A method and system for issuing coupons to users of a second-hand e-commerce platform

Abstract

The application relates to a method and system for issuing coupons to users of a second-hand e-commerce platform, wherein the method comprises: obtaining a coupon issuing index of a target user and judging whether a coupon issuing condition is met; in response to the target user meeting the coupon issuing condition, determining a mapping relationship between a corresponding coupon type and an issuing index according to the coupon issuing index of the target user; and determining a coupon type according to the mapping relationship between the coupon type and the issuing index and the coupon issuing index of the target user, and issuing the coupon to the target user. The application realizes targeted and fine coupon issuing, can improve the order placing enthusiasm of the user, can bring benefits to the platform, and fully plays the role of the coupon.

Description

Technical Field

[0001] This invention relates to the field of internet application technology, and in particular to a method and system for issuing coupons to users of a second-hand e-commerce platform. Background Technology

[0002] Coupons are an effective promotional tool, and depending on the transaction method, they can be divided into offline coupons and online coupons. Offline coupons are usually paper-based vouchers, while online coupons are typically obtained by users actively applying for them during online transactions or are proactively pushed to users by merchants. The coupon amount is deducted from the total order amount at checkout. Regarding the user-application-based coupon distribution model, there are situations where users miss out on coupons due to a lack of understanding of the distribution model or application requirements. In such cases, most users are dissatisfied, usually resulting in them contacting customer service for a replacement coupon or experiencing decreased user satisfaction and reduced spending on the platform. Either outcome negatively impacts the trading platform. Therefore, some platforms adopt a proactive coupon distribution model to avoid these situations. Regarding proactive coupon distribution, most trading platforms typically launch coupon pushes during events with specific timeframes, such as 618 promotions and Double 11 promotions, to increase sales. These pushes may involve issuing coupons of a fixed value at once, or issuing coupons of several denominations matching the total amount of a user's current order. However, neither of these two models is well-suited for secondhand e-commerce platforms. For example, a large proportion of sellers on these platforms are individuals, and their goods are typically unique, with transactions usually involving only one item at a time. Issuing coupons of a fixed value all at once would be ineffective if the value is too small to encourage transactions, while a large value would negatively impact platform profits. Conversely, matching coupons of different values ​​to different order totals limits the scope of application due to the single-item nature of the transactions. Therefore, a suitable coupon distribution model has yet to be found for the secondhand e-commerce platform scenario. Summary of the Invention

[0003] In response to the technical problems existing in the prior art, this invention proposes a method and system for issuing coupons to users of second-hand e-commerce platforms. This method and system can adapt to the transaction model of second-hand e-commerce platforms, issue coupons in a targeted manner, and enhance the effectiveness of coupons.

[0004] To address the aforementioned technical problems, according to one aspect of the present invention, a method for distributing coupons to users of a second-hand e-commerce platform is provided, comprising the following steps: obtaining coupon distribution indicators for a target user and determining whether the conditions for coupon distribution are met; in response to the target user meeting the conditions for coupon distribution, determining a mapping relationship between the corresponding coupon type and the distribution indicator based on the coupon distribution indicators of the target user; and determining the coupon type based on the mapping relationship between the coupon type and the distribution indicator and the coupon distribution indicators of the target user, and distributing the coupon to the target user.

[0005] According to another aspect of the present invention, the present invention provides a system for distributing coupons to users of a second-hand e-commerce platform, comprising a distribution indicator acquisition module, a distribution decision module, a matching module, and a distribution module, wherein the distribution indicator acquisition module is configured to acquire coupon distribution indicators for target users; the distribution decision module is connected to the distribution indicator acquisition module and is configured to determine whether coupon distribution conditions are met based on the coupon distribution indicators of the target users; the matching module is connected to the distribution indicator acquisition module and the distribution decision module and is configured to determine, when the coupon distribution conditions are met, a coupon type matching the coupon distribution indicators of the target users based on the mapping relationship between the coupon distribution indicators of the target users and the corresponding coupon types; the distribution module is connected to the matching module and is configured to distribute the determined coupon types to the target users according to a distribution strategy.

[0006] This invention provides a method and system for issuing coupons to second-hand e-commerce platforms. By using the mapping relationship between different coupon types and issuance indicators to determine the corresponding coupon types, targeted and precise coupon issuance is achieved. This not only increases users' enthusiasm for placing orders but also brings revenue to the platform, fully enhancing the role of coupons. At the same time, it can control the average order value while ensuring the order rate. Attached Figure Description

[0007] The preferred embodiments of the present invention will now be described in further detail with reference to the accompanying drawings, wherein:

[0008] Figure 1 This is a flowchart of a method for issuing coupons to users of a second-hand e-commerce platform according to Embodiment 1 of the present invention;

[0009] Figure 2 This is a flowchart of a method for obtaining the mapping relationship between coupon types and order rate intervals according to an embodiment of the present invention;

[0010] Figure 3 This is a flowchart illustrating a method for issuing coupons to users of a second-hand e-commerce platform according to Embodiment 2 of the present invention.

[0011] Figure 4 This is a flowchart of a method for issuing coupons to users of a second-hand e-commerce platform according to Embodiment 3 of the present invention;

[0012] Figure 5 This is a flowchart of user segmentation and the establishment of a mapping table between coupon types and user segments according to an embodiment of the present invention;

[0013] Figure 6 This is a flowchart of a method for issuing coupons to users of a second-hand e-commerce platform according to Embodiment 4 of the present invention;

[0014] Figure 7 This is a flowchart of training an order rate prediction model according to an embodiment of the present invention;

[0015] Figure 8 This is a system principle block diagram for issuing coupons to users of a second-hand e-commerce platform according to an embodiment of the present invention;

[0016] Figure 9 This is a partial block diagram illustrating the principle of a system for issuing coupons to users of a second-hand e-commerce platform according to an embodiment of the present invention.

[0017] Figure 10 This is a partial block diagram of a system for issuing coupons to users of a second-hand e-commerce platform according to another embodiment of the present invention; and

[0018] Figure 11 This is a schematic block diagram of a model module in a system according to another embodiment of the present invention. Detailed Implementation

[0019] 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.

[0020] In the following detailed description, reference can be made to the accompanying drawings, which form part of this application and illustrate specific embodiments of the present application. In the drawings, similar reference numerals describe substantially similar components in different figures. Specific embodiments of the present application are described in sufficient detail below to enable those skilled in the art to implement the technical solutions of the present application. It should be understood that other embodiments may also be utilized, or structural, logical, or electrical changes may be made to the embodiments of the present application.

[0021] As a secondhand e-commerce platform providing secondhand goods transactions, its users are divided into sellers and buyers. Sellers, when posting product information on the platform, need to place an order with the platform and become sellers according to an agreement, thus listing their products on the platform (such as a website or app). Buyers transact with sellers through the platform, placing orders to purchase goods. To facilitate successful transactions, the platform can offer coupons to sellers to encourage them to sell goods, and it can also offer coupons to buyers to encourage them to purchase goods. In the following description, unless otherwise specified, "user" refers to both sellers and buyers. Furthermore, "placing an order" can also refer to a user's action that generates revenue for the platform, such as purchasing or recycling goods.

[0022] Example 1

[0023] Figure 1 This is a flowchart of a method for issuing coupons to users of a second-hand e-commerce platform according to Embodiment 1 of the present invention. In this embodiment, the method includes the following steps:

[0024] Step S1a: Detect user behavior on the platform. This includes detecting user activities such as browsing products, adding products / sellers to favorites, contacting sellers / the platform, and placing orders.

[0025] Step S2a: Determine whether the first condition for issuing a coupon to the current target user is met. If it is met, proceed to step S3a; otherwise, return to step S1a. In one embodiment, the first condition for issuing a coupon is that the current user's activity involves an activity that issues coupons, or in other embodiments, the page the user is currently visiting is a page that needs to issue coupons to the user.

[0026] Step S3a involves obtaining the predicted order rate values ​​for both issuing and not issuing coupons to the target user. In this embodiment, the user's order rate is used as the coupon issuance indicator. When platform resources are sufficient, an online order rate prediction model can be invoked to predict the order rate, thereby obtaining the predicted order rate value as the coupon issuance indicator. To predict the order rate, the historical data and current behavior data of the target user are first obtained. For example, the current user's identity is first determined. If the user is logged into the platform, the user's identity identifier can be read. If the current user is not logged into the platform, the user's device identifier can be obtained through user data, and then the user's identity can be identified based on the device identifier. After identifying the user's identity, the user characteristics of the target user are obtained. These user characteristics include historically saved user characteristics and user characteristics obtained after processing real-time data, such as gender, age, number of logins within a week, and number of transactions within a week. The user characteristics are then input into the order rate prediction model to obtain the predicted order rate value. In this model, "coupon" is used as a user feature input to the order rate prediction model. When its feature value is set to 1, the model outputs the order rate prediction value with coupons issued, which is called the first order rate prediction value. When its feature value is set to 0, the model outputs the order rate prediction value without coupons issued, which is called the second order rate prediction value.

[0027] Step S4a: Compare the predicted first order rate with the first threshold, and compare the predicted second order rate with the second threshold; wherein the first threshold is less than the second threshold. The first and second thresholds represent an order rate range where issuing coupons can increase the order rate, such as [0.2, 0.8]. If the order rate after issuing a coupon to a user is still less than the minimum value of this range, i.e., the first threshold, it means that issuing coupons to that user is not effective in increasing the order rate, and therefore, it is not necessary to issue coupons to that user. On the other hand, if it is not necessary to issue coupons to a user because their order rate is already high enough, such as exceeding the second threshold, then it is also not necessary to issue coupons to that user.

[0028] Step S5a: Determine whether the second condition for issuing coupons is met: the first predicted order rate is greater than or equal to the first threshold, and the second predicted order rate is less than the second threshold. In other words, determine whether the user is a user for whom issuing coupons can effectively increase the order rate. If the first predicted order rate is greater than or equal to the first threshold, and the second predicted order rate is less than the second threshold, then proceed to step S6a. Otherwise, in step S10a, determine that coupons will not be issued to the target user.

[0029] Step S6a: Calculate the difference between the first predicted order rate and the second predicted order rate. This difference determines whether the order rate after issuing a coupon is higher than the order rate without a coupon, and whether this difference is sufficiently high. In other words, does issuing a coupon effectively increase the order rate? If it does, and the platform's revenue after a user places an order exceeds the subsidy paid by the platform when issuing the coupon (i.e., the amount represented by the coupon's face value), then the coupon is issued. If the increase in the order rate is minimal, or even insufficient to cover the subsidy paid by the platform when issuing the coupon, then the coupon should not be issued. Based on the product category and experience, a suitable difference threshold can be determined. When the difference between the first and second predicted order rate is greater than or equal to this threshold, issuing a coupon to the user can bring positive revenue to the platform.

[0030] Step S7a: Determine whether the difference between the first order rate prediction value and the second order rate prediction value is greater than or equal to the difference threshold. If yes, proceed to step S8a. If no, the target user does not meet the issuance conditions, and in step S10a, determine that the coupon will not be issued to him / her.

[0031] Step S8a: Match coupon types to the target user using the mapping relationship between coupon types and order rates. In this embodiment, a first predicted order rate value is used to match the coupon type with the mapping relationship. The coupon type refers to different face values, with face value as the type. In one embodiment, coupon types and standard order rate scores correspond one-to-one, as shown in Table 1. When the first predicted order rate value for the target user is obtained, the mapping table is queried, and the standard order rate scores in the table are compared with the predicted order rate value to obtain the order rate score closest to the predicted order rate value, thus obtaining the corresponding coupon type.

[0032] Table 1: Mapping Table of Coupon Types and Standard Score for Order Rate

[0033] Type 1 Five yuan 0.3 Type 2 Eight Yuan 0.4 Category 3 Ten yuan 0.5 …… …… ……

[0034] In another embodiment, coupon types correspond one-to-one with order rate intervals, as shown in the mapping table in Table 2. When the first order rate prediction value for the target user is obtained, the mapping table is queried to obtain the order rate interval where the first order rate prediction value falls, and thus the corresponding coupon type is obtained.

[0035] Table 2: Mapping Table of Coupon Types and Order Rate Ranges

[0036] Type 1 Three Yuan (0.001,0.09] Type 2 Five yuan (0.09,0.14] Category 3 Eight Yuan (0.14,0.25] Type 4 Ten yuan (0.25,0.31] Type 5 Twelve yuan (0.31,0.52] …… …… ……

[0037] Step S9a: Distribute the coupons of the corresponding denominations to the target users.

[0038] In the above embodiments, the mapping tables in Tables 1 and 2 are calculated by the system based on user historical data. Taking Table 2 as an example, ... Figure 2 As shown, the process of obtaining the mapping table shown in Table 2 is as follows:

[0039] Step S1b involves obtaining the order rate for multiple users within a preset historical time period. For example, for seller users, the order rate is obtained by dividing the number of price estimates and orders placed by the number of orders placed by the number of price estimates. For buyer users, the order rate is obtained by dividing the number of inquiries, views, and orders placed by the buyer users regarding products by the number of inquiries and views placed by the number of orders placed by the sum of the number of inquiries and views placed.

[0040] Step S2b: Obtain the total order rate range based on the minimum and maximum order rates of the multiple users;

[0041] Step S3b: Divide the total order rate range into corresponding order rate ranges based on the number of coupon types available that day. Coupon distribution typically requires ensuring that the total subsidy amount (i.e., the amount corresponding to the coupon denomination) within a marketing cycle does not exceed a certain percentage of total sales revenue, such as 5%. Some distribution strategies set a daily upper limit on the subsidy amount corresponding to the coupons issued, and divide these subsidy amounts into multiple denominations, i.e., the number of coupon types, and then calculate the number of copies for each denomination (type).

[0042] Step S4b establishes a one-to-one mapping relationship between each type of coupon for the day and a specific order rate range, as shown in Table 2. Here, the coupon type refers to the coupon denomination, with smaller denominations corresponding to lower order rate ranges.

[0043] Example 2

[0044] Figure 3 This is a flowchart illustrating a method for issuing coupons to users of a second-hand e-commerce platform according to Embodiment 2 of the present invention. In this embodiment, the difference between a first predicted order rate when coupons are issued and a second predicted order rate when coupons are not issued is used as the coupon issuance indicator. Specifically, the method includes the following steps:

[0045] Step S1c: Detect the user's behavior on the platform to determine if the current target user meets the first condition for issuing coupons. For example, the current target user is engaged in an activity that involves issuing coupons, or the page the target user is currently visiting is a page that requires issuing coupons to users.

[0046] Step S2c: Obtain the order rate prediction values ​​when coupons are issued to target users and when coupons are not issued, i.e., the first order rate prediction value when coupons are issued to target users and the second order rate prediction value when coupons are not issued to target users.

[0047] Step S3c: Compare the predicted value of the first order rate with the first threshold, and compare the predicted value of the second order rate with the second threshold.

[0048] Step S4c: Determine whether the second condition for issuing coupons is met: the first predicted order rate is greater than or equal to the first threshold, and the second predicted order rate is less than the second threshold. In other words, determine whether the user is someone for whom issuing coupons can effectively increase the order rate. If the first predicted order rate is greater than or equal to the first threshold, and the second predicted order rate is less than the second threshold, then proceed to step S5c. Otherwise, in step S9c, determine that coupons will not be issued to the target user.

[0049] Step S5c: Calculate the difference between the first order rate prediction value and the second order rate prediction value.

[0050] Step S6c: Determine whether the difference between the first predicted order rate and the second predicted order rate is greater than or equal to the difference threshold, such as whether it is greater than or equal to 0.1. If yes, proceed to step S7c; otherwise, the target user does not meet the issuance conditions, and in step S9c, determine that the coupon will not be issued to them.

[0051] Step S7c involves matching the coupon order rate of the target user using the mapping relationship between coupon type and the order rate difference. In this embodiment, the difference between the first order rate prediction value and the second order rate prediction value is used to match the appropriate coupon type. The mapping relationship between coupon type and order rate difference is shown in Table 3. This embodiment uses the difference between the first order rate prediction value and the second order rate prediction value as the distribution indicator; the larger the difference, the higher the coupon value. This can better promote the user's order rate.

[0052] Table 3: Mapping Table of Coupon Types and Order Rate Difference Range

[0053] Type 1 One hundred yuan (0.1,0.3] Type 2 150 yuan (0.3,0.5] Type 3 Two hundred yuan (0.5,0.7] Type 4 After 240 yuan (0.7,0.8] Type 5 280 yuan (0.8,0.9] …… …… ……

[0054] Step S8c: Send the matched coupons to the target user.

[0055] Example 3

[0056] Figure 4A flowchart illustrating a method for issuing coupons to users of a second-hand e-commerce platform according to Embodiment 3 of the present invention. In this embodiment, the predicted order rate when issuing coupons to target users is used as the coupon issuance indicator. The mapping relationship between coupon types and issuance indicators is that the coupon with the smallest face value corresponds to the predicted order rate that meets the issuance conditions. See the following steps for details:

[0057] Step S1d involves obtaining the predicted order rate values ​​when coupons are distributed to target users in ascending order of coupon value. For example, there are currently 5 coupon values: 10 yuan, 15 yuan, 20 yuan, 25 yuan, and 30 yuan. Using the coupon value as a feature, along with other user features of the target users, the order rate prediction model is input to obtain the corresponding 5 predicted values ​​λ1-λ5.

[0058] Step S2d: Compare each predicted order rate value λ1-λ5 with the order rate threshold λ. v The size of the order rate threshold λ. v To meet the order rate requirements for issuing coupons.

[0059] Step S3d: Obtain the order rate threshold λ. v The predicted order rate values, such as λ4 and λ5, are used as the predicted qualified order rate values.

[0060] Step S4d: Obtain the coupon amount used when obtaining the predicted value of qualified order rate, such as 25 yuan and 30 yuan.

[0061] Step S5d: Send the coupon with the smallest face value among the multiple coupons to the target user, that is, send the coupon of 25 yuan to the target user.

[0062] In this embodiment, the system stores coupons of various denominations. When different denomination coupons are used as input features of the prediction model, different order rates can be predicted. When the predicted order rate is greater than the order rate threshold, it indicates that the platform's revenue can be satisfied when the corresponding denomination coupons are issued to the target users. This embodiment further issues the minimum denomination coupons that meet the requirements to the users, which not only increases user enthusiasm but also brings good revenue to the platform.

[0063] Example 4

[0064] In this embodiment, when the resources of the second-hand e-commerce platform are scarce, real-time calculation using the order rate prediction model would consume excessive resources and cause system crashes. To address this situation, this embodiment provides a solution: by using the order rate prediction model offline to predict the order rate of platform users, the platform users are grouped, and a first user group meeting the conditions for coupon issuance is established. Within the first user group, multiple second user groups are further divided according to the order rate, order rate range, or order rate difference range. A mapping table between coupon types and second user groups is established. Thus, when the target users meeting the first condition for coupon issuance are obtained in real time, real-time calculation is not required. It is only necessary to query based on the target user's group to determine whether to issue coupons and the type of coupon to be issued.

[0065] Figure 5 This is a flowchart for segmenting users and creating a mapping table between coupon types and user segments, as shown below:

[0066] Step S1e: Obtain the training dataset. This involves acquiring user data from multiple users on the platform within a preset historical time period. For example, acquiring user data from the most recent week and user profile data stored on the platform. The user profile data includes, for example, age, gender, and device identifier. User data also includes user behavior data on the platform. For seller users, this includes login time and frequency, product sales form completion time and frequency, platform-provided product estimates for sellers' sales forms, and the number of products sold. For buyer users, this includes login time and frequency, number of inquiries, number of purchases, and purchased products.

[0067] Step S2e involves performing user feature mining based on the dataset to obtain multiple user features. For example, firstly, a large number of candidate features are generated from the available user data. Then, a feature mining model is used to calculate the effect of each feature on the model. If the current feature improves the model's performance, it is used; otherwise, it is discarded. Mined user features include whether there was a price estimate before the current estimate, whether there was a price estimate in the previous three days, whether there has been a historical sale of goods, user profile features, and features of the goods sold by the user, such as mobile phone model and condition.

[0068] Step S3e involves iterating through all user features to obtain all possible feature combinations, so that any user can be represented by a feature combination. For example, all user features are arranged and combined to obtain multiple combinations, and contradictory features are eliminated. For example, for a seller user, the three features are "the number of valuations in the last three days is 1", "the number of product sales in the last three days is 2", and "the number of product sales in the last three days is 0". Obviously, the first two features are contradictory, while the first and third features are not contradictory. Therefore, the first and third features can be placed in a feature combination.

[0069] Step S4e involves using each feature combination as input to the trained order rate prediction model to obtain the first and second order rate prediction values ​​for both the presence and absence of coupons. Specifically, by setting the feature values ​​of the coupon feature in the feature combination to 1 and 0, the corresponding order rate prediction values ​​for the presence and absence of coupons can be obtained.

[0070] Step S5e: Take the first order rate prediction value and the second order rate prediction value corresponding to one of the feature combinations as calculation parameters.

[0071] Step S6e: Determine whether the predicted first order rate is greater than or equal to the first threshold. In this embodiment, an order rate range that meets the conditions for issuing coupons is determined by the first threshold and the second threshold. When the predicted first order rate is less than the first threshold, it means that even if coupons are issued, the conditions for issuing them will not be met. Therefore, in step S13e, the users corresponding to the feature combination are determined as the feature combination corresponding to the second type of users, that is, the user group that does not need to issue coupons. When the predicted first order rate is greater than or equal to the first threshold, step S7e is executed.

[0072] Step S7e: Determine whether the predicted second order rate is greater than or equal to the second threshold. If the predicted second order rate is greater than or equal to the second threshold, it indicates that coupons are not needed and the order rate is already high enough. Therefore, in step S13e, the users corresponding to the feature combination are identified as the feature combination corresponding to the second type of users, which is the user group that does not need coupons. If the predicted second order rate is less than the second threshold, it indicates that the users corresponding to the feature combination meet the conditions for issuing coupons. The feature combination is identified as the first type of users, and step 8e is executed.

[0073] Step S8e: Determine if there are any unpredicted feature combinations. If so, return to step S5e; otherwise, proceed to step S9e. The following processing applies to the first type of user.

[0074] Step S9e: Sort the order rate sequence from low to high according to the first order rate prediction value.

[0075] Step S10e: Divide the order rate sequence into a corresponding number of order rate subsequences according to the number of coupon types for the day.

[0076] Step S11e: Establish a one-to-one correspondence between coupon types and order rate subsequences according to the various coupons with face values ​​from smallest to largest and the various order rate subsequences with numerical values ​​from smallest to largest.

[0077] Step S12e: Group the feature combinations classified into the first type of users according to the order rate subsequence. Each group of feature combinations corresponds to a user subgroup, thus obtaining a one-to-one correspondence between coupon types and user subgroups, as shown in Table 4.

[0078] Table 4: Mapping Table of Coupon Types and User Segments

[0079] [0.001,0.09] Type 1 Five yuan 1 (0.14,0.25] Type 2 Eight Yuan 2 (0.25,0.31] Type 3 Ten yuan 3 …… …… ……

[0080] Figure 6 This is a flowchart of a method for issuing coupons to users of a second-hand e-commerce platform according to Embodiment 4 of the present invention. The method includes the following steps:

[0081] Step S1f involves obtaining the user segment to which the target user belongs. In one embodiment, the platform system stores the characteristics of all users who have historically logged into the platform (such as a website or app). Once the target user is identified, their user characteristics can be obtained based on their identity identifier, and these characteristics are updated according to current real-time data. In another embodiment, user data can be processed in real-time to obtain corresponding user characteristics. By matching the feature combinations of all the target user's user characteristics with the feature combinations of user segments, the user segment corresponding to the closest feature combination is determined as the target user's user segment.

[0082] Step S2f: Determine whether the user group belongs to the second type of user, i.e., users who do not need coupons. If so, in step S5f, determine that coupons will not be issued to the target users. If they do not belong to the second type of user, proceed to step S3f.

[0083] Step S3f: Query the mapping table between coupon types and user groups to obtain the matching coupon types.

[0084] Step S4f: The matched coupons are issued to the target users.

[0085] In this embodiment, even without real-time prediction of the target user's order rate, it is still possible to determine whether to issue coupons and the types of coupons to be issued based on user characteristics. This method involves minimal computation and consumes few system resources.

[0086] In the foregoing embodiments, the order rate prediction model used is a deep neural network model, such as the Wide&Deep model, the DeepFM model, the DCN (Deep&Cross Network) model, or the DIN (Deep Interest Network) model. In one embodiment, the order rate prediction model uses the DeepFM neural network model, whose input consists of various user features. The DeepFM neural network model includes an input layer, a Deep layer, an FM layer, and a fully connected layer. User features are embedded in the input layer and then simultaneously input into the Deep layer and the FM layer. Their respective outputs are combined and input into the fully connected layer. After passing through the fully connected layer, the output is used to obtain the order rate score.

[0087] In this invention, assuming a secondhand e-commerce platform continuously distributes coupons to users, the order rate prediction model is trained using user data from a pre-defined time period during the coupon distribution period. Furthermore, the model is periodically updated based on user data after coupon distribution, thereby improving prediction accuracy. The specific process for training the order rate prediction model is as follows: Figure 7 As shown, it includes the following steps:

[0088] Step S1g: Obtain user data from a predetermined time period during the coupon distribution period as the training data source. For example, user data from the most recent two weeks.

[0089] Step S2g: The training data source is processed into a user feature set, with each user's user feature set serving as a sample.

[0090] Step S3g involves labeling each sample using whether a user has placed an order as the sample label. A portion of the labeled samples is used as the training sample set, and the other portion as the test sample set.

[0091] Step S4g: Train a deep neural network model using samples from the training sample set.

[0092] Step S5g: Test the trained order rate prediction model using samples from the test sample set.

[0093] Step S6g: Evaluate the test results according to preset evaluation metrics to determine if they meet the evaluation requirements. If they do, training is complete; otherwise, return to step S4g to continue training until the requirements are met. Step S7g: The model is ready for use after training is complete. To improve the model's prediction accuracy, it needs frequent updates. In this embodiment, the model is updated periodically, such as once a day. Therefore, timing is initiated each time the model is completed and put into use.

[0094] Step S8g: Determine if the timing period (e.g., 24 hours) has been reached. If it has, and if step S9g determines that the model still needs updating, return to step S1g to retrieve user data from a more recent period (e.g., the last 3 days) as the training data source and retrain the query. If the timing period has not been reached, wait and monitor whether it has been reached. When step S9g determines that no further updates are needed, end the model training and update process.

[0095] After updating the order rate prediction model, relevant calculations are performed on Example 4 according to the updated order rate prediction model, thereby updating the user groups and corresponding mapping relationships.

[0096] On the other hand, the present invention also provides a system for distributing coupons to users of second-hand e-commerce platforms. For example... Figure 8 The diagram shown is a system principle block diagram of one embodiment of the system. The system includes a distribution indicator acquisition module 1, a distribution decision module 2, a matching module 3, and a distribution module 4. The distribution indicator acquisition module 1 is used to acquire coupon distribution indicators for target users. The distribution decision module 2 is connected to the distribution indicator acquisition module 1 and is used to determine whether the coupon distribution conditions are met based on the target user's user data. The matching module 3 is connected to the distribution indicator acquisition module 1 and the distribution decision module 2, and is used to determine the coupon type that matches the target user's coupon distribution indicator based on the target user's coupon distribution indicator and the mapping relationship between the corresponding coupon type and the distribution indicator. The distribution module 4 is connected to the matching module 3 and is used to distribute the determined coupon type to the target user according to the distribution strategy.

[0097] like Figure 9 As shown, in this invention, depending on the coupon distribution metric, the distribution metric acquisition module 1 includes an order rate prediction unit 11, a calculation unit 12, and a user segmentation determination unit 13. When platform resources are sufficient, the order rate prediction unit 11 calls an online order rate prediction model to predict the first and second order rate prediction values ​​for the target user when coupons are distributed and / or not distributed. The calculation unit 12 is connected to the order rate prediction unit 11 and is configured to calculate the order rate difference between the first order rate prediction value when coupons are distributed and the second order rate prediction value when coupons are not distributed.

[0098] In one embodiment, the predicted first order rate when coupons are issued to the target user is used as the coupon issuance indicator. In this case, the mapping relationship between the corresponding coupon types and the issuance indicator is a one-to-one mapping relationship between multiple coupons with face values ​​ranging from small to large and multiple standard order rate / order rate ranges ranging from small to large.

[0099] In another embodiment, the difference between the order rate prediction of the first order rate and the order rate prediction of the second order rate can be used as the coupon distribution indicator. Correspondingly, the mapping relationship between coupon types and distribution indicators is a one-to-one mapping relationship between various coupons with face values ​​ranging from small to large and order rate difference intervals ranging from small to large.

[0100] The user segmentation determination unit 13 is used to query the user segment to which the target user belongs based on the target user's identity identifier or user characteristics; correspondingly, the mapping relationship between coupon types and distribution indicators is a one-to-one mapping relationship between coupon types and user segments.

[0101] The distribution decision module 2 includes a decision unit 21, a first comparison unit 22, and a second comparison unit 23. The decision unit 21 is connected to the calculation unit 12, the first comparison unit 22, and the second comparison unit 23, respectively. It determines whether the difference between the first order rate prediction value and the second order rate prediction value is positive, that is, whether the first order rate prediction value is greater than the second order rate prediction value. If the first order rate prediction value is greater than the second order rate prediction value, it further determines whether the difference between the two is greater than or equal to a threshold, that is, whether the first order rate prediction value is significantly greater than the second order rate prediction value. When the first order rate prediction value is significantly greater than the second order rate prediction value, it determines that the target user meets the coupon distribution conditions, and then sends a matching instruction to the matching module 3, so that the matching module 3 queries the mapping relationship table of multiple standard order rates / order rate intervals for coupon types or the mapping relationship table of coupon types and order rate difference intervals.

[0102] The first comparison unit 22 is connected to the first order rate prediction unit and is used to compare the first order rate prediction value with the first threshold, and send the comparison result to the decision unit 21. The second comparison unit 23 is connected to the first order rate prediction unit and is used to compare the second order rate prediction value with the second threshold, wherein the first threshold is less than the second threshold; and send the comparison result to the decision unit 21.

[0103] The decision unit 21 determines not to issue a coupon to the user when the first predicted order rate is less than the first threshold or the second predicted order rate is greater than the second threshold. Conversely, when the first predicted order rate is greater than or equal to the first threshold and the second predicted order rate is less than or equal to the second threshold, it sends a calculation notification to the calculation unit. Upon receiving the calculation instruction, the first calculation unit calculates the difference between the first and second predicted order rates.

[0104] The decision unit 21 is connected to the user group determination unit 13. Based on the user group to which the target user belongs as determined by the user group determination unit 13, it queries the user classification stored in the system. When the target user belongs to the second type of user who does not receive coupons, it determines that no coupon will be issued to him. When the target user belongs to the first type of user who receives coupons, it sends an instruction to the matching module 3 to query the mapping table between coupon types and user groups.

[0105] In another embodiment, decision unit 21 instructs order rate prediction unit 11 to predict the value of each coupon denomination, compare the predicted value with a threshold, determine the coupon corresponding to the predicted value greater than the threshold as an available coupon, and send a matching instruction to matching module 3. Correspondingly, matching module 3 selects the coupon with the smallest denomination from the available coupons as the coupon to be issued.

[0106] like Figure 10 As shown, the system further includes a parameter preset module 5, which includes a mapping relationship construction unit 51, a user feature mining unit 52, and a user segmentation creation unit 53. The mapping relationship construction unit 51 is used to construct a one-to-one correspondence between order rate / order rate intervals and coupon types. For example, it uses... Figure 2 The process shown establishes a one-to-one correspondence between coupon types and order rate ranges or order rates; the specific process will not be elaborated here. The user feature mining unit 52 is used to mine multiple user features from user data within a preset time period during the coupon distribution period. For example… Figure 5 Step 2e in the above steps will not be repeated here. The user segmentation creation unit 53 uses the feature combinations constituting a user as input to the order rate model to obtain the predicted order rate value corresponding to the feature combinations. Based on the order rate subsequences segmented by the predicted order rate values, it determines the corresponding feature combinations for grouping. Each group's feature combinations correspond to a user segment. For details, please refer to [link to previous section]. Figure 5 The process and explanations shown will not be repeated here.

[0107] In one embodiment of the present invention, the system further includes a model module 6, such as Figure 11As shown, it includes a training set construction unit 61, a sample processing unit 62, and a model training unit 63. The training set construction unit 61 obtains user data from a database during a preset time period of coupon distribution as the training data source. The sample processing unit 62, connected to the training set acquisition unit 61, processes the training data source to obtain a user feature set, treating each user's feature set as a sample, using whether the user placed an order as the sample label, and dividing it into a training sample set and a test sample set. The model training unit 63, connected to the sample processing unit 62, is configured to train a deep neural network model using samples from the training sample set and test the trained order rate prediction model using samples from the test sample set. The test results are evaluated using evaluation metrics; if the results do not meet the evaluation metrics, retraining is performed until they do. Further, the model module 6 includes a timing unit 64 and a model update unit 65. The timing unit 64, connected to the model training unit 63, starts a timer when the order rate prediction model is obtained and triggers the model update unit 65 when the timing period is reached. The model update unit 65 is connected to the timing unit 64, the training set construction unit 61, the sample processing unit 62, and the model training unit 63. After receiving the trigger information from the timing unit 64, it sends update instructions to the training set construction unit 61, the sample processing unit 62, and the model training unit 63, respectively. The training set construction unit 61 obtains user data from the most recent second preset time period (e.g., 3 days) during the coupon distribution period as a new training data source. The sample processing unit 62 processes the new training data source to obtain samples. The model training unit 63 trains a deep neural network model with the new samples to obtain an updated order rate prediction model.

[0108] In secondhand e-commerce scenarios, such as mobile phone recycling and home appliance recycling, it is impossible to accumulate data on random coupon issuance. Furthermore, coupon issuance needs to be long-term and stable, ensuring a consistent coupon issuance strategy for each user at different times. This invention uses data generated by the coupon issuance strategy as model training data. By continuously deploying the model and updating it based on usage data, it solves the problem of inaccurate order rate prediction caused by the lack of random coupon issuance data. By establishing a direct or indirect correlation between coupons and order rates, the platform's subsidy rate (the ratio of coupon amount to the original price of the product) can be adjusted, and the total subsidy budget can be controlled. This promotes user transaction enthusiasm while maintaining the platform's revenue.

[0109] The above embodiments are for illustrative purposes only and are not intended to limit the invention. Those skilled in the art can make various changes and modifications without departing from the scope of the invention. Therefore, all equivalent technical solutions should also fall within the scope of the invention.

Claims

1. A method for distributing coupons to users of a second-hand e-commerce platform, comprising: Obtain the coupon distribution metrics for the target users and determine whether the coupon distribution conditions are met; In response to a target user meeting the coupon distribution conditions, the mapping relationship between the corresponding coupon type and the distribution index is determined based on the target user's coupon distribution index; and The coupon types are determined based on the mapping relationship between the coupon types and the distribution indicators and the coupon distribution indicators of the target users, and then distributed to the target users. The coupon types are different denominations. This also includes: Obtain real-time data of the target user and historical data within a preset time period; The real-time and historical data are processed to obtain multiple user characteristics, including gender, age, number of logins, and number of transactions. The order rate prediction model is invoked, and the multiple user features are used as inputs to obtain the predicted order rate value for the target user. When the feature values ​​of the coupon feature are set to 1 and 0 respectively in the user features, the order rate prediction model obtains the first order rate prediction value and the second order rate prediction value accordingly. In response to the first order rate prediction value being greater than or equal to the first threshold and the second order rate prediction value being less than the second threshold, the difference between the first order rate prediction value and the second order rate prediction value is calculated, where the first threshold is less than the second threshold. If the difference between the first order rate prediction value and the second order rate prediction value is positive and greater than or equal to the difference threshold, it is determined that the target user meets the conditions for coupon issuance.

2. The method according to claim 1, wherein the coupon distribution indicator is the order rate when coupons are distributed to target users, and the coupons include multiple types in ascending order of face value. Correspondingly, the mapping relationship between coupon types and distribution indicators is a one-to-one mapping relationship between multiple coupons with face value from small to large and multiple standard order rates with numerical values ​​from small to large.

3. The method according to claim 2, further wherein the multiple standard order rates constitute multiple order rate intervals in ascending order, and the mapping relationship between coupon types and order rates is a one-to-one mapping relationship between multiple coupons with face values ​​from small to large and multiple order rate intervals with values ​​from small to large.

4. The method according to claim 1, wherein the coupon issuance indicator is the difference between the first predicted order rate when coupons are issued and the second predicted order rate when coupons are not issued; the coupons include multiple types arranged from smallest to largest face value, and the mapping relationship between coupon types and issuance indicators is a one-to-one mapping relationship between multiple coupons with face values ​​from smallest to largest and the order rate difference range from smallest to largest.

5. The method according to claim 1, wherein the coupon distribution indicator is the predicted order rate when issuing coupons to target users, and the mapping relationship between coupon types and distribution indicators is that the coupon with the smallest face value corresponds to the predicted order rate that meets the distribution conditions; The method further includes: Obtain the predicted order rate values ​​for issuing coupons to target users in ascending order of coupon value; Compare each predicted order rate value with the order rate threshold in turn; Obtain the order rate prediction value that is greater than or equal to the order rate threshold as the qualified order rate prediction value; The coupon amount used when obtaining the predicted value of the qualified order rate is used as the coupon that can be issued. as well as The smallest available coupon is issued to the target user.

6. The method of claim 1, further comprising the step of training an order rate prediction model: Use user data from a pre-defined time period during the coupon distribution period as the training data source; The training data source is processed into a user feature set, with each user's user feature set serving as a sample. The label used to represent whether a user has placed an order; as well as The sample is used to train a deep neural network model to obtain the order rate prediction model.

7. The method according to claim 6, wherein the deep neural network model is a Wide&Deep model, a DeepFM model, a DCN model, or a DIN model.

8. The method of claim 6, further comprising: Timing begins when the order rate prediction model is obtained; as well as In response to reaching the preset timing period, user data from the most recent second preset time period during the coupon distribution period is obtained as the training data source to retrain the model.

9. The method according to claim 1, wherein the coupon distribution indicator is the user segment to which the target user belongs; correspondingly, the mapping relationship between coupon types and distribution indicators is a one-to-one mapping relationship between coupon types and user segments.

10. The method of claim 9, further comprising the step of obtaining the mapping relationship between coupon types and user segments: Obtain user data for a pre-defined time period during the coupon distribution period; Multiple second user characteristics were extracted from the user data; The multiple second user features are traversed to obtain multiple feature combinations; The order rate is obtained by using feature combinations as input to the order rate model; Sort the multiple order rates in ascending order to obtain the total order rate sequence; The total order rate sequence is divided into a corresponding number of order rate subsequences according to the number of available coupon types. The feature combinations are grouped according to the order rate subsequence, and each group of feature combinations corresponds to a user segment; and Establish a one-to-one mapping relationship between coupon types and user groups.

11. A system for issuing coupons to users of a second-hand e-commerce platform, comprising: The coupon distribution indicator acquisition module is configured to obtain coupon distribution indicators for target users. The distribution decision module is connected to the distribution indicator acquisition module and is configured to determine whether the coupon distribution conditions are met based on the coupon distribution indicators of the target user. A matching module, which is connected to the distribution indicator acquisition module and the distribution decision module, is configured to determine the type of coupon that matches the coupon distribution indicator of the target user when the coupon distribution conditions are met, based on the coupon distribution indicator of the target user and the mapping relationship between the corresponding coupon type and the distribution indicator. The coupon type is a different face value. as well as The distribution module, which is connected to the matching module, is configured to distribute coupons of a certain type to the target user in accordance with the distribution strategy; The issuance decision module is specifically configured as follows: Obtain real-time data of the target user and historical data within a preset time period; The real-time and historical data are processed to obtain multiple user characteristics, including gender, age, number of logins, and number of transactions. The order rate prediction model is invoked, and the multiple user features are used as inputs to obtain the predicted order rate value for the target user. When the feature values ​​of the coupon feature are set to 1 and 0 respectively in the user features, the order rate prediction model obtains the first order rate prediction value and the second order rate prediction value accordingly. In response to the first order rate prediction value being greater than or equal to the first threshold and the second order rate prediction value being less than the second threshold, the difference between the first order rate prediction value and the second order rate prediction value is calculated, where the first threshold is less than the second threshold. If the difference between the first order rate prediction value and the second order rate prediction value is positive and greater than or equal to the difference threshold, it is determined that the target user meets the conditions for coupon issuance.

12. The system according to claim 11, wherein the distribution indicator acquisition module comprises: The order rate prediction unit is configured to predict the first order rate prediction value and / or the second order rate prediction value of the target user when coupons are issued and / or not issued. The predicted first order rate prediction value when coupons are issued to the target user is used as the coupon issuance indicator. Correspondingly, the mapping relationship between coupon types and issuance indicators is a one-to-one mapping relationship between multiple coupons with face values ​​from small to large and multiple standard order rates / order rate ranges with values ​​from small to large. or A calculation unit, connected to the order rate prediction unit, is configured to calculate the difference between the order rate prediction value when coupons are issued and the order rate prediction value when no coupons are issued, using this difference as the coupon issuance indicator. Correspondingly, the mapping relationship between coupon types and issuance indicators is a one-to-one mapping relationship between various coupons with denominations ranging from small to large and the order rate difference ranges ranging from small to large; or... The user segmentation determination unit is configured to query the user segment to which the target user belongs based on the target user's identity or user characteristics; correspondingly, the mapping relationship between coupon types and distribution indicators is a one-to-one mapping relationship between coupon types and user segments.

13. The system according to claim 11, wherein the distribution decision module further includes a decision unit connected to the calculation unit, configured to determine that the target user meets the coupon distribution conditions in response to a positive difference between the first order rate prediction value and the second order rate prediction value being greater than a threshold.

14. The system of claim 12, wherein the disbursement decision module further comprises: The first comparison unit is connected to the order rate prediction unit and is configured to compare the first order rate prediction value with the first threshold. The second comparison unit is connected to the order rate prediction unit and is configured to compare the second order rate prediction value with the second threshold; wherein the first threshold is less than the second threshold. Correspondingly, the decision unit is connected to the first comparison unit and the second comparison unit respectively. When the first order rate prediction value is greater than or equal to the first threshold and the second order rate prediction value is less than the second threshold, the decision unit sends a calculation instruction to the calculation unit. Correspondingly, when the calculation unit receives the calculation instruction, it calculates the difference between the first order rate prediction value and the second order rate prediction value.

15. The system according to claim 11 or 14, further comprising a parameter preset module, which includes: The mapping relationship construction unit is configured to build a one-to-one correspondence between order rate / order rate range and coupon type; The user feature mining unit is configured to mine multiple user features from user data within a preset time period during the coupon distribution period; and The user segmentation creation unit is configured to use the feature combination that constitutes a user as input to the order rate model to obtain the order rate prediction value corresponding to the feature combination. Based on the order rate subsequence segmented by the order rate prediction value, the corresponding feature combination is determined and grouped. Each group of feature combinations corresponds to a user segment.

16. The system of claim 11, further comprising a model module: The training set building unit is configured to use user data from a preset time period during the coupon distribution period as the training data source; A sample processing unit, which is connected to the training set acquisition unit, is configured to process the training data source to obtain a user feature set, taking each user's user feature set as a sample, and using whether the user has placed an order as the label of the sample. as well as A model training unit, connected to the sample processing unit, is configured to train a deep neural network model using the samples to obtain the order rate prediction model.

17. The system of claim 15, wherein the model module further comprises: A timing unit, which is connected to the model training unit, is configured to time when the order rate prediction model is obtained; The model update unit, connected to the timing unit, training set construction unit, sample processing unit, and model training unit, is configured to send update commands to the training set construction unit, sample processing unit, and model training unit respectively in response to reaching a preset timing period. The training set construction unit obtains user data from the most recent second preset time period during the coupon distribution period as a new training data source. The sample processing unit processes the new training data source to obtain samples. The model training unit trains a deep neural network model with the new samples to obtain an updated order rate prediction model.

Images