Discount information recommendation system and method, electronic equipment and storage medium

Abstract

According to the preferential information recommendation system and method, the electronic equipment and the storage medium provided by the embodiment of the invention, the preferential information recommendation system can effectively solve the inherent defects of an extensive preferential strategy of a current e-commerce platform through a modular collaboration mechanism, on the whole, the user feature data realizes operation upgrading from'thousands of persons' to'thousands of persons', and the user experience is improved. Through accurate matching of user demands and preferential resources, the ROI of promotion budget is optimized, a virtuous cycle is formed in the aspects of activation of dormant users, cultivation of brand loyalty, mining of incremental markets and the like, and finally the traditional trapping of high investment and low conversion is broken through.

Description

Technical Field

[0001] The embodiments of the present invention relate to the field of computer technology, and in particular to a discount information recommendation system, method, electronic device and storage medium. Background Technology

[0002] Currently, e-commerce platforms generally employ a "one-size-fits-all" discount strategy in their promotional activities. This means that the platform generates and distributes the same discounts or spending threshold reductions to all users. This operational model, lacking user segmentation, has revealed significant efficiency shortcomings. Data shows that high-spending users have a lower redemption rate for small discounts like "10 off for every 100 spent," while low-spending users have an even lower redemption rate for high-threshold discounts like "5,000 off for every 30,000 spent." This resource mismatch leads to a double waste of resources—failing to stimulate incremental demand from high-spending groups and ineffectively consuming budgets allocated to potential users. A deeper problem lies in the fact that standardized discount mechanisms completely ignore user differences. This extensive operation not only causes a continuous decline in the ROI (Return on Investment) of promotional resources but also causes e-commerce platforms to miss multiple opportunities to leverage precise discounts to increase average order value, activate dormant users, and cultivate consumer loyalty, ultimately falling into a promotional predicament of "high investment, low conversion." Summary of the Invention

[0003] In view of this, embodiments of the present invention provide a discount information recommendation system, method, electronic device, and storage medium, which can effectively alleviate the above-mentioned technical problems.

[0004] In a first aspect, embodiments of the present invention provide a discount information recommendation system, which includes a data feature module, a product sorting module, a discount information calculation module, and a discount information delivery module connected in sequence, wherein the data feature module is also connected to the discount information calculation module;

[0005] The data feature module is used to collect user behavior data of target users and product feature data of each product in the product pool in real time, generate user feature data based on user behavior data, and send the user feature data and product feature data to the product sorting module, as well as send the user feature data to the discount information calculation module.

[0006] The product sorting module is used to determine the matching degree value for each product based on user feature data and product feature data, generate a product recommendation list according to the matching degree value of each product, and select at least one target product from the product recommendation list to send to the discount information calculation module; wherein, the target product is the product that needs to be given discount information to the target user;

[0007] The discount information calculation module is used to generate discount information corresponding to each of the at least one target products based on user feature data, and send the discount information to the discount information delivery module.

[0008] The promotional information delivery module is used to calculate the channel activity ratio of target users across various communication channels, determine the target communication channels to reach based on the channel activity ratio, and send promotional information to target users through the target communication channels.

[0009] In one possible implementation, the data feature module includes a data acquisition unit, a feature generation unit, and a data transmission unit connected in sequence.

[0010] The data acquisition unit is used to collect user behavior data of target users and product feature data of each product in the product pool in real time through a distributed data acquisition architecture, and send the user behavior data to the feature generation unit and the product feature data to the data sending unit.

[0011] The feature generation unit is used to clean user behavior data, perform feature clustering analysis on the cleaned user behavior data, generate user feature data of the target user, and send the user feature data to the data sending unit.

[0012] The data sending unit is used to send user feature data and product feature data to the product sorting module, and to send user feature data to the discount information calculation module.

[0013] In one possible implementation, the product sorting module includes a matching degree generation unit, a product list generation unit, and a first sending unit connected in sequence.

[0014] The matching degree generation unit is used to input user feature data and product feature data into a pre-trained matching degree value generation model for each product. The matching degree value generation model outputs a basic matching degree value, extracts the remaining validity period of the product from the product feature data, determines the timeliness coefficient based on the remaining validity period of the product, generates a matching degree value based on the basic matching degree value and the timeliness coefficient, and sends the matching degree value of each product to the product list generation unit.

[0015] The product list generation unit is used to sort the products according to the matching degree value of each product to generate a product recommendation list, send the product recommendation list to the display interface to display to the target user, and send the product recommendation list to the first sending unit;

[0016] The first sending unit is used to select at least one target product from the product recommendation list and send the at least one target product to the discount information calculation module.

[0017] In one possible implementation, the timeliness coefficient is determined based on the remaining shelf life of the goods, including:

[0018] Obtain product circulation information;

[0019] When the commodity circulation information is set as accelerated commodity circulation information, the timeliness coefficient is determined based on the remaining shelf life of the commodity and the first timeliness coefficient calculation formula.

[0020] When the commodity circulation information is set to not accelerate the circulation of the commodity, the timeliness coefficient is determined based on the remaining shelf life of the commodity and the second timeliness coefficient calculation formula.

[0021] In one possible implementation, the discount information calculation module includes a calculation unit and a second sending unit connected in sequence.

[0022] The computing unit is used to determine the user type of the target user based on user feature data using a clustering algorithm, extract business data from the user feature data, input the user type and business data into a pre-trained discount information generation model, the discount information generation model outputs the discount information corresponding to the target product, and sends the discount information to the second sending unit; wherein, the business data includes at least user price sensitivity data, consumption potential data, and churn risk data;

[0023] The second sending unit is used to send the promotional information to the promotional information delivery module.

[0024] In one possible implementation, the promotional information delivery module includes a statistical calculation unit and a channel adjustment delivery unit connected in sequence.

[0025] The statistical calculation unit is used to count the interaction frequency of target users on each communication channel within a specified time range, calculate the channel activity ratio of each communication channel based on the interaction frequency, and send the channel activity ratio of each communication channel to the channel adjustment determination unit.

[0026] The channel adjustment and outreach unit is used to use the channel activity ratio of each communication channel as the context feature of the dynamic resource allocation algorithm. The dynamic resource allocation algorithm dynamically adjusts the channel weight of each communication channel and determines the communication channel with the channel weight exceeding the preset weight threshold as the target communication channel. The promotional information is asynchronously pushed to the message topic of the target communication channel using a message queue, so as to send the promotional information to the target user.

[0027] In one possible implementation, the discount information recommendation system further includes a tracking module connected to the discount information calculation module;

[0028] The tracking module is used to track the target user's full-link response, obtain the weight value of each link in the full-link response, and send the weight value of each link to the calculation unit so that each weight value can be used as the sample weight of the loss function to update the discount information generation model. Here, the full-link response refers to the quantitative tracking of the target user's complete behavioral path from product display to final repurchase.

[0029] Secondly, embodiments of the present invention provide a method for recommending preferential information, which is applied to the aforementioned preferential information recommendation system. The method includes:

[0030] Real-time collection of user behavior data of target users and product feature data of each product in the product pool; and generation of user feature data based on user behavior data.

[0031] For each product, a matching degree value is determined based on user feature data and product feature data. A product recommendation list is generated according to the matching degree value corresponding to each product, and at least one target product is selected from the product recommendation list. The target product is the product that needs to be delivered to the target user with preferential information.

[0032] For each of the at least one target products, generate corresponding discount information based on user feature data;

[0033] Statistically analyze the channel activity ratio of target users across various communication channels, determine the target communication channels to reach based on the channel activity ratio, and use the target communication channels to deliver promotional information to target users.

[0034] Thirdly, embodiments of the present invention provide a discount information recommendation device, the device comprising:

[0035] The first module is used to collect user behavior data of target users and product feature data of each product in the product pool in real time, and generate user feature data based on user behavior data.

[0036] The second module is used to determine the matching degree value for each product based on user feature data and product feature data, generate a product recommendation list according to the matching degree value corresponding to each product, and select at least one target product from the product recommendation list; wherein, the target product is the product that needs to be delivered preferential information to the target user;

[0037] The third module is used to generate corresponding discount information for each of the at least one target products based on user feature data.

[0038] The fourth module is used to statistically analyze the channel activity ratio of target users across various communication channels, determine the target communication channels to reach based on the channel activity ratio, and use the target communication channels to deliver promotional information to the target users. In a fourth aspect, embodiments of the present invention provide an electronic device, comprising: a processor and a memory, wherein the processor is used to execute a program storing promotional information recommendations in the memory to implement the aforementioned promotional information recommendation method.

[0039] Fifthly, embodiments of the present invention provide a storage medium, wherein the storage medium stores one or more programs, which can be executed by one or more processors to implement the above-described method for recommending preferential information.

[0040] The discount information recommendation system, method, electronic device, and storage medium provided in this invention effectively address the inherent defects of the current extensive discount strategies of e-commerce platforms through a modular collaborative mechanism. Its core beneficial effects are reflected in the following: Through user characteristic data generated in real time by the data feature module, the system can accurately identify the preferences of users at different consumption levels; the product sorting module filters high-potential target products corresponding to users based on multi-dimensional feature matching, fundamentally avoiding the mismatch of discount resources; the discount information calculation module generates differentiated discount strategies (such as exclusive benefits for high-net-worth users or tiered discounts for potential users) through user characteristic data, significantly improving discount redemption rates and average order value; and the discount information delivery module, based on an intelligent distribution mechanism of communication channel activity ratios, ensures that discount information reaches users through the most efficient path, reducing invalid exposure while enhancing user stickiness. Overall, user characteristic data has enabled an operational upgrade from "one-size-fits-all" to "personalized" user experience. By accurately matching user needs with preferential resources, it has not only optimized the ROI of promotional budgets, but also created a virtuous cycle in activating dormant users, cultivating brand loyalty, and exploring incremental markets, ultimately breaking the traditional dilemma of "high investment and low conversion". Attached Figure Description

[0041] Figure 1 This is a schematic diagram of the structure of a discount information recommendation system provided in an embodiment of the present invention;

[0042] Figure 2 This is a schematic diagram of another discount information recommendation system provided in an embodiment of the present invention;

[0043] Figure 3 A flowchart illustrating an embodiment of a discount information recommendation method provided by this invention;

[0044] Figure 4 A block diagram illustrating an embodiment of a discount information recommendation device provided by the present invention;

[0045] Figure 5This is a schematic diagram of the structure of an electronic device provided in an embodiment of the present invention. Detailed Implementation

[0046] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0047] To facilitate understanding of the embodiments of the present invention, further explanations and descriptions will be provided below with reference to the accompanying drawings and specific embodiments. These embodiments do not constitute a limitation on the embodiments of the present invention.

[0048] This invention provides a discount information recommendation system, such as... Figure 1 The diagram shows a structural schematic of a discount information recommendation system. The discount information recommendation system includes: a data feature module 100, a product sorting module 101, a discount information calculation module 102, and a discount information delivery module 103 connected in sequence. The data feature module 100 is also connected to the discount information calculation module 102.

[0049] In specific implementation, the data feature module is used to collect user behavior data of target users and product feature data of each product in the product pool in real time, generate user feature data based on user behavior data, and send the user feature data and product feature data to the product sorting module and the discount information calculation module; the product sorting module is used to determine the matching degree value of each product based on the user feature data and product feature data, generate a product recommendation list according to the matching degree value of each product, and select at least one target product from the product recommendation list to send to the discount information calculation module; wherein, the target product is the product that needs to be delivered discount information to the target user; the discount information calculation module is used to generate the discount information corresponding to the target product based on the user feature data for each of the at least one target product, and send the discount information to the discount information delivery module; the discount information delivery module is used to count the channel activity ratio of the target user in each communication channel, determine the target communication channel to be reached based on the channel activity ratio, and send the discount information to the target user using the target communication channel.

[0050] This discount information recommendation system can be deployed on service platforms within the product operation phase of a membership-based business ecosystem, covering areas such as video platforms (e.g., movie membership packages), e-commerce memberships (e.g., exclusive discounted products), and service memberships (e.g., annual gym memberships, food delivery coupons). Its core applications include personalized display of membership products, dynamic pricing, and precise outreach throughout the entire process. It is particularly suitable for platforms with highly volatile user behavior and a wide variety of product categories (e.g., the "movie + sports + children's" multi-category membership system of a comprehensive video platform).

[0051] Among them, the user behavior data collected in real time by the data feature module is the core data source for quantitatively representing users' personal preferences, decision-making patterns and psychological states. In this embodiment, user behavior data includes, but is not limited to: 1. Micro-behavior: dwell time on product details page (≥2 seconds triggers recording), clicking "Add to Cart" but not paying (marked as "hesitation behavior"), coupons not used within 72 hours after being claimed (marked as "invalid coupon"); 2. Macro-behavior: daily distribution of member product browsing categories, weekly average number of purchases, cross-device login trajectory (such as switching from mobile phone to tablet to PC); 3. Environmental data: current network status (Wi-Fi / 4G), device model (affecting display adaptation), geographical location (such as pushing "region-exclusive discounts" to non-member service areas).

[0052] This product pool serves as a standardized product repository for the centralized service platform, integrating all products required for the platform's current operations. Each product within the service platform is configured with a multi-dimensional feature data system, including static attributes such as membership type (movie / sports / general) and service period (monthly / quarterly / yearly), as well as dynamic attributes such as real-time updated operational metrics (e.g., 24-hour sales growth rate, inventory turnover warning threshold).

[0053] User characteristic data generated based on user behavior data is a collection of digital profiles that can systematically reflect the behavior patterns and consumption characteristics of individual users. This type of data includes at least real-time interests (such as the weight of categories browsed in the past hour), historical preferences (the percentage of categories purchased in the past 30 days), and price tolerance (the highest historical purchase amount). Essentially, it constitutes the core data asset for service platforms to identify user needs, predict behavioral tendencies, and formulate precise preferential operation strategies.

[0054] The matching score determined by the product sorting module is used to measure the degree of fit between the target user's needs and the product's needs. It can be understood that the matching score, as a core quantitative indicator, directly reflects the strength of the demand association between the target user and each product. The matching score is positively correlated with the user's willingness to buy. That is, the higher the matching score, the greater the user's potential purchase probability, while the lower the matching score, the weaker the demand match.

[0055] To attract target users, a product recommendation list can be generated by sorting the products in descending order according to their matching scores. This list can be clearly presented to target users, and at least one product that the target user has a strong purchase intention can be selected from the sorted product recommendation list as the target product for which to recommend discount information. This serves as the core basis for formulating discount strategies and ensures that marketing resources accurately reach users with high conversion potential.

[0056] The discount information calculation module intelligently generates differentiated discount information based on user characteristic data, ensuring that the discount information received by target users is accurately matched with their consumption preferences. The discount information delivery module realizes intelligent distribution of discount information through cross-channel marketing automation platforms (such as APP push, SMS, social media and other communication channels), effectively improving user reach and conversion efficiency.

[0057] This promotional information recommendation system, through a modular collaborative mechanism, effectively addresses the inherent flaws of current e-commerce platforms' extensive promotional strategies. Its core benefits are: First, the system accurately identifies the preferences of users at different consumption levels by generating real-time user characteristic data through the data feature module. Second, the product sorting module filters high-potential target products for users based on multi-dimensional feature matching, fundamentally avoiding mismatch of promotional resources. Third, the promotional information calculation module generates differentiated promotional strategies (such as exclusive benefits for high-net-worth users or tiered discounts for potential users) based on user characteristic data, significantly improving discount redemption rates and average order value. Fourth, the promotional information delivery module uses an intelligent distribution mechanism based on the active percentage of communication channels to ensure that promotional information reaches users through the most efficient path, reducing ineffective exposure while enhancing user stickiness. Overall, user characteristic data enables an operational upgrade from "one-size-fits-all" to "personalized" experiences. By accurately matching user needs with promotional resources, it not only optimizes the ROI of promotional budgets but also creates a virtuous cycle in activating dormant users, cultivating brand loyalty, and tapping into incremental markets, ultimately breaking the traditional predicament of "high investment, low conversion."

[0058] exist Figure 1 On this basis, Figure 2 This diagram illustrates the structure of another discount information recommendation system, such as... Figure 2 As shown, the data feature module 100 includes a data acquisition unit 201, a feature generation unit 202, and a data transmission unit 203 connected in sequence.

[0059] The data acquisition unit is used to collect user behavior data of target users and product feature data of each product in the product pool in real time through a distributed data acquisition architecture, and send the user behavior data to the feature generation unit and the product feature data to the data sending unit. The feature generation unit is used to clean the user behavior data, perform feature clustering analysis on the cleaned user behavior data, generate user feature data of the target users, and send the user feature data to the data sending unit. The data sending unit is used to send the user feature data and product feature data to the product sorting module and the discount information calculation module.

[0060] Specifically, the data acquisition unit can adopt a multi-node deployment acquisition agent, which captures the various user behavior data and product feature data described in the above embodiments through the data collection SDK (Software Development Kit).

[0061] The feature generation unit cleans the user behavior data, removing invalid and abnormal user behavior data, and then uses clustering algorithms such as K-means, DBSCAN (Density-Based Spatial Clustering of Applications with Noise), or large language models to cluster the user behavior data to obtain user feature data.

[0062] As the final output stage of the data feature module, the data sending unit sends user feature data and product feature data to the product sorting module on the one hand, and user feature data to the discount information calculation module on the other hand.

[0063] like Figure 2As shown, the product sorting module 101 includes a matching degree generation unit 204, a product list generation unit 205, and a first sending unit 206 connected in sequence. The matching degree generation unit is used to input user feature data and product feature data into a pre-trained matching degree value generation model for each product. The matching degree value generation model outputs a basic matching degree value, extracts the remaining validity period of the product from the product feature data, determines the timeliness coefficient based on the remaining validity period, generates a matching degree value based on the basic matching degree value and the timeliness coefficient, and sends the matching degree value of each product to the product list generation unit. The product list generation unit is used to sort the products according to the size of their matching degree values ​​to generate a product recommendation list, sends the product recommendation list to the display interface for display to the target user, and sends the product recommendation list to the first sending unit. The first sending unit is used to select at least one target product from the product recommendation list and send the at least one target product to the discount information calculation module.

[0064] The aforementioned matching score generation model employs the DeepFM (Deep Factorization Machine) deep learning framework. It is jointly trained by fusing historical user feature data (real-time interests (e.g., category weights viewed in the past hour), historical preferences (percentage of categories purchased in the past 30 days), price tolerance (highest historical purchase amount), etc.) and historical product feature data (static attributes (member type, duration), dynamic attributes (sales growth rate in the past 24 hours, remaining inventory warning value)). This model innovatively achieves: 1. low-order feature cross-referencing (e.g., direct association between user interest tags and product categories), and 2. high-order feature combination (e.g., deep non-linear interaction of multi-dimensional features such as purchase frequency and price sensitivity). Ultimately, it outputs a quantified basic user-product matching score, providing data support for precision marketing.

[0065] It is important to note that historical user feature data and historical product feature data need to be converted into vector form before they can be used as training data.

[0066] In practical applications, the user feature data of the target user and the product feature data of each product are transformed into vectors and then input into the matching degree value generation model. The matching degree value generation model outputs the basic matching degree value between the target user and each product.

[0067] The remaining valid period of the product extracted from the product feature data; it refers to the difference in natural days between the current date (August 1, 2025) and the planned off-shelf date of the product (August 15, 2025). This indicator is calculated as follows: Remaining valid period of the product (days) = Off-shelf date of the product - Current date. The above remaining valid period of the product is 14 days. The unit of the specific remaining valid period of the product can be set according to actual needs and is not limited here.

[0068] The timeliness system is used to reflect the impact of product timeliness on the basic matching degree value. The determination process of the specific timeliness coefficient can be achieved through steps A1 to A3:

[0069] Step A1, obtain the product circulation information;

[0070] The product circulation information is a key parameter in the service platform and is information used to dynamically adjust the product sales strategy.

[0071] Step A2, when the product circulation information is set as the product accelerated circulation information, determine the timeliness coefficient based on the remaining valid period of the product and the first timeliness coefficient calculation formula;

[0072] As the core control parameter, the product circulation information realizes the differential control of the sales strategy through binary state setting: when set as "product accelerated circulation information", it will automatically trigger the process of dealing with near-expiry products (i.e., step A2), and enhance the weight value of the product in the supply-demand matching algorithm by dynamically increasing the timeliness coefficient, which can be calculated by substituting the remaining valid period of the product into the first timeliness coefficient calculation formula. The above first timeliness coefficient calculation formula is A + B × Remaining valid period of the product, (where A represents the basic guarantee coefficient, and its value is in the positive integer domain; B is the attenuation adjustment factor, and its value range is limited to a decimal of <0 < B < 1). This design not only ensures that near-expiry products obtain sufficient traffic tilt but also avoids the risk of numerical overflow through parameter constraints.

[0073] Step A3, when the product circulation information is set as the product non-accelerated circulation information, determine the timeliness coefficient based on the remaining valid period of the product and the second timeliness coefficient calculation formula.

[0074] Conversely, when the commodity circulation information is set to "non-accelerated commodity circulation information", the conventional inventory management strategy (i.e., step A3) is executed. By dynamically adjusting the timeliness coefficient, the weight ratio of the commodity in the supply-demand matching algorithm is reduced to maintain the standard sales rhythm. The timeliness coefficient can be calculated by substituting the remaining validity period of the commodity into the second timeliness coefficient calculation formula. The above second timeliness coefficient calculation formula is B × the remaining validity period of the commodity (where B is the decay adjustment factor, and the value range is limited to a decimal of 0 < B < 1). This design not only ensures that the commodity maintains a normal sales rhythm but also avoids excessive interference in the market price system through a pure multiplier calculation model, which is particularly suitable for inventory control scenarios of high-value commodities or limited-edition commodities.

[0075] Among them, the matching degree value can be obtained by multiplying the basic matching degree value and the timeliness coefficient. The corresponding calculation formula is: matching degree value = basic matching degree value × timeliness coefficient.

[0076] After sorting each commodity in descending order of the matching degree value to obtain a commodity recommendation list, the first commodities with a preset display quantity ranked at the top can be screened from the commodity recommendation list, and the first commodities can be dynamically rendered to the display interface. For example, the top 3 commodities are displayed at the member entrance on the home page, and 2 potential commodities (not purchased by the user but frequently purchased by similar groups) ranked 4th and 5th are displayed at the secondary entrance, and the commodity ranked 6th is displayed at the bottom.

[0077] In actual application, the number of at least one target commodity selected from the commodity recommendation list can be set according to actual needs and is not limited here. For example, when selecting one target commodity, the commodity ranked first in the commodity recommendation list can be selected as the target commodity. When selecting multiple target commodities, multiple commodities ranked at the top in the commodity recommendation list can be selected as the target commodities.

[0078] The commodity sorting module can accurately quantify the degree of fit between the user's needs and the commodity based on the user characteristic data and commodity characteristic data of the target user, and can accurately identify the target commodities with strong user purchase intentions through the matching degree value, which is used as the core basis for formulating preferential strategies. The dynamic adjustment mechanism of the timeliness coefficient can be intelligently optimized for special scenarios such as approaching expiration commodities, which not only avoids inventory backlogs but also guarantees reasonable profits. The commodity sorting module can accurately quantify the degree of fit between the user's needs and the commodity based on the user characteristic data and commodity characteristic data of the target user, and can accurately identify the target commodities with strong user purchase intentions through the matching degree value, which is used as the core basis for formulating preferential strategies. The dynamic adjustment mechanism of the timeliness coefficient can be intelligently optimized for special scenarios such as approaching expiration commodities, which not only avoids inventory backlogs but also guarantees reasonable profits.

[0079] Such as Figure 2As shown, the discount information calculation module 102 includes a calculation unit 207 and a second sending unit 208 connected in sequence. The calculation unit is used to determine the user type of the target user based on user feature data using a clustering algorithm, extract business data from the user feature data, input the user type and business data into a pre-trained discount information generation model, the discount information generation model outputs the discount information corresponding to the target product, and sends the discount information to the second sending unit. The business data includes at least user price sensitivity data, consumption potential data, and churn risk data. The second sending unit is used to send the discount information to the discount information delivery module.

[0080] In practice, the computing unit first performs Z-score standardization on the original user feature data to eliminate dimensional differences between multidimensional features. Then, it uses the K-Means++ algorithm for iterative clustering, evaluates the fit between individuals and clusters using the silhouette score, and verifies the clustering effectiveness using the Calinski-Harabasz index. The dual indicators work together to determine the optimal K value. The algorithm terminates when it reaches the preset maximum number of iterations or the centroid shift threshold, thereby classifying target users into typical user types such as price-sensitive, high-frequency loyal, high-potential silent, new user conversion, or balanced user. The specific definition criteria for each type can be flexibly configured according to the characteristics of the business scenario.

[0081] Next, user types and business data extracted from user characteristic data (including user price sensitivity data, consumption potential data, and churn risk data) are input into a pre-trained discount information generation model, which then outputs the discount information corresponding to the target product.

[0082] The discount information generation model is an intelligent decision-making system built on reinforcement learning algorithms (such as Q-Learning). Its framework design includes complete Markov decision process elements: user types and business data are set in the state space; multiple discount combinations are set in the action space, and exploration and utilization are balanced through an ε-greedy strategy; the reward function adopts a dual-objective user response value optimization design of purchase = average order value - discount cost, and no purchase = -0.1 × display cost, and parameter updates are achieved through TD (Temporal Difference) error backpropagation. In the online inference stage, the model receives the user type generated in step 301, calculates the optimal discount information (such as "70% off + 15 off for purchases over 50") through the policy network, and this dynamic adjustment mechanism enables the service platform to achieve Pareto optimality between user incentives and resource management.

[0083] Finally, the second sending unit sends the discount information corresponding to each of the at least one target products to the discount information delivery module, so as to deliver the discount information to the target user.

[0084] like Figure 2 As shown, the discount information delivery module 103 includes a statistical calculation unit 209 and a channel adjustment delivery unit 210 connected in sequence. The statistical calculation unit is used to count the interaction frequency of target users with each communication channel within a specified time range, calculate the channel activity ratio of each communication channel based on the interaction frequency, and send the channel activity ratio of each communication channel to the channel adjustment determination unit. The channel adjustment determination unit is used to use the channel activity ratio of each communication channel as the context feature of the dynamic resource allocation algorithm, dynamically adjust the channel weight of each communication channel, determine the communication channel with the channel weight exceeding the preset weight threshold as the target communication channel, and use a message queue to asynchronously push the discount information to the message topic of the target communication channel to achieve the delivery of discount information to the target user.

[0085] Statistically track the number of times target users open and click on communication channels such as SMS, app push notifications, and emails on a daily / weekly (specific settings available) basis, calculate the standardized interaction frequency (number of open times + number of clicks), and then calculate the channel activity ratio based on the interaction frequency (e.g., interaction frequency of SMS channel / total interaction frequency).

[0086] The aforementioned dynamic resource allocation algorithm can be any algorithm capable of dynamically allocating reach resources, such as multi-armed slot machine algorithms or reinforcement learning algorithms; no specific limitation is made here. Its core logic uses the activity ratio of each communication channel as the initial weight, and dynamically adjusts the channel weights (i.e., allocation ratios) through continuous iterative feedback optimization. Communication channels whose weights exceed a preset weight threshold (which can be set according to actual needs) are designated as target communication channels, achieving optimal allocation of promotional information across multiple reach channels to deliver it to the target user.

[0087] To ensure the accuracy of the promotional information generation model, such as Figure 2 As shown, the tracking module 211, which is connected to the discount information calculation module 102, can be used to update the discount information generation model online and offline.

[0088] The tracking module is used to track the target user's full-link response, obtain the weight value of each link in the full-link response, and send the weight value of each link to the calculation unit so that each weight value can be used as the sample weight of the loss function to update the discount information generation model. Here, the full-link response refers to the quantitative tracking of the target user's complete behavioral path from product display to final repurchase.

[0089] The above end-to-end response can be set as display → click → add to cart → purchase → repeat purchase, with each step assigned a weight (purchase = 10 points, click = 2 points, ignore = -1 point). For online updates, the weight values ​​of each step are obtained within a short period of time (e.g., 1 hour, which can be set specifically). The parameters of the model are generated by incrementally updating the discount information based on the weight values ​​of each step using the FTRL (Follow The Regularized Leader) algorithm.

[0090] Offline updates involve acquiring the weight values ​​of each stage over a long period (e.g., 24 hours, configurable), retraining the model based on the full feedback data, and replacing the low-precision online model (triggered when the test set AUC (Area Under the Curve) improves by ≥2%). This online model update ensures rapid response to behavioral changes, while the offline iterative model guarantees long-term model accuracy.

[0091] This invention provides a method for recommending discount information, which is applied to the aforementioned discount information recommendation system. (See attached image.) Figure 3 , Figure 3 This is a flowchart illustrating an embodiment of a discount information recommendation method provided by the present invention. Figure 3 The process shown may include the following steps:

[0092] Step 301: Collect user behavior data of the target user and product feature data of each product in the product pool in real time, and generate user feature data based on user behavior data;

[0093] Step 302: For each product, determine the matching degree value based on user feature data and product feature data, generate a product recommendation list according to the matching degree value corresponding to each product, and select at least one target product from the product recommendation list; wherein, the target product is the product that needs to be delivered preferential information to the target user;

[0094] Step 303: For each of the at least one target products, generate the corresponding discount information based on user feature data;

[0095] Step 304: Calculate the channel activity ratio of target users on each communication channel, determine the target communication channels to reach based on the channel activity ratio, and use the target communication channels to deliver promotional information to target users.

[0096] The implementation process of the above method can be found in the description of the preferential information recommendation system, and will not be repeated here.

[0097] The discount information recommendation method provided in this invention effectively addresses the inherent flaws of current e-commerce platforms' extensive discount strategies. Its core benefits are: by generating real-time user characteristic data, it can accurately identify the preferences of users at different consumption levels, and filter high-potential target products corresponding to users based on multi-dimensional feature matching, thus avoiding mismatch of discount resources at the source; by generating differentiated discount strategies through user characteristic data (such as exclusive benefits for high-net-worth users or tiered discounts for potential users), it significantly improves discount redemption rates and average order value; and by employing an intelligent distribution mechanism based on the active percentage of communication channels, it ensures that discount information reaches users through the most efficient path, reducing ineffective exposure while enhancing user stickiness. Overall, user characteristic data achieves an operational upgrade from "one-size-fits-all" to "personalized" approaches. By accurately matching user needs with discount resources, it not only optimizes the ROI of promotional budgets but also creates a virtuous cycle in activating dormant users, cultivating brand loyalty, and tapping into incremental markets, ultimately breaking the traditional predicament of "high investment, low conversion."

[0098] See Figure 4 This is a block diagram illustrating an embodiment of a discount information recommendation device provided by the present invention. Figure 4 As shown, the device includes:

[0099] The first module 401 is used to collect user behavior data of target users and product feature data of each product in the product pool in real time, and generate user feature data based on user behavior data.

[0100] The second module 402 is used to determine the matching degree value for each product based on user feature data and product feature data, generate a product recommendation list according to the matching degree value corresponding to each product, and select at least one target product from the product recommendation list; wherein, the target product is the product that needs to be delivered preferential information to the target user;

[0101] The third module 403 is used to generate discount information corresponding to each of the at least one target products based on user feature data.

[0102] The fourth module, 404, is used to calculate the channel activity ratio of target users across various communication channels, determine the target communication channels to reach based on the channel activity ratio, and use the target communication channels to deliver promotional information to target users.

[0103] This solution effectively addresses the inherent flaws of current e-commerce platforms' extensive discount strategies. Its core benefits are: real-time generated user characteristic data allows for precise identification of preferences among users at different consumption levels; multi-dimensional feature matching enables the selection of high-potential target products for each user, fundamentally preventing misallocation of discount resources; differentiated discount strategies are generated based on user characteristic data (such as exclusive benefits for high-net-worth users or tiered discounts for potential users), significantly improving discount redemption rates and average order value; and an intelligent distribution mechanism based on the activity ratio of communication channels ensures that discount information reaches users through the most efficient path, reducing ineffective exposure while enhancing user stickiness. Overall, user characteristic data enables an operational upgrade from "one-size-fits-all" to "personalized" solutions. By accurately matching user needs with discount resources, it not only optimizes the ROI of promotional budgets but also creates a virtuous cycle in activating dormant users, cultivating brand loyalty, and tapping into incremental markets, ultimately breaking the traditional predicament of "high investment, low conversion."

[0104] Figure 5 This is a schematic diagram of the structure of an electronic device provided in an embodiment of the present invention. Figure 5 The illustrated electronic device 1200 includes at least one processor 1201, a memory 1202, at least one network interface 1204, and other user interfaces 1203. The various components in the electronic device 1200 are coupled together via a bus system 1205. It is understood that the bus system 1205 is used to implement communication between these components. In addition to a data bus, the bus system 1205 also includes a power bus, a control bus, and a status signal bus. However, for clarity, ... Figure 5 The general labeled all buses as Bus System 1205.

[0105] The user interface 1203 may include a display, keyboard, or clicking device (e.g., mouse, trackball, touchpad, or touchscreen).

[0106] It is understood that the memory 1202 in the embodiments of the present invention can be volatile memory or non-volatile memory, or may include both volatile and non-volatile memory. The non-volatile memory can be read-only memory (ROM), programmable read-only memory (PROM), erasable programmable read-only memory (EPROM), electrically erasable programmable read-only memory (EEPROM), or flash memory. The volatile memory can be random access memory (RAM), which is used as an external cache. By way of example, but not limitation, many forms of RAM are available, such as Static Random Access Memory (SRAM), Dynamic Random Access Memory (DRAM), Synchronous DRAM (SDRAM), Double Data Rate Synchronous DRAM (DDRSDRAM), Enhanced Synchronous DRAM (ESDRAM), Synchronous Link DRAM (SLDRAM), and Direct Rambus RAM (DRRAM). The memory 1202 described herein is intended to include, but is not limited to, these and any other suitable types of memory.

[0107] In some implementations, memory 1202 stores elements, executable units or data structures, or subsets thereof, or extended sets thereof: operating system 12021 and application program 12022.

[0108] The operating system 12021 includes various system programs, such as the framework layer, core library layer, and driver layer, used to implement various basic business functions and handle hardware-based tasks. The application program 12022 includes various applications, such as a media player and a browser, used to implement various application functions. The program implementing the method of this embodiment can be included in the application program 12022.

[0109] In this embodiment of the invention, the processor 1201 executes the method steps provided in each method embodiment by calling the program or instructions stored in the memory 1202, specifically the program or instructions stored in the application program 12022.

[0110] The methods disclosed in the above embodiments of the present invention can be applied to processor 1201, or implemented by processor 1201. Processor 1201 may be an integrated circuit chip with signal processing capabilities. In the implementation process, each step of the above method can be completed by the integrated logic circuit of the hardware in processor 1201 or by instructions in the form of software. The processor 1201 may be a general-purpose processor, a digital signal processor (DSP), an application-specific integrated circuit (ASIC), a field-programmable gate array (FPGA), or other programmable logic devices, discrete gate or transistor logic devices, or discrete hardware components. It can implement or execute the methods, steps, and logic block diagrams disclosed in the embodiments of the present invention. The general-purpose processor may be a microprocessor or any conventional processor. The steps of the methods disclosed in the embodiments of the present invention can be directly embodied in the execution of a hardware decoding processor, or executed by a combination of hardware and software units in the decoding processor. The software units may be located in random access memory, flash memory, read-only memory, programmable read-only memory, electrically erasable programmable memory, registers, or other mature storage media in the art. The storage medium is located in memory 1202. Processor 1201 reads the information in memory 1202 and completes the steps of the above method in conjunction with its hardware.

[0111] It is understood that the embodiments described herein can be implemented in hardware, software, firmware, middleware, microcode, or a combination thereof. For hardware implementation, the processing unit can be implemented in one or more application-specific integrated circuits (ASICs), digital signal processors (DSPs), digital signal processing devices (DSPDs), programmable logic devices (PLDs), field-programmable gate arrays (FPGAs), general-purpose processors, controllers, microcontrollers, microprocessors, other electronic units for performing the functions described herein, or combinations thereof.

[0112] For software implementation, the techniques described herein can be implemented by units that perform the functions described herein. The software code can be stored in memory and executed by a processor. The memory can be implemented in the processor or external to the processor.

[0113] The electronic device provided in this embodiment may be as follows: Figure 5 The electronic device shown can perform the following: Figure 3 All steps of the method for recommending preferential information in China, thereby achieving Figure 3 For details on the technical effectiveness of the promotional information recommendation method shown, please refer to [link / reference]. Figure 3 The relevant descriptions are presented concisely and will not be elaborated upon here.

[0114] This invention also provides a storage medium (computer-readable storage medium). This storage medium stores one or more programs. The storage medium may include volatile memory, such as random access memory; the memory may also include non-volatile memory, such as read-only memory, flash memory, hard disk, or solid-state drive; the memory may also include combinations of the above types of memory.

[0115] When one or more programs in the storage medium can be executed by one or more processors to implement the above-mentioned preferential information recommendation method.

[0116] The processor is used to execute a program that generates and distributes discount information stored in the memory, in order to implement the steps of the discount information recommendation method.

[0117] Those skilled in the art will further recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, computer software, or a combination of both. To clearly illustrate the interchangeability of hardware and software, the components and steps of the various examples have been generally described in terms of functionality in the foregoing description. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementations should not be considered beyond the scope of this invention.

[0118] The steps of the methods or algorithms described in conjunction with the embodiments disclosed herein can be implemented in hardware, a software module executed by a processor, or a combination of both. The software module can be located in random access memory (RAM), main memory, read-only memory (ROM), electrically programmable ROM, electrically erasable programmable ROM, registers, hard disk, removable disk, CD-ROM, or any other form of storage medium known in the art.

[0119] The specific embodiments described above further illustrate the purpose, technical solution, and beneficial effects of the present invention. It should be understood that the above description is only a specific embodiment of the present invention and is not intended to limit the scope of protection of the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A discount information recommendation system, characterized in that, The discount information recommendation system includes a data feature module, a product sorting module, a discount information calculation module, and a discount information delivery module connected in sequence. The data feature module is also connected to the discount information calculation module. The data feature module is used to collect user behavior data of target users and product feature data of each product in the product pool in real time, generate user feature data based on the user behavior data, and send the user feature data and the product feature data to the product sorting module, and send the user feature data to the discount information calculation module. The product sorting module is used to determine a matching degree value for each product based on the user feature data and the product feature data, generate a product recommendation list according to the matching degree value corresponding to each product, and select at least one target product from the product recommendation list and send it to the discount information calculation module; wherein, the target product is the product for which discount information needs to be delivered to the target user; The discount information calculation module is used to generate discount information corresponding to each of the at least one target product based on the user feature data, and send the discount information to the discount information delivery module. The discount information delivery module is used to calculate the channel activity ratio of target users on each communication channel, determine the target communication channel to be reached based on the channel activity ratio, and send the discount information to the target user through the target communication channel.

2. The discount information recommendation system according to claim 1, characterized in that, The data feature module includes a data acquisition unit, a feature generation unit, and a data transmission unit connected in sequence. The data acquisition unit is used to collect user behavior data of the target user and product feature data of each product in the product pool in real time through a distributed data acquisition architecture, and send the user behavior data to the feature generation unit and the product feature data to the data sending unit. The feature generation unit is used to perform data cleaning on the user behavior data, perform feature clustering analysis on the cleaned user behavior data, generate user feature data of the target user, and send the user feature data to the data sending unit. The data sending unit is used to send the user feature data and the product feature data to the product sorting module, and to send the user feature data to the discount information calculation module.

3. The discount information recommendation system according to claim 1, characterized in that, The product sorting module includes a matching degree generation unit, a product list generation unit, and a first sending unit connected in sequence. The matching degree generation unit is used to input the user feature data and the product feature data into a pre-trained matching degree value generation model for each product. The matching degree value generation model outputs a basic matching degree value, extracts the remaining validity period of the product from the product feature data, determines the timeliness coefficient based on the remaining validity period of the product, generates a matching degree value according to the basic matching degree value and the timeliness coefficient, and sends the matching degree value of each product to the product list generation unit. The product list generation unit is used to sort the products according to the matching degree value of each product to generate a product recommendation list, send the product recommendation list to the display interface to display to the target user, and send the product recommendation list to the first sending unit. The first sending unit is used to select at least one target product from the product recommendation list and send at least one target product to the discount information calculation module.

4. The discount information recommendation system according to claim 3, characterized in that, The determination of the timeliness coefficient based on the remaining shelf life of the product includes: Obtain product circulation information; When the commodity circulation information is set as commodity accelerated circulation information, the timeliness coefficient is determined based on the remaining shelf life of the commodity and the first timeliness coefficient calculation formula. When the commodity circulation information is set to non-accelerated commodity circulation information, the timeliness coefficient is determined based on the remaining shelf life of the commodity and the second timeliness coefficient calculation formula.

5. The discount information recommendation system according to claim 1, characterized in that, The discount information calculation module includes a calculation unit and a second sending unit connected in sequence. The computing unit is used to determine the user type of the target user based on the user feature data using a clustering algorithm, extract business data from the user feature data, input the user type and business data into a pre-trained discount information generation model, the discount information generation model outputs discount information corresponding to the target product, and sends the discount information to the second sending unit; wherein, the business data includes at least user price sensitivity data, consumption potential data, and churn risk data; The second sending unit is used to send the discount information to the discount information delivery module.

6. The discount information recommendation system according to claim 1, characterized in that, The promotional information delivery module includes a statistical calculation unit and a channel adjustment delivery unit connected in sequence; The statistical calculation unit is used to, within a specified time range, count the interaction frequency of the target user with each communication channel, calculate the channel activity ratio of each communication channel based on the interaction frequency, and send the channel activity ratio of each communication channel to the channel adjustment determination unit. The channel adjustment and outreach unit is used to use the channel activity ratio of each of the communication channels as the context feature of the dynamic resource allocation algorithm. The dynamic resource allocation algorithm dynamically adjusts the channel weight of each of the communication channels, determines the communication channels whose channel weight exceeds a preset weight threshold as the target communication channels, and uses a message queue to asynchronously push the discount information to the message topic of the target communication channel, so as to send the discount information to the target user.

7. The discount information recommendation system according to claim 5, characterized in that, The discount information recommendation system also includes a tracking module connected to the discount information calculation module; The tracking module is used to track the target user's full-link response, obtain the weight value of each link in the full-link response, and send the weight value of each link to the calculation unit so as to use each weight value as the sample weight of the loss function to update the discount information generation model; wherein, the full-link response refers to the complete behavioral path of the target user from product display to final repurchase by quantitatively tracking the entire behavioral path.

8. A method for recommending preferential information, characterized in that, The method is applied to the discount information recommendation system according to any one of claims 1 to 7, and the method includes: Real-time collection of user behavior data of target users and product feature data of each product in the product pool; and generation of user feature data based on the user behavior data. For each of the aforementioned products, a matching degree value is determined based on the user feature data and the product feature data. A product recommendation list is generated according to the matching degree value corresponding to each of the aforementioned products, and at least one target product is selected from the product recommendation list; wherein, the target product is the product for which preferential information needs to be delivered to the target user. For each of the at least one of the target products, generate corresponding discount information for the target product based on the user feature data; The target user's channel activity percentage across various communication channels is statistically analyzed. Based on the channel activity percentage, the target communication channel for reaching the target user is determined, and the promotional information is delivered to the target user through the target communication channel.

9. An electronic device, characterized in that, include: A processor and a memory, the processor being configured to execute a program for generating and allocating discount information stored in the memory, to implement the discount information recommendation method of claim 8.

10. A storage medium, characterized in that, The storage medium stores one or more programs, which can be executed by one or more processors to implement the discount information recommendation method as described in claim 8.

Images