CN113239230B

CN113239230B - Service recommendation method, behavior data increment prediction model generation method and device

Info

Publication number: CN113239230B
Application number: CN202110296920.2A
Authority: CN
Inventors: 彭泽阳
Original assignee: Beijing Dajia Internet Information Technology Co Ltd
Current assignee: Beijing Dajia Internet Information Technology Co Ltd
Priority date: 2021-03-19
Filing date: 2021-03-19
Publication date: 2024-05-17
Anticipated expiration: 2041-03-19
Also published as: CN113239230A

Abstract

The disclosure relates to a service recommendation method, a behavior data increment prediction model generation method and a behavior data increment prediction model generation device. The service recommendation method comprises the following steps: acquiring user characteristics of a plurality of user accounts to be identified; inputting the user characteristics of each user account to be identified into a pre-trained behavior data increment prediction model corresponding to the first service, and obtaining a behavior data increment prediction value of each user account to be identified; acquiring target user accounts from a plurality of user accounts to be identified according to the behavior data increment predicted value of each user account to be identified; and recommending the first service to the client logging in the target user account. The embodiment of the disclosure can improve the recommendation effect.

Description

Service recommendation method, behavior data increment prediction model generation method and device

Technical Field

The disclosure relates to the field of short videos, and in particular relates to a service recommendation method, a behavior data increment prediction model generation method, a device, a server and a storage medium.

Background

With the popularization of the mobile internet and the maturity of communication technology, more and more people participate in the trend of social interaction by sharing short videos. Short video refers to high frequency pushed video content played on various new media platforms, suitable for viewing in a mobile state and a short leisure state, varying from a few seconds to a few minutes.

In order to continuously improve user experience and meet personalized demands of users, the short video platform can issue corresponding services to short video users along with the development of the services. In the related art, a short video platform typically pushes a service to be delivered to all users using the short video application. However, such a recommendation method has a problem of poor recommendation effect.

Disclosure of Invention

The disclosure provides a service recommendation method, a behavior data increment prediction model generation method, a device, a server and a storage medium, so as to at least solve the problems of poor recommendation effect and the like of a recommendation mode in the related technology. The technical scheme of the present disclosure is as follows:

according to a first aspect of an embodiment of the present disclosure, there is provided a service recommendation method, including:

Acquiring user characteristics of a plurality of user accounts to be identified;

inputting the user characteristics of each user account to be identified into a pre-trained behavior data increment prediction model corresponding to a first service, and obtaining a behavior data increment prediction value of each user account to be identified; wherein the incremental behavior data prediction model is generated based on first behavior data of a first sample user account corresponding to a first service and second behavior data of a second sample user account corresponding to a second service, the first service being a new version of the second service;

Acquiring target user accounts from the plurality of user accounts to be identified according to the behavior data increment predicted value of each user account to be identified; and

Recommending the first service to a client logging in the target user account.

According to some embodiments of the present disclosure, the behavioral data increment prediction model is trained by:

Acquiring first behavior data of a first sample user account corresponding to a first service, and acquiring second behavior data of a second sample user account corresponding to a second service; wherein, the first service is a new version service of the second service;

training a first machine learning model according to the user characteristics of the second sample user account and the second behavior data to obtain model parameters, and generating a correction model based on the model parameters;

Acquiring a behavior data increment corresponding to the first sample user account according to the user characteristics of the first sample user account, the first behavior data and the deviation correction model;

and generating a behavior data increment prediction model corresponding to the first service according to the user characteristics of the first sample user account and the behavior data increment.

According to some embodiments of the present disclosure, the obtaining, according to the user characteristics of the first sample user account, the first behavior data, and the correction model, the behavior data increment corresponding to the first sample user account includes:

Inputting the user characteristics of the first sample user account into the deviation rectification model, and predicting a behavior data predicted value of the first sample user account corresponding to the second service;

and acquiring the behavior data increment corresponding to the first sample user account according to the difference value between the first behavior data corresponding to the first service of the first sample user account and the behavior data predicted value.

According to some embodiments of the disclosure, the generating a behavior data increment prediction model corresponding to the first business according to the user feature of the first sample user account and the behavior data increment includes:

taking the user characteristics of the first sample user account as training data, and taking the behavior data increment as a label corresponding to the training data;

and training a second machine learning model based on the training data and the label to obtain model parameters, and generating a behavior data increment prediction model corresponding to the first service based on the model parameters.

According to some embodiments of the disclosure, the first machine learning model and the second machine learning model are bayesian optimization LightGBM models.

According to some embodiments of the disclosure, the obtaining the target user account from the plurality of user accounts to be identified according to the behavior data increment predicted value of each user account to be identified includes:

Comparing the behavior data increment predicted value of each user account to be identified with a target threshold value;

Identifying the user account to be identified, of which the behavior data increment predicted value is greater than the target threshold, from the plurality of user accounts to be identified;

And determining the user account to be identified, of which the behavior data increment predicted value is larger than the target threshold value, as the target user account.

Sequencing the behavior data increment predicted values of each user account to be identified according to the sequence from large to small, and determining the user account to be identified corresponding to the behavior data increment predicted value of the previous N row as the target user account; or alternatively

And sequencing the plurality of user accounts to be identified according to the sequence of the behavior data increment predicted value of each user account to be identified from large to small, and determining the user account to be identified with the previous N row as the target user account.

According to a second aspect of the embodiments of the present disclosure, there is provided a behavior data increment prediction model generation method, including:

acquiring user characteristics of the first sample user account and user characteristics of the second sample user account;

According to a third aspect of the embodiments of the present disclosure, there is provided a service recommendation device, including:

the first acquisition module is configured to acquire user characteristics of a plurality of user accounts to be identified;

The second acquisition module is configured to input the user characteristics of each user account to be identified into a pre-trained behavior data increment prediction model corresponding to the first service, and acquire a behavior data increment predicted value of each user account to be identified; wherein the incremental behavior data prediction model is generated based on first behavior data of a first sample user account corresponding to a first service and second behavior data of a second sample user account corresponding to a second service, the first service being a new version of the second service;

The third acquisition module is configured to acquire target user accounts from the plurality of user accounts to be identified according to the behavior data increment predicted value of each user account to be identified; and

And the recommending module is configured to recommend the first service to a client logging in the target user account.

According to some embodiments of the disclosure, the apparatus further comprises:

a model training module configured to pre-train the behavioral data increment prediction model; wherein the model training module is specifically configured to:

According to some embodiments of the disclosure, the model training module is specifically configured to:

According to some embodiments of the disclosure, the third acquisition module is specifically configured to:

According to a fourth aspect of the embodiments of the present disclosure, there is provided an incremental predictive model generating apparatus for behavioral data, including:

The first acquisition module is configured to acquire first behavior data of a first sample user account corresponding to a first service and acquire second behavior data of a second sample user account corresponding to a second service; wherein, the first service is a new version service of the second service;

A second acquisition module configured to acquire user characteristics of the first sample user account and user characteristics of the second sample user account;

The first generation module is configured to train a first machine learning model according to the user characteristics of the second sample user account and the second behavior data to obtain model parameters, and generate a deviation correction model based on the model parameters;

the third acquisition module is configured to acquire the behavior data increment corresponding to the first sample user account according to the user characteristics of the first sample user account, the first behavior data and the deviation rectification model;

and the second generation module is configured to generate a behavior data increment prediction model corresponding to the first service according to the user characteristics of the first sample user account and the behavior data increment.

According to some embodiments of the disclosure, the second generation module is specifically configured to:

According to a fifth aspect of embodiments of the present disclosure, there is provided a server comprising:

A processor;

A memory for storing the processor-executable instructions;

wherein the processor is configured to execute the instructions to implement the service recommendation method according to the embodiment of the first aspect of the present disclosure, or implement the behavior data increment prediction model generation method according to the embodiment of the second aspect of the present disclosure.

According to a sixth aspect of embodiments of the present disclosure, there is provided a storage medium, which when executed by a processor of a server, enables the server to perform the service recommendation method according to the embodiment of the first aspect of the present disclosure, or to implement the behavior data increment prediction model generation method according to the embodiment of the second aspect of the present disclosure.

According to a seventh aspect of embodiments of the present disclosure, there is provided a computer program product comprising a computer program which, when executed by a processor, implements the service recommendation method according to the embodiments of the first aspect of the present disclosure, or implements the behavior data increment prediction model generation method according to the embodiments of the second aspect of the present disclosure.

The technical scheme provided by the embodiment of the disclosure at least brings the following beneficial effects:

The method comprises the steps of inputting the user behavior characteristics of each user account to be identified into a pre-trained behavior data increment prediction model corresponding to a first service, obtaining a behavior data increment prediction value of each user account to be identified, and obtaining target user accounts from a plurality of user accounts to be identified according to the behavior data increment prediction value of each user account to be identified, so that the target users are a group of users with preference for the first service, namely, the users with preference for the first service can be effectively screened out, and in this way, the first service is recommended to a client logged in the target user account, so that the service takes effect on a user group with preference for the services, and the recommendation effect can be improved.

It is to be understood that both the foregoing general description and the following detailed description are exemplary and explanatory only and are not restrictive of the disclosure.

Drawings

The accompanying drawings, which are incorporated in and constitute a part of this specification, illustrate embodiments consistent with the disclosure and together with the description, serve to explain the principles of the disclosure and do not constitute an undue limitation on the disclosure.

Fig. 1 is a flowchart illustrating a service recommendation method according to an exemplary embodiment.

FIG. 2 is a flowchart illustrating a method of training a behavioral data increment prediction model, according to an example embodiment.

FIG. 3 is a flowchart illustrating a method of incremental predictive model generation of behavioral data according to an example embodiment.

FIG. 4 is a flowchart illustrating yet another behavior data delta prediction model generation method, according to an example embodiment.

Fig. 5 is a flowchart illustrating an example of a business recommendation method according to an example embodiment.

Fig. 6 is an exemplary second flowchart illustrating a business recommendation method according to an exemplary embodiment.

Fig. 7 is a block diagram illustrating a service recommender in accordance with an exemplary embodiment.

FIG. 8 is a block diagram illustrating an incremental predictive model generation apparatus for behavioral data according to an example embodiment.

Fig. 9 is a block diagram of a server, according to an example embodiment.

Detailed Description

In order to enable those skilled in the art to better understand the technical solutions of the present disclosure, the technical solutions of the embodiments of the present disclosure will be clearly and completely described below with reference to the accompanying drawings.

It should be noted that the terms "first," "second," and the like in the description and claims of the present disclosure and in the foregoing figures are used for distinguishing between similar objects and not necessarily for describing a particular sequential or chronological order. It is to be understood that the data so used may be interchanged where appropriate such that the embodiments of the disclosure described herein may be capable of operation in sequences other than those illustrated or described herein. The implementations described in the following exemplary examples are not representative of all implementations consistent with the present disclosure. Rather, they are merely examples of apparatus and methods consistent with some aspects of the present disclosure as detailed in the accompanying claims.

Fig. 1 is a flowchart illustrating a service recommendation method according to an exemplary embodiment. The application scenario of the service recommendation method in the embodiment of the present disclosure may be: the short video server pushes a service to some users using the short video application to enable recommendation of the service to a group of people interested in the service. As shown in fig. 1, the method for recommending industry is used in the server according to the embodiment of the present disclosure, and includes the following steps.

In step S11, user characteristics of a plurality of user accounts to be identified are acquired.

In embodiments of the present disclosure, the user account may include, but is not limited to, any one or more of a user ID (Identity Document, identification number), a user name, a user nickname, and the like. A plurality of users to be identified can be understood as all users using the short video application. For example, the plurality of users to be identified may be understood as a full number of users of the short video application, for example, all users who download and install the short video application may be regarded as the users to be identified.

Optionally, a storage module is provided in the server for storing user behavior data of the short video application user account. Wherein user behavior data for each user account to be identified may be stored in the storage module. In this step, user behavior data of a plurality of user accounts to be identified may be obtained from the storage module, and user features of the user accounts to be identified may be extracted from the user behavior data of each user account to be identified.

In some embodiments of the present disclosure, the user behavior data may be understood as behavior data generated by a user when using a short video application. For example, the user behavior data may include behavior data generated when the user views a short video using a short video application, such as a name of the short video viewed by the user, a viewing duration or number of times, etc.; as another example, the user behavior data may also include behavior data generated when the user uses the short video application to make a short video work, such as a name, a duration, etc. of the short video work made by the user.

In some embodiments of the present disclosure, the user features described above may include, but are not limited to, one or more of user portraits, content consumption and content throughput, among others. Wherein the user profile may include, but is not limited to, one or more of the user's occupation, the user's gender, the user's age, the user's interests, etc. The content consumption amount refers to the number of short video contents in the short video application being watched, and/or the length of time the short video is watched using the short video application, and/or the number of times the short video is watched using the short video application, and the like. The content throughput refers to the number of short video products produced and/or the length of time that a short video work is produced using a short video application and/or the number of times a short video is produced using a short video application, etc.

In step S12, inputting the user characteristics of each user account to be identified into a pre-trained behavior data increment prediction model corresponding to the first service, and obtaining a behavior data increment prediction value of each user account to be identified; the behavior data increment prediction model is generated based on first behavior data of a first sample user account corresponding to a first service and second behavior data of a second sample user account corresponding to a second service, and the first service is a new version service of the second service.

In an embodiment of the disclosure, the behavior data increment prediction model has learned a mapping relationship between a behavior data increment of a user account corresponding to a first service and a user feature, where the behavior data increment of the user account corresponding to the first service refers to a behavior data increment generated by a user using a client logged into the user account due to the first service. Since the behavior data increment prediction model has already learned the mapping relationship, in practical application, when the short video server needs to recommend the service to the user using the short video application, by inputting the user characteristics of the full user accounts of the short video application into the behavior data increment prediction model, it can be predicted by the behavior data increment prediction model which target user accounts in the full user accounts will increase the corresponding behavior data due to the service, and the target user accounts can be considered to be interested in the recommended service, so that the service can be recommended to the client logging in the target user accounts.

In this step, the behavior data increment prediction value may be understood by assuming that, after recommending the first service to the client logging in the user account to be identified, it is predicted that the user account to be identified will have corresponding behavior data increased due to the first service. The behavior data increment prediction model can predict the behavior data increment of each user account to be identified by using the user characteristics, so that the behavior data increment predicted value of each user account to be identified can be output.

It should be noted that, due to different services, the corresponding incremental behavior data prediction model may also be different. Therefore, when determining to recommend a certain service to the client of the login target user account, training a behavior data increment prediction model corresponding to the service in advance, and identifying an audience user account of the service from the audience user accounts based on the corresponding behavior data increment prediction model, so as to recommend the service to the client of the login target user account.

In step S13, a target user account is obtained from a plurality of user accounts to be identified according to the behavior data increment predicted value of each user account to be identified.

That is, after obtaining the predicted value of the increment of the behavior data of the user account to be identified, it may be determined whether the predicted value of the increment of the behavior data of the user account to be identified meets a preset condition, and if so, the user accounts meeting the preset condition may be obtained from the plurality of user accounts to be identified, and the user accounts meeting the preset condition may be used as target user accounts. As an example, the satisfaction of the preset condition may include, but is not limited to: the behavior data increment predictor is greater than the target threshold, or the behavior data increment predictor is ranked top N, etc. Wherein, N can be a positive integer, and the value of N can be determined according to practical conditions.

In one embodiment of the disclosure, the behavior data increment predicted value of each user account to be identified can be compared with the target threshold value, the user accounts to be identified, of which the behavior data increment predicted value is greater than the target threshold value, are identified, and the user accounts to be identified, of which the behavior data increment predicted value is greater than the target threshold value, are determined to be target user accounts.

In another embodiment of the present disclosure, the behavior data increment predicted values of each user account to be identified may be ranked from large to small, and the user account to be identified corresponding to the behavior data increment predicted value of the previous N is determined as the target user account; or sequencing the plurality of user accounts to be identified according to the sequence of the behavior data increment predicted value of each user account to be identified from large to small, and determining the user account to be identified with the top N as a target user account.

In step S14, the first service is recommended to the client logging in the target user account.

Optionally, the first service is sent to the client of the login target user account, so that the client can perform corresponding service update after receiving the first service, thereby realizing the purpose of recommending the first service to the client of the login target user account, and realizing recommending the service to the crowd interested in the service.

According to the service recommendation method disclosed by the embodiment of the invention, the behavior data increment predicted value of each user account to be identified can be obtained by inputting the user behavior characteristics of each user account to be identified into the pre-trained behavior data increment predicted model corresponding to the first service, and the target user accounts are obtained from a plurality of user accounts to be identified according to the behavior data increment predicted value of each user account to be identified, so that the target users are a group of users who prefer the first service, namely, the users who prefer the first service can be effectively screened out, and therefore, the first service is recommended to the client side logging in the target user account, and the recommendation effect can be improved.

It should be noted that the incremental predictive model of behavioral data of embodiments of the present disclosure may be pre-trained. Since the corresponding incremental behavior data prediction model is different for different services, when determining that a first service is to be recommended to a client that logs in to a target user account, it is necessary to train the incremental behavior data prediction model corresponding to the first service in advance. In some embodiments of the present disclosure, as shown in fig. 2, the training method of the behavior data delta prediction model may include the following steps.

In step S21, first behavior data of a first sample user account corresponding to a first service is obtained, and second behavior data of a second sample user account corresponding to a second service is obtained; the first service is a new version service of the second service.

Wherein, in the embodiment of the present disclosure, the first behavior data of the first sample user account corresponding to the first service may be understood as behavior data generated by the first service of the user using the client logged in to the first sample user account. The second behavior data of the second sample user account corresponding to the second service may be understood as behavior data generated by the second service of the user using the client logged into the second sample user account.

In step S22, the first machine learning model is trained according to the user characteristics of the second sample user account and the second behavior data, model parameters are obtained, and a correction model is generated based on the model parameters.

In some embodiments of the present disclosure, the user features may include, but are not limited to, one or more of user portraits, content consumption amounts, and content throughputs, among others. Wherein the user profile may include, but is not limited to, one or more of the user's occupation, the user's gender, the user's age, the user's interests, etc. The content consumption amount refers to the number of short video contents in the short video application being watched, and/or the length of time the short video is watched using the short video application, and/or the number of times the short video is watched using the short video application, and the like. The content throughput refers to the number of short video products produced and/or the length of time that a short video work is produced using a short video application and/or the number of times a short video is produced using a short video application, etc.

In the embodiment of the present disclosure, the first machine learning model may be a bayesian optimization LightGBM model, or may also be other neural network algorithms or machine learning algorithms, which are not particularly limited in this disclosure. As an example, taking the first machine learning model as the bayesian optimization LightGBM model, the user characteristics of the second sample user account may be obtained, and the LightGBM model is trained by using the user characteristics of the second sample user account and the behavior data generated by the user using the client logged in to the second sample user account due to the second service, for example, the user characteristics of the second sample user account are taken as the input of the model, the behavior data generated by the user using the client logged in to the second sample user account is taken as the output of the model, and then the trained LightGBM model is taken as the correction model, so that the correction model has learned the mapping relationship between the user characteristics and the behavior data generated by the user using the client due to the second service.

In step S23, according to the user characteristics of the first sample user account, the first behavior data, and the correction model, the behavior data increment corresponding to the first sample user account is obtained.

Optionally, the user characteristics of the first sample user account are input into the deviation rectification model, the behavior data predicted value of the first sample user account corresponding to the second service is predicted, and the behavior data increment corresponding to the first sample user account is obtained according to the difference value between the first behavior data of the first sample user account corresponding to the first service and the behavior data predicted value.

That is, the user characteristics of the first sample user account are input into the deviation correction model for prediction, so as to obtain a behavior data predicted value of the first sample user account corresponding to the second service, and subtraction operation is performed on the first behavior data of the first sample user account corresponding to the first service and the behavior data predicted value, and the obtained difference value is the behavior data increment corresponding to the first sample user account.

In step S24, a behavior data increment prediction model corresponding to the first service is generated according to the user characteristics and the behavior data increment of the first sample user account.

Optionally, taking the user characteristics of the first sample user account as training data, taking the behavior data increment as a label corresponding to the training data, training the second machine learning model based on the training data and the label to obtain model parameters, and generating a behavior data increment prediction model corresponding to the first service based on the model parameters. Wherein, in the embodiment of the present disclosure, the second machine learning model may be a bayesian optimization LightGBM model, or may also be other neural network algorithms or machine learning algorithms, which is not specifically limited in the embodiment of the present disclosure.

As an example, taking the second machine learning model as the bayesian optimization LightGBM model, the LightGBM model may be understood as a correspondence model between user features and behavior data increments generated by a user using a client due to a first service, that is, taking the user features as input to the model and the behavior data increments as output to the model. At this time, the user characteristics of the first sample user account may be used as training data, the behavior data increment corresponding to the first sample user account is used as a tag corresponding to the training data (where the tag may be understood as a real value corresponding to the behavior data increment corresponding to the first sample user account), the user characteristics of the first sample user account are input to the LightGBM model to predict the behavior data increment, a predicted value is obtained, the predicted value is compared with the real value corresponding to the tag in a difference manner, and parameters of the LightGBM model are adjusted based on the difference degree, so as to realize training of the LightGBM model. And until training of the LightGBM model is completed, obtaining model parameters in the LightGBM model after training is completed, and taking a model formed by the model parameters as the behavior data increment prediction model, so that the behavior data increment prediction model learns to obtain a mapping relation between user characteristics and behavior data increment generated by a user using a client due to first service.

It will be appreciated that the incremental predictive model of behavioral data described above may be built using experimental data from an AB experiment, where the AB experiment may be understood as: and sending the first service to the client held by the experiment group user, and performing experiments under the condition that the first service is not sent to the client held by the comparison group user. The behavioral data delta predictive model may be established based on experimental data of the AB experiment. The manner in which the incremental predictive model of behavioral data is generated will be described in detail below in conjunction with FIG. 3.

For example, as shown in fig. 3, the behavior data increment prediction model generation method may include the following steps.

In step S31, the user characteristics of the pre-experiment reference group user account and the user characteristics of the experiment group user account are acquired.

In step S32, actual behavior data corresponding to the user account of the comparison group after the experiment and actual behavior data corresponding to the user account of the experiment group are obtained.

In step S33, training the first machine learning model according to the user characteristics of the user account of the comparison group and the actual behavior data corresponding to the user account of the comparison group after the experiment to obtain model parameters, and generating a correction model based on the model parameters.

Taking a first machine learning model as an example, taking a Bayesian optimization LightGBM model as an example, taking user characteristics of a comparison group user account as input of the model, taking actual behavior data corresponding to the comparison group user account after an experiment as output of the model, training the LightGBM model, and taking a LightGBM model which is completely trained as a correction model. For example, model parameters in the LightGBM models after training are acquired, and a LightGBM model formed by the model parameters is determined as the correction model.

In step S34, according to the user characteristics of the experimental group user account, the actual behavior data corresponding to the experimental group user account after the experiment, and the deviation correction model, the behavior data increment corresponding to the experimental group user account is obtained.

In some embodiments of the present disclosure, user characteristics of an experimental group user account may be input to a correction model to obtain a behavior data prediction value generated by a user using a client logging in the experimental group user account without using a first service; and acquiring the behavior data increment corresponding to the experimental group user account according to the difference value between the actual behavior data corresponding to the experimental group user account after the experiment and the behavior data predicted value.

In step S35, a behavior data increment prediction model corresponding to the first business is generated according to the user characteristics and the behavior data increment of the experimental group user account.

In some embodiments of the present disclosure, user features of the experimental group user account are used as training data, and the behavior data increment is used as a label corresponding to the training data; training the second machine learning model based on the training data and the labels to obtain model parameters, and generating a behavior data increment prediction model corresponding to the first service based on the model parameters.

As an example, taking the second machine learning model as the bayesian optimization LightGBM model, the LightGBM model may be understood as a correspondence model between user features and behavior data increments generated by a user using a client due to a first service, that is, taking the user features as input to the model and the behavior data increments as output to the model. At this time, the user characteristics of the user account of the experiment group can be used as training data, the behavior data increment corresponding to the user account of the experiment group is used as a tag corresponding to the training data (wherein the tag can be understood as a true value corresponding to the behavior data increment corresponding to the user account of the experiment group), the user characteristics of the user account of the experiment group are input into the LightGBM model to predict the behavior data increment, a predicted value is obtained, the predicted value is compared with the true value corresponding to the tag in a difference mode, and parameters of the LightGBM model are adjusted based on the difference degree so as to train the LightGBM model. And until training of the LightGBM model is completed, obtaining model parameters in the LightGBM model after training is completed, and taking the model formed by the model parameters as the behavior data increment prediction model.

According to the technical scheme of the embodiment of the disclosure, the behavior data increment prediction model is established by utilizing the data before and after the comparison group and the experiment group in the AB experiment, so that the potential preference service corresponding to each user can be predicted by utilizing the behavior data increment prediction model, and further the recommendation effect of different crowds on the recommended service can be effectively screened out according to the predicted potential preference service, so that crowds with preference to the service, namely the preference users of the service, can be effectively identified.

In order to implement the above embodiment, the present disclosure further proposes a behavior data increment prediction model generation method.

FIG. 4 is a flowchart illustrating a method of incremental predictive model generation of behavioral data according to an example embodiment. As shown in fig. 4, the behavior data delta prediction model generation method may include the following steps.

In step S41, first behavior data of a first sample user account corresponding to a first service is acquired, and second behavior data of a second sample user account corresponding to a second service is acquired; the first service is a new version service of the second service.

In the embodiment of the present disclosure, step S41 may be implemented in any manner of each embodiment of the present disclosure, which is not limited to this and is not described in detail.

In step S42, the user characteristics of the first sample user account and the user characteristics of the second sample user account are acquired.

In the embodiment of the present disclosure, step S42 may be implemented in any manner of each embodiment of the present disclosure, which is not limited to this embodiment, and is not described in detail.

In step S43, the first machine learning model is trained according to the user characteristics of the second sample user account and the second behavior data, model parameters are obtained, and a correction model is generated based on the model parameters.

In the embodiment of the present disclosure, step S43 may be implemented in any manner of each embodiment of the present disclosure, which is not limited to this embodiment, and is not described in detail.

In step S44, according to the user characteristics of the first sample user account, the first behavior data, and the correction model, the behavior data increment corresponding to the first sample user account is obtained.

In the embodiment of the disclosure, user characteristics of a first sample user account are input into a deviation rectification model, and a behavior data predicted value of the first sample user account corresponding to a second service is predicted; and acquiring the behavior data increment corresponding to the first sample user account according to the difference value between the first behavior data corresponding to the first service and the behavior data predicted value of the first sample user account.

In the embodiment of the present disclosure, step S44 may be implemented in any manner of each embodiment of the present disclosure, which is not limited to this embodiment, and is not described in detail.

In step S45, a behavior data increment prediction model corresponding to the first service is generated according to the user characteristics and the behavior data increment of the first sample user account.

In the embodiment of the disclosure, the user characteristics of the user account of the first sample are used as training data, and the behavior data increment is used as a label corresponding to the training data; training the second machine learning model based on the training data and the labels to obtain model parameters, and generating a behavior data increment prediction model corresponding to the first service based on the model parameters.

In the embodiment of the present disclosure, step S45 may be implemented in any manner of each embodiment of the present disclosure, which is not limited to this embodiment, and is not described in detail.

In some embodiments of the present disclosure, the first machine learning model and the second machine learning model are both bayesian optimization LightGBM models.

According to the technical scheme of the embodiment of the disclosure, a deviation correction model can be generated through the user characteristics of the second sample user account and the second behavior data of the second sample user account corresponding to the second service, and the user characteristics of the first sample user account, the first behavior data of the first sample user account corresponding to the first service and the deviation correction model are utilized to obtain the behavior data increment corresponding to the first sample user account, and further, the behavior data increment prediction model corresponding to the first service is generated according to the user characteristics of the first sample user account and the behavior data increment, so that the behavior data increment prediction model has learned the mapping relation between the user characteristics and the behavior data increment generated by the user using the client under the condition of using the first service, and therefore, the potential preference service corresponding to each user can be predicted by utilizing the behavior data increment prediction model, and further, the recommendation effect of different crowds on the recommended service can be effectively screened according to the predicted potential preference service, and therefore, the crowd with preference for the service can be effectively identified.

In order to enable those skilled in the art to more clearly understand the aspects of the present disclosure, two examples will be given below for description.

Taking a first service of "adding praise to the work published by the author" as an example, wherein the experimental group has praise more than the work of the control group, as shown in fig. 5, the service recommendation method may include the following steps:

In step S51, an AB experiment is performed on a sample user, the post-base-group-experiment work amount is predicted by the pre-base-group-experiment user behavior characteristics, the model is denoted as model1, that is, the post-base-group-experiment priori work amount is not recommended to the base group under the condition of the first service, and the model is a correction model of the exp group.

In step S52, the model1 is used to calculate the predicted work amount after the exp group experiment if the first business is not recommended to the exp group, and the actual work amount after the exp group experiment is subtracted from the exp group predicted work amount to calculate the work amount delta of the exp group due to the experiment.

In step S53, delta is predicted using the user behavior characteristics of the exp set of sample users, denoted model2. The model is an incremental predictive model of behavioral data corresponding to the first business given in example one. The delta is the work volume increment obtained in the step S52 and caused by the experiment, so that the modeling in the step can directly predict which users can generate the business volume increment for the experiment.

In step S54, all users using the short video application are predicted and ranked by the model2, and the user whose predicted value is high (for example, the predicted value is ranked in the top N) is the target user satisfying the preset condition.

In step S55, the first service is pushed to the client held by the target user.

Example two, take the first business "add' people who may be aware" next to the short video application discovery page user name. That is, the short video service provider wants to add "people who may know" next to the short video application discovery page user name, to know which users have the greatest content consumption, and to send the service to the clients held by that portion of the users. Assuming that the total content consumption amount of a user on a short video application is C (where C is an output value of a function of consumption indexes such as a duration of using the short video application, a duration of viewing the video, a duration of viewing the live broadcast, and a retention), if the total content consumption amount of the user after implementing the "person who may know" service is C _exp, the content consumption amount of the service to the person is Δc=c _exp -C. The short video service provider wants to find a user with Δc >0 and implements the service for a user with Δc >0 so that the increment of the experiment for each person is positive. Specifically, as shown in fig. 6, the service recommendation method may include the steps of:

In step S61, an AB experiment is performed on the sample user, and the correction model1 is calculated from the data before and after the base group experiment. The calculation method is to firstly model lightgbm by using user behavior characteristics (represented by X _base, such as including portrait, consumption, production and the like) before the base group experiment to predict the benefit C _{base_pred} after the base group experiment. Since the base group has no first service, C _{base_pred} can be considered a priori content consumption by the user.

In step S62, model1 is used to predict that the increment after exp-group sample user experiments is C _{exp_pred} by the user behavior feature X _exp before exp-group experiments, and Δc=c _{exp_actual}-C_{exp_pred} is then recorded as the experimental increment due to the addition of "people likely to be recognized" next to the found page user name. Wherein C _{exp_actual} is the actual delta generated by the experimental group sample user in the case of recommending the first service to the client held by the experimental group sample user.

In step S63, a Δc prediction model needs to be established. The model predictive deltac is built lightgbm by using the user behavior characteristics before exp group experiments, and the model is marked as model2 and is a behavior data increment predictive model corresponding to the first service given in the example two.

In step S64, all users using the short video application are predicted by the model2, ΔC of each user is predicted, and a user with ΔC >0 is found, and the user with ΔC >0 is taken as a target user satisfying a preset condition.

In step S65, the first service is recommended to the client where the target user is located. I.e. to implement a "possibly recognized people" service for the target user obtained in step S64, such expected increments would be larger than the overall increment obtained by recommending the first service for the full number of users.

Fig. 7 is a block diagram illustrating a service recommender in accordance with an exemplary embodiment. Referring to fig. 7, the service recommendation device 700 includes: a first acquisition module 701, a second acquisition module 702, a third acquisition module 703, and a recommendation module 704.

Wherein the first obtaining module 701 is configured to obtain user characteristics of a plurality of user accounts to be identified.

The second obtaining module 702 is configured to input the user characteristics of each user account to be identified into a pre-trained behavior data increment prediction model corresponding to the first service, and obtain a behavior data increment predicted value of each user account to be identified; the behavior data increment prediction model is generated based on first behavior data of a first sample user account corresponding to a first service and second behavior data of a second sample user account corresponding to a second service, wherein the first service is a new version service of the second service.

The third obtaining module 703 is configured to obtain a target user account from a plurality of user accounts to be identified according to the behavior data increment prediction value of each user account to be identified.

In some embodiments of the present disclosure, the third obtaining module 703 compares the behavior data increment predicted value of each user account to be identified with a target threshold; identifying a user account to be identified, of which the behavior data increment predicted value is greater than a target threshold value, from a plurality of user accounts to be identified; and determining the user account to be identified, of which the behavior data increment predicted value is larger than the target threshold value, as the target user account.

In other embodiments of the present disclosure, the third obtaining module 703 sorts the behavior data increment predicted values of each user account to be identified according to the order from large to small, and determines the user account to be identified corresponding to the behavior data increment predicted value of the previous N row as the target user account; or sequencing the plurality of user accounts to be identified according to the sequence of the behavior data increment predicted value of each user account to be identified from large to small, and determining the user account to be identified with the top N as a target user account.

The recommendation module 704 is configured to recommend the first service to a client logging into the target user account.

In some embodiments of the present disclosure, the service recommendation device may further include: and a model training module. The model training module is configured to pre-train the behavioral data delta predictive model. Wherein, in the embodiments of the present disclosure, the model training module is specifically configured to: acquiring first behavior data of a first sample user account corresponding to a first service, and acquiring second behavior data of a second sample user account corresponding to a second service; the first service is a new version service of the second service; training a first machine learning model according to the user characteristics of the second sample user account and the second behavior data to obtain model parameters, and generating a correction model based on the model parameters; acquiring a behavior data increment corresponding to the first sample user account according to the user characteristics of the first sample user account, the first behavior data and the deviation rectification model; and generating a behavior data increment prediction model corresponding to the first business according to the user characteristics and the behavior data increment of the first sample user account.

In the embodiment of the present disclosure, the specific implementation process of the model training module obtaining, according to the user characteristics of the first sample user account, the first behavior data, and the correction model, the behavior data increment of the first sample user account corresponding to the first service may be as follows: inputting the user characteristics of the user account of the first sample into the deviation rectification model, and predicting the behavior data predicted value of the user account of the first sample corresponding to the second service; and acquiring the behavior data increment corresponding to the first sample user account according to the difference value between the first behavior data corresponding to the first service and the behavior data predicted value of the first sample user account.

In the embodiment of the present disclosure, the specific implementation process of the model training module for generating the behavior data increment prediction model corresponding to the first service according to the user characteristics and the behavior data increment of the first sample user account may be as follows: taking the user characteristics of the user account of the first sample as training data, and taking the behavior data increment as a label corresponding to the training data; training the second machine learning model based on the training data and the labels to obtain model parameters, and generating a behavior data increment prediction model corresponding to the first service based on the model parameters. Wherein, as an example, the first machine learning model and the second machine learning model are both bayesian optimization LightGBM models.

The specific manner in which the various modules perform the operations in the apparatus of the above embodiments have been described in detail in connection with the embodiments of the method, and will not be described in detail herein.

According to the service recommending device disclosed by the embodiment of the disclosure, the behavior data increment predicted value of each user account to be identified can be obtained by inputting the user behavior characteristic of each user account to be identified into the pre-trained behavior data increment predicted model corresponding to the first service, and the target user accounts are obtained from a plurality of user accounts to be identified according to the behavior data increment predicted value of each user account to be identified, so that the target users are a group of users capable of improving certain service volume due to the first service, namely, the users capable of improving certain service volume due to the first service can be effectively screened out, and in this way, the first service is recommended to the client side logging in the target user accounts, so that the service takes effect on the user groups capable of improving certain service volume, and the total service volume of the service can be improved, and the recommending effect can be improved.

FIG. 8 is a block diagram illustrating an incremental predictive model generation apparatus for behavioral data according to an example embodiment. As shown in fig. 8, the behavior data increment prediction model generation apparatus 800 may include: a first acquisition module 801, a second acquisition module 802, a first generation module 803, a third acquisition module 804, and a second generation module 805.

Wherein the first obtaining module 801 is configured to obtain first behavior data of a first sample user account corresponding to a first service, and obtain second behavior data of a second sample user account corresponding to a second service; the first service is a new version service of the second service.

The second acquisition module 802 is configured to acquire user characteristics of the first sample user account and user characteristics of the second sample user account.

The first generation module 803 is configured to train the first machine learning model according to the user characteristics of the second sample user account and the second behavior data, obtain model parameters, and generate a correction model based on the model parameters.

The third obtaining module 804 is configured to obtain the behavior data increment corresponding to the first sample user account according to the user feature of the first sample user account, the first behavior data, and the deskew model.

In some embodiments of the present disclosure, the third acquisition module 804 is specifically configured to: inputting the user characteristics of the user account of the first sample into the deviation rectification model, and predicting the behavior data predicted value of the user account of the first sample corresponding to the second service; and acquiring the behavior data increment corresponding to the first sample user account according to the difference value between the first behavior data corresponding to the first service and the behavior data predicted value of the first sample user account.

The second generation module 805 is configured to generate a behavioral data increment prediction model corresponding to the first business based on the user characteristics and behavioral data increments of the first sample user account.

In some embodiments of the present disclosure, the second generation module 805 is specifically configured to: taking the user characteristics of the user account of the first sample as training data, and taking the behavior data increment as a label corresponding to the training data; training the second machine learning model based on the training data and the labels to obtain model parameters, and generating a behavior data increment prediction model corresponding to the first service based on the model parameters. As one example, the first machine learning model and the second machine learning model are both bayesian optimization LightGBM models.

According to the technical scheme of the embodiment of the disclosure, the deviation correction model can be generated through the user characteristics of the second sample user account and the second behavior data of the second sample user account corresponding to the second service, and the behavior data increment corresponding to the first sample user account is obtained by utilizing the user characteristics of the first sample user account, the first behavior data of the first sample user account corresponding to the first service and the deviation correction model, so that the behavior data increment prediction model corresponding to the first service is generated according to the user characteristics of the first sample user account and the behavior data increment, and the mapping relation between the user characteristics and the behavior data increment generated by the user using the client under the condition of using the first service is already learned by the behavior data increment prediction model, so that the potential service volume corresponding to each user can be predicted by utilizing the behavior data increment prediction model, and further, the service volume prediction conditions of different crowds for the recommended service can be effectively distinguished according to the predicted potential service volume, and thus the audience users sensitive to the service can be effectively identified.

Fig. 9 is a block diagram of a server 200, according to an example embodiment. As shown in fig. 9, the server 200 may include:

a memory 210 and a processor 220, a bus 230 connecting the different components (including the memory 210 and the processor 220), the memory 210 storing processor 220 executable instructions; wherein the processor 220 is configured to execute the instructions to implement the business recommendation method or the behavior data increment prediction model generation method according to the embodiments of the present disclosure.

Bus 230 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, and not limitation, such architectures include Industry Standard Architecture (ISA) bus, micro channel architecture (MAC) bus, enhanced ISA bus, video Electronics Standards Association (VESA) local bus, and Peripheral Component Interconnect (PCI) bus.

Server 200 typically includes a variety of electronic device readable media. Such media can be any available media that is accessible by server 200 and includes both volatile and nonvolatile media, removable and non-removable media. Memory 210 may also include computer system readable media in the form of volatile memory, such as Random Access Memory (RAM) 240 and/or cache memory 250. Server 200 may further include other removable/non-removable, volatile/nonvolatile computer system storage media. By way of example only, storage system 260 may be used to read from or write to non-removable, nonvolatile magnetic media (not shown in FIG. 9, commonly referred to as a "hard disk drive"). Although not shown in fig. 9, a magnetic disk drive for reading from and writing to a removable non-volatile magnetic disk (e.g., a "floppy disk"), and an optical disk drive for reading from or writing to a removable non-volatile optical disk (e.g., a CD-ROM, DVD-ROM, or other optical media) may be provided. In such cases, each drive may be coupled to bus 230 via one or more data medium interfaces. Memory 210 may include at least one program product having a set (e.g., at least one) of program modules configured to carry out the functions of the various embodiments of the disclosure.

Program/utility 280 having a set (at least one) of program modules 270 may be stored in, for example, memory 210, such program modules 270 including, but not limited to, an operating system, one or more application programs, other program modules, and program data, each or some combination of which may include an implementation of a network environment. Program modules 270 generally perform the functions and/or methods in the embodiments described in this disclosure.

The server 200 may also communicate with one or more external devices 290 (e.g., keyboard, pointing device, display 291, etc.), one or more devices that enable a user to interact with the server 200, and/or any device (e.g., network card, modem, etc.) that enables the server 200 to communicate with one or more other computing devices. Such communication may occur through an input/output (I/O) interface 292. Also, server 200 may communicate with one or more networks such as a Local Area Network (LAN), a Wide Area Network (WAN) and/or a public network, such as the Internet, via network adapter 293. As shown, network adapter 293 communicates with other modules of server 200 over bus 230. It should be appreciated that although not shown, other hardware and/or software modules may be used in connection with server 200, including, but not limited to: microcode, device drivers, redundant processing units, external disk drive arrays, RAID systems, tape drives, data backup storage systems, and the like.

The processor 220 executes various functional applications and data processing by running programs stored in the memory 210.

It should be noted that, the implementation process and the technical principle of the server in this embodiment refer to the foregoing explanation of the service recommendation method or the behavior data increment prediction model generation method in the embodiments of the present disclosure, and are not repeated herein.

In order to implement the above-described embodiments, the present disclosure also proposes a storage medium.

Wherein the instructions in the storage medium, when executed by the processor of the server, enable the server to perform the business recommendation method or the behavior data increment prediction model generation method as described above.

To achieve the above embodiments, the present disclosure also provides a computer program product. The computer program product comprises a computer program which, when executed by a processor of a server, implements a business recommendation method or a behavior data increment prediction model generation method as described above.

Other embodiments of the disclosure will be apparent to those skilled in the art from consideration of the specification and practice of the disclosure disclosed herein. This application is intended to cover any adaptations, uses, or adaptations of the disclosure following, in general, the principles of the disclosure and including such departures from the present disclosure as come within known or customary practice within the art to which the disclosure pertains. It is intended that the specification and examples be considered as exemplary only, with a true scope and spirit of the disclosure being indicated by the following claims.

It is to be understood that the present disclosure is not limited to the precise arrangements and instrumentalities shown in the drawings, and that various modifications and changes may be effected without departing from the scope thereof. The scope of the present disclosure is limited only by the appended claims.

Claims

1. A business recommendation method, comprising:

Recommending the first service to a client for logging in the target user account;

The behavior data increment prediction model is obtained through training by the following steps:

acquiring first behavior data of a first sample user account corresponding to a first service, and acquiring second behavior data of a second sample user account corresponding to a second service;

2. The service recommendation method according to claim 1, wherein the obtaining the behavior data increment corresponding to the first sample user account according to the user feature of the first sample user account, the first behavior data, and the correction model includes:

3. The business recommendation method according to claim 1, wherein the generating a behavior data increment prediction model corresponding to the first business according to the user characteristics of the first sample user account and the behavior data increment comprises:

4. The business recommendation method of claim 3, wherein the first machine learning model and the second machine learning model are bayesian optimization LightGBM models.

5. The method according to any one of claims 1 to 4, wherein the obtaining a target user account from the plurality of user accounts to be identified according to the behavior data increment predicted value of each user account to be identified includes:

6. The method according to any one of claims 1 to 4, wherein the obtaining a target user account from the plurality of user accounts to be identified according to the behavior data increment predicted value of each user account to be identified includes:

7. A method for generating an incremental predictive model of behavioral data, comprising:

8. The method for generating a predictive model for incremental behavior data according to claim 7, wherein the obtaining the incremental behavior data corresponding to the first sample user account according to the user characteristics of the first sample user account, the first behavior data, and the correction model comprises:

9. The method for generating an incremental predictive model of behavioral data according to claim 7, wherein said generating an incremental predictive model of behavioral data corresponding to said first business based on said user characteristics of said first sample user account and said incremental behavioral data comprises:

10. The method of claim 9, wherein the first machine learning model and the second machine learning model are bayesian optimized LightGBM models.

11. A service recommendation device, comprising:

A recommending module configured to recommend the first service to a client logging in the target user account;

Wherein the apparatus further comprises:

a model training module configured to pre-train the behavioral data increment prediction model; wherein,

The model training module is specifically configured to:

12. The business recommendation device of claim 11, wherein the model training module is specifically configured to:

13. The business recommendation device of claim 11, wherein the model training module is specifically configured to:

14. The traffic recommendation device of claim 13, wherein the first machine learning model and the second machine learning model are bayesian optimization LightGBM models.

15. The service recommendation device according to any one of claims 11 to 14, wherein the third acquisition module is specifically configured to:

16. The service recommendation device according to any one of claims 11 to 14, wherein the third acquisition module is specifically configured to:

17. An incremental predictive model generation apparatus for behavioral data, comprising:

18. The behavioral data increment prediction model generation apparatus of claim 17, wherein the third acquisition module is specifically configured to:

19. The behavioral data increment prediction model generation apparatus of claim 17, wherein the second generation module is specifically configured to:

20. The behavioral data increment prediction model generation apparatus of claim 19 wherein the first and second machine learning models are bayesian optimization LightGBM models.

21. A server, comprising:

A processor;

A memory for storing the processor-executable instructions;

wherein the processor is configured to execute the instructions to implement the business recommendation method of any one of claims 1 to 6 or to implement the behavioral data increment prediction model generation method of any one of claims 7 to 10.

22. A storage medium, which when executed by a processor of a server, causes the server to perform the business recommendation method of any one of claims 1 to 6, or to implement the behavioral data increment prediction model generation method of any one of claims 7 to 10.