Marketing Information Recommendation Method, Device, Electronic Device and Storage Medium

Abstract

This application is applicable to the field of marketing technology, and provides a marketing information recommendation method, device, electronic device, and storage medium, including: obtaining user information and first behavior data of basic users, where the basic users include users who have not consumed in the target business format, and the first behavior data reflects the number of times of business format consumption of the basic users in business formats other than the target business format; using the user information and the first behavior data as inputs to a pre-trained prediction model to obtain the cross-business-format consumption probability output by the prediction model, where the prediction model is used to predict the probability of a basic user consuming in the target business format based on the user information and the first behavior data, so as to obtain the cross-business-format consumption probability; determining target users from the basic users according to the cross-business-format consumption probability; and pushing marketing information to the target users, where the marketing information is used to guide the target users to consume in the target business format. This application can improve the recommendation effect of marketing information.

Description

Technical Field

[0001] This application belongs to the field of marketing technology, and particularly relates to a marketing information recommendation method, apparatus, electronic device, and computer-readable storage medium. Background Art

[0002] To promote the sales of goods or the promotion of services, etc., each platform usually attracts users to consume through various marketing methods. Among them, push marketing information refers to a marketing method of actively pushing marketing information such as coupons to users in the form of emails, APP notifications, or social media messages, so as to promote product sales or maintain user activity.

[0003] Currently, when pushing marketing information to users, a large amount of marketing information is usually randomly pushed to a large number of users. It is difficult to control the marketing cost, and when it is necessary to conduct targeted marketing for a certain business format, the marketing effect of the random push method is poor. Summary of the Invention

[0004] The embodiments of this application provide a marketing information recommendation method, apparatus, electronic device, and storage medium, which can improve the recommendation effect of marketing information.

[0005] In a first aspect, the embodiments of this application provide a marketing information recommendation method, including:

[0006] Obtain the user information and first behavior data of basic users, where the basic users include users who have not consumed in the target business format, and the first behavior data reflects the number of business format consumption times of the basic users in business formats other than the target business format;

[0007] Use the user information and the first behavior data as the input of a pre-trained prediction model to obtain the cross-business format consumption probability output by the prediction model. The prediction model is used to predict the probability of the basic user consuming in the target business format based on the user information and the first behavior data, so as to obtain the cross-business format consumption probability;

[0008] Determine target users from the basic users according to the cross-business format consumption probability;

[0009] Push marketing information to the target users, where the marketing information is used to guide the target users to consume in the target business format.

[0010] In a second aspect, the embodiments of this application provide a marketing information recommendation apparatus, including:

[0011] An obtaining module, configured to obtain the user information and first behavior data of basic users, where the basic users include users who have not consumed in the target business format, and the first behavior data reflects the number of business format consumption times of the basic users in business formats other than the target business format;

[0012] A prediction module, configured to use the user information and the first behavior data as inputs to a pre-trained prediction model, and obtain a cross-format consumption probability output by the prediction model. The prediction model is used to predict the probability of the basic user consuming in the target format based on the user information and the first behavior data, so as to obtain the cross-format consumption probability;

[0013] A screening module, configured to determine target users from the basic users according to the cross-format consumption probability;

[0014] A pushing module, configured to push marketing information to the target users, where the marketing information is used to guide the target users to consume in the target format.

[0015] In a third aspect, an embodiment of the present application provides an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, the steps of the marketing information recommendation method described in the first aspect above are implemented.

[0016] In a fourth aspect, an embodiment of the present application provides a computer-readable storage medium, where the computer storage medium stores a computer program, and when the computer program is executed by a processor, the steps of the marketing information recommendation method described in the first aspect above are implemented.

[0017] In a fifth aspect, an embodiment of the present application provides a computer program product. When the computer program product runs on an electronic device, the electronic device is enabled to execute the marketing information recommendation method described in the first aspect above.

[0018] The beneficial effects of the embodiments of the present application compared with the prior art are as follows:

[0019] In the embodiments of the present application, when marketing for the target format is required, the user information and the first behavior data of the basic users are obtained. The first behavior data reflects the number of times the basic user consumes in each format other than the target format. Therefore, based on the user information and the first behavior data, the prediction model can better predict the probability of the basic user consuming in the target format according to the formats the basic user has consumed and the number of times of consumption in that format, and obtain a cross-format consumption probability with higher accuracy. Furthermore, based on the predicted cross-format consumption probability, target users are screened and then marketing information is pushed, which can target users with a greater probability of consuming in the target format in the future, improve the recommendation effect of the marketing information, and effectively control the marketing cost. At the same time, the basic users are users who have not consumed in the target format. Marketing by screening target users based on the basic users can achieve cross-format marketing and better explore potential users of the target format. Brief Description of the Drawings

[0020] In order to more clearly illustrate the technical solutions in the embodiments of the present application, the following will briefly introduce the drawings required for use in the description of the embodiments or the prior art.

[0021] Figure 1 It is a schematic flowchart of a marketing information recommendation method provided by an embodiment of the present application;

[0022] Figure 2 It is a schematic flowchart of another marketing information recommendation method provided by an embodiment of the present application;

[0023] Figure 3 It is a schematic structural diagram of a marketing information recommendation device provided by an embodiment of the present application;

[0024] Figure 4 It is a schematic structural diagram of an electronic device provided by an embodiment of the present application. Detailed Description of the Embodiments

[0025] In the following description, specific details such as specific system structures and technologies are proposed for the purpose of illustration rather than limitation, so as to thoroughly understand the embodiments of the present application. However, those skilled in the art should clearly understand that the present application can also be implemented in other embodiments without these specific details. In other cases, detailed descriptions of well-known systems, devices, circuits, and methods are omitted to avoid unnecessary details from interfering with the description of the present application.

[0026] It should be understood that when used in the description of the present application specification and the appended claims, the term "comprising" indicates the presence of the described features, wholes, steps, operations, elements, and / or components, but does not exclude the presence or addition of one or more other features, wholes, steps, operations, elements, components, and / or their combinations.

[0027] It should also be understood that the term "and / or" used in the description of the present application specification and the appended claims refers to any combination and all possible combinations of one or more of the associated listed items, and includes these combinations.

[0028] In addition, in the description of the present application specification and the appended claims, the terms "first", "second", "third", etc. are only used for distinguishing descriptions and cannot be understood as indicating or implying relative importance.

[0029] References to "one embodiment" or "some embodiments" etc. described in the specification of this application mean that specific features, structures, or characteristics described in connection with that embodiment are included in one or more embodiments of this application. Thus, statements such as "in one embodiment", "in some embodiments", "in other some embodiments", "in still other embodiments", etc. that appear in different places in this specification do not necessarily all refer to the same embodiment, but mean "one or more but not all embodiments", unless otherwise specifically emphasized in other ways.

[0030] Embodiment 1:

[0031] Figure 1 The flowchart of a marketing information recommendation method provided by an embodiment of the present invention is shown and described in detail as follows:

[0032] Step S101, obtain the user information and first behavior data of basic users. The above basic users include users who have never consumed in the target business format, and the above first behavior data reflects the number of consumption times of the basic users in business formats other than the target business format.

[0033] The above user information at least includes the user's identity identifier (such as ID number or user number, etc.). Optionally, the user information may also include one or more of the user's gender, age, location, monthly average consumption amount, consumption level and other information.

[0034] It should be noted that the number of consumption times in a business format reflects the number of consumption times of a basic user in a single business format, that is, when a basic user has consumed in multiple business formats, the first behavior data includes the number of consumption times corresponding to multiple business formats.

[0035] In some embodiments, the above basic users may be users who have only consumed in one business format, that is, users who have not had cross-business format consumption. At this time, the above target business format may include one or more business formats other than the business format (historical business format) consumed by the basic user. Through the above processing, marketing can be carried out for users who have not yet had cross-business format consumption, improving the attractiveness of the selling platform to basic users while expanding the breadth of the user market, thereby enhancing user value.

[0036] In some embodiments, the above marketing information recommendation method can be applied to a selling platform, which includes commodities in multiple business formats. At this time, the basic users are the users of the selling platform. Optionally, each business format of the selling platform may include an online channel and an offline channel. When obtaining the first behavior data, the number of consumption times of the business format is determined according to the online consumption times of the basic user in the online channel of the business format and the offline consumption times of the offline channel.

[0037] In some embodiments, the above first-row behavior data may further reflect one or more pieces of information such as the number of times a basic user searches for a business format, views a business format, adds a business format to the cart, collects a business format, the number of times of consumption through a channel, the total number of times of consumption through a channel, and the number of times and days of login of the basic user to the sales platform. It can be understood that a basic user is a user who has not consumed in the target business format. When the first-row behavior data further reflects the number of times of consumption through a channel and the total number of times of consumption through a channel of the basic user in all business formats, the number of times of consumption through the channel and the total number of times of consumption through the channel corresponding to the target business format in the first-row behavior data are both 0.

[0038] For example, the first user behavior data may include a user behavior sequence, and the user behavior sequence includes the product identifiers of the products corresponding to the consumption behaviors and add-to-cart behaviors of the user in each business format. According to the consumption behaviors and add-to-cart behaviors in the user behavior sequence, the number of times of consumption of the user in each business format including the target business format and the number of times of adding the business format to the cart can be obtained. At the same time, the user behavior sequence can also reflect the products preferred by the user in each business format, so that when predicting the cross-business-format consumption probability later, the products preferred by the basic user in each business format can be combined for marketing, improving the matching degree of the marketing information with the basic user.

[0039] In the embodiments of the present application, user information that can reflect information such as the identity of the user is obtained, as well as the number of times of consumption of the basic user in the historical business format, so as to better predict the probability that the basic user will consume in a target business format different from the historical business format based on the personal identity of the basic user and the number of times of consumption in the preferred business format later.

[0040] Step S102, use the above user information and the above first-row behavior data as the input of a pre-trained prediction model to obtain the cross-business-format consumption probability output by the prediction model. The prediction model is used to predict the probability that the basic user will consume in the above target business format according to the above user information and the above first-row behavior data, and obtain the above cross-business-format consumption probability.

[0041] Optionally, the above prediction model may be other models such as an extreme gradient boosting (XGboost) model, an attention-based neural network model, or a large language model, and no specific limitation is made here.

[0042] In the embodiments of the present application, since the model can better mine the deep relationship between the user information and the number of times of consumption of the business format, therefore, the pre-trained prediction model can better predict the probability that the basic user will consume in the target business format according to the user information and the first-row behavior data, ensuring the accuracy of the obtained cross-business-format consumption probability.

[0043] Step S103: Determine target users from the above-mentioned basic users according to the above cross-format consumption probability.

[0044] Specifically, since the marketing budget of the sales platform is usually limited, in order to improve the marketing effect as much as possible, after determining the cross-format consumption probability of each basic user, first screen the basic users according to this cross-format consumption probability, and reduce the marketing cost by screening out the basic users with a relatively low probability of consuming in the target format, so as to be able to improve the marketing effect as much as possible under the condition of meeting the marketing budget.

[0045] Optionally, the N (such as 10,000) basic users with the largest cross-format consumption probability can be used as target users, or the basic users with a cross-format consumption probability greater than or equal to a threshold (such as 75%) can be used as target users.

[0046] In the embodiment of the present application, the basic users are screened to preferentially push marketing information to the basic users with a relatively high cross-format consumption probability. At the same time, the number of target users obtained by screening and control can effectively control the marketing cost.

[0047] Step S104: Push marketing information to the above-mentioned target users, and the above-mentioned marketing information is used to guide the above-mentioned target users to consume in the above-mentioned target format.

[0048] Optionally, the above-mentioned marketing information may include one or more pieces of information such as product information, coupon information, and event information. The product information may include information on target products (such as products participating in preferential activities such as flash sales) in the target format, and the coupon information may include information on coupons applicable to all products or target products in the target format.

[0049] Optionally, when pushing marketing information to target users, the marketing information can be pushed to target users through one or more push methods such as sending messages to target users in the sales platform, pushing links in the background of target users, displaying banners at the top of the sales platform, specific push pages, and sending text messages, and no specific limitation is made here.

[0050] In the embodiments of the present application, since the first behavioral data of the basic users can reflect the number of times the basic users consume in each business format other than the target business format, based on the user information and the first behavioral data of the basic users, the prediction model can better predict the probability that the basic users will consume in the target business format according to the business formats the basic users have consumed and the number of times they have consumed in that business format, and obtain a cross-business-format consumption probability with relatively high accuracy. Furthermore, by screening target users based on the predicted cross-business-format consumption probability and then pushing marketing information, it is possible to conduct marketing for users with a relatively high probability of consuming in the target business format in the future, improving the recommendation effect of the marketing information while effectively controlling the marketing cost. Moreover, the above basic users are users who have not consumed in the target business format. By screening target users based on these basic users and pushing marketing information, it is possible to better explore potential users in the target business format and improve the user value of the basic users.

[0051] In some embodiments, before the above step S102, it further includes:

[0052] Obtain the second behavioral data of the reference users. The above reference users include users who have not consumed in the above target business format within the first time period. The above second behavioral data reflects the number of times the reference users consume in the above business formats other than the above target business format within the above first time period, and the number of times the reference users consume in all the above business formats within the second time period. The second time period is later than the first time period.

[0053] Among them, since consumption behavior can most intuitively reflect users' shopping preferences, the second behavioral data at least reflects the number of times the reference users consume in business formats within the first time period and the second time period.

[0054] It can be understood that since the reference users are users who have not consumed in the target business format within the first time period, the number of times of consumption in business formats within the first time period is the number of times of consumption in business formats other than the target business format.

[0055] It should be noted that the duration of the above first time period can be greater than, equal to, or less than the duration of the second time period. In the embodiments of the present application, in order to fully learn the influence of the number of times of consumption in business formats that the users have consumed within the first time period on the business formats and the number of times of consumption that the users consume within the second time period, the duration of the first time period is at least twice the duration of the second time period.

[0056] For example, assuming that the current time is March 28th and the duration of the second time period is set to 7 days. At this time, the period from March 22nd to March 28th can be used as the second time period, and then the first time period is at least 14 days. At this time, the period from March 8th to March 21st can be used as the first time period.

[0057] Construct training samples based on the second-line data, and determine the labels of the above training samples according to the consumption times of the target business format. The consumption times of the target business format are the consumption times of the corresponding business format within the second time period. One training sample includes the user information of one reference user and the second-line data.

[0058] Train the constructed prediction model according to the above training samples to obtain the pre-trained prediction model.

[0059] Specifically, since the model needs to predict the probability that a user who has not consumed in the target business format will consume in the target business format in the future, when constructing the training samples of the prediction model, users who have not consumed in the target business format within the first time period can be used as reference users, and the behavior data of the reference users within the first time period and after the second time period can be obtained as the second-line data, so that the prediction model can learn the consumption rules of users according to the changes in the consumption times of the business format of users within the first time period and the second time period, and thus can better predict the probability that users will consume in the target business format in the future.

[0060] After constructing the training samples according to the user information and the second-line data of the reference users, the consumption times of the users in the target business format within the second time period are also determined according to the second-line data of the training samples to obtain the consumption times of the target business format, and then the labels of the training samples are determined according to the consumption times of the target business format, so that the model can be optimized according to the error between the predicted probability corresponding to the training samples and the true labels during the training process.

[0061] Optionally, when determining the labels of the above training samples according to the consumption times of the target business format, for the training samples with the consumption times of the target business format greater than 0, the label value can be set to 1, and for the training samples with the consumption times of the target business format equal to 0, the label value can be set to 0.

[0062] Optionally, the loss function of the model can be defined as the total error rate of the predicted probabilities and the true label values of all users in the training samples. During the training process, the above loss function is reduced by optimizing the model parameters, so that the predicted probabilities output by the model better fit the true label values of the training samples.

[0063] In the embodiments of the present application, since the reference user is a user who has not consumed in the target business type during the first time period, and the obtained second behavior data can reflect the number of times the reference user has consumed in the business type during the first time period and the number of times the reference user has consumed in the business type during the second time period that is later than the first time period, therefore, based on the training samples constructed from these second behavior data, the prediction model can better learn the consumption pattern of the reference user according to the changes in the number of times the reference user has consumed in each business type during the first time period and the second time period. Furthermore, the prediction model can better predict the probability that the user will consume in the target business type in the future based on the first behavior data of the basic user.

[0064] In some embodiments, the label of the above training sample is used to divide the above training sample into positive samples and negative samples. The training of the prediction model based on the above training samples includes:

[0065] Respectively use the above positive samples and the above negative samples as the inputs of the above constructed prediction model to obtain the prediction probabilities output by the above constructed prediction model.

[0066] For each of the above positive samples, determine a first loss value based on the loss function, the above prediction probability corresponding to the positive sample, and a first weight. The first weight is determined according to the number of times of consuming the target business type corresponding to the positive sample.

[0067] For each negative sample, determine a second loss value based on the above loss function and the above prediction probability corresponding to the negative sample.

[0068] Update the parameters of the above constructed prediction model according to the above first loss value and the above second loss value to obtain the above pre-trained prediction model.

[0069] Specifically, in order to further improve the accuracy of the prediction model, during the training process of the prediction model, after inputting the training sample into the prediction model to obtain the prediction probability corresponding to the training sample, calculate the loss value corresponding to the training sample based on the type of the training sample, so as to optimize the prediction model according to the loss value and the corresponding optimization algorithm (such as the gradient descent algorithm).

[0070] Since the number of times different reference users consume in the target business format usually varies during the second time period, when calculating the loss value corresponding to a positive sample (i.e., a training sample with the number of times of consuming in the target business format greater than 0, and its label value is 1), the loss value can be calculated by combining the number of times of consuming in the target business format in this positive sample, the error between the predicted probability and the label value of this positive sample, so as to obtain the final first loss value. By using the number of times of consuming in the target business format to amplify the error between the predicted probability and the true label value of the positive sample, when training the prediction model, more attention can be paid to the error rate of positive samples with a large number of cross-business format consumption times, increasing their loss values, thereby enhancing the influence of these positive samples on the total loss value.

[0071] Among them, the first weight can be determined first according to the number of times of consuming in the target business format in this positive sample. After calculating the loss value of this positive sample based on the set loss function, the predicted probability and the label value corresponding to the positive sample, the error rate can be directly weighted by the first weight to obtain the final first loss value.

[0072] Optionally, when calculating the first weight corresponding to each positive sample according to the number of times of consuming in the target business format, the number of times of consuming in each target business format can be normalized first, and then the first weight can be calculated according to the normalized number of times of consuming in the target business format. Through the above processing, the number of times of consuming in the target business format is kept on the same scale, avoiding problems such as numerical instability, so as to improve the stability of the optimization algorithm and the accuracy of the prediction model.

[0073] For negative samples (i.e., training samples with the number of times of consuming in the target business format being 0, and their label values are 0), the reference users in the negative samples did not consume in the target business format during the second time period. At this time, the loss value of the negative sample can be directly calculated according to the loss function and the error between the predicted probability and the label value corresponding to the negative sample, so as to obtain the second loss value.

[0074] When optimizing the prediction model, the parameters of the prediction model are updated according to the set optimization algorithm, the first loss value corresponding to each positive sample, and the second loss value corresponding to each negative sample, so as to obtain the optimized prediction model, that is, the pre-trained prediction model.

[0075] Optionally, if the optimized prediction model does not meet the requirements (such as the accuracy is lower than the threshold of 0.9, or the number of iterations does not reach the target number of 100), the optimized prediction model can be continuously trained until the optimized prediction model meets the requirements, so as to obtain the pre-trained prediction model.

[0076] In the embodiments of the present application, since the greater the number of consumption times of the target business format in the positive samples indicates that the probability of the reference user corresponding to the positive sample consuming in the target business format is greater, therefore, when optimizing the prediction model, the error between the predicted probability and the true label value is weighted by the first weight determined according to the number of consumption times of the target business format, which can increase the influence of the error of the prediction result of the positive sample, and can accelerate the convergence speed of the prediction model and improve the accuracy of the prediction model.

[0077] In some embodiments, after the above step S104, the following is further included:

[0078] If the current marketing stage does not belong to the last one of the above marketing stages in the marketing cycle, before starting the next one of the above marketing stages, a new above first time period and a new above second time period are determined according to the start time of the next one of the above marketing stages.

[0079] Based on the new above first time period and the new above second time period, new above second behavior data is obtained, and a new above pre-trained prediction model is trained based on the new above second behavior data, where the new above second behavior data also reflects whether the above reference user has been pushed the above marketing information.

[0080] When starting the next one of the above marketing stages, based on the new above pre-trained prediction model, the above marketing information recommendation method is re-executed.

[0081] In the embodiments of the present application, the marketing cycle usually includes multiple marketing stages, and the selling platform will push marketing information to the determined target users in each marketing stage. In order to further improve the recommendation effect of the marketing information, when ending the current marketing stage, if the current marketing stage does not belong to the last marketing stage in the marketing cycle, then before starting the next marketing stage, the first time period and the second time period can be re-determined according to the start time of the next marketing stage.

[0082] As Figure 2 shown, after determining the new first time period and the new second time period, the second behavior data can be re-obtained based on the new first time period and the second time period to obtain new second behavior data, and then according to the new second behavior data, a training sample is re-constructed. Finally, based on the new training sample, the constructed prediction model or the current pre-trained prediction model is trained to obtain the latest pre-trained prediction model.

[0083] After obtaining the latest pre-trained prediction model, when the start time of the next marketing stage arrives, the above-mentioned marketing information recommendation method can be re-executed based on the latest pre-trained prediction model, that is, the basic users are re-determined, and the cross-format consumption probabilities of the new basic users are predicted through the latest pre-trained prediction model. Then, target users are screened from the basic users according to the cross-format consumption probabilities, and the marketing information of this marketing stage is pushed to the new target users.

[0084] For example, assume that the marketing cycle includes three marketing stages. The current marketing stage is the first marketing stage, and its corresponding time period is from March 1st to March 6th. The corresponding time period of the next marketing stage (the second marketing stage) is from March 7th to March 13th. Then, after the first marketing stage ends, according to the start time of the second marketing stage, which is March 7th, a new second time period can be set as from March 1st to March 6th, and a new first time period can be set as from February 1st to February 28th. Then, according to the new first time period and the new second time period, new second behavior data is obtained.

[0085] Among them, since the new first time period and the second time period may include the time corresponding to the previous marketing stage, when re-obtaining the second behavior data of the reference users according to the new first time period and the new second time period, the reference users may be the target users of the first marketing stage, that is, the reference users may be the users who have received the marketing information of the previous marketing stage. Therefore, in order to fully consider the influence of the marketing information of the previous marketing stage on the reference users, the obtained new second behavior data also reflects whether the reference user has been pushed marketing information. In some embodiments, the new second behavior data can also reflect in which marketing stages within the marketing cycle the reference user has been pushed marketing information.

[0086] In the embodiments of the present application, since before starting the next marketing stage, the first time period and the second time period are re-determined based on the start time of the next marketing stage, and then new second behavior data is obtained for training the prediction model, and the new second behavior data usually includes users who have received marketing information within the previous marketing stage. Therefore, the new prediction model can better learn the consumption patterns of the target users of the previous marketing stage after being pushed marketing information, improve the accuracy of predicting the cross-format consumption probability, and thus improve the recommendation accuracy of the marketing information of the new marketing stage.

[0087] In some embodiments, the above-mentioned marketing information includes coupon information. Before the above-mentioned step S103, it further includes:

[0088] In the case where there is a previous above-mentioned marketing stage, the actual marketing budget is determined according to the remaining marketing budget of the previous above-mentioned marketing stage and the above-mentioned marketing budget of the current above-mentioned marketing stage.

[0089] Determine the target quantity based on a preset write-off rate, the above-mentioned actual marketing budget, and the coupon amount of the coupon.

[0090] Correspondingly, determining the target users from the above-mentioned basic users according to the above-mentioned cross-format consumption probability includes:

[0091] Select the above-mentioned target users of the above-mentioned target quantity from the above-mentioned basic users according to the above-mentioned cross-format consumption probability.

[0092] The write-off rate refers to the ratio of the number of coupons used by users to the number of coupons issued.

[0093] Since when the sales platform plans a marketing activity, corresponding marketing budgets are usually set in each marketing stage. When pushing marketing information to target users, coupons will be pushed to users in combination with the marketing budget. However, some users may not use the coupons after obtaining them, resulting in a surplus of the marketing budget in this marketing stage. Therefore, in order to further improve the marketing effect, before screening the target users of the current marketing stage according to the cross-format consumption summary, it can be first determined whether there is a surplus of the marketing budget in the previous marketing stage. If there is a surplus of the marketing budget, then jointly determine the actual marketing budget based on the surplus of the marketing budget in the previous marketing stage and the marketing budget of the current marketing stage.

[0094] Among them, since the available issuance amount of the coupon is determined by the marketing budget, and some users who are pushed marketing information may not receive and use the coupon. Therefore, when determining the issuance quantity of the coupon, the target quantity is inversely deduced according to the preset write-off rate, the actual marketing budget, and the coupon amount of the coupon (for example, assuming the coupon is a coupon of 20 off for every 200 spent, then its coupon amount is 20), taking full account of the situation that users do not receive and use the coupon after it is pushed to them, ensuring the write-off effect of the coupon, and thus being able to further improve the recommendation effect of the marketing information.

[0095] Optionally, before calculating the target quantity, the average consumption level can be calculated according to the consumption levels of each basic user, and then the coupon amount of the coupon can be determined according to the average consumption level.

[0096] Finally, when screening the target users, select the target users of the target quantity from the basic users according to the cross-format consumption probability of the basic users.

[0097] For example, assume the marketing budget is 1 million yuan, the marketing cycle is 28 days, and every 7 days is a marketing stage. Then the start times of the marketing stages are the 1st day, the 8th day, the 15th day, and the 22nd day. Among them, assume that the marketing budgets planned to be used in each marketing stage are: 400,000 yuan, 200,000 yuan, 200,000 yuan, and 200,000 yuan respectively.

[0098] Assume that the preset write-off rate is set at 20% based on historical experience. When issuing coupons on the first day, if the planned marketing budget to be used this time is 400,000 yuan and the preset write-off rate is 20%, then coupons with a quota of 2,000,000 yuan (400,000 / 20%) can be issued.

[0099] After the end of the first marketing stage and before starting a new marketing stage on the eighth day, we count the actual usage of the 2,000,000 yuan worth of coupons. If the actual coupon write-off rate of users is exactly 20%, then exactly 400,000 yuan of the marketing budget is used at this time, that is, there is no remaining marketing budget in the previous marketing stage. Then, in the new marketing stage, the actual marketing budget is 200,000 yuan, that is, coupons with a quota of 1,000,000 yuan (200,000 / 20%) can be issued.

[0100] If the actual write-off rate in the previous marketing stage exceeds 20%, assume it is 22%, then actually 2,000,000 * 22% = 440,000 yuan of the marketing budget has been used at this time. Then the actual marketing budget for the new marketing stage is 160,000 yuan (400,000 + 200,000 - 440,000), that is, the available issuance quota for the new marketing stage is 800,000 yuan (160,000 / 20%).

[0101] If the actual coupon write-off rate of users is lower than 20%, assume it is 18%, then actually 2,000,000 * 18% = 360,000 yuan of the marketing budget has been used at this time. Then the actual marketing budget for the new marketing stage is 240,000 yuan (400,000 + 200,000 - 360,000), that is, the available issuance quota for the new marketing stage is 1,200,000 yuan (240,000 / 20%).

[0102] In the embodiments of the present application, adding the remaining marketing budget of the previous marketing stage to the marketing budget of the current marketing stage, on the basis of meeting the marketing budget of the entire marketing cycle, can improve the marketing budget of the current marketing stage as much as possible, which is beneficial to improving the marketing effect. Moreover, before determining the target users, first determine the number of target users by combining the preset write-off rate, the actual marketing budget, and the amount of the coupon, fully considering the situation where users do not receive and use the coupon after it is pushed to them, so as to ensure the write-off effect of the coupon, thereby improving the recommendation effect of the marketing information.

[0103] In some embodiments, the above-mentioned marketing information includes coupon information. Before the above-mentioned step S103, it further includes:

[0104] Weight the above-mentioned cross-format consumption probability according to the second weight corresponding to the above-mentioned basic users to obtain the weighted cross-format consumption probability, and the second weight reflects the sensitivity of the above-mentioned basic users to the coupon.

[0105] Correspondingly, the above-mentioned determining target users from the above-mentioned basic users according to the above-mentioned cross-format consumption probability includes:

[0106] Determine target users from the above basic users according to the weighted cross-format consumption probability.

[0107] The sensitivity of a user to a coupon refers to the degree of response and preference tendency of the user to the coupon, which can reflect the probability of the user making a consumption after obtaining the coupon.

[0108] Specifically, since different users have different sensitivities to coupons, users with a higher sensitivity to coupons have a higher probability of making a consumption after receiving the coupons. Therefore, in order to further improve the marketing effect, when issuing coupons to users for marketing, the second weight corresponding to the basic user can be determined according to the sensitivity of the basic user to the coupon. After obtaining the cross-format consumption probability output by the prediction model, the cross-format consumption model can be weighted according to the second weight corresponding to the basic user, and then the basic users can be screened according to the weighted cross-format consumption probability to obtain the final target users.

[0109] In some embodiments, if it is currently necessary to recommend marketing information during a holiday, at this time, the corresponding third weight can also be determined according to the sensitivity of the user to holidays of the same type as the current holiday, and then the cross-format consumption probability output by the prediction model can be weighted according to the determined third weight. Since the sensitivity of users to holidays of each type can reflect the probability of users making a consumption during holidays of each type, through the above processing, the probability of different users making a consumption during the current holiday can be fully considered, and the accuracy of the determined cross-format consumption probability can be improved.

[0110] In the embodiments of the present application, the cross-format consumption probability output by the prediction model is weighted based on the probability of the user making a consumption after obtaining the coupon, that is, the probability of the basic user making a consumption using the coupon is fully considered in the process of determining the target user, so as to further improve the recommendation effect of the marketing information.

[0111] In some embodiments, before the above-mentioned cross-format consumption probability is weighted according to the second weight corresponding to the above-mentioned basic user to obtain the weighted above-mentioned cross-format consumption probability, it further includes:

[0112] Based on the age and gender of the users, multiple user groups are divided, and different user groups correspond to different age ranges and different genders.

[0113] Push trial coupons to multiple users in each of the above user groups.

[0114] For each of the above user groups, obtain the number of consumption behaviors of the first user and the number of consumption behaviors of the second user within a preset duration after pushing the above trial coupons. The first user is the user who is pushed the above trial coupons, and the second user is the user who is not pushed the above trial coupons.

[0115] Determine the above second weight corresponding to each of the above user groups according to the number of consumption behaviors of the first user and the number of consumption behaviors of the second user.

[0116] Determine the above second weight corresponding to the above basic user according to the above age group in which the age of the above basic user is located and the above user group corresponding to the gender of the above basic user.

[0117] The above number of consumption behaviors refers to the total number of consumption behaviors of the user in all business formats.

[0118] Since the consumption of the same user before and after receiving the coupon is also affected by the time variable, in order to ensure the accuracy of the second weight, the sensitivity of the basic users with the same characteristics (such as the same age or the same location, etc.) to the coupon can be analyzed through the control group method to obtain the second weight corresponding to the basic user.

[0119] Among them, in the embodiments of the present application, users of the same gender within the same age group can be divided into the same user group according to the age and gender of the user, and user groups corresponding to different age groups and different genders can be obtained. That is, considering the differences in shopping preferences of users of different age groups and different genders, different users are divided from two characteristics of age and gender.

[0120] For example, assuming that the set age groups include the following 5 age groups: 0 to 18, 19 to 24, 25 to 35, 35 to 60, and over 60, then after dividing each user according to the age and gender of the user, 10 corresponding user groups can be obtained.

[0121] After dividing the user groups, within each user group, analyze the sensitivity of the users in the user group to the coupon through the control group method to calculate the second weight corresponding to the user group, that is, obtain the second weight corresponding to the age group of the user group.

[0122] Among them, when analyzing the sensitivity of users in this user group to coupons by means of a control group, experimental coupons can be pushed to some users in the user group, and then, within a preset time period after the experimental coupons are pushed (such as within 3 days after the experimental coupons are pushed), the number of consumption behaviors of the users who are pushed the coupons (i.e., the first users) and the number of consumption behaviors of the users who are not pushed the coupons (i.e., the second users) can be obtained. Then, the second weight corresponding to this user group, that is, the second weight corresponding to the corresponding age group and gender, can be calculated based on the number of consumption behaviors of the first users and the number of consumption behaviors of the second users.

[0123] After determining the second weights corresponding to each age group, when it is necessary to determine the second weight of the basic users, the user group (target user group) corresponding to the target age group and target gender can be determined according to the age group (target age group) and gender (target gender) of the basic users. Thus, based on the second weight corresponding to the target user group, the second weight corresponding to this basic user can be obtained.

[0124] In some embodiments, the above second weight can be expressed in the following form:

[0125]

[0126] lift is the second weight, N1 is the average number of consumption behaviors of the first users, and N2 is the average number of consumption behaviors of the second users. That is, the sensitivity of users of the same gender in this age group to coupons is analyzed based on the average consumption times of the users who are pushed the experimental coupons and the users who are not pushed the experimental coupons in the user group within the same time period, rather than analyzing the sensitivity of users to coupons based on the consumption times of a single user before and after being pushed the coupons, excluding the influence of the time variable on the user's consumption decision. And each user group usually includes a large number of users. Calculating the second weight based on the average consumption times of these users within the same time period can better ensure the reliability of the obtained second weight.

[0127] In some embodiments, when dividing each user according to the age group and gender of the user, one or more characteristics such as the user's location, average consumption level, preferred commodity application scenario, and activity can also be combined to divide the user, so that the users in the obtained user group are users with multiple identical characteristics. By limiting the number of identical characteristics, the reliability of the determined second weight is further improved.

[0128] In the embodiments of the present application, after dividing multiple user groups according to the age group and gender of the users, test coupons are pushed to some users in each user group, and then the sensitivity of the users in the user group to the coupons is analyzed based on the number of consumption behaviors of the users who have been pushed the test coupons and those who have not been pushed the coupons in the specified time period. That is, by means of a control group, the number of consumption behaviors of users with the same characteristics in the same time period is analyzed, avoiding the influence of time variables on users' consumption decisions and ensuring the accuracy of the obtained second weight.

[0129] It should be understood that the magnitudes of the sequence numbers of the steps in the above embodiments do not mean the order of execution. The execution order of each process should be determined according to its function and internal logic, and should not constitute any limitation to the implementation process of the embodiments of the present application.

[0130] Embodiment 2:

[0131] Corresponding to the above-mentioned marketing information recommendation method in the foregoing embodiments, Figure 3 The structural block diagram of the marketing information recommendation device provided by the embodiments of the present application is shown. For the sake of convenience of description, only the parts related to the embodiments of the present application are shown.

[0132] Referring to Figure 3 , the device includes: an acquisition module 31, a prediction module 32, a screening module 33, and a push module 34. Among them,

[0133] The acquisition module 31 is configured to acquire the user information and the first behavior data of the basic users. The basic users include users who have not consumed in the target business format, and the first behavior data reflects the number of consumption times of the basic users in the business formats other than the target business format;

[0134] The prediction module 32 is configured to use the user information and the first behavior data as the input of a pre-trained prediction model, and obtain the cross-business-format consumption probability output by the prediction model. The prediction model is used to predict the probability of the basic users consuming in the target business format according to the user information and the first behavior data, so as to obtain the cross-business-format consumption probability;

[0135] The screening module 33 is configured to determine target users from the basic users according to the cross-business-format consumption probability;

[0136] The push module 34 is configured to push marketing information to the target users, and the marketing information is used to guide the target users to consume in the target business format.

[0137] In the embodiment of the present application, since the first behavior data of the basic user can reflect the number of times the basic user consumes in various formats other than the target format, based on the user information and the first behavior data of the basic user, the prediction model can better predict the probability of the basic user consuming in the target format according to the formats that the basic user has consumed and the number of times the basic user has consumed in the format, and obtain a cross-format consumption probability with high accuracy; further, based on the predicted cross-format consumption probability, the target user is screened and the marketing information is pushed, which can be targeted at users with a high probability of consuming in the target format in the future, and the recommendation effect of the marketing information can be improved while effectively controlling the marketing cost. In addition, the above-mentioned basic users are users who have not consumed in the target format. Based on these basic users, the target users are screened and the marketing information is pushed, which can better tap the potential users of the target format and improve the user value of the basic users.

[0138] In some embodiments, the marketing information recommendation device further includes:

[0139] The second behavior data acquisition module is used to obtain the second behavior data of reference users, the above-mentioned reference users include users who have not consumed in the above-mentioned target business format within the first time period, and the above-mentioned second behavior data reflects the number of consumption of the above-mentioned business format by the above-mentioned reference users in the above-mentioned business format other than the above-mentioned target business format within the above-mentioned first time period, and the number of consumption of the above-mentioned business format by the above-mentioned reference users in all the above-mentioned business formats within the second time period, and the above-mentioned second time period is later than the above-mentioned first time period.

[0140] A sample construction module is used to construct a training sample based on the above-mentioned second behavior data, and determine the label of the above-mentioned training sample according to the number of consumption times of the target business format. The above-mentioned number of consumption times of the target business format is the number of consumption times of the above-mentioned business format corresponding to the above-mentioned target business format in the above-mentioned second time period. One of the above-mentioned training samples includes the above-mentioned user information of the above-mentioned reference user and the above-mentioned second behavior data.

[0141] The training module is used to train the constructed prediction model according to the above training samples to obtain the above pre-trained prediction model.

[0142] In some embodiments, the labels of the training samples are used to divide the training samples into positive samples and negative samples, and the marketing information recommendation device further includes:

[0143] The training prediction module is used to use the positive sample and the negative sample as inputs of the prediction model constructed above, respectively, to obtain the prediction probability output by the prediction model constructed above.

[0144] The positive sample loss value calculation module is used to determine, for each of the above positive samples, a first loss value based on a loss function, the predicted probability corresponding to the above positive sample, and a first weight, where the first weight is determined according to the number of consumption times of the target business type corresponding to the above positive sample.

[0145] The negative sample loss value calculation module is used to determine, for each negative sample, a second loss value based on the above loss function and the predicted probability corresponding to the above negative sample.

[0146] The model optimization module is used to update the parameters of the above constructed prediction model according to the above first loss value and the above second loss value to obtain the above pre-trained prediction model.

[0147] In some embodiments, the marketing information recommendation device further includes:

[0148] The time period determination module is used to, if the current marketing stage does not belong to the last one of the above marketing stages in the marketing cycle, determine a new above first time period and a new above second time period according to the start time of the next above marketing stage before starting the next above marketing stage.

[0149] The re-training module is used to obtain new above second behavior data based on the new above first time period and the new above second time period, and train a new above pre-trained prediction model based on the new above second behavior data, where the new above second behavior data also reflects whether the above reference user has been pushed the above marketing information.

[0150] The execution module is used to, when starting the next above marketing stage, re-execute the above marketing information recommendation method based on the new above pre-trained prediction model.

[0151] In some embodiments, the above marketing information includes coupon information, and the marketing information recommendation device further includes:

[0152] The actual marketing budget determination module is used to, in the case of the existence of a previous above marketing stage, determine the actual marketing budget according to the remaining marketing budget of the previous above marketing stage and the above marketing budget of the current above marketing stage.

[0153] The target quantity determination module is used to determine the target quantity based on a preset write-off rate, the above actual marketing budget, and the coupon amount of the coupon.

[0154] Correspondingly, the screening module 33 includes:

[0155] The first screening unit is used to select the above target quantity of the above target users from the above basic users according to the above cross-business type consumption probability.

[0156] In some embodiments, the above marketing information includes coupon information, and the marketing information recommendation device further includes:

[0157] A weighting module, configured to perform weighting processing on the above cross-format consumption probability according to the second weight corresponding to the above basic user to obtain the weighted above cross-format consumption probability, where the second weight reflects the sensitivity of the above basic user to coupons.

[0158] Correspondingly, the screening module 33 includes:

[0159] A second screening unit, configured to determine target users from the above basic users according to the weighted above cross-format consumption probability.

[0160] In some embodiments, the marketing information recommendation device further includes:

[0161] A user grouping module, configured to divide multiple users based on the age and gender of the users to obtain multiple user groups, where different above user groups correspond to different age ranges and different genders.

[0162] A coupon pushing module, configured to push trial coupons to multiple users in each of the above user groups.

[0163] A control data acquisition module, configured to, for each of the above user groups, acquire the number of consumption behaviors of a first user and the number of consumption behaviors of a second user within a preset duration after pushing the above trial coupon, where the first user is the user to whom the above trial coupon is pushed, and the second user is the user who is not pushed the above trial coupon.

[0164] A second weight calculation module, configured to determine the above second weight corresponding to each of the above user groups according to the number of consumption behaviors of the first user and the number of consumption behaviors of the second user.

[0165] A second weight matching module, configured to determine the above second weight corresponding to the above basic user according to the above age range in which the age of the above basic user is located and the above user group corresponding to the gender of the above basic user.

[0166] It should be noted that for the information interaction, execution process, etc. between the above devices / units, since they are based on the same concept as the method embodiments of the present application, their specific functions and the technical effects brought are specifically described in the method embodiment part, and will not be elaborated here.

[0167] Embodiment 3:

[0168] Figure 4 This is a schematic structural diagram of an electronic device provided by an embodiment of the present application. As Figure 4 shown, the electronic device 4 of this embodiment includes: at least one processor 40 (Figure 4 only shows one processor), a memory 41, and a computer program 42 stored in the memory 41 and executable on the at least one processor 40. When the processor 40 executes the computer program 42, the steps in any of the above method embodiments are implemented.

[0169] The electronic device 4 can be a computing device such as a desktop computer, a notebook, a palm computer, and a cloud server. The electronic device may include, but is not limited to, a processor 40 and a memory 41. Those skilled in the art can understand that Figure 4 merely examples of the electronic device 4, and do not constitute a limitation on the electronic device 4. It may include more or fewer components than shown in the figure, or combine certain components, or different components. For example, it may also include input / output devices, network access devices, etc.

[0170] The so-called processor 40 may be a central processing unit (CPU), and the processor 40 may also be other general-purpose processors, digital signal processors (DSPs), application specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The general-purpose processor may be a microprocessor or the processor may also be any conventional processor, etc.

[0171] The memory 41 may be an internal storage unit of the electronic device 4 in some embodiments, such as the hard disk or memory of the electronic device 4. The memory 41 may also be an external storage device of the electronic device 4 in some other embodiments, such as a plug-in hard disk, a smart media card (SMC), a secure digital (SD) card, a flash card, etc. equipped on the electronic device 4. Further, the memory 41 may also include both the internal storage unit and the external storage device of the electronic device 4. The memory 41 is used to store an operating system, application programs, a boot loader, data, and other programs, such as the program code of the computer program. The memory 41 may also be used to temporarily store data that has been output or will be output.

[0172] Those skilled in the art can clearly understand that, for the convenience and brevity of description, only the above division of each functional unit and module is used as an example. In actual applications, the above functions can be allocated to different functional units and modules according to needs, that is, the internal structure of the device can be divided into different functional units or modules to complete all or part of the functions described above. Each functional unit and module in the embodiment can be integrated in a processing unit, or each unit can exist physically alone, or two or more units can be integrated in one unit. The above integrated unit can be implemented in the form of hardware or in the form of a software functional unit. In addition, the specific names of each functional unit and module are only for the convenience of mutual distinction and do not limit the protection scope of this application. The specific working processes of the units and modules in the above system can refer to the corresponding processes in the foregoing method embodiments and will not be elaborated here.

[0173] An embodiment of this application also provides a network device, which includes: at least one processor, a memory, and a computer program stored in the memory and executable on the at least one processor. When the processor executes the computer program, the steps in any of the foregoing method embodiments are implemented.

[0174] An embodiment of this application also provides a computer-readable storage medium, which stores a computer program. When the computer program is executed by a processor, the steps in each of the foregoing method embodiments can be implemented.

[0175] An embodiment of this application provides a computer program product. When the computer program product runs on an electronic device, the electronic device can be made to execute the steps in each of the foregoing method embodiments when executed.

[0176] When the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, to implement all or part of the processes in the above method embodiments of this application, a computer program can be used to instruct relevant hardware to complete. The computer program can be stored in a computer-readable storage medium. When the computer program is executed by a processor, the steps of the above method embodiments can be implemented. Among them, the computer program includes computer program code, and the computer program code can be in the form of source code, object code, executable file or some intermediate form, etc. The computer-readable medium can at least include: any entity or device, recording medium, computer memory, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), electrical carrier signal, telecommunication signal, and software distribution medium that can carry the computer program code to the photographing device / electronic device. For example, a USB flash drive, a mobile hard disk, a magnetic disk or an optical disc, etc. In some jurisdictions, according to legislation and patent practice, the computer-readable medium cannot be an electrical carrier signal and a telecommunication signal.

[0177] In the above embodiments, the descriptions of the respective embodiments have their own emphases. For parts not detailed or recorded in a certain embodiment, reference can be made to the relevant descriptions of other embodiments.

[0178] Those of ordinary skill in the art can realize that the units and algorithm steps of the examples described in conjunction with the embodiments disclosed herein can be implemented by electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are executed in a hardware or software manner depends on the specific application and design constraints of the technical solution. A professional technician can use different methods to implement the described functions for each specific application, but such implementation should not be considered to exceed the scope of this application.

[0179] In the embodiments provided in this application, it should be understood that the disclosed device / network device and method can be implemented in other ways. For example, the device / network device embodiments described above are only illustrative. For example, the division of the modules or units is only a logical function division. In actual implementation, there can be other division methods. For example, multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the displayed or discussed coupling or direct coupling or communication connection to each other can be through some interfaces. The indirect coupling or communication connection of the device or unit can be in an electrical, mechanical or other form.

[0180] The unit described as a separation component may or may not be physically separated. The component shown as a unit may or may not be a physical unit, that is, it may be located in one place or distributed across multiple network units. Some or all of the units can be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0181] The above-described embodiments are only used to illustrate the technical solutions of the present application, and are not intended to limit the same; although the present application has been described in detail with reference to the foregoing embodiments, those of ordinary skill in the art should understand that they can still modify the technical solutions described in the foregoing embodiments, or perform equivalent replacements for some of the technical features; and these modifications or replacements do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the various embodiments of the present application, and should all be included in the protection scope of the present application.

Claims

1. A marketing information recommendation method, characterized in that, Including: Obtain the user information and first behavior data of basic users, where the basic users include users who have never consumed in the target business format. The first behavior data reflects the number of times the basic users have consumed in business formats other than the target business format, as well as the number of times of adding to the cart for each business format; Use the user information and the first behavior data as the input of a pre-trained prediction model to obtain the cross-business-format consumption probability output by the prediction model. The prediction model is used to predict the probability of the basic users consuming in the target business format based on the user information and the first behavior data, so as to obtain the cross-business-format consumption probability; Perform weighted processing on the cross-business-format consumption probability according to the second weight corresponding to the basic user. The second weight corresponding to the basic user is determined according to the second weight corresponding to the target user group, and is used to reflect the sensitivity of the basic user to coupons. The target user group is the user group corresponding to the age range and gender of the basic user. The second weight corresponding to the user group is determined by the control group method, and is expressed as the ratio of the difference between the average consumption behavior times of the first user and the average consumption behavior times of the second user in the user group to the average consumption behavior times of the second user. The first user is the user who is pushed coupons, and the second user is the user in the user group who is not pushed coupons; If it is currently necessary to recommend marketing information during holidays, perform weighted processing on the weighted cross-business-format consumption probability according to the third weight to obtain the weighted cross-business-format consumption probability. The third weight is determined according to the sensitivity of the basic user to holidays of the same type as the current holiday; Determine target users from the basic users according to the weighted cross-business-format consumption probability. The target users include the basic users whose weighted cross-business-format consumption probability is greater than or equal to the threshold; Push marketing information to the target users. The marketing information is used to guide the target users to consume in the target business format.

2. The marketing information recommendation method according to claim 1, wherein Before using the user information and the first behavior data as the input of a pre-trained prediction model to obtain the cross-business-format consumption probability output by the prediction model, it further includes: Obtain the second behavior data of reference users, where the reference users include users who have never consumed in the target business format during the first time period. The second behavior data reflects the number of times the reference users have consumed in business formats other than the target business format during the first time period, as well as the number of times the reference users have consumed in all business formats during the second time period. The second time period is later than the first time period; Construct training samples based on the second behavior data and determine the labels of the training samples according to the number of times of consuming in the target business format. The number of times of consuming in the target business format is the number of times of consuming in the business format corresponding to the target business format during the second time period. One training sample includes the user information and the second behavior data of one reference user; Train the constructed prediction model according to the training samples to obtain the pre-trained prediction model.

3. The marketing information recommendation method according to claim 2, characterized in that, The labels of the training samples are used to divide the training samples into positive samples and negative samples. The training of the constructed prediction model according to the training samples to obtain the pre-trained prediction model includes: Respectively use the positive samples and the negative samples as the inputs of the constructed prediction model to obtain the prediction probabilities output by the constructed prediction model; For each positive sample, determine a first loss value based on a loss function, the prediction probability corresponding to the positive sample, and a first weight, where the first weight is determined according to the number of consumption times of the target business type corresponding to the positive sample; For each negative sample, determine a second loss value based on the loss function and the prediction probability corresponding to the negative sample; Update the parameters of the constructed prediction model according to the first loss value and the second loss value to obtain the pre-trained prediction model.

4. The marketing information recommendation method according to claim 2, wherein After pushing the marketing information to the target user, it further includes: If the current marketing stage does not belong to the last marketing stage of the marketing cycle, before starting the next marketing stage, determine a new first time period and a new second time period according to the start time of the next marketing stage; Obtain new second behavior data based on the new first time period and the new second time period, and train a new pre-trained prediction model based on the new second behavior data, where the new second behavior data also reflects whether the reference user has been pushed the marketing information; When starting the next marketing stage, based on the new pre-trained prediction model, re-execute the marketing information recommendation method as described in claim 1.

5. The marketing information recommendation method according to claim 4, wherein The marketing information includes coupon information. Before determining the target user from the basic users according to the cross-business consumption probability, it further includes: In the case of a previous marketing stage, determine the actual marketing budget according to the remaining marketing budget of the previous marketing stage and the marketing budget of the current marketing stage; Determine the target quantity based on a preset write-off rate, the actual marketing budget, and the face value of the coupon; Correspondingly, the determination of the target user from the basic users according to the cross-business consumption probability includes: Select the target quantity of target users from the basic users according to the cross-business consumption probability.

6. The marketing information recommendation method according to claim 1, wherein Before weighting the cross-business consumption probability according to the second weight corresponding to the basic user to obtain the weighted cross-business consumption probability, it further includes: Divide multiple users based on the age and gender of the users to obtain multiple user groups, where different user groups correspond to different age ranges and different genders; Push trial coupons to multiple users in each user group. For each of the said user groups, obtain the number of consumption behaviors of the first user and the number of consumption behaviors of the second user within a preset duration after pushing the trial coupon. The first user is the user who is pushed the trial coupon, and the second user is the user who is not pushed the trial coupon. Determine the second weight corresponding to each of the said user groups according to the ratio of the difference between the average number of consumption behaviors of the first user and the average number of consumption behaviors of the second user to the average number of consumption behaviors of the second user.

7. A marketing information recommendation device, characterized in that, Including: An acquisition module, configured to acquire user information and first behavior data of basic users. The basic users include users who have never consumed in the target business format. The first behavior data reflects the number of consumption times of the basic users in business formats other than the target business format, and the number of add-to-cart times for each business format. A prediction module, configured to use the user information and the first behavior data as inputs to a pre-trained prediction model, and obtain the cross-business-format consumption probability output by the prediction model. The prediction model is used to predict the probability of the basic users consuming in the target business format according to the user information and the first behavior data, so as to obtain the cross-business-format consumption probability. A second weight weighting module, configured to weight the cross-business-format consumption probability according to the second weight corresponding to the basic users, so as to obtain the weighted cross-business-format consumption probability. The second weight corresponding to the basic users is determined according to the second weight corresponding to the target user group, and is used to reflect the sensitivity of the basic users to coupons. The target user group is the user group corresponding to the age range and gender of the basic users. The second weight corresponding to the user group is determined by the control group method, and is expressed as the ratio of the difference between the average number of consumption behaviors of the first user and the average number of consumption behaviors of the second user in the user group to the average number of consumption behaviors of the second user. The first user is the user who is pushed the coupon, and the second user is the user in the user group who is not pushed the coupon. A third weight weighting module, configured to, if it is necessary to recommend marketing information during holidays currently, weight the weighted cross-business-format consumption probability according to the third weight, so as to obtain the weighted cross-business-format consumption probability. The third weight is determined according to the sensitivity of the basic users to holidays of the same type as the current holiday. A screening module, configured to determine target users from the basic users according to the weighted cross-business-format consumption probability. The target users include the basic users whose weighted cross-business-format consumption probability is greater than or equal to a threshold. A pushing module, configured to push marketing information to the target users. The marketing information is used to guide the target users to consume in the target business format.

8. An electronic device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the computer program, it implements the method according to any one of claims 1 to 6.

9. A computer-readable storage medium storing a computer program, characterized in that, When the computer program is executed by the processor, it implements the method according to any one of claims 1 to 6.

Images