A commercial pre-paid card internet advertisement marketing method and system

Abstract

The application relates to a commercial prepayment card Internet advertisement marketing method and system, transaction information from different payment channels is acquired, basic associated feature information is extracted, when a new transaction occurs, the associated feature information of the new transaction is compared with the basic associated feature information according to an associated rule, and combined with payment amount information, transaction records belonging to the same consumption behavior are recognized and aggregated to form a consumption event; whether the consumption event is a valid consumption event is determined; if the consumption event is a valid consumption event, commodity preference information of a user is extracted, wherein when the valid consumption event lacks specific commodity details, commodity preference information is obtained by combining merchant identity information and payment amount information for inference; and advertisement matching is carried out by using an updated user portrait according to the commodity preference information. Through the technical scheme, the application effectively solves the problems of transaction record fragmentation and lack of specific commodity details of a new payment interface in the prior art.

Description

Technical Field

[0001] This application relates to the field of internet advertising and marketing technology, and more specifically, to a method and system for internet advertising and marketing of prepaid commercial cards. Background Technology

[0002] With the increasing popularity of new prepaid card payment methods, consumers have gradually developed the habit of mixed payment. At the same merchant, or even in the same transaction, users often use both traditional prepaid card balances and new digital prepaid vouchers or mini-program coupons simultaneously. This shift in consumer payment behavior means that a single, logical "purchase event" no longer corresponds to a single, complete transaction record, but may instead be fragmented into multiple transaction records from different payment systems or different data formats. Under this prevalent mixed payment model, marketing systems face significant challenges in attempting to link user identities with transaction records from different payment channels. This is mainly due to the lack of a unified transaction identifier that can span traditional and new payment systems, and the absence of an effective cross-channel data bridging mechanism.

[0003] Furthermore, the technical integration solutions between new payment methods and merchants' existing POS systems are typically quite basic. When users redeem coupons using mini-programs, the POS system's data interface often only receives simple confirmation of successful redemption and the final discount amount. Crucially, it usually cannot obtain information on which specific products the coupon was applied to. This limitation results in a natural lack of detailed consumption information at the data source. Even if the marketing system successfully links fragmented transactions, its underlying data lacks the granularity needed to infer specific product preferences. The system cannot determine whether a user prefers a particular brand of beverage; it can only know that they used a coupon. This fundamental data gap severely hinders the system's ability to build accurate user profiles based on actual product consumption.

[0004] Ultimately, due to the increasingly diverse sources of user consumption behavior data, the heterogeneity of data structures across different channels, the complexity brought about by hybrid payment models, and the lack of some key consumption details, existing systems are unable to effectively identify users' true consumption preferences and potential needs when building user profile tags. This fragmented, incomplete, and often misclassified data makes it difficult for the system to accurately learn and assign meaningful attributes to users. As a direct consequence, the accuracy of the matching degree calculation between advertisements and user profiles decreases significantly. This leads to advertisements frequently not matching users' actual interests, resulting in low user engagement, unsatisfactory click-through rates, and ultimately a significant reduction in advertising conversion efficiency, despite the system's initial goal of achieving precise targeting. The information the system obtains about users becomes vague, incomplete, and often misleading, making truly targeted and effective marketing virtually impossible.

[0005] To address the aforementioned issues, existing technologies urgently need improvement. Summary of the Invention

[0006] This application discloses a method and system for internet advertising marketing of commercial prepaid cards, which aims to address the shortcomings of existing internet advertising marketing systems for commercial prepaid cards in processing diverse payment channel data, aggregating fragmented transaction records, inferring specific product preferences, and building accurate user profiles, thereby improving the accuracy and conversion efficiency of advertising placement.

[0007] Firstly, this application discloses a method for internet advertising marketing of prepaid commercial cards, including: Obtain transaction information from different payment channels and extract basic related feature information from the transaction information. The related feature information includes at least one of the following: payment time information, payment amount information, merchant identity information, and user identity fragment information. Set association rules, which include time proximity rules, merchant consistency rules, and user identity fragment association rules; When a new transaction occurs, the actual association feature information and the basic association feature information of the new transaction are compared according to the association rules, and logical inference is made in combination with the payment amount information to identify and aggregate transaction records belonging to the same consumption behavior to form a consumption event. Conduct credibility assessments on consumption events and determine whether a consumption event is a valid consumption event based on the credibility assessment results; If the consumption event is valid, then based on the valid consumption event, the user's product preference information is extracted. Specifically, when a valid consumption event lacks specific product details, product preference information is inferred by combining merchant identity information and payment amount information; and Update user profiles based on product preference information, and use the updated user profiles for ad matching.

[0008] In some preferred implementations, when a new transaction occurs, the steps of comparing the actual association feature information and the basic association feature information of the new transaction according to association rules, and combining payment amount information to make logical inferences, to identify and aggregate transaction records belonging to the same consumption behavior and form a consumption event include: According to the association rules, the actual payment time information and basic payment time information of the new transaction, the actual merchant identity information and basic merchant identity information of the new transaction, and the actual user identity fragment information and basic user identity fragment information of the new transaction are compared to obtain a preliminary aggregation set. Obtain the merchant's currently active set of promotional rules, which includes bundled sales discounts, tiered discounts, and discounts for specific product combinations; Based on the promotional rule set, the merchant's product list is preprocessed to generate a dynamic discounted product combination list that includes product combinations with different discount rules applied and their corresponding discounted prices. The payment amount of the initial aggregated set is compared with the discounted price in the dynamic discount product bundle list; and Based on the comparison results, the combination of goods that best matches the payment amount is selected as the inference result to identify and aggregate transaction records belonging to the same consumption behavior, thus forming a consumption event.

[0009] In some preferred embodiments, the steps of performing a credibility assessment on the consumption event and determining whether the consumption event is a valid consumption event based on the credibility assessment results include: Identify the scenario type in which the current consumption event occurred, including merchant type, transaction time period, and geographical region; Based on the scenario type, select the weight configuration corresponding to the scenario type from the preset weight configuration set; By utilizing weighted configuration, a weighted sum of the matching scores of each associated feature in a consumption event is calculated to obtain the integrated credibility score of the consumption event; and The system compares the integrated credibility score with a preset event aggregation threshold to determine whether a consumption event is a valid consumption event. If the integrated credibility score is greater than the preset event aggregation threshold, then the consumed event is determined to be a valid consumed event.

[0010] In some preferred implementations, if the consumption event is a valid consumption event, then based on the valid consumption event, the user's product preference information is extracted. When a valid consumption event lacks specific product details, the product preference information is inferred by combining merchant identity information and payment amount information. The steps for obtaining the product preference information include: Acquire user interaction information from online channels and extract non-transactional behavioral clues from the interaction information. These non-transactional behavioral clues include the duration of time a user browses a specific product page, the frequency of clicking on ads for a specific product category, the history of collecting coupons, and interactions with specific product topics. A virtual product display list is created for each merchant in a mixed business format. This virtual product display list includes the goods or services offered by the merchant and the sub-business format to which they belong. Based on the merchant's identity information and payment amount information, combined with the virtual product display list and known discount rules, multiple product combination schemes are generated. Each product combination scheme includes the product category and the total price after applying discounts. Based on non-transactional behavioral cues, a user interest attraction field is constructed, which includes the intensity of user interest in specific product categories or service topics. Multiple product combination schemes are compared with the user's interest attraction field, and the product combination scheme with the highest degree of consistency with the user's interest attraction field is selected as the product preference information; and Update user profiles based on product preference information, and use the updated user profiles for ad matching.

[0011] In some preferred embodiments, the steps of updating user profiles based on product preference information and using the updated user profiles for ad matching include: Obtain new product preference information; Assign initial short-term interest weights to new product preference information and include them in the short-term interest weight pool; Based on the consistency between the new product preference information and the user's historical long-term preferences, adjust the accumulation intensity of the new product preference information in the long-term preference accumulation area; When there is a first preference in the short-term interest weight pool whose weight reaches the preset preference fusion threshold, and a preference that is related to the accumulation strength of the first preference is found in the long-term preference accumulation area, the short-term interest is converted into a long-term preference. Update the user profile based on the current weight of preferences in the short-term interest weight pool and the accumulated strength of preferences in the long-term preference accumulation zone; and Use the updated user profiles to perform ad matching.

[0012] In some preferred embodiments, the step of extracting basic association feature information from transaction information includes: The basic association feature information is standardized and combined to generate a behavioral structure feature.

[0013] This technical solution standardizes and combines basic association feature information to generate unified behavioral structure features, providing a more standardized and easier-to-process data foundation for subsequent transaction comparison and logical inference, thereby improving the efficiency and accuracy of data processing.

[0014] In some preferred embodiments, the steps of standardizing and combining basic association feature information to generate a behavioral structure feature include: The payment amount is classified according to the payment amount information. The payment amount in the first amount range is divided into the low amount range, the payment amount in the second amount range is divided into the medium amount range, and the payment amount in the third amount range is divided into the high amount range. Obtain the merchant's latitude and longitude coordinates or administrative region information based on the merchant's identity information, and hash the latitude and longitude coordinates or administrative region information to generate a short string of the merchant's location; Based on the user's identity fragment information, obtain the user's prepaid card number, digital platform user ID, and mobile phone number, and then combine and hash the user's prepaid card number, digital platform user ID, and mobile phone number.

[0015] Through this technical solution, this application achieves refined and standardized processing of key related feature information by classifying payment amounts, hashing merchant location information, and combining hashes of user identity fragment information. This further enhances the expressive power and recognition efficiency of behavioral structure features, providing a more reliable basis for subsequent transaction aggregation.

[0016] In some preferred embodiments, the step of logically inferring the actual and basic association characteristics of new transactions, combined with payment amount information, to identify and aggregate transaction records belonging to the same consumption behavior, thus forming a consumption event, includes: Identify the actual associated feature information and behavioral structure features.

[0017] This technical solution simplifies the comparison logic in the transaction aggregation process by identifying actual related feature information with pre-generated behavioral structure features, thereby improving the efficiency and accuracy of identifying and aggregating transaction records belonging to the same consumption behavior.

[0018] In some preferred embodiments, prior to the step of performing a credibility assessment of the consumption event, the following steps are included: Based on the consumption event, obtain multiple original transaction information, including payment time information and payment amount information; Calculate the average or median of multiple payment time information as a calibration timestamp for the consumption event; Add up the payment amounts from all the original transaction information to get a total amount; The total amount is matched with the merchant's product or service price, and inferences are made based on the known semantics of the coupon to obtain products or services that match the amount logically.

[0019] Secondly, this application also discloses a commercial prepaid card internet advertising and marketing system, which includes: The information acquisition module is used to acquire transaction information from different payment channels and extract basic related feature information from the transaction information. The related feature information includes at least one of the following: payment time information, payment amount information, merchant identity information, and user identity fragment information. The rule setting module is used to set association rules, which include time proximity rules, merchant consistency rules, and user identity fragment association rules. The event aggregation module, when a new transaction occurs, compares the actual association feature information and the basic association feature information of the new transaction according to the association rules, and performs logical inference in combination with the payment amount information to identify and aggregate transaction records belonging to the same consumption behavior, forming a consumption event; The credibility assessment module is used to assess the credibility of consumption events and determine whether a consumption event is a valid consumption event based on the credibility assessment results. The preference extraction module is used to extract the user's product preference information based on a valid consumption event. Specifically, when a valid consumption event lacks detailed product information, it infers the product preference information by combining merchant identity information and payment amount information. The ad matching module is used to update user profiles based on product preference information and then use the updated user profiles to perform ad matching.

[0020] This application provides a commercial prepaid card internet advertising and marketing system. Through modular design, it effectively supports commercial prepaid card internet advertising and marketing methods. The system can automatically complete core functions such as information acquisition, rule setting, event aggregation, credibility assessment, preference extraction, and ad matching, providing comprehensive technical support for solving the challenges of existing systems in data processing, user profile construction, and precise ad delivery. Beneficial effects

[0021] The commercial prepaid card internet advertising marketing method disclosed in this application obtains transaction information from different payment channels and extracts basic correlation features. It sets rules for correlation based on time proximity, merchant consistency, and user identity fragments. When a new transaction occurs, it compares correlation features according to these rules and performs logical inference based on the payment amount. This effectively identifies and aggregates fragmented transaction records belonging to the same consumption behavior, forming a complete consumption event. Based on this, the credibility of the consumption event is assessed to ensure its validity. For valid consumption events, even without specific product details, it can infer and extract user product preference information by combining merchant identity information and payment amount information. Finally, it updates the user profile based on this product preference information and uses the updated user profile for precise advertising matching.

[0022] Through the above technical solution, this application effectively solves the dilemma in the prior art where the diversification of consumer data sources, heterogeneous data structures, fragmented transaction records due to mixed payment modes, and the lack of specific product details in new payment interfaces make it difficult for the system to build a comprehensive and detailed user profile, thereby affecting the accuracy and conversion efficiency of advertising. Attached Figure Description

[0023] Figure 1 This is a flowchart illustrating a commercial prepaid card internet advertising marketing method proposed in this application.

[0024] Figure 2 yes Figure 1 The flowchart of step S3 shown is a schematic diagram of an embodiment.

[0025] Figure 3 yes Figure 1 The flowchart shown is a schematic diagram of an embodiment of step S4.

[0026] Figure 4 yes Figure 1 The flowchart shown is a schematic diagram of an embodiment of step S5.

[0027] Figure 5 yes Figure 1 The flowchart shown is a schematic diagram of an embodiment of step S6.

[0028] Figure 6 This is a flowchart illustrating another commercial prepaid card internet advertising marketing method proposed in this application. (Figure)

[0029] Figure 7 This is a flowchart illustrating another commercial prepaid card internet advertising marketing method proposed in this application. (Figure)

[0030] Figure 8 This is a schematic diagram of the structure of a commercial prepaid card internet advertising and marketing system provided in this application. Detailed Implementation

[0031] The technical solutions of this application will now be clearly and completely described with reference to the accompanying drawings. Obviously, the described embodiments are merely some embodiments of this application, and not all embodiments. The components of this application described and shown in the accompanying drawings can generally be arranged and designed in various different configurations. Therefore, the following detailed description of the embodiments of this application provided in the accompanying drawings is not intended to limit the scope of the claimed application, but merely to illustrate selected embodiments of this application. All other embodiments obtained by those skilled in the art based on the embodiments of this application without inventive effort are within the scope of protection of this application.

[0032] It should be noted that similar reference numerals and letters in the following figures indicate similar items; therefore, once an item is defined in one figure, it does not need to be further defined and explained in subsequent figures. Furthermore, in the description of this application, terms such as "first," "second," etc., are used only to distinguish descriptions and should not be construed as indicating or implying relative importance.

[0033] Traditional online advertising and marketing systems for prepaid cards have significant limitations in processing data from diverse payment channels, identifying consumer behavior under mixed payment models, and obtaining detailed product information. These issues make it difficult for the system to build comprehensive and detailed user profiles, thereby affecting the accuracy and conversion efficiency of ad targeting.

[0034] In response, this application proposes a method for internet advertising marketing of prepaid commercial cards. Please refer to [link / reference]. Figure 1 This includes the following steps: Step S1: Obtain transaction information from different payment channels and extract basic related feature information from the transaction information. The related feature information includes at least one of the following: payment time information, payment amount information, merchant identity information, and user identity fragment information. Step S2: Set association rules, which include time proximity rules, merchant consistency rules, and user identity fragment association rules; Step S3: When a new transaction occurs, the actual association feature information and the basic association feature information of the new transaction are compared according to the association rules, and logical inference is made in combination with the payment amount information to identify and aggregate transaction records belonging to the same consumption behavior to form a consumption event. Step S4: Conduct a credibility assessment of the consumption event, and determine whether the consumption event is a valid consumption event based on the credibility assessment results; Step S5: If the consumption event is a valid consumption event, extract the user's product preference information based on the valid consumption event. When the valid consumption event lacks specific product details, inference is made by combining the merchant's identity information and payment amount information to obtain the product preference information. Step S6: Update the user profile based on product preference information, and use the updated user profile for ad matching.

[0035] This application firstly proposes several approaches to acquiring transaction information from different payment channels and extracting basic related feature information. For example, the system can be configured with multiple data interfaces to connect with bank card payment systems, third-party payment platforms (such as Alipay and WeChat Pay), and merchants' own mini-programs or prepaid card verification systems. These interfaces receive transaction data streams in real time when a transaction occurs. For each transaction, the system parses its data structure to extract payment time information, payment amount information, merchant identity information, and user identity fragments. For example, payment time information can be accurate to the second, payment amount information can be the actual currency value paid, merchant identity information can be the merchant's unified social credit code or store ID, and user identity fragments can be an encrypted user ID, the last four digits of a mobile phone number, or a specific field of a prepaid card number. Another approach is to periodically import transaction logs in batches from the databases of various payment channels, and then use data cleaning and standardization modules to unify the data format before extracting feature information. For example, the system will convert the different date formats that may exist between different payment channels into a standard timestamp; for non-standardized merchant names, the system will map them through a preset merchant dictionary to ensure the consistency of merchant identity information.

[0036] Secondly, regarding the setting of association rules, these rules are crucial for identifying the same consumer behavior. Time proximity rules can be set such that if the payment time interval between two transactions is within a preset threshold (e.g., 5 minutes), they are considered to be close in time. Merchant consistency rules can be set such that if the merchant identity information of two transactions is completely identical, they are considered to have occurred at the same merchant. User identity fragment association rules can be set such that if the user identity fragment information of two transactions matches after hashing, they are considered to likely belong to the same user. These rules can be pre-configured and adjusted by operations personnel in the system backend to adapt to the needs of different business scenarios. For example, in a fast-food restaurant scenario, the time proximity threshold can be set shorter, while in a large shopping mall, the threshold can be appropriately relaxed.

[0037] Furthermore, when a new transaction occurs, the system compares the actual and basic association characteristics of the new transaction according to association rules, and performs logical inference based on payment amount information to identify and aggregate transaction records belonging to the same consumption behavior, forming a consumption event. Specifically, when the system receives a new transaction record, it first extracts its association characteristic information. Subsequently, the system compares the payment time information of the new transaction with the payment time information of historical transaction records to determine whether the time proximity rule is met. At the same time, it also compares merchant identity information and user identity fragment information to meet merchant consistency rules and user identity fragment association rules. For example, if a user makes two consecutive payments at a merchant, one for 50 yuan via prepaid card and the other for 20 yuan via WeChat, and both transactions occur within the same minute, the system will initially determine, based on time proximity, merchant consistency, and user identity fragment association rules, that these two transactions may belong to the same consumption behavior. On this basis, logical inference is performed in conjunction with payment amount information. For example, if a merchant has a promotion of "10 yuan off for purchases over 70 yuan", the system will infer that the two transactions (50 yuan + 20 yuan = 70 yuan) may constitute a single purchase with the discount. In this way, the system can aggregate fragmented transaction records into a complete consumption event.

[0038] Next, the credibility of the consumption event is assessed, and based on the assessment results, it is determined whether the consumption event is valid. Credibility assessment can be implemented in several ways. One approach is to assign a weight to the matching degree of each related feature; for example, the closer the time proximity, the higher the weight; and if the merchant identities are completely identical, the weight is the highest. These weights are then combined with the actual matching results to calculate a comprehensive credibility score. For example, if all related features in a consumption event are highly matched, its credibility score will be high. A preset event aggregation threshold is set; if the aggregate credibility score of a consumption event is higher than this threshold, it is determined to be a valid consumption event.

[0039] If the consumption event is valid, the user's product preference information is extracted based on it. When a valid consumption event lacks specific product details, product preference information is inferred by combining merchant identity information and payment amount information.

[0040] Finally, based on product preference information, the user profile is updated, and the updated user profile is used for ad matching. When new product preference information is obtained, the system integrates it into the user's existing profile. For example, if a user is inferred to prefer "coffee-latte," the system will add or strengthen the "coffee lover" tag in their user profile. User profile updates can be real-time or periodic. The updated user profile will be used for ad matching. For example, if the user profile shows a strong preference for "outdoor sports equipment," the system will prioritize pushing ads from outdoor sports brands to them. The ad matching algorithm can be based on collaborative filtering, content matching, or deep learning models to ensure a high relevance between ads and user preferences.

[0041] Traditional systems rely heavily on transaction data from a single source, making it difficult to effectively handle fragmented information from different payment channels and inaccurately inferring user preferences in the absence of detailed product information. This application, by introducing multi-dimensional association rules and a logical inference mechanism based on merchants and payment amounts, successfully aggregates scattered transaction records into meaningful consumption events and can still extract user product preferences even in the absence of detailed product information. For example, in traditional systems, a user purchasing goods at the same merchant using a prepaid card and WeChat Pay might be identified as two independent transactions, with no information on what was purchased. This application, however, intelligently associates these two transactions and infers the possible product categories the user might have purchased based on the total amount and merchant type, thus constructing a more comprehensive and detailed user profile. This capability allows this application to overcome the limitations of existing technologies in terms of data fragmentation and information gaps, significantly improving the accuracy of user profiles and enabling more efficient and accurate ad matching, bringing significant progress to internet advertising marketing for prepaid cards.

[0042] Please see Figure 2 The technical solution in step S3 above, which states that "when a new transaction occurs, the actual association feature information and the basic association feature information of the new transaction are compared according to the association rules, and logical inference is performed in conjunction with the payment amount information to identify and aggregate transaction records belonging to the same consumption behavior to form a consumption event," further includes the following steps: Step S31: According to the association rules, compare the actual payment time information and basic payment time information of the new transaction, the actual merchant identity information and basic merchant identity information of the new transaction, and the actual user identity fragment information and basic user identity fragment information of the new transaction to obtain a preliminary aggregation set; Step S32: Obtain the merchant's currently effective promotion rule set, which includes bundled sales discounts, tiered discounts, and specific product combination discounts; Step S33: Based on the promotion rule set, preprocess the merchant's product list to generate a dynamic discounted product combination list containing product combinations with different discount rules and their corresponding discounted prices; Step S34: Compare the payment amount of the initial aggregate with the discounted price in the dynamic discount product combination list; Step S35: Based on the comparison results, select the product combination that best matches the payment amount as the inference result to identify and aggregate transaction records belonging to the same consumption behavior, forming a consumption event.

[0043] Specifically, when identifying and aggregating transaction records belonging to the same consumption behavior to form a consumption event, firstly, according to preset association rules, the actual payment time information, merchant identity information, and user identity fragment information of the newly occurring transaction are compared with the corresponding information in the existing basic transaction records. This comparison process aims to initially screen transaction records with high consistency in time, merchant, and user identity fragments, thereby forming a preliminary aggregation set. This preliminary aggregation set contains candidate transaction records that may belong to the same consumption behavior.

[0044] The promotional rules set can be understood as various preferential strategies set by merchants to attract consumers. Specific forms may include, but are not limited to, bundled sales discounts (e.g., a combined discount price for purchasing product A and product B), tiered discounts (e.g., a discount of 20 yuan for purchases over 100 yuan, and 50 yuan for purchases over 200 yuan), and discounts on specific product combinations (e.g., a 20% discount for purchasing specific styles of clothing and shoes). These rules are dynamically changing and require real-time access to the latest rules currently in effect from the merchant.

[0045] In practical applications, to accurately reflect the impact of promotional activities on product prices, it is necessary to preprocess the product list provided by the merchant based on the obtained set of promotional rules. This preprocessing process simulates the application of various promotional rules, generating a dynamic list of discounted product combinations. This list not only lists the products or services provided by the merchant but also includes the various combinations of these products or services under different promotional rules and their corresponding discounted prices. For example, if a product's original price is 100 yuan and there is a "spend 100 yuan, get 20 yuan off" promotion, the list will include an entry for "original price 100 yuan, discounted price 80 yuan." This list is dynamic and updates in real time as the promotional rules change.

[0046] Furthermore, the payment amounts of all transactions in the initial aggregate are summed and compared with the discounted prices listed in the dynamic discount product bundle list. The purpose of this comparison is to find the product bundle scheme that best matches the actual payment amount. For example, if the total payment amount in the initial aggregate is 80 yuan, and there is a product bundle in the dynamic discount product bundle list with an original price of 100 yuan and a price of 80 yuan after applying the "100 yuan off 20 yuan" discount, then the bundle is considered a match.

[0047] Therefore, based on the comparison results, the product combination that best matches the payment amount of the initial aggregated set is selected as the inference result. Here, "best match" can mean that the payment amount is exactly the same as the discounted price, or is closest within a preset error range. In this way, even if the transaction amount changes due to promotions, the system can accurately identify the actual consumption content and aggregate transaction records belonging to the same consumption behavior to ultimately form a complete consumption event.

[0048] This application's solution effectively addresses the problem of basic solutions struggling to accurately aggregate transaction records into consumption events when handling complex commercial promotional scenarios by introducing merchant promotional rules and a dynamic list of discounted product combinations. Specifically, it first uses association rules for initial aggregation, narrowing the scope of transactions to be analyzed. Second, it acquires and utilizes the merchant's currently active promotional rule set, enabling the system to understand and simulate actual commercial discount logic. Based on this, the merchant's product list is preprocessed to generate a dynamic list of discounted product combinations including discounted prices, providing a realistic and comprehensive reference for subsequent amount comparisons. By comparing the actual payment amount with the discounted price that takes promotional factors into account, the system can accurately infer the product combination actually purchased by the user, overcoming potential errors from comparing only the raw amount and ensuring that transaction records belonging to the same consumption behavior can be correctly identified and aggregated when multiple payment methods or promotional activities exist.

[0049] Please see Figure 3 The technical solution for "conducting a credibility assessment of the consumption event and determining whether the consumption event is a valid consumption event based on the credibility assessment results" in step S4 above includes the following steps: Step S41: Identify the scenario type of the current consumption event, including merchant type, transaction time period, and geographical region; Step S42: Based on the scene type, select the weight configuration corresponding to the scene type from the preset weight configuration set; Step S43: Using weight configuration, calculate the weighted sum of the matching scores of each associated feature in the consumption event to obtain the integrated credibility score of the consumption event; and Step S44: Compare the integrated credibility score with the preset event aggregation threshold to determine whether the consumption event is a valid consumption event; Step S45: If the integration credibility score is greater than the preset event aggregation threshold, then the consumed event is determined to be a valid consumed event.

[0050] Specifically, identifying the scenario type of the current consumption event means that the system automatically determines the specific environmental characteristics of the event based on the transaction information contained within it. For example, merchant type can refer to industry categories such as catering, retail, and entertainment; transaction time can refer to time periods such as weekdays, weekends, holidays, daytime, or nighttime; and geographical area can refer to specific business districts, administrative divisions, or city areas. These scenario types help to more finely categorize consumer behavior.

[0051] The process involves selecting a weight configuration corresponding to the scenario type from a pre-defined set of weight configurations. This can be understood as the system maintaining a database containing various weight configurations, each associated with a specific scenario type. For example, in a restaurant scenario, payment time information might have a higher weight, while in a retail scenario, merchant consistency rules might have a higher weight. Once the scenario type of the current consumption event is identified, the system automatically loads the corresponding weight configuration to ensure the evaluation is targeted.

[0052] In practical applications, the integrated credibility score of a consumption event is obtained by calculating the weighted sum of the matching scores of each associated feature in the consumption event using weight configuration. This involves quantifying the matching degree of each associated feature in the consumption event (such as payment time proximity, merchant consistency, and user identity fragment association) into a score, and then summing these scores according to the selected weight configuration. For example, if the payment time proximity score is 0.9, the merchant consistency score is 1.0, and the user identity fragment association score is 0.8, with corresponding weights of 0.4, 0.3, and 0.3 respectively, then the integrated credibility score is 0.9*0.4 + 1.0*0.3 + 0.8*0.3 = 0.36 + 0.3 + 0.24 = 0.9.

[0053] Furthermore, determining whether a consumption event is a valid consumption event involves comparing the integrated credibility score with a preset event aggregation threshold. This threshold can be adjusted based on business needs and data characteristics to distinguish between high-credibility valid consumption events and low-credibility invalid consumption events. If the integrated credibility score is greater than the preset event aggregation threshold, the consumption event is determined to be a valid consumption event, indicating that the aggregation result of this consumption event has high reliability and can be used for subsequent user preference extraction and ad matching.

[0054] This application's solution addresses the inaccuracy of traditional single evaluation criteria in assessing credibility in complex and ever-changing consumption scenarios by introducing scenario type identification and dynamic weight configuration mechanisms. Specifically, after a consumption event is aggregated, its specific scenario type is first identified, such as whether it belongs to a restaurant or retail business, occurs during a weekday lunch break, or is located in a specific commercial area. Based on the identified scenario type, the system selects the most suitable weight configuration from a pre-set set of weight configurations. Because the contribution of various related features (such as payment time, merchant, and user identity fragments) to consumption event aggregation differs across scenarios, dynamically adjusting the weights of these features can more accurately reflect the true credibility of the consumption event in that scenario. Subsequently, using these scenario-specific weights, the matching scores of each related feature in the consumption event are weighted and summed to obtain a more scenario-adaptive integrated credibility score. Finally, comparing this integrated credibility score with a pre-set event aggregation threshold allows for a more accurate determination of whether a consumption event is valid. This mechanism makes credibility assessment no longer a "one-size-fits-all" approach, but rather allows for intelligent adjustments based on the actual context, thereby significantly improving the accuracy and reliability of the assessment.

[0055] In some of the embodiments described above in this application, a method for extracting user product preference information based on valid consumption events was proposed. However, in its implementation, when specific product details are lacking in valid consumption events, inference based solely on merchant identity information and payment amount information may result in insufficiently refined or comprehensive extraction of product preference information. Therefore, this application further proposes a more refined method for inferring product preference information by introducing multi-dimensional user behavior cues and merchant product structure information to more accurately identify user product preferences.

[0056] Please see Figure 4 The technical solution for step S5 above, "If the consumption event is a valid consumption event, then based on the valid consumption event, extract the user's product preference information. Where, when the valid consumption event lacks specific product details, inference is made by combining merchant identity information and payment amount information to obtain product preference information," includes the following steps: Step S51: Obtain user interaction information from online channels and extract non-transactional behavior clues from the interaction information. Non-transactional behavior clues include the duration of user browsing specific product pages, the frequency of clicking on specific category advertisements, the record of collecting coupons, and interaction with specific product topics. Step S52: Construct a virtual product display list for each composite business type merchant. The virtual product display list includes the goods or services provided by the merchant and the sub-business type to which they belong. Step S53: Based on the merchant's identity information and payment amount information, combined with the virtual product display list and known discount rules, generate multiple product combination schemes. The product combination scheme includes product categories and the total price after applying discounts. Step S54: Based on non-transactional behavioral cues, construct a user interest attraction field, which includes the intensity of user interest in specific product categories or service themes; Step S55: Compare multiple product combination schemes with the user's interest gravity field, and select the product combination scheme with the highest degree of consistency with the user's interest gravity field as the product preference information; Step S56: Update the user profile based on product preference information, and use the updated user profile for ad matching.

[0057] Specifically, non-transactional behavioral cues refer to users' interests and preferences for goods or services displayed on online platforms, beyond actual purchasing behavior. For example, the length of time a user spends browsing a specific product page on an e-commerce platform or content community can reflect their level of interest in that product; the frequency of clicking on ads for a specific category indicates a user's potential demand for that category of goods; records of collecting coupons can reveal a user's intention to purchase a specific product or service; and interactions with specific product topics, such as liking, commenting, or sharing, can reflect the intensity of a user's interest in and engagement with related products. These cues are used to more comprehensively understand users' potential interests, compensating for the lack of detailed product information in transaction data.

[0058] Among them, multi-format merchants refer to businesses that offer multiple categories of goods or services, such as large supermarkets, department stores, or comprehensive service platforms. Creating a virtual product display list for these merchants can be understood as creating a digital catalog that includes all the goods or services that the merchant may offer and their respective sub-categories. For example, a supermarket's virtual product display list might include sub-categories such as "fresh food," "daily necessities," and "home appliances," as well as specific goods or services within each category. The purpose of this list is to provide an inferable range of goods when specific transaction details are lacking.

[0059] In practical applications, multiple product combination schemes can be generated based on merchant identity information and payment amount information, combined with the virtual product display list and known discount rules. For example, if a user pays 150 yuan at a supermarket, the system will infer several possible product combinations based on the supermarket's virtual product display list and currently effective discount rules (such as minimum purchase amount reduction, discounts, etc.), such as "purchased 100 yuan of fresh produce and 50 yuan of dairy products" or "purchased baked goods originally priced at 100 yuan, discounted to 80 yuan, and 70 yuan of fresh produce." Each product combination scheme includes its product categories and the total price after applying discounts to match the actual payment amount.

[0060] Furthermore, the user interest gravity field can be understood as a distribution model of the intensity of user interest in different product categories or service themes. This model is constructed based on non-transactional behavioral cues obtained by users. For example, if a user frequently browses digital product pages and clicks on related advertisements, their interest gravity field strength in the "digital products" category will be relatively high. This gravity field is used to quantify users' potential preferences in different product areas, providing a basis for subsequent product combination selection.

[0061] Therefore, multiple product combination schemes are compared with the user's interest gravity field to find the product combination that best matches the user's potential interests. For example, if a user spends 150 yuan at a supermarket, and their interest gravity field shows a much stronger interest in "fresh food" than "daily necessities," then among the multiple product combination schemes, the one containing more fresh food will be considered to have a higher degree of relevance. Through this comparison, the actual categories of goods purchased by the user can be more accurately inferred, thereby extracting the user's product preference information.

[0062] Finally, based on the inferred product preference information, the user profile will be updated, and the updated user profile will be used for ad matching. This means that information such as user interest tags and consumption tendencies will be updated in real time or near real time, enabling subsequent ads to reach users more accurately and improving ad effectiveness.

[0063] This application's solution overcomes the shortcomings of traditional methods in inferring value when specific product details are lacking by incorporating user interaction information from online channels and extracting non-transactional behavioral clues. Specifically, when a valid consumption event only includes merchant identity information and payment amount information without specific product details, the system no longer relies solely on rough amount matching. Instead, it first obtains the user's interest preferences displayed in other online scenarios, such as browsing time, click frequency, collection history, and interaction behavior. These non-transactional behavioral clues are used to construct a multi-dimensional user interest attraction field, thereby more comprehensively characterizing the user's potential product preferences. Simultaneously, for merchants with diverse business formats, this application constructs a virtual product display list. This list details the products or services that the merchant may offer and their sub-formats, providing a structured product range for inference. By combining merchant identity information and payment amount information with this virtual product display list and known discount rules, the system can generate multiple reasonable product combination schemes. Finally, these product combination schemes are compared with the user interest attraction field, and the scheme with the highest matching degree is selected as the user's product preference information. This method takes into account users’ explicit transaction behavior (payment amount, merchant) and implicit interests (non-transactional behavior cues), as well as the merchant’s product structure, thereby achieving a more accurate and detailed inference of users’ product preferences in the absence of direct product details.

[0064] Traditional user profile update methods often directly overlay new preference information into the user profile or employ simple decay mechanisms when processing user product preference information. However, this approach may fail to effectively distinguish between short-term fluctuations in user interest and long-term stable preferences, resulting in insufficient accuracy and timeliness of user profiles. For example, suppose a user purchases a product they don't usually buy due to a chance promotion. Directly updating the user profile might incorrectly assume the user has a long-term preference for that product, leading to the delivery of irrelevant ads. If this problem is not addressed, the accuracy of ad matching will be affected, potentially resulting in inefficient ad delivery and a degraded user experience. To address this, this application proposes a more refined user profile update mechanism that introduces a strategy combining short-term interests and long-term preferences to more accurately reflect the evolution of users' true interests.

[0065] Please see Figure 5 The technical solution in step S6 above, "updating the user profile based on product preference information and using the updated user profile for ad matching," further includes the following steps: Step S61: Obtain new product preference information; Step S62: Assign initial short-term interest weights to the new product preference information and include them in the short-term interest weight pool; Step S63: Based on the consistency between the new product preference information and the user's historical long-term preferences, adjust the accumulation intensity of the new product preference information in the long-term preference accumulation area; Step S64: When there is a first preference in the short-term interest weight pool whose weight reaches the preset preference fusion threshold, and a preference that is related to the accumulation strength of the first preference is found in the long-term preference accumulation area, the short-term interest is converted into a long-term preference. Step S65: Update the user profile based on the current weight of preferences in the short-term interest weight pool and the accumulated strength of preferences in the long-term preference accumulation area; Step S66: Use the updated user profile to perform ad matching.

[0066] Specifically, acquiring new product preference information refers to data extracted by the system from valid consumption events that reflects a user's inclination towards specific products or services. This information may include product categories, brands, price ranges, etc. Assigning initial short-term interest weights to this new product preference information and including it in a short-term interest weight pool can be understood as assigning an initial activity or importance score to a user's newly generated interests and placing them in a dynamically changing set. This short-term interest weight pool is used to temporarily store and manage a user's recent, potentially unstable interests. The initial short-term interest weights can be set based on factors such as the source, intensity, or timeliness of the preference information. In practical applications, adjusting the accumulation strength of new product preference information in the long-term preference accumulation area based on its consistency with the user's historical long-term preferences means that the system assesses the degree of correlation between the new preference and the user's existing, time-tested stable preferences. If the new preference is highly consistent with long-term preferences, its accumulation strength in the long-term preference accumulation area will be enhanced; conversely, if the consistency is low, the accumulation strength will be weakened accordingly. The long-term preference accumulation area is a region that stores and manages a user's stable and lasting interests, and its accumulation strength reflects the stability of the preferences. Furthermore, when a first preference with a weight reaching the preset preference fusion threshold exists in the short-term interest weight pool, and a preference with a correlation to the accumulation strength of the first preference is found in the long-term preference accumulation area, the short-term interest is transformed into a long-term preference. This indicates that when a short-term interest, after a period of observation and accumulation, reaches a certain weight and has some correlation with the user's existing long-term preferences, it is considered stable and worthy of being included in the category of long-term preferences. The preset preference fusion threshold is a standard for judging whether a short-term interest is stable enough to be transformed into a long-term preference. Therefore, the user profile is updated based on the current weight of preferences in the short-term interest weight pool and the accumulation strength of preferences in the long-term preference accumulation area. This means that updating the user profile is a comprehensive process, simultaneously considering the user's currently active and potentially changing interests (short-term interests) and the user's stable and persistent interests (long-term preferences), constructing a comprehensive and dynamic user profile through the combination of both. Finally, the updated user profile is used for ad matching, aiming to push highly relevant ad content to users based on their latest and most accurate preference information, thereby improving ad click-through rates and conversion rates.

[0067] This application's solution achieves refined management and dynamic updating of user product preference information by introducing a short-term interest weight pool and a long-term preference accumulation zone. When new product preference information is acquired, it is first treated as a short-term interest and assigned an initial weight to be included in the short-term interest weight pool. This mechanism allows the system to capture the user's latest interests, even if these interests may be temporary. Simultaneously, the system evaluates the consistency of these new preferences with the user's historical long-term preferences and adjusts their accumulation intensity in the long-term preference accumulation zone accordingly. This dual-track processing avoids the distortion of the user profile that may result from simply adding new preferences, ensuring the stability of long-term preferences. When a short-term interest remains active in the weight pool and reaches a preset preference fusion threshold, and is associated with a long-term preference, it is identified and converted into a long-term preference. This conversion process is based on in-depth analysis of user behavior patterns, ensuring that only truly stable interests are solidified into long-term preferences. In this way, updating the user profile is no longer a simple accumulation of information, but a dynamic modeling of the evolution of user interests, enabling the user profile to more accurately reflect the user's real, current, and potential product preferences.

[0068] Specifically, the step of extracting basic association feature information from transaction information in step S1 above further includes standardizing and combining the basic association feature information to generate a behavioral structure feature.

[0069] The basic association feature information refers to the raw data directly extracted from transaction information, such as payment time information, payment amount information, merchant identity information, and fragmented user identity information. Standardizing this information aims to eliminate differences caused by different data sources or formats, ensuring data consistency and comparability. For example, payment time information can be standardized to a unified timestamp format; payment amount information can be standardized to a unified currency unit and precision. The combination operation logically integrates these standardized association feature information to form a more representative and structured feature representation. For example, payment time information, payment amount information, merchant identity information, and fragmented user identity information can be concatenated or encoded according to preset rules to generate a "behavioral structure feature" that comprehensively reflects a single transaction. This behavioral structure feature, as a whole, can be more effectively utilized by subsequent association rule comparisons and logical inferences.

[0070] This application's solution standardizes and combines basic association feature information, transforming raw, scattered transaction data into unified, structured behavioral structure features. This transformation makes subsequent association rule comparison more efficient and accurate. Specifically, standardization ensures the comparability of features between different transaction records, avoiding matching errors caused by inconsistent data formats; combination integrates multiple independent features into a whole, enabling a comprehensive judgment of multiple key dimensions at once when comparing new transactions with basic transactions, thereby more accurately identifying transaction records belonging to the same consumption behavior. Thus, the generation of behavioral structure features provides a solid data foundation for subsequent consumption event aggregation.

[0071] In some of the embodiments described above in this application, after extracting basic correlation feature information from transaction information, this basic correlation feature information needs to be standardized and combined to generate a behavioral structure feature. For details, please refer to... Figure 6 The technical solution for standardizing and combining basic relational feature information to generate a behavioral structure feature can further include the following steps: Step S71: Based on the payment amount information, classify the payment amount into different levels: the payment amount in the first amount range is divided into the low amount range, the payment amount in the second amount range is divided into the medium amount range, and the payment amount in the third amount range is divided into the high amount range. Step S72: Obtain the merchant's latitude and longitude coordinates or administrative region information based on the merchant's identity information, and hash the latitude and longitude coordinates or administrative region information to generate a short string of the merchant's location; Step S73: Obtain the user's prepaid card number, digital platform user ID, and mobile phone number based on the user's identity fragment information, and perform a combined hash of the user's prepaid card number, digital platform user ID, and mobile phone number.

[0072] Specifically, the tiered processing of payment amount information aims to discretize continuous monetary data into categories that are easier to process and compare. For example, the first amount range can be set to 0-50 yuan, representing low-value consumption; the second amount range can be set to 51-500 yuan, representing medium-value consumption; and the third amount range can be set to above 501 yuan, representing high-value consumption. These ranges can be flexibly adjusted according to actual business scenarios and data distribution, with the aim of improving the efficiency and robustness of subsequent comparisons by simplifying monetary information.

[0073] The processing of merchant identity information involves obtaining the merchant's geographical location information, such as latitude and longitude coordinates or administrative region information. This geographical information is then hashed to generate a short string representing the merchant's location. The purpose of hashing is to transform complex geographical coordinates or administrative division information into fixed-length, easy-to-store and compare identifiers, while protecting the merchant's privacy. In this way, it is possible to quickly determine whether different transactions occurred at geographically proximate merchants, thereby aiding in determining the correlation between transactions.

[0074] In practical applications, the processing of user identity fragments specifically involves obtaining the user's prepaid card number, digital platform user ID, and mobile phone number. These different user identifiers are then subjected to combined hashing. The purpose of combined hashing is to unify the user's identity information generated on different platforms or through different methods, generating a unique and anonymous user identifier. This helps to identify different transaction records belonging to the same user without directly exposing sensitive user information, thereby achieving cross-channel user behavior aggregation.

[0075] This application's solution transforms raw, heterogeneous transaction data into structured behavioral features by standardizing and combining basic correlation features such as payment amount, merchant identity, and user identity fragments. The tiered processing of payment amounts shifts the focus from precise numerical values ​​to range-based matching of spending levels, thus tolerating minor differences in transaction amounts and improving aggregation flexibility. The hashing of merchant locations provides an efficient and privacy-friendly method for identifying geographic correlations. The combined hashing of user identity fragments resolves the issue of inconsistent user identities across different payment channels or platforms, ensuring accurate correlation of transactions by the same user in different scenarios.

[0076] In some of the embodiments described above in this application, when a new transaction occurs, it is necessary to compare the actual association feature information and the basic association feature information of the new transaction according to the association rules, and to perform logical inference in combination with the payment amount information, so as to identify and aggregate transaction records belonging to the same consumption behavior and form a consumption event. However, in practical applications, directly comparing and logically inferring based on the original association feature information may face problems such as data fragmentation, inconsistent feature information, or low inference efficiency, thereby affecting the accuracy and robustness of consumption event aggregation.

[0077] In response, this application further proposes the above-mentioned steps of comparing the actual related characteristic information and basic related characteristic information of new transactions, and combining payment amount information to make logical inferences in order to identify and aggregate transaction records belonging to the same consumption behavior to form a consumption event, including: identifying the actual related characteristic information and behavioral structure characteristics.

[0078] Specifically, upon receiving new transaction information, the system first extracts its actual related feature information, which may include payment time, payment amount, merchant identity information, and fragmented user identity information. Simultaneously, the system has pre-processed and standardized the basic related feature information using the aforementioned methods to generate behavioral structure features. Behavioral structure features are highly abstracted and structured representations of the original related feature information. For example, payment amounts are categorized into low, medium, or high ranges; merchant location information is hashed into a short string representing the merchant's location; and user identity fragments are combined and hashed. Here, "identification" refers to the process of matching and comparing the actual related feature information of the new transaction with the pre-generated behavioral structure features. This may involve standardizing and combining the actual related feature information of the new transaction in the same way as generating the behavioral structure features, forming a temporary behavioral structure feature, and then performing similarity calculations or pattern matching between this temporary behavioral structure feature and existing behavioral structure features. In this way, it is possible to efficiently determine whether a new transaction is highly related to existing transaction records belonging to the same consumption behavior.

[0079] In some of the embodiments described above in this application, when assessing the credibility of a consumption event, it is possible that the event is composed of multiple original transaction records, resulting in inaccurate timestamps and payment amount information, which may affect the accuracy of subsequent credibility assessments. For example, when a consumption behavior is broken down into multiple transactions, simply aggregating these transactions may not provide a unified and accurate point in time or total amount, nor may it be easy to directly infer the actual goods or services purchased, which poses a challenge to credibility assessment.

[0080] For this, please refer to Figure 7 This application further proposes the following steps prior to the credibility assessment of a consumer event: Step S81: Obtain multiple original transaction information corresponding to the consumption event. The transaction information includes payment time information and payment amount information. Step S82: Calculate the average or median of multiple payment time information as the calibration timestamp for the consumption event; Step S83: Add up the payment amounts of all the original transaction information to get a total amount; Step S84: Match the total amount with the merchant's product or service price, and infer the product or service that matches the known coupon semantics.

[0081] Specifically, before assessing the credibility of a consumption event, it is necessary to first obtain multiple sets of original transaction information corresponding to that event. This original transaction information forms the foundational data of the consumption event, and includes at least payment time and payment amount information. Obtaining this raw, fine-grained transaction data is essential for more accurate calibration and semantic inference of the aggregated consumption event.

[0082] Furthermore, to provide a unified and representative time reference for consumption events, the average or median of the payment time information of these original transaction details is calculated and used as the calibration timestamp for the consumption event. This effectively eliminates the time dispersion caused by multiple transactions occurring within a short period, providing a more accurate and stable time benchmark for subsequent credibility assessments.

[0083] Furthermore, to obtain the true total value of a consumption event, the payment amounts from all original transaction information are summed to arrive at a total amount. This total amount represents the actual expenditure involved in the consumption event and is a key basis for inferring the value of goods or services.

[0084] Based on this, the total amount obtained above is matched with the price of the goods or services offered by the merchant. During the matching process, known coupon semantics are also considered for inference. For example, if the total amount matches the original price of a product plus tax, minus the amount of a coupon, then it can be inferred that the consumption event likely involves that product. In this way, even if a consumption event lacks specific product details, goods or services that match the monetary logic of the consumption event can be inferred through monetary logic, thus endowing the consumption event with richer semantic information.

[0085] This application's solution effectively addresses the potential ambiguity in time and amount that may arise from aggregated transaction records by preprocessing and calibrating consumption events before credibility assessment. By acquiring original transaction information, calculating calibration timestamps and total amounts, and combining merchant pricing and coupon semantics to infer goods or services, each consumption event possesses more precise time positioning, more accurate total amount, and clearer identification of goods or services. This provides a more solid and reliable data foundation for subsequent credibility assessment, significantly improving the accuracy and effectiveness of the evaluation.

[0086] Traditional online advertising and marketing systems for prepaid cards have significant limitations in processing data from diverse payment channels, identifying consumer behavior under mixed payment models, and obtaining detailed product information. These issues make it difficult for the system to build comprehensive and detailed user profiles, thereby affecting the accuracy and conversion efficiency of ad targeting.

[0087] For this, please refer to Figure 8 The specific implementation of this application also discloses a commercial prepaid card internet advertising and marketing system, which is applied to the method described above, specifically including: The information acquisition module 81 is used to acquire transaction information from different payment channels and extract basic related feature information from the transaction information. The related feature information includes at least one of the following: payment time information, payment amount information, merchant identity information, and user identity fragment information. The rule setting module 82 is used to set association rules, which include time proximity rules, merchant consistency rules, and user identity fragment association rules. The event aggregation module 83, when a new transaction occurs, is used to compare the actual association feature information and the basic association feature information of the new transaction according to the association rules, and to make logical inferences in combination with the payment amount information, so as to identify and aggregate transaction records belonging to the same consumption behavior and form a consumption event. The credibility assessment module 84 is used to assess the credibility of consumption events and determine whether a consumption event is a valid consumption event based on the credibility assessment results. The preference extraction module 85 is used to extract the user's product preference information based on a valid consumption event, provided that the valid consumption event lacks specific product details. Specifically, when a valid consumption event lacks detailed product information, the user's preference information is inferred by combining merchant identity information and payment amount information. The ad matching module 86 is used to update the user profile based on product preference information and to perform ad matching using the updated user profile.

[0088] It should be understood that the above modules can also execute the marketing methods described above.

[0089] Compared to existing technologies, the core innovation of the commercial prepaid card internet advertising and marketing system proposed in this application lies in its modular design and intelligent processing capabilities for multi-source heterogeneous transaction data. Traditional systems often struggle to effectively integrate fragmented transaction data from different payment channels, and are even less capable of accurately inferring user preferences in the absence of specific product details. This application overcomes the limitations of existing technologies in terms of data fragmentation and information gaps by using an information acquisition module to uniformly process heterogeneous data, an event aggregation module to intelligently identify and integrate scattered transactions under mixed payment models, and a preference extraction module to perform logical inference when information is missing. For example, in a traditional system, if a user uses a prepaid card and a digital coupon to pay at the same merchant, it may be identified as two independent transactions, and it is impossible to know what was purchased. However, the system in this application can intelligently link these two transactions through the event aggregation module, and the preference extraction module can infer the category of goods the user may have purchased based on the total amount and merchant type, thereby constructing a more comprehensive and detailed user profile. This capability enables the system in this application to significantly improve the accuracy of user profiling, thereby achieving more efficient and accurate ad matching, bringing significant progress to internet advertising marketing for prepaid commercial cards.

[0090] The above description is merely an embodiment of this application and is not intended to limit the scope of protection of this application. Various modifications and variations can be made to this application by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this application should be included within the scope of protection of this application.

Claims

1. A method for internet advertising and marketing of prepaid commercial cards, characterized in that, include: Transaction information from different payment channels is obtained, and basic related feature information is extracted from the transaction information. The related feature information includes at least one of the following: payment time information, payment amount information, merchant identity information, and user identity fragment information. Set association rules, including time proximity rules, merchant consistency rules, and user identity fragment association rules; When a new transaction occurs, the actual association feature information and the basic association feature information of the new transaction are compared according to the association rules, and logical inference is performed in combination with the payment amount information to identify and aggregate transaction records belonging to the same consumption behavior to form a consumption event. The credibility of the consumption event is assessed, and based on the credibility assessment result, it is determined whether the consumption event is a valid consumption event. If the consumption event is a valid consumption event, then based on the valid consumption event, the user's product preference information is extracted. When the valid consumption event lacks specific product details, the product preference information is obtained by combining the merchant identity information and the payment amount information. as well as Based on the product preference information, update the user profile and use the updated user profile for ad matching.

2. The method for internet advertising and marketing of prepaid commercial cards according to claim 1, characterized in that, When a new transaction occurs, the steps of comparing the actual association feature information and the basic association feature information of the new transaction according to the association rules, and performing logical inference in conjunction with the payment amount information to identify and aggregate transaction records belonging to the same consumption behavior to form a consumption event include: According to the association rules, the actual payment time information and basic payment time information of the new transaction, the actual merchant identity information and basic merchant identity information of the new transaction, and the actual user identity fragment information and basic user identity fragment information of the new transaction are compared to obtain a preliminary aggregation set. Obtain the merchant's currently effective set of promotional rules, which includes bundled sales discounts, tiered discounts, and discounts for specific product combinations; Based on the promotion rule set, the merchant's product list is preprocessed to generate a dynamic discounted product combination list containing product combinations with different discount rules applied and their corresponding discounted prices. The payment amount of the initial aggregated set is compared with the discounted price in the dynamic discount product combination list; and Based on the comparison results, the combination of goods that best matches the payment amount is selected as the inference result to identify and aggregate transaction records belonging to the same consumption behavior, thus forming a consumption event.

3. The method for internet advertising and marketing of prepaid commercial cards according to claim 1, characterized in that, The step of performing a credibility assessment on the consumption event and determining whether the consumption event is a valid consumption event based on the credibility assessment result includes: Identify the scenario type in which the current consumption event occurs, including merchant type, transaction time period, and geographical region; Based on the scenario type, select the weight configuration corresponding to the scenario type from the preset weight configuration set; Using the aforementioned weight configuration, a weighted sum of the matching scores of each associated feature in the consumption event is calculated to obtain the integrated credibility score of the consumption event; and The integration credibility score is compared with a preset event aggregation threshold to determine whether the consumption event is a valid consumption event. If the integration credibility score is greater than the preset event aggregation threshold, then the consumption event is determined to be a valid consumption event.

4. The method for internet advertising and marketing of prepaid commercial cards according to claim 1, characterized in that, If the consumption event is a valid consumption event, then based on the valid consumption event, the user's product preference information is extracted. Specifically, when the valid consumption event lacks detailed product information, the product preference information is inferred by combining the merchant's identity information and the payment amount information. Obtain user interaction information on online channels and extract non-transactional behavior clues from the interaction information. The non-transactional behavior clues include the duration of user browsing specific product pages, the frequency of clicking on specific category advertisements, the record of collecting coupons, and the interaction with specific product topics. A virtual product display list is constructed for each merchant in a complex business format. The virtual product display list includes the goods or services provided by the merchant and the sub-business format to which they belong. Based on the merchant identity information and the payment amount information, combined with the virtual product display list and known discount rules, multiple product combination schemes are generated. The product combination scheme includes product categories and the total price after applying discounts. Based on the non-transactional behavioral clues, a user interest attraction field is constructed, which includes the intensity of a user's interest in a specific product category or service theme. The multiple product combination schemes are compared with the user interest gravity field, and the product combination scheme with the highest degree of consistency with the user interest gravity field is selected as the product preference information; and Based on the product preference information, update the user profile and use the updated user profile for ad matching.

5. The method for internet advertising and marketing of prepaid commercial cards according to claim 1, characterized in that, The step of updating the user profile based on the product preference information and using the updated user profile for ad matching includes: Obtain new product preference information; The new product preference information is assigned an initial short-term interest weight and included in the short-term interest weight pool. Based on the consistency between the new product preference information and the user's historical long-term preferences, the accumulation intensity of the new product preference information in the long-term preference accumulation area is adjusted; When there is a first preference in the short-term interest weight pool whose weight reaches the preset preference fusion threshold, and a preference that is associated with the accumulation intensity of the first preference is found in the long-term preference accumulation area, the short-term interest is converted into a long-term preference. Update the user profile based on the current weights of preferences in the short-term interest weight pool and the accumulated strength of preferences in the long-term preference accumulation area; and The updated user profile is used for ad matching.

6. The commercial prepaid card internet advertising marketing method according to claim 1, characterized in that, The step of extracting basic association feature information from the transaction information includes: The basic associated feature information is standardized and combined to generate a behavioral structure feature.

7. A commercial prepaid card internet advertising marketing method according to claim 6, characterized in that, The step of standardizing and combining the basic associated feature information to generate a behavioral structure feature includes: The payment amount is classified into three categories based on the payment amount information: the payment amount in the first amount range is classified into the low amount range, the payment amount in the second amount range is classified into the medium amount range, and the payment amount in the third amount range is classified into the high amount range. The merchant's latitude and longitude coordinates or administrative region information are obtained based on the merchant's identity information, and the latitude and longitude coordinates or administrative region information are hashed to generate a short string of the merchant's location. Based on the user's identity fragment information, obtain the user's prepaid card number, digital platform user ID, and mobile phone number, and then combine and hash the user's prepaid card number, digital platform user ID, and mobile phone number.

8. A commercial prepaid card internet advertising marketing method according to claim 7, characterized in that, The steps of logically inferring the actual and basic association characteristics of a new transaction, combined with the payment amount information, to identify and aggregate transaction records belonging to the same consumption behavior, thus forming a consumption event, include: The actual associated feature information and the behavioral structure features are identified.

9. A commercial prepaid card internet advertising marketing method according to claim 1, characterized in that, Prior to the step of performing a credibility assessment on the consumption event, the following steps are included: Based on the consumption event, obtain multiple original transaction information, including payment time information and payment amount information; Calculate the average or median of multiple payment time information as the calibration timestamp of the consumption event; Add up the payment amounts of all the original transaction information to get a total amount; The total amount is matched with the merchant's product or service price, and inferences are made based on known coupon semantics to obtain products or services with logically consistent amounts.

10. A commercial prepaid card internet advertising and marketing system, characterized in that, The system includes: The information acquisition module is used to acquire transaction information from different payment channels and extract basic related feature information from the transaction information. The related feature information includes at least one of the following: payment time information, payment amount information, merchant identity information, and user identity fragment information. The rule setting module is used to set association rules, which include time proximity rules, merchant consistency rules, and user identity fragment association rules. The event aggregation module, when a new transaction occurs, is used to compare the actual association feature information and the basic association feature information of the new transaction according to the association rules, and to make logical inferences in combination with the payment amount information, so as to identify and aggregate transaction records belonging to the same consumption behavior and form a consumption event. The credibility assessment module is used to assess the credibility of the consumption event and determine whether the consumption event is a valid consumption event based on the credibility assessment result. The preference extraction module is used to extract the user's product preference information based on the valid consumption event if the consumption event is a valid consumption event. Specifically, when the valid consumption event lacks detailed product information, the product preference information is inferred by combining the merchant's identity information and the payment amount information. The ad matching module is used to update the user profile based on the product preference information, and to perform ad matching using the updated user profile.

Images