Target user determination method and device, equipment and storage medium

Abstract

The embodiment of the invention discloses a target user determination method and device, equipment and a storage medium, and relates to the technical field of computers, and the method comprises the steps: determining a to-be-optimized business target and user behavior characteristics of each user; for each user, inputting the user behavior characteristics of the current user into a pre-trained service score prediction model, so that the service score prediction model determines a prediction score of each to-be-optimized service target corresponding to the current user according to the user behavior characteristics of the current user; and determining a target user according to the quantile corresponding to each to-be-optimized service target and the predicted score of each to-be-optimized service target corresponding to each user. According to the technical scheme, the target user deeply related to the to-be-optimized business target is determined.

Description

Technical Field

[0001] Embodiments of the present invention relate to the field of computer technology, and in particular to a target user determination method, apparatus, device and storage medium. Background Art

[0002] With the rapid development of the Internet, Internet promotion has become one of the mainstream marketing methods. In Internet promotion, it is necessary to identify the target users and send the promotion information to the client corresponding to the target users.

[0003] In the prior art, the promotion platform can determine each user's preference for each type of information based on the user's user behavior data, determine the target user based on the information category to which the promoter's promotional information belongs and the user's preference for each type of information, and then send the promotional information to the client corresponding to the target user.

[0004] In the process of implementing the present invention, the inventors found that there are at least the following technical problems in the prior art:

[0005] It is difficult to identify target users who are deeply relevant to the promotional information, resulting in the inability to meet the promotion needs of the promoter. Summary of the invention

[0006] The present invention provides a target user determination method, device, equipment and storage medium, which can realize accurate determination of target users who meet promotion needs.

[0007] In a first aspect, an embodiment of the present invention provides a method for determining a target user, comprising:

[0008] Determine the business objectives to be optimized and the user behavior characteristics of each user;

[0009] For each of the users, inputting the user behavior characteristics of the current user into a pre-trained business score prediction model, so that the business score prediction model determines the prediction score of each of the business objectives to be optimized corresponding to the current user according to the user behavior characteristics of the current user;

[0010] The target user is determined according to the quantile corresponding to each of the business objectives to be optimized and the predicted score of each of the business objectives to be optimized corresponding to each of the users.

[0011] In a second aspect, an embodiment of the present invention further provides a target user determination device, including:

[0012] A determination module is used to determine the business objectives to be optimized and the user behavior characteristics of each user;

[0013] A prediction module, for each of the users, inputting the user behavior characteristics of the current user into a pre-trained business score prediction model, so that the business score prediction model determines the prediction score of each of the business objectives to be optimized corresponding to the current user according to the user behavior characteristics of the current user;

[0014] The execution module is used to determine the target user according to the quantile corresponding to each of the business objectives to be optimized and the predicted score of each of the business objectives to be optimized corresponding to each of the users.

[0015] In a third aspect, an embodiment of the present invention further provides a computer device, the computer device comprising:

[0016] one or more processors;

[0017] a storage device for storing one or more programs,

[0018] When the one or more programs are executed by the one or more processors, the one or more processors implement the target user determination method as described in any one of the first aspects.

[0019] In a fourth aspect, an embodiment of the present invention further provides a storage medium comprising computer executable instructions, wherein the computer executable instructions, when executed by a computer processor, are used to execute a target user determination method as described in any one of the first aspects.

[0020] The embodiments of the above invention have the following advantages or beneficial effects:

[0021] An embodiment of the present invention provides a method for determining a target user, the method comprising: determining a business target to be optimized and user behavior characteristics of each user; for each of the users, inputting the user behavior characteristics of the current user into a pre-trained business score prediction model, so that the business score prediction model determines the prediction score of each of the business targets to be optimized corresponding to the current user according to the user behavior characteristics of the current user; determining the target user according to the quantile corresponding to each of the business targets to be optimized and the prediction score of each of the business targets to be optimized corresponding to each of the users. The above technical solution, firstly, can determine the business target to be optimized according to the promotion needs, and can also determine the user behavior characteristics of each user according to the user behavior data of each user within a period of time. Secondly, the user behavior characteristics of each user can be input into a pre-trained business score prediction model as input information. Since the business score prediction model has learned the user behavior characteristics, the node status of the business node related to the business target to be optimized, and the correlation and difference of the predicted scores of the business target to be optimized, the business score prediction model can determine the predicted scores of each business target to be optimized corresponding to each user according to the user behavior characteristics of each user. The predicted score is deeply related to the business target to be optimized. Furthermore, it can be determined whether the predicted scores of each business target to be optimized corresponding to each user meet the preset conditions according to the predicted scores of each business target to be optimized corresponding to each user and the quantiles corresponding to each business target to be optimized. When it is determined that the predicted scores of each business target to be optimized corresponding to the user meet the preset conditions, the user can be determined as a target user. Otherwise, the user can be determined as an excluded user, so as to divide the users into target users and excluded users, and determine the target users deeply related to the business target to be optimized. BRIEF DESCRIPTION OF THE DRAWINGS

[0022] Figure 1 A flow chart of a method for determining a target user provided by an embodiment of the present invention;

[0023] Figure 2 A flowchart of another target user determination method provided by an embodiment of the present invention;

[0024] Figure 3 A schematic diagram of a multi-task learning model in another target user determination method provided by an embodiment of the present invention;

[0025] Figure 4 A schematic diagram of the structure of a target user determination device provided by an embodiment of the present invention;

[0026] Figure 5 A schematic diagram of the structure of a computer device provided in an embodiment of the present invention. DETAILED DESCRIPTION

[0027] The present invention will be further described in detail below in conjunction with the accompanying drawings and embodiments. It is to be understood that the specific embodiments described herein are only used to explain the present invention, rather than to limit the present invention. It should also be noted that, for ease of description, only parts related to the present invention, rather than all structures, are shown in the accompanying drawings.

[0028] It should be mentioned before discussing exemplary embodiments in more detail that some exemplary embodiments are described as processes or methods depicted as flow charts. Although the flow charts describe various operations (or steps) as sequential processes, many operations therein can be implemented in parallel, concurrently or simultaneously. In addition, the order of various operations can be rearranged. The process can be terminated when its operation is completed, but can also have additional steps not included in the accompanying drawings. The process can correspond to methods, functions, procedures, subroutines, subprograms, etc. In addition, the embodiments in the present invention and the features in the embodiments can be combined with each other without conflict.

[0029] With the rapid development of the Internet, Internet promotion has become one of the mainstream marketing methods. Internet promotion often involves three roles, namely users, promotion platforms, and promoters. After the promoter sets the promotion information, the promotion platform can send the promotion information to the client of the target user that matches the promotion information. Specifically, the promotion platform can determine the user's preference for various types of information based on the user's user behavior data. For example, the user's preference for various types of information can be learned based on the user's browsing data, click data, and search data. Furthermore, the promotion platform can determine the target user based on the information category to which the promotion information belongs and the user's preference for various types of information, and send the promotion information to the client corresponding to the target user.

[0030] The effect of promotional information delivery mainly depends on the target users determined by the promotion platform. The accuracy of the target users determined by the promotion platform through the information category to which the promotional information belongs and the degree of preference of each user for each type of information determined by shallow user behavior data such as browsing data, click data and search data of each user is not high, making it difficult to achieve accurate delivery of promotional information, which in turn makes it difficult to meet the promotion needs of the promoter. There are also many solutions that use machine learning and deep learning to solve the problem of accurate delivery. For example, in a method for accurate delivery of Internet online advertisements based on multi-task learning, click-through rate and conversion rate are selected as the goals of multi-task learning for learning to optimize the delivery effect. In a method for accurate delivery of advertisements based on user portraits, user behavior data of the advertising platform is used to build user portraits to improve delivery efficiency.

[0031] However, the purpose of the promoter to release promotional information is often to obtain users who can be deeply converted. The above solutions all determine the target users from the perspective of the promotion platform. It is difficult to determine the target users who are deeply related to the promotional information, and it is also difficult to meet the promotion needs of the promoter.

[0032] Therefore, the prior art proposes a target user determination method to accurately determine the target users that meet the promotion needs.

[0033] Figure 1 The flowchart of a method for determining a target user provided by an embodiment of the present invention is applicable to the situation where it is necessary to accurately determine the target users who meet the promotion needs. The method can be executed by a target user determination device, which can be implemented by software and / or hardware. Figure 1 The method specifically comprises the following steps:

[0034] Step 110: Determine the business objectives to be optimized and the user behavior characteristics of each user.

[0035] The business target to be optimized can be understood as the promotion demand for promoting applications. In order to optimize the business target to be optimized, it is also necessary to determine the business nodes related to the business target to be optimized.

[0036] User behavior characteristics are determined based on user request data, and may specifically be the number of user requests, the number of requests in each time period, the number of requests in each time period, the number of requests based on different device types, and the like.

[0037] Specifically, first, the business target to be optimized can be determined as at least one of clicks on the promoted application, registrations on the promoted application, applications for preset functions in the promoted application, activations of preset functions in the promoted application, and payments in the promoted application according to the promotion needs of the promoter. User behavior characteristics can also be determined based on the user behavior data of each user over a period of time. Considering the issue of information confidentiality, user behavior data can be understood as the user's device information and request information. The device information can include device identification and device type, and the request information can be the request time. Furthermore, feature extraction can be performed on the user behavior data over a period of time to obtain user behavior characteristics, for example, the number of user requests, user request time, user request device, etc. can be obtained.

[0038] It should be noted that the specific duration of a period of time can be determined based on actual needs without specific limitation. Of course, the longer the specific duration of a period of time, the more accurate the determined user behavior characteristics.

[0039] In the embodiment of the present invention, the business target to be optimized is determined according to the promotion demand, and the user behavior characteristics are determined according to the user behavior data of the users within a period of time, so as to provide a data basis for determining the target users by combining the business target to be optimized and the user behavior characteristics.

[0040] Step 120: For each of the users, the user behavior characteristics of the current user are input into a pre-trained business score prediction model, so that the business score prediction model determines the prediction score of each of the business objectives to be optimized corresponding to the current user according to the user behavior characteristics of the current user.

[0041] The business score prediction model is obtained by training the multi-task learning model, so the business score prediction model can determine the predicted score of at least one business target. The business score prediction model is trained by the historical user behavior characteristics of each user in the first historical time period and the node status of each business node related to each business target to be optimized corresponding to each user. The node status of each business node related to the business target to be optimized can determine the predicted score of the business target to be optimized. Therefore, the business score prediction model learns the correlation and difference between the user behavior characteristics, the node status of the business nodes related to the business target to be optimized, and the predicted score of the business target to be optimized, and can determine the predicted score of each business target to be optimized based on the user behavior characteristics.

[0042] The input of the business score prediction model may be the user behavior characteristics of the user, and the output may be the predicted score of the business target to be optimized, that is, the business score prediction model is used to determine the predicted score of the business target to be optimized corresponding to the user according to the user behavior characteristics of the user.

[0043] Specifically, the user behavior characteristics of each user may be input as input information into a pre-trained business score prediction model, and the business score prediction model may determine the prediction score of the business goal to be optimized corresponding to each user according to the user behavior characteristics of each user.

[0044] In an embodiment of the present invention, the business score prediction model learns the correlation between user behavior characteristics and the predicted scores of business objectives to be optimized, and can determine the predicted scores of each business objective to be optimized based on the user behavior characteristics. The predicted scores of business objectives to be optimized corresponding to users corresponding to the user behavior characteristics determined by the business score prediction model are more accurate.

[0045] Step 130: determine the target user according to the quantile corresponding to each of the business objectives to be optimized and the predicted score of each of the business objectives to be optimized corresponding to each of the users.

[0046] The quantile of the business objective to be optimized is used to determine whether the predicted score of the business objective to be optimized corresponding to the user meets the preset condition or does not meet the preset condition. For example, when it is determined that the predicted score of the business objective to be optimized corresponding to the user is greater than the quantile of the business objective to be optimized, it can be determined that the predicted score of the business objective to be optimized corresponding to the user meets the preset condition.

[0047] Specifically, after determining the predicted scores of the business objectives to be optimized corresponding to each user, it is possible to determine whether the predicted scores of the business objectives to be optimized corresponding to each user meet the preset conditions according to the predicted scores of the business objectives to be optimized corresponding to each user and the quantiles corresponding to the business objectives to be optimized. If it is determined that the predicted scores of the business objectives to be optimized corresponding to the user meet the preset conditions, the user can be determined as a target user, otherwise, the user can be determined as an excluded user. In this way, the determination of the target user is achieved.

[0048] In an embodiment of the present invention, after determining the predicted scores of each business objective to be optimized corresponding to each user, the users are divided into target users and excluded users according to the predicted scores of each business objective to be optimized corresponding to each user and the quantile points of each business objective to be optimized, thereby accurately determining the target users.

[0049] The target user determination method provided by an embodiment of the present invention includes: determining a business goal to be optimized and user behavior characteristics of each user; for each of the users, inputting the user behavior characteristics of the current user into a pre-trained business score prediction model, so that the business score prediction model determines the prediction score of each of the business goals corresponding to the current user according to the user behavior characteristics of the current user; determining the target user according to the quantile corresponding to each of the business goals and the prediction score of each of the business goals corresponding to each of the users. The above technical solution, firstly, can determine the business target to be optimized according to the promotion needs, and can also determine the user behavior characteristics of each user according to the user behavior data of each user within a period of time. Secondly, the user behavior characteristics of each user can be input into a pre-trained business score prediction model as input information. Since the business score prediction model has learned the user behavior characteristics, the node status of the business node related to the business target to be optimized, and the correlation and difference of the predicted scores of the business target to be optimized, the business score prediction model can determine the predicted scores of each business target to be optimized corresponding to each user according to the user behavior characteristics of each user. The predicted score is deeply related to the business target to be optimized. Furthermore, it can be determined whether the predicted scores of each business target to be optimized corresponding to each user meet the preset conditions according to the predicted scores of each business target to be optimized corresponding to each user and the quantiles corresponding to each business target to be optimized. When it is determined that the predicted scores of each business target to be optimized corresponding to the user meet the preset conditions, the user can be determined as a target user. Otherwise, the user can be determined as an excluded user, so as to divide the users into target users and excluded users, and determine the target users deeply related to the business target to be optimized.

[0050] Figure 2A flowchart of another target user determination method provided for an embodiment of the present invention, the embodiment of the present invention is applicable to situations where it is necessary to accurately determine target users who meet promotion needs. Based on the above embodiment, before inputting the user behavior characteristics of the current user into a pre-trained business score prediction model, the embodiment of the present invention adds "constructing a multi-task learning model according to the number of the business objectives to be optimized." and "determining the business nodes related to each of the business objectives to be optimized; for each of the users, obtaining the historical user behavior characteristics of the current user in the first historical time period and the node status of the business nodes related to each of the business objectives to be optimized corresponding to the current user; using the historical user behavior characteristics of each of the users in the first historical time period and the node status of the business nodes related to each of the business objectives to be optimized corresponding to the users as training data sets for the multi-task learning model. The business learning model is trained to obtain the business score prediction model. "After obtaining the business score prediction model, "for each user, the historical user behavior characteristics of the current user in the second historical time period are obtained; the historical user behavior characteristics of each user in the second historical time period are input into the business score prediction model, so that the business score prediction model determines the prediction score of each business target to be optimized corresponding to each user; according to the prediction score of each business target to be optimized corresponding to each user in the second historical time period, the quantile corresponding to each business target to be optimized is determined. ", wherein the explanations of the terms that are the same or corresponding to the above embodiments are not repeated here. See Figure 2 , the target user determination method provided by the embodiment of the present invention includes:

[0051] Step 210: Determine the business objectives to be optimized and the user behavior characteristics of each user.

[0052] Specifically, the business goal to be optimized can be the promotion needs of the current application, for example, it can be users clicking on the current application, users registering for the current application, users applying for preset functions in the current application, users activating preset functions in the current application, users making payments in the current application, etc.

[0053] The conversion path of the business target to be optimized is usually very long, for example, users make payments in the current application. When advertising platforms deliver promotional information, due to the long conversion path of the business target to be optimized and the shallow conversion behavior that the advertising platform can obtain, the advertising platform usually intelligently selects the business target to be optimized, which can usually only select shallow goals such as users registering for the current application and users applying for preset functions in the current application.

[0054] In order to optimize the deep goals and achieve precise delivery effects, after determining the business goals to be optimized, it is also necessary to determine the business nodes related to the business goals to be optimized, for example, determine the business node related to the user's click on the current application as click, determine the business node related to the user's registration of the current application as click and registration, determine the business node related to the user's application for the preset functions in the current application as click, registration and application, determine the business node related to the user's activation of the preset functions in the current application as click, registration, application and activation, and determine the business node related to the user's payment in the current application as click, registration, application, activation and payment.

[0055] In one implementation, determining the user behavior characteristics of each user includes:

[0056] Obtain user behavior data of each of the users, the user behavior data including user device information and user request information; perform feature extraction on the user behavior data of each of the users to obtain the user behavior features of each of the users, the user behavior features including at least the number of user requests, the user request time, and the user request device.

[0057] Specifically, the user behavior data of each user within a period of time can be obtained from the advertiser's current application, or the user behavior data of each user within a period of time can be obtained from the advertising platform. Considering the confidentiality of information, the user behavior data may include the user's device information and request information. The device information may include the device identification and device type, and the request information may include the request time. Furthermore, the user behavior data of each user within a period of time can be subjected to feature extraction to obtain the user behavior features of each user. The user behavior features may include the number of user requests, the user request time, the user request device, etc. The number of user requests can be understood as the number of requests made by the user within a period of time, the user request time can be understood as the number of requests made by the user in each time period within a period of time, and the user request device can be understood as the number of requests made by the user based on different types of devices within a period of time.

[0058] There is a natural correlation between user behavior characteristics and the business goals to be optimized for the following reasons: the business goals to be optimized are correlated with the user's activity; the number of user requests and the time of user requests can both reflect the user's activity; the type of device used by the user often also affects the business goals to be optimized.

[0059] In the embodiment of the present invention, the business target to be optimized is determined according to the promotion demand, and the user behavior characteristics are determined according to the user behavior data of the users within a period of time, so as to provide a data basis for determining the target users by combining the business target to be optimized and the user behavior characteristics.

[0060] Step 220: construct a multi-task learning model according to the number of business objectives to be optimized.

[0061] The multi-task learning model may be MMOE, which is composed of Expert modules and Gate modules. The number of Gate modules is determined by the number of business objectives to be optimized.

[0062] Specifically, after determining the number of business objectives to be optimized, the number of Gate modules in the multi-task learning model can be determined according to the number of business objectives to be optimized, thereby constructing a multi-task learning model suitable for the business objectives to be optimized.

[0063] Figure 3 A schematic diagram of a multi-task learning model in another target user determination method provided by an embodiment of the present invention, such as Figure 3 As shown, the multi-task learning model includes two Gate modules, and there are also two business objectives to be optimized.

[0064] It should be noted that each Expert module is composed of a multilayer perceptron (MLP), and each Expert module is independent of each other and has different parameters. Each Gate module is also composed of an MLP, and the activation function is softmax.

[0065] In the embodiment of the present invention, a multi-task learning model suitable for the business objectives to be optimized is constructed according to the number of business objectives to be optimized.

[0066] Step 230: training the multi-task learning model based on the historical user behavior characteristics of each user in the first historical time period and the node status of the business nodes corresponding to each user and related to each business target to be optimized to obtain a business score prediction model.

[0067] In one implementation, step 230 may specifically include:

[0068] Determine the business nodes related to each of the business objectives to be optimized; for each of the users, obtain the historical user behavior characteristics of the current user in the first historical time period and the node status of the business nodes related to each of the business objectives to be optimized corresponding to the current user; use the historical user behavior characteristics of each of the users in the first historical time period and the node status of the business nodes related to each of the business objectives to be optimized corresponding to the users as training data sets to train the multi-task learning model to obtain the business score prediction model.

[0069] like Figure 3As shown, when the multi-task learning model includes two Gate modules, there are two business objectives to be optimized, for example, users can activate preset functions in the current application and users can pay in the current application. Therefore, it is necessary to improve the payment rate while improving the activation rate.

[0070] Specifically, the business nodes related to the user activating the preset function in the current application are determined to be click, register, apply, and activate, and the business nodes related to the user making a payment in the current application are determined to be click, register, apply, activate, and pay. Therefore, the business nodes related to each business goal to be optimized can be determined to be click, register, apply, activate, and pay.

[0071] The historical user behavior characteristics of each user in the first historical time period and the corresponding click status, registration status, application status, activation status and payment status of each user are obtained; the historical user behavior characteristics of each user in the first historical time period and the corresponding click status, registration status, application status, activation status and payment status of each user are used as training data sets to train the multi-task learning model, and the multi-task learning model can learn the correlation between the user behavior characteristics and the click status, registration status, application status, activation status and payment status, and can also learn the correlation and difference between the click status, registration status, application status, activation status and payment status. Therefore, the business score prediction model obtained by training can determine the prediction score of each business target to be optimized according to the user behavior characteristics.

[0072] In the embodiment of the present invention, a business score prediction model is obtained by training a multi-task learning model.

[0073] Step 240: Based on the business score prediction model, the historical user behavior characteristics of each user in the second historical time period are processed to obtain the predicted scores of each business goal to be optimized corresponding to each user, and the quantile corresponding to each business goal to be optimized is determined.

[0074] In one embodiment, step 240 may specifically include:

[0075] For each of the users, the historical user behavior characteristics of the current user within the second historical time period are obtained; the historical user behavior characteristics of each of the users within the second historical time period are input into the business score prediction model, so that the business score prediction model determines the prediction scores of each of the business objectives to be optimized corresponding to each of the users; based on the prediction scores of each of the business objectives to be optimized corresponding to each of the users within the second historical time period, the quantiles corresponding to each of the business objectives to be optimized are determined.

[0076] Specifically, the historical user behavior characteristics of each user in the second historical time period are obtained, and the historical user behavior characteristics of each user in the second historical time period are input into the business score prediction model, and the business score prediction model can determine the prediction score of each business target to be optimized corresponding to each user. Furthermore, the quantile corresponding to each business target to be optimized can be determined according to the prediction score of the business target to be optimized corresponding to each user.

[0077] In one implementation, determining the quantile corresponding to each of the business objectives to be optimized according to the predicted scores of each of the business objectives to be optimized corresponding to each of the users in the second historical time period includes:

[0078] For each of the business objectives to be optimized, the predicted scores of the business objectives to be optimized corresponding to each of the users in the second historical time period are sorted to obtain a sorting result; a predicted critical value is determined in the sorting result according to a preset percentage, and the predicted critical value is determined as the quantile corresponding to the business objective to be optimized.

[0079] Specifically, the quantile corresponding to the business target to be optimized can be understood as the dividing point between the first preset percentage of predicted scores and the remaining predicted scores after sorting the predicted scores of the business targets to be optimized corresponding to each user. In order to determine the quantile of the business target to be optimized, first, it is necessary to sort the predicted scores of the business targets to be optimized corresponding to each user in the second historical time period to obtain the sorting results, and then, the prediction critical value can be determined in the sorting results according to the preset percentage, and the prediction critical value is determined as the quantile of the business target to be optimized. When the preset percentage is 20%, the predicted scores that divide the first 20% of the predicted scores of the business targets to be optimized and the remaining predicted scores of the business targets to be optimized can be determined as the prediction critical value, and the prediction critical value is determined as the quantile of the business target to be optimized.

[0080] When it is determined that the predicted score of the business target to be optimized is greater than the quantile of the business target to be optimized, the state of the business target to be optimized can be determined to be running. For example, when it is determined that the predicted score of the user making a payment in the current application is greater than the quantile of the user making a payment in the current application, the state of the user making a payment in the current application can be determined to be running, that is, it can be determined that the user made a payment in the current application in the second historical time period. Therefore, the state of each business target to be optimized can be determined according to the predicted score and quantile of each business target to be optimized, and then the target user can be determined according to the state of each business target to be optimized.

[0081] In the embodiment of the present invention, it is achieved to determine the quantile corresponding to each business goal to be optimized.

[0082] Step 250: For each of the users, the user behavior characteristics of the current user are input into a pre-trained business score prediction model, so that the business score prediction model determines the prediction score of each of the business objectives to be optimized corresponding to the current user according to the user behavior characteristics of the current user.

[0083] Specifically, the user behavior feature x=(x 1 , x 2 , ..., x n ) is input into the pre-trained business score prediction model, and each Expert module can predict the user behavior feature x = (x 1 , x 2 , ..., x n ) is processed to obtain f i (x), where i represents the ith Expert module. Each Gate module can determine the weight g of each Expert module when determining each business target to be optimized. k (x) i =softmax(W gk x), where k is the kth business objective to be optimized, W gk is the parameter information of each Expert module. Since the weight of each Expert module determined by the Gate module represents the probability of each Expert module being selected when determining each business target to be optimized, Where m is the number of Expert modules. Then, the weights determined by each Gate module can be added to the weighted sum of each Expert module to obtain the input of each Tower module. Each Tower module can be based on the input f k (x) to get the output y k =h k (f k (x)), the output is the predicted score of each business goal to be optimized corresponding to the user. Based on this, the business score prediction model determines the predicted score of each business goal to be optimized corresponding to each user according to the user behavior characteristics of each user.

[0084] It should be noted that each Tower module is also composed of MLP.

[0085] like Figure 3 As shown, when the business score prediction model includes three Expert modules and two Gate modules, it can be determined that two Tower modules are included. 1 , x 2 , ..., x n ) is input into the business score prediction model, the first Expert module can perform user behavior feature x = (x1 , x 2 , ..., x n ) is processed to obtain f i1 (x), the second Expert module can analyze the user behavior feature x=(x 1 , x 2 , ..., x n ) is processed to obtain f i2 (x), the third Expert module can analyze the user behavior feature x=(x 1 , x 2 , ..., x n ) is processed to obtain f i3 (x), the first Gate module can determine the weight g of each Expert module when determining the first service target to be optimized k1 (x) i =softmax(W gk1 x), the second Gate module can determine the weight g of each Expert module when determining the second service target to be optimized k2 (x) i =softmax(W gk2 x) The weight g determined by the first Gate module k1 (x) i f determined by the first Expert module i1 (x), f determined by the second Expert module i2 (x), f determined by the third Expert module i3 (x) Perform weighted summation to obtain the input of the first Tower module The weight g determined by the second Gate module k2 (x) i f determined by the first Expert module i1 (x), f determined by the second Expert module i2 (x), f determined by the third Expert module i3 (x) Perform weighted summation to obtain the input of the second Tower module The first Tower module can be based on the input Determine the output y k1 =h k1 (f k1 (x)), that is, the predicted score y of the first business goal to be optimized can be determined k1 =h k1 (f k1 (x)), the second Tower module can be based on the input Determine the output y k2 =hk2 (f k2 (x)), that is, the predicted score y of the second business goal to be optimized can be determined k2 =h k2 (f k2 (x)). Furthermore, the business score prediction model determines the prediction scores of the two business objectives to be optimized corresponding to each user according to the user behavior characteristics of each user.

[0086] In the embodiment of the present invention, the prediction score of the to-be-optimized business target corresponding to the user corresponding to the user behavior feature is determined based on the business score prediction model.

[0087] Step 260: Determine the target user according to the quantile corresponding to each of the business objectives to be optimized and the predicted score of each of the business objectives to be optimized corresponding to each of the users.

[0088] In one implementation, step 260 may specifically include:

[0089] When it is determined that the relationship between the predicted scores of the business objectives to be optimized corresponding to the user and the quantile points of the business objectives to be optimized meets a preset condition, the user is determined to be a target user.

[0090] Specifically, after determining the predicted scores of the business objectives to be optimized corresponding to each user, it is possible to determine whether the predicted scores of the business objectives to be optimized corresponding to each user meet the preset conditions according to the predicted scores of the business objectives to be optimized corresponding to each user and the quantiles corresponding to the business objectives to be optimized. If it is determined that the predicted scores of the business objectives to be optimized corresponding to the user meet the preset conditions, the user can be determined as a target user, otherwise, the user can be determined as an excluded user. Thus, the determination of the target user can be achieved.

[0091] In actual applications, advertisers can send promotional information to the client of target users, increase the activation rate while improving the payment rate, and achieve deep conversion of users based on the promotional information.

[0092] In an embodiment of the present invention, after determining the predicted scores of each business objective to be optimized corresponding to each user, the users are divided into target users or excluded users according to the predicted scores of each business objective to be optimized corresponding to each user and the quantiles of each business objective to be optimized, thereby accurately determining the target users.

[0093] The target user determination method provided by the embodiment of the present invention includes: determining the business objectives to be optimized and the user behavior characteristics of each user; constructing a multi-task learning model according to the number of business objectives to be optimized; training the multi-task learning model based on the historical user behavior characteristics of each user in a first historical time period and the node status of the business nodes corresponding to each user and related to each business objective to be optimized, to obtain a business score prediction model; based on the business score prediction model, the historical user behavior characteristics of each user in a second historical time period are processed to obtain the predicted scores of each business objective to be optimized corresponding to each user, and determining the quantile corresponding to each business objective to be optimized; for each of the users, inputting the user behavior characteristics of the current user into a pre-trained business score prediction model, so that the business score prediction model determines the predicted scores of each of the business objectives to be optimized corresponding to the current user according to the user behavior characteristics of the current user; and determining the target user according to the quantile corresponding to each of the business objectives to be optimized and the predicted scores of each of the business objectives to be optimized corresponding to each of the users. The above technical scheme can, firstly, determine the business target to be optimized according to the promotion demand, and can also determine the user behavior characteristics of each user according to the user behavior data of each user within a period of time. Secondly, the number of Gate modules in the multi-task learning module can be determined according to the number of business targets to be optimized, and a multi-task learning model can be constructed. Thirdly, the historical user behavior characteristics of each user in the first historical time period and the node status of the business nodes corresponding to each user and related to each business target to be optimized can be used as training data sets to train the multi-task learning model to obtain a business score prediction model. The historical user behavior characteristics of each user in the second historical time period can be processed based on the business score prediction model to obtain the predicted scores of each business target to be optimized corresponding to each user, and the quantile corresponding to each business target to be optimized can be determined. Then, the user behavior characteristics of each user are input as input information into the trained business score prediction model. In the model, since the business score prediction model has learned the user behavior characteristics, the node status of the business nodes related to the business objectives to be optimized, and the correlation and difference of the predicted scores of the business objectives to be optimized, the business score prediction model can determine the predicted scores of the business objectives to be optimized corresponding to each user according to the user behavior characteristics of each user. The predicted scores are deeply related to the business objectives to be optimized. Furthermore, it can be determined whether the predicted scores of the business objectives to be optimized corresponding to each user meet the preset conditions according to the predicted scores of the business objectives to be optimized corresponding to each user and the quantiles corresponding to the business objectives to be optimized. When it is determined that the predicted scores of the business objectives to be optimized corresponding to the user meet the preset conditions, the user can be determined as a target user. Otherwise, the user can be determined as an excluded user, thereby realizing the division of users into target users and excluded users, and determining the target users who are deeply related to the business objectives to be optimized.

[0094] In addition, advertisers can send promotional information to the target users' clients, which can not only achieve personalized and accurate delivery of promotional information, but also achieve multiple business goals to be optimized at the same time and realize deep conversion of users.

[0095] Figure 4 This is a schematic diagram of the structure of a target user determination device provided by an embodiment of the present invention. The device and the target user determination methods of the above embodiments belong to the same inventive concept, and the details not described in detail in the embodiment of the target user determination device can refer to the embodiment of the above target user determination method.

[0096] The specific structure of the target user determination device is as follows: Figure 4 As shown, including:

[0097] A determination module 410 is used to determine the business target to be optimized and the user behavior characteristics of each user;

[0098] The prediction module 420 is used to input the user behavior characteristics of the current user into a pre-trained business score prediction model for each of the users, so that the business score prediction model determines the prediction score of each of the business objectives to be optimized corresponding to the current user according to the user behavior characteristics of the current user;

[0099] The execution module 430 is used to determine the target user according to the quantile corresponding to each of the business objectives to be optimized and the predicted score of each of the business objectives to be optimized corresponding to each of the users.

[0100] Based on the above embodiment, the determination module 410 is specifically used for:

[0101] Determine the business goal to be optimized; obtain user behavior data of each of the users, the user behavior data including user device information and user request information; perform feature extraction on the user behavior data of each of the users to obtain the user behavior features of each of the users, the user behavior features including at least the number of user requests, the user request time, and the user request device.

[0102] Based on the above embodiment, the device further includes:

[0103] A building module is used to build a multi-task learning model according to the number of business objectives to be optimized.

[0104] Based on the above embodiment, the device further includes:

[0105] A training module is used to determine the business nodes related to each of the business objectives to be optimized; for each of the users, the historical user behavior characteristics of the current user in the first historical time period and the node status of the business nodes related to each of the business objectives to be optimized corresponding to the current user are obtained; the historical user behavior characteristics of each of the users in the first historical time period and the node status of the business nodes related to each of the business objectives to be optimized corresponding to each of the users are used as training data sets to train the multi-task learning model to obtain the business score prediction model.

[0106] Based on the above embodiment, the device further includes:

[0107] The quantile determination module is used to obtain, for each of the users, the historical user behavior characteristics of the current user within the second historical time period; input the historical user behavior characteristics of each of the users within the second historical time period into the business score prediction model, so that the business score prediction model determines the predicted scores of each of the business objectives to be optimized corresponding to each of the users; and determine the quantile corresponding to each of the business objectives to be optimized based on the predicted scores of each of the business objectives to be optimized corresponding to each of the users within the second historical time period.

[0108] In one implementation, determining the quantile corresponding to each of the business objectives to be optimized according to the predicted scores of each of the business objectives to be optimized corresponding to each of the users in the second historical time period includes:

[0109] For each of the business objectives to be optimized, the predicted scores of the business objectives to be optimized corresponding to each of the users in the second historical time period are sorted to obtain a sorting result; a predicted critical value is determined in the sorting result according to a preset percentage, and the predicted critical value is determined as the quantile corresponding to the business objective to be optimized.

[0110] Based on the above embodiment, the execution module 430 is specifically used for:

[0111] When it is determined that the relationship between the predicted scores of the business objectives to be optimized corresponding to the user and the quantile points of the business objectives to be optimized meets a preset condition, the user is determined to be a target user.

[0112] The target user determination device provided in the embodiment of the present invention can execute the target user determination method provided in any embodiment of the present invention, and has the corresponding functional modules and beneficial effects of executing the target user determination method.

[0113] It is worth noting that in the embodiment of the above-mentioned target user determination device, the various units and modules included are only divided according to functional logic, but are not limited to the above-mentioned division, as long as the corresponding functions can be achieved; in addition, the specific names of the functional units are only for the convenience of distinguishing each other, and are not used to limit the scope of protection of the present invention.

[0114] Figure 5 A schematic diagram of the structure of a computer device provided in an embodiment of the present invention. Figure 5 A block diagram of an exemplary computer device 5 suitable for use in implementing embodiments of the present invention is shown. Figure 5 The computer device 5 shown is only an example and should not bring any limitation to the functions and scope of use of the embodiments of the present invention.

[0115] like Figure 5 As shown, the computer device 5 is in the form of a general-purpose computing electronic device. The components of the computer device 5 may include, but are not limited to: one or more processors or processing units 16, a system memory 28, and a bus 18 connecting different system components (including the system memory 28 and the processing unit 16).

[0116] Bus 18 represents one or more of several types of bus structures, including a memory bus or memory controller, a peripheral bus, an accelerated graphics port, a processor or a local bus using any of a variety of bus architectures. By way of example, these architectures include, but are not limited to, an Industry Standard Architecture (ISA) bus, a Micro Channel Architecture (MAC) bus, an Enhanced ISA bus, a Video Electronics Standards Association (VESA) local bus, and a Peripheral Component Interconnect (PCI) bus.

[0117] The computer device 5 typically includes a variety of computer system readable media. These media can be any available media that can be accessed by the computer device 5, including volatile and non-volatile media, removable and non-removable media.

[0118] The system memory 28 may include computer system readable media in the form of volatile memory, such as random access memory (RAM) 30 and / or cache memory 32. The computer device 5 may further include other removable / non-removable, volatile / non-volatile computer system storage media. By way of example only, the storage system 34 may be used to read and write non-removable, non-volatile magnetic media ( Figure 5 not shown, usually called a "hard drive"). Although Figure 5Not shown in the figure, a disk drive for reading and writing to a removable non-volatile disk (e.g., a "floppy disk"), and an optical disk drive for reading and writing to a removable non-volatile optical disk (e.g., a CD-ROM, a DVD-ROM, or other optical media) may be provided. In these cases, each drive may be connected to the bus 18 via one or more data medium interfaces. The system memory 28 may include at least one program product having a set (e.g., at least one) of program modules that are configured to perform the functions of the various embodiments of the present invention.

[0119] A program / utility 40 having a set (at least one) of program modules 42 may be stored, for example, in system memory 28, such program modules 42 including, but not limited to, an operating system, one or more application programs, other program modules, and program data, each of which or some combination may include an implementation of a network environment. Program modules 42 generally perform the functions and / or methods of the embodiments described herein.

[0120] The computer device 5 may also communicate with one or more external devices 14 (e.g., keyboard, pointing device, display 24, etc.), one or more devices that enable a user to interact with the computer device 5, and / or any device that enables the computer device 5 to communicate with one or more other computing devices (e.g., network card, modem, etc.). Such communication may be performed through an input / output (I / O) interface 22. Furthermore, the computer device 5 may also communicate with one or more networks (e.g., a local area network (LAN), a wide area network (WAN), and / or a public network, such as the Internet) through a network adapter 20. Figure 5 As shown, the network adapter 20 communicates with other modules of the computer device 5 via the bus 18. It should be understood that although Figure 5 Not shown, other hardware and / or software modules may be used in conjunction with computer device 5, including but not limited to: microcode, device drivers, redundant processing units, external disk drive arrays, RAID systems, tape drives, and data backup storage systems.

[0121] The processing unit 16 executes various functional applications and page displays by running the program stored in the system memory 28, for example, implementing the target user determination method provided in the embodiment of the present invention, which includes:

[0122] Determine the business objectives to be optimized and the user behavior characteristics of each user;

[0123] For each of the users, inputting the user behavior characteristics of the current user into a pre-trained business score prediction model, so that the business score prediction model determines the prediction score of each of the business objectives to be optimized corresponding to the current user according to the user behavior characteristics of the current user;

[0124] The target user is determined according to the quantile corresponding to each of the business objectives to be optimized and the predicted score of each of the business objectives to be optimized corresponding to each of the users.

[0125] Of course, those skilled in the art can understand that the processor can also implement the technical solution of the target user determination method provided by any embodiment of the present invention.

[0126] An embodiment of the present invention provides a computer-readable storage medium on which a computer program is stored. When the program is executed by a processor, a target user determination method provided by an embodiment of the present invention is implemented. The method includes:

[0127] Determine the business objectives to be optimized and the user behavior characteristics of each user;

[0128] For each of the users, inputting the user behavior characteristics of the current user into a pre-trained business score prediction model, so that the business score prediction model determines the prediction score of each of the business objectives to be optimized corresponding to the current user according to the user behavior characteristics of the current user;

[0129] The target user is determined according to the quantile corresponding to each of the business objectives to be optimized and the predicted score of each of the business objectives to be optimized corresponding to each of the users.

[0130] The computer storage medium of the embodiment of the present invention can adopt any combination of one or more computer-readable media. The computer-readable medium can be a computer-readable signal medium or a computer-readable storage medium. The computer-readable storage medium can be, for example, but not limited to: an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, device or device, or any combination of the above. More specific examples (non-exhaustive list) of computer-readable storage media include: an electrical connection with one or more wires, a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the above. In this document, a computer-readable storage medium can be any tangible medium containing or storing a program, which can be used by an instruction execution system, device or device or used in combination with it.

[0131] Computer-readable signal media may include data signals propagated in baseband or as part of a carrier wave, which carry computer-readable program code. Such propagated data signals may take a variety of forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination of the above. Computer-readable signal media may also be any computer-readable medium other than a computer-readable storage medium, which may send, propagate, or transmit a program for use by or in conjunction with an instruction execution system, apparatus, or device.

[0132] The program code embodied on the computer readable medium may be transmitted using any appropriate medium, including but not limited to wireless, wireline, optical fiber cable, RF, etc., or any suitable combination of the foregoing.

[0133] Computer program code for performing the operations of the present invention may be written in one or more programming languages ​​or a combination thereof, including object-oriented programming languages ​​such as Java, Smalltalk, C++, and conventional procedural programming languages ​​such as "C" or similar programming languages. The program code may be executed entirely on the user's computer, partially on the user's computer, as a separate software package, partially on the user's computer and partially on a remote computer, or entirely on a remote computer or server. In the case of a remote computer, the remote computer may be connected to the user's computer through any type of network, including a local area network (LAN) or a wide area network (WAN), or may be connected to an external computer (e.g., via the Internet using an Internet service provider).

[0134] It should be understood by those skilled in the art that the modules or steps of the present invention described above can be implemented by a general-purpose computing device, they can be concentrated on a single computing device, or distributed on a network composed of multiple computing devices, optionally, they can be implemented by a program code executable by a computer device, so that they can be stored in a storage device and executed by the computing device, or they can be made into individual integrated circuit modules, or multiple modules or steps therein can be made into a single integrated circuit module for implementation. Thus, the present invention is not limited to any specific combination of hardware and software.

[0135] It should be noted that the collection, collection, updating, analysis, processing, use, transmission, storage and other aspects of user personal information involved in the technical solution of this disclosure are in compliance with the provisions of relevant laws and regulations, are used for legitimate purposes, and do not violate public order and good morals. Necessary measures are taken for user personal information to prevent illegal access to user personal information data and maintain the security of user personal information, network security and national security.

[0136] For example: After obtaining user behavior data, we will de-identify the user behavior data through technical means.

[0137] Note that the above are only preferred embodiments of the present invention and the technical principles used. Those skilled in the art will understand that the present invention is not limited to the specific embodiments herein, and that various obvious changes, readjustments and substitutions can be made by those skilled in the art without departing from the scope of protection of the present invention. Therefore, although the present invention has been described in more detail through the above embodiments, the present invention is not limited to the above embodiments, and may include more other equivalent embodiments without departing from the concept of the present invention, and the scope of the present invention is determined by the scope of the appended claims.

Claims

1. A method for determining a target user, characterized in that: include: Determine the business objectives to be optimized and the user behavior characteristics of each user; For each of the users, inputting the user behavior characteristics of the current user into a pre-trained business score prediction model, so that the business score prediction model determines the prediction score of each of the business objectives to be optimized corresponding to the current user according to the user behavior characteristics of the current user; The target user is determined according to the quantile corresponding to each of the business objectives to be optimized and the predicted score of each of the business objectives to be optimized corresponding to each of the users.

2. The target user determination method according to claim 1, characterized in that: Determine the user behavior characteristics of each user, including: Acquire user behavior data of each of the users, wherein the user behavior data includes user device information and user request information; Feature extraction is performed on the user behavior data of each of the users to obtain the user behavior features of each of the users, wherein the user behavior features at least include the number of user requests, the user request time, and the user request device.

3. The target user determination method according to claim 1, characterized in that: Before inputting the user behavior characteristics of the current user into the pre-trained business score prediction model, it also includes: A multi-task learning model is constructed according to the number of business objectives to be optimized.

4. The target user determination method according to claim 3, characterized in that: Before inputting the user behavior characteristics of the current user into the pre-trained business score prediction model, it also includes: Determine the business nodes related to each of the business objectives to be optimized; For each of the users, obtaining historical user behavior characteristics of the current user in a first historical time period and node states of service nodes corresponding to the current user and related to each of the service objectives to be optimized; The multi-task learning model is trained using the historical user behavior characteristics of each of the users in the first historical time period and the node status of the business nodes corresponding to each of the users and related to each of the business objectives to be optimized as a training data set to obtain the business score prediction model.

5. The target user determination method according to claim 4, characterized in that: After obtaining the business score prediction model, the method further includes: For each of the users, obtaining historical user behavior characteristics of the current user in a second historical time period; Inputting the historical user behavior characteristics of each of the users in the second historical time period into the business score prediction model, so that the business score prediction model determines the prediction score of each of the business objectives to be optimized corresponding to each of the users; According to the predicted scores of the business objectives to be optimized corresponding to the users in the second historical time period, the quantiles corresponding to the business objectives to be optimized are determined.

6. The target user determination method according to claim 5, characterized in that: Determining, according to the predicted scores of the business objectives to be optimized corresponding to the users in the second historical time period, the quantiles corresponding to the business objectives to be optimized, including: For each of the business objectives to be optimized, the predicted scores of the business objectives to be optimized corresponding to each of the users in the second historical time period are sorted to obtain a sorting result; A prediction critical value is determined in the sorting results according to a preset percentage, and the prediction critical value is determined as a quantile corresponding to the business objective to be optimized.

7. The target user determination method according to claim 1, characterized in that: Determining a target user according to the quantiles corresponding to the business objectives to be optimized and the predicted scores of the business objectives to be optimized corresponding to the users includes: When it is determined that the relationship between the predicted scores of the business objectives to be optimized corresponding to the user and the quantile points of the business objectives to be optimized meets a preset condition, the user is determined to be a target user.

8. A target user determination device, characterized in that: include: A determination module is used to determine the business objectives to be optimized and the user behavior characteristics of each user; A prediction module, for each of the users, inputting the user behavior characteristics of the current user into a pre-trained business score prediction model, so that the business score prediction model determines the prediction score of each of the business objectives to be optimized corresponding to the current user according to the user behavior characteristics of the current user; The execution module is used to determine the target user according to the quantile corresponding to each of the business objectives to be optimized and the predicted score of each of the business objectives to be optimized corresponding to each of the users.

9. A computer device, characterized in that: The computer device comprises: one or more processors; a storage device for storing one or more programs, When the one or more programs are executed by the one or more processors, the one or more processors implement the target user determination method as described in any one of claims 1 to 7.

10. A storage medium containing computer executable instructions, characterized in that: When the computer executable instructions are executed by a computer processor, they are used to perform the target user determination method according to any one of claims 1 to 7.

Images