, which encapsulates a pre-trained recommendation model and related matching, prediction, and evaluation algorithms, used to generate the core decision in step 107; and a result generation and interaction module, used to format recommendation results, generate recommendation reasons, and receive feedback data in step 109. These modules can be logically divided in program 230 and executed sequentially or in parallel by processor 220 to jointly complete the intelligent recommendation task.

[0072] Alternatively, this application can also be implemented as a computer-readable storage medium storing information for causing a computer to perform, such as Figure 1 The procedure of the steps in the method shown. Here, computer-readable storage media can be various types of computer-readable storage media, such as disks (e.g., magnetic disks, optical disks, etc.), cards (e.g., memory cards, optical cards, etc.), semiconductor memory (e.g., ROM, non-volatile memory, etc.), and tapes (e.g., magnetic tape, cassette tape, etc.).

[0073] This disclosure can also be implemented as a computer program product comprising a computer program that, when executed by a processor, implements, as described above. Figure 1 The procedure for the steps in the method shown.

[0074] Where applicable, the various embodiments provided in this application may be implemented using hardware, software, or a combination of hardware and software. Furthermore, where applicable, without departing from the scope of this application, the various hardware and / or software components described herein may be combined into composite components comprising software, hardware, and / or both. Where applicable, without departing from the scope of this application, the various hardware and / or software components described herein may be divided into sub-components comprising software, hardware, or both. Additionally, where applicable, it is contemplated that software components may be implemented as hardware components, and vice versa.

[0075] The software (such as program code and / or data) according to this application may be stored on one or more computer-readable storage media. It is also contemplated that the software identified herein may be implemented using one or more networked and / or otherwise general-purpose or special-purpose computers and / or computer systems. Where applicable, the order of the various steps described herein may be changed, combined into compound steps, and / or divided into sub-steps to provide the features described herein.

[0076] The embodiments and examples presented herein are provided to best illustrate embodiments of this application and its particular applications, thereby enabling those skilled in the art to implement and use this application. However, those skilled in the art will understand that the above description and examples are provided for ease of illustration and example only. The descriptions presented are not intended to cover all aspects of this application or to limit this application to the precise forms disclosed.

Claims

1. A method for intelligently recommending payment tools, characterized in that, The method includes: In response to a user's payment request, obtain real-time scenario data for the current payment scenario; Based on the real-time scene data and user historical behavior data, dynamic user profile features are generated. Access the rights and benefits knowledge base, which contains information on rights and benefits marketing activities corresponding to multiple payment tools; Based on the user dynamic profile features and the rights knowledge base, a payment recommendation decision is generated through a pre-trained intelligent recommendation model; and Based on the payment recommendation decision, a recommendation result containing the recommended payment tool and the reasons for the recommendation is output.

2. The method as described in claim 1, wherein, The real-time scenario data includes at least one of the following: current transaction amount, merchant type, geographical location, payment time, and terminal device information that triggered the payment request.

3. The method as described in claim 1, wherein, The user's historical behavior data includes at least one of the following: a list of payment tools linked to the user, historical transaction records, records of participation in historical benefit marketing activities, and geolocation access history.

4. The method of claim 1, wherein, The features for generating the user dynamic profile include: Based on the user's historical behavior data, the user's basic preference weights across multiple preset dimensions are determined, wherein the multiple preset dimensions include at least two of the following: consumption category, payment time period, geographical location, merchant type, and benefit marketing activity category; and Based on the real-time scene data, the basic preference weights are adjusted to generate a comprehensive feature set that reflects the user's current state as the user dynamic profile feature.

5. The method of claim 4, wherein, The basic preference weight of a user in a specific consumption category is determined by statistically analyzing the ratio of the user's historical consumption frequency in that specific consumption category to the user's total historical consumption frequency.

6. The method of claim 1, wherein, The rights and benefits knowledge base is connected to the marketing systems of bank card organizations and banks through a real-time data interface, and the rights and benefits marketing activity information stored in the rights and benefits knowledge base includes at least one of the following: activity rules, current stage of the activity, remaining participation slots, list of merchants associated with the activity, rights and benefits type, rights and benefits eligibility conditions, rights and benefits value, and activity validity period.

7. The method of claim 6, wherein, The types of benefits include at least one of the following: instant discounts, dynamic achievement benefits, task-based marketing activities, and points rewards.

8. The method of claim 1, wherein, Generating payment recommendation decisions through pre-trained intelligent recommendation models includes: The user dynamic profile features are dynamically matched with the rights and benefits marketing activity information using multiple tags; Based on the matching results, the feasibility of the user completing the target conditions for each benefit marketing activity is predicted.

9. The method of claim 8, wherein, Generating payment recommendation decisions through pre-trained intelligent recommendation models also includes: Based on the user dynamic profile characteristics, the user's preference for each benefit marketing activity is determined; A comprehensive evaluation of the preference and the feasibility is conducted to obtain a comprehensive evaluation result. Based on the comprehensive evaluation results, at least one recommended payment tool is identified.

10. The method of claim 9, wherein, The comprehensive evaluation is achieved through the following weighting method: the preference degree and the feasibility are multiplied by preset weight coefficients respectively and then summed to obtain a comprehensive score; and at least one recommended payment tool is determined based on the comprehensive score and according to preset selection rules.

11. The method according to claim 10, wherein, The preset selection rules include at least one of the following: The comprehensive score is compared with a preset score threshold, and a payment tool whose comprehensive score reaches the score threshold is selected. Based on the comprehensive scores, from highest to lowest, select the predetermined number of payment tools that rank highest. When multiple payment tools have the same overall score, the payment tool with higher benefit value will be given priority.

12. The method of claim 8 or 9, wherein, The feasibility prediction is also based on risk control and business rationality assessment, including: Determine whether the current transaction amount complies with the risk control rules of the payment instrument; and / or Determine whether this recommendation complies with the business rules of the payment tool itself.

13. The method of claim 1, wherein, The recommendation reason is generated using large model language generation technology, which generates text information including the following content by parsing the matching tags and recommendation logic on which the payment recommendation decision is based: the name of the recommended payment tool, the name of the associated rights and benefits marketing campaign, and the relevant benefits that users can obtain from it.

14. The method of claim 1, wherein, The method further includes: The system receives feedback data from the user regarding the recommendation results, wherein the feedback data includes: whether the user accepts the recommendation, the user's satisfaction rating with the recommendation results, and whether the user successfully obtains the relevant rights after payment.

15. The method of claim 14, wherein, The method further includes: Based on the feedback data, the parameters of the pre-trained intelligent recommendation model are dynamically optimized using a reinforcement learning algorithm; and / or The logic for generating the user dynamic profile features is optimized using the feedback data.

16. The method of claim 1, wherein, The intelligent recommendation model employs a collaborative filtering strategy. By analyzing users' historical behavior and preferences, it matches the current user with similar user groups and generates payment recommendation decisions based on the usage of payment tools by these similar user groups.

17. A payment tool intelligent recommendation system, characterized in that, It includes: a memory; a processor; and a computer program stored in the memory and executable on the processor, the execution of which causes the following operations: In response to a user's payment request, obtain real-time scenario data for the current payment scenario; Based on the real-time scene data and user historical behavior data, dynamic user profile features are generated. Access the rights and benefits knowledge base, which contains information on rights and benefits marketing activities corresponding to multiple payment tools; Based on the user dynamic profile features and the rights knowledge base, a payment recommendation decision is generated through a pre-trained intelligent recommendation model; as well as Based on the payment recommendation decision, a recommendation result containing the recommended payment tool and the reasons for the recommendation is output.

18. The system of claim 17, wherein, The real-time scenario data includes at least one of the following: current transaction amount, merchant type, geographical location, payment time, and terminal device information that triggered the payment request.

19. The system of claim 17, wherein, The user's historical behavior data includes at least one of the following: a list of payment tools linked to the user, historical transaction records, records of participation in historical benefit marketing activities, and geolocation access history.

20. The system of claim 17, wherein, The features for generating the user dynamic profile include: Based on the user's historical behavior data, the user's basic preference weights across multiple preset dimensions are determined, wherein the multiple preset dimensions include at least two of the following: consumption category, payment time period, geographical location, merchant type, and benefit marketing activity category; and Based on the real-time scene data, the basic preference weights are adjusted to generate a comprehensive feature set that reflects the user's current state as the user dynamic profile feature.

21. The system of claim 20, wherein, The basic preference weight of a user in a specific consumption category is determined by statistically analyzing the ratio of the user's historical consumption frequency in that specific consumption category to the user's total historical consumption frequency.

22. The system of claim 17, wherein, The rights and benefits knowledge base is connected to the marketing systems of bank card organizations and banks through a real-time data interface, and the rights and benefits marketing activity information stored in the rights and benefits knowledge base includes at least one of the following: activity rules, current stage of the activity, remaining participation slots, list of merchants associated with the activity, rights and benefits type, rights and benefits eligibility conditions, rights and benefits value, and activity validity period.

23. The system of claim 22, wherein, The types of benefits include at least one of the following: instant discounts, dynamic achievement benefits, task-based marketing activities, and points rewards.

24. The system of claim 17, wherein, Generating payment recommendation decisions through pre-trained intelligent recommendation models includes: The user dynamic profile features are dynamically matched with the rights and benefits marketing activity information using multiple tags; Based on the matching results, the feasibility of the user completing the target conditions for each benefit marketing activity is predicted.

25. The system of claim 24, wherein, Generating payment recommendation decisions through pre-trained intelligent recommendation models also includes: Based on the user dynamic profile characteristics, the user's preference for each benefit marketing activity is determined; A comprehensive evaluation of the preference and the feasibility is conducted to obtain a comprehensive evaluation result. Based on the comprehensive evaluation results, at least one recommended payment tool is identified.

26. The system of claim 25, wherein, The comprehensive evaluation is achieved through the following weighting method: the preference degree and the feasibility are multiplied by preset weight coefficients respectively and then summed to obtain a comprehensive score; and at least one recommended payment tool is determined based on the comprehensive score and according to preset selection rules.

27. The system according to claim 26, wherein, The preset selection rules include at least one of the following: The comprehensive score is compared with a preset score threshold, and a payment tool whose comprehensive score reaches the score threshold is selected. Based on the comprehensive scores, from highest to lowest, select the predetermined number of payment tools that rank highest. When multiple payment tools have the same overall score, the payment tool with higher benefit value will be given priority.

28. The system of claim 24 or 25, wherein, The feasibility prediction is also based on risk control and business rationality assessment, including: Determine whether the current transaction amount complies with the risk control rules of the payment instrument; and / or Determine whether this recommendation complies with the business rules of the payment tool itself.

29. The system of claim 17, wherein, The recommendation reason is generated using large model language generation technology, which generates text information including the following content by parsing the matching tags and recommendation logic on which the payment recommendation decision is based: the name of the recommended payment tool, the name of the associated rights and benefits marketing campaign, and the relevant benefits that users can obtain from it.

30. The system of claim 17, wherein, The operation of the computer program also results in the following operations: receiving feedback data from the user regarding the recommendation results, wherein the feedback data includes: whether the user accepted the recommendation, the user's satisfaction rating with the recommendation results, and whether the user successfully obtained the relevant rights after payment.

31. The system of claim 30, wherein, The execution of the computer program also results in the following operations: dynamically tuning the parameters of the pre-trained intelligent recommendation model using a reinforcement learning algorithm based on the feedback data; and / or optimizing the logic for generating the user dynamic profile features using the feedback data.

32. The system as described in claim 17, wherein the intelligent recommendation model employs a collaborative filtering strategy, analyzes the user's historical behavior and preferences, matches the current user with similar user groups, and generates the payment recommendation decision based on the usage of payment tools by the similar user groups.

33. A computer-readable storage medium, characterized in that, The computer-readable storage medium includes instructions that, when executed, perform the payment tool intelligent recommendation method according to any one of claims 1-16.

34. A computer program product, characterized in that, It includes a computer program that, when executed by a processor, implements the payment instrument intelligent recommendation method according to any one of claims 1-16.

Images