CN114223012A

CN114223012A - Push object determination method and device, terminal equipment and storage medium

Info

Publication number: CN114223012A
Application number: CN201980099270.3A
Authority: CN
Inventors: 郭子亮
Original assignee: Guangdong Oppo Mobile Telecommunications Corp Ltd; Shenzhen Huantai Technology Co Ltd
Current assignee: Guangdong Oppo Mobile Telecommunications Corp Ltd; Shenzhen Huantai Technology Co Ltd
Priority date: 2019-10-31
Filing date: 2019-10-31
Publication date: 2022-03-22
Also published as: WO2021081914A1

Abstract

The embodiment of the application discloses a pushed object determining method, a pushed object determining device, terminal equipment and a storage medium, and relates to the technical field of computers. The method comprises the following steps: acquiring user characteristics of a plurality of users; inputting the user characteristics of the users into a pre-trained classification model to obtain a pushing probability corresponding to each user, wherein the classification model is constructed based on the user characteristics, and the pushing probability is used for representing the probability that the user is a pushing object; and determining a pushing object from the plurality of users according to the pushing probability. The method and the device can effectively determine a certain number of users needing to be pushed from a large number of user groups to serve as the pushing target user group, so that the number of the users pushing the target user group is better expanded, the coverage of the effective pushing users in the user groups is increased, and the pushing efficiency is improved.

Description

Push object determination method and device, terminal equipment and storage medium

Technical Field

The present application relates to the field of computer technologies, and in particular, to a method and an apparatus for determining a pushed object, a terminal device, and a storage medium.

Background

With the rapid development of the internet, internet users have reached hundreds of millions of scales, and how to select effective push objects from a plurality of internet users obviously becomes a hot direction of current research.

Disclosure of Invention

In view of the above problems, the present application provides a method, an apparatus, a terminal device and a storage medium for determining a push object, so as to solve the above problems.

In a first aspect, an embodiment of the present application provides a method for determining a pushed object, where the method includes: acquiring user characteristics of a plurality of users; inputting the user characteristics of a plurality of users into a pre-trained classification model to obtain the pushing probability corresponding to each user, wherein the classification model is constructed based on the user characteristics, and the pushing probability is used for representing the probability that the user is a pushing object; and determining a push object from a plurality of users according to the push probability.

In a second aspect, an embodiment of the present application provides a push object determining apparatus, where the apparatus includes: the system comprises a user characteristic acquisition module, a push probability acquisition module and a push object selection module, wherein the user characteristic acquisition module is used for acquiring user characteristics of a plurality of users; the push probability acquisition module is used for inputting the user characteristics of a plurality of users into a pre-trained classification model to obtain the push probability corresponding to each user, the classification model is constructed based on the user portrait, and the push probability is used for representing the probability that the user is a push object; the push object selection module is used for determining a push object from a plurality of users according to the push probability.

In a third aspect, an embodiment of the present application provides a terminal device, including one or more processors, a memory, and one or more applications, where the one or more applications are stored in the memory and configured to be executed by the one or more processors, and the one or more applications are configured to execute the above-mentioned push object determination method.

In a fourth aspect, an embodiment of the present application provides a computer-readable storage medium, where a program code is stored in the computer-readable storage medium, and the program code may be called by a processor to execute the above pushed object determining method.

According to the pushed object determining method, the pushed object determining device, the terminal device and the storage medium, the classification model is built based on the user characteristics, when the user characteristics of the user are input into the classification model, the probability of whether the user is the user needing to be pushed can be rapidly and accurately output, a certain number of users needing to be pushed can be effectively determined from a large number of levels of user groups to serve as a pushing target user group based on the probability of the user needing to be pushed, the number of the users pushing the target user group is further well expanded, the coverage of the users effectively pushing the user in the user groups is increased, and the pushing efficiency is improved.

Drawings

In order to more clearly illustrate the technical solutions in the embodiments of the present application, the drawings needed to be used in the description of the embodiments are briefly introduced below, and it is obvious that the drawings in the following description are only some embodiments of the present application, and it is obvious for those skilled in the art to obtain other drawings based on these drawings without creative efforts.

Fig. 1 shows a flowchart of a push object determination method according to an embodiment of the present application.

Fig. 2 is a flowchart illustrating a push object determining method according to another embodiment of the present application.

Fig. 3 shows a flowchart of a method according to an embodiment of step S230 in the method for determining a push object shown in fig. 2 of the present application.

FIG. 4 is a flow chart of a pushed object determination method according to another embodiment of the present application

Fig. 5 is a flowchart of a method according to an embodiment of step S340 in the method for determining a push object shown in fig. 4.

Fig. 6 is a flowchart illustrating a method according to an embodiment of step S341 in the method for determining a push object illustrated in fig. 5.

Fig. 7 is a functional block diagram of a push object determining apparatus according to an embodiment of the present application.

Fig. 8 shows a block diagram of a terminal device according to an embodiment of the present application.

Fig. 9 is a storage medium storing or carrying program code implementing a push object determining method according to an embodiment of the present application.

Detailed Description

In order to make the technical solutions better understood by those skilled in the art, the technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the drawings in the embodiments of the present application.

With the rapid development of the internet, users of the internet are increasingly proliferated, and for a huge internet user group, how to select an effective push object from the large internet user group becomes crucial.

At present, machine learning algorithms are widely applied to various fields, and user pushing is no exception. The inventor finds that because the portrait characteristics of each user in the user group have certain similarity, a small-magnitude suitable target user group is obtained through a rule method, model training can be performed on the user group through a machine learning algorithm, a corresponding classification model is established, the classification model is applied to an unknown large-scale user group, a large-magnitude user group similar to the small-magnitude user group is mined, and then the pushing efficiency can be effectively improved by pushing the user groups.

However, the inventor of the present invention has found that, through research, the recommendation method based on user images has a single rule, and although some users can be found to push according to the similarity of user image features, the features of each user have respective features, so that only a small number of users can be found as objects to be pushed. The product which needs to be popularized in a large range on the new market cannot play a good popularization effect and recommendation role. Therefore, the inventor provides the push object determining method, the push object determining device, the terminal device and the storage medium, which are provided by the embodiment of the present application, so that a certain number of users needing to be pushed can be effectively determined from a large number of user groups to serve as a push target user group, the number of users pushing the target user group is further expanded, the coverage of the users effectively pushing the user group is increased, and the push efficiency is improved.

The following describes in detail a push object determination method, a push object determination apparatus, a mobile terminal, and a storage medium according to embodiments of the present application.

Referring to fig. 1, fig. 1 is a flowchart illustrating a method for determining a pushed object according to an embodiment of the present application, where the method may include the following steps:

s110, user characteristics of a plurality of users are obtained.

The user feature source may also be referred to as user profile feature, where the data is related to the user. In some embodiments, the user features may be obtained from a series of data such as browsing information, purchasing information, collecting information, clicking information, searching information, evaluating information, etc. of the logged-in user on a website or a platform, or obtained from identity information disclosed by the user, and registration information filled in by the website or an e-commerce platform.

And S120, inputting the user characteristics of a plurality of users into a pre-trained classification model to obtain the pushing probability corresponding to each user, wherein the classification model is constructed based on the user characteristics, and the pushing probability is used for representing the probability that the user is a pushing object.

The pre-trained classification model can be applied to a large-magnitude user group, and the classification model is constructed in advance based on user characteristics, so that when the user characteristics of a plurality of users in the large-magnitude user group are input into the pre-trained classification model, the probability that the user corresponding to each input characteristic is a push object is calculated by the classification model.

It should be noted that the user may refer to a service object in the business logic, and the user may be represented by a user identifier. Taking an instant messaging program as an example, each user is represented by an instant messaging account; taking some applications on the mobile phone as an example, each user is represented by a mobile phone number. In the present application, users, user accounts, and people can be regarded as the same concept, and a user group can be regarded as an account set.

S130, determining a push object from a plurality of users according to the push probability.

The user with the higher push probability can be selected to be determined as the push object, and the user with the lower push probability is ignored, so that the push accuracy is improved. The determined push objects are also called diffusion crowds predicted from a plurality of users according to the classification model, and each diffusion crowd comprises a plurality of users.

In some embodiments, the determining of the push object from the plurality of users according to the push probability may be that a preset number of users are selected from the plurality of users according to a descending order of the push probability, and then the preset number of users are determined as the push object. As an example, if 500 multiple users need to be selected from a large-scale user group for pushing, the multiple pushing probabilities obtained in S120 may be arranged into a list in descending order, then the first 500 ten thousand pushing probabilities in the list are selected, and the 500 multiple users corresponding to the first 500 ten thousand pushing probabilities are determined as the pushing objects.

In the embodiment, the users with the preset number are selected from the plurality of users according to the sequence of the pushing probability from large to small, so that the higher probability of the selected users with the preset number is the pushing object, and the pushing accuracy of the users is improved.

In other embodiments, the push object is determined from a plurality of users according to a push probability, and at least one target push probability is selected from the plurality of push probabilities, where the target push probability is greater than or equal to a probability threshold; and determining a pushing object for the user corresponding to the target pushing probability. As an example, if the probability threshold is 70%, it may be determined whether each of the push probabilities obtained in S120 is greater than or equal to 70%, determine a push probability greater than or equal to 70% as a target push probability, and then determine a user corresponding to the target push probability as a push target.

In the embodiment, by setting the probability threshold, the user with the push probability greater than or equal to the probability threshold is determined as the push object, and the user with the higher push probability can be effectively selected from any number of user groups as the push object, so that the push accuracy is ensured, and the push method and the device have wide applicability.

In the embodiment, the classification model is constructed based on the user characteristics, when the user characteristics of the user are input into the classification model, the probability of whether the user is the user needing to be pushed can be rapidly and accurately output, a certain number of users needing to be pushed can be effectively determined from a large number of user groups based on the probability of the user needing to be pushed to serve as a pushing target user group, the number of the users pushing the target user group is further better expanded, the coverage of the effective pushing users in the user groups is increased, and the pushing efficiency is improved.

Referring to fig. 2, fig. 2 is a flowchart illustrating a method for determining a pushed object according to another embodiment of the present application, where the method may include the following steps:

s210, obtaining user characteristics of a plurality of sample users in the sample user group.

Wherein, each sample user group comprises a plurality of sample users. Each sample user population may also be roughly considered a group of users having the same or similar characteristics. As an example, the sample user population may be first hand user data provided by an advertising publisher.

In some implementations, the user characteristics can include at least one of user attribute characteristics, user behavior characteristics, and user interest characteristics.

Compared with the conventional pushing party which is usually based on unilateral user portrait and also called user characteristic, different types of content pushing can be carried out based on different user portraits, and the relation among different portraits of a user can not be comprehensively considered.

The user attribute feature may also be referred to as a user basic feature, and may be used to indicate a feature on a basic attribute of a character.

In one embodiment, when the user characteristic is a user attribute characteristic, the user characteristic of the sample user is obtained by obtaining attribute information of the sample user, wherein the attribute information includes at least one of gender, birth date, occupation and education background; and determining the user attribute characteristics of the sample user according to the attribute information. When the user attribute characteristics of the sample user are determined according to the attribute information, the attribute information of the sample user can be represented by a vector to obtain the user attribute characteristics of the sample user.

Optionally, the attribute information may also include address, income, marital status, and the like. Optionally, when obtaining the attribute information of the sample user, the attribute information may be obtained from some websites or platforms, for example, many websites or e-commerce platforms currently require to fill some public identity information when the user registers, for example, a microblog, the user needs to fill in a birth date, a gender, a province, and the like, and these registered information may be used as the attribute information of the user.

In the embodiment, the attribute characteristics of the sample user can be comprehensively and accurately represented by determining the attribute characteristics of the sample user according to the attribute information such as gender, birth date, occupation, education background and the like.

The user behavior characteristics can be used for indicating characteristics generated by various behaviors of the user when the user uses the internet service.

In one embodiment, when the user characteristics include user behavior characteristics, the user characteristics of the sample user are obtained, and an operation log of the sample user within a preset time length is obtained, wherein the operation log records operation behaviors generated when the user uses a network, and the operation behaviors include at least one of purchasing behaviors, accessing behaviors and browsing behaviors; and determining the user behavior characteristics of the sample user according to the operation log. Illustratively, the record of the purchase behavior of the sample user in one month can be obtained from the e-commerce platform, the record of the browsing behavior of the sample user in one month can be obtained, and the record of the access behavior of the sample user in one month can also be obtained. When the user behavior characteristics of the sample user are determined according to the operation log, the operation behavior of the sample user can be represented by a vector to obtain the user behavior characteristics of the sample user. Optionally, the preset duration may be determined according to actual conditions, and since the user behavior characteristics are not as obvious as the user attribute characteristics and require a certain time to be found, generally, the longer the preset duration is, the more accurately the user behavior characteristics of the user can be reflected.

In the embodiment, the behavior characteristics of the sample user are determined according to the operation behaviors such as the purchasing behavior, the accessing behavior and the browsing behavior, so that the user behavior characteristics of the sample user can be comprehensively and accurately embodied.

In another embodiment, when the user characteristics are user interest characteristics, the user characteristics of the sample user are obtained, and social records of the sample user can be obtained, wherein the social records comprise at least one of search records, comment records and attention records; and determining user interest characteristics according to the social records. When the user interest characteristics of the sample user are determined according to the social records, the social records of the sample user can be represented by the vectors to obtain the user interest characteristics of the sample user.

In the embodiment, the user interest characteristics of the sample user can be comprehensively and accurately embodied by determining the user interest characteristics of the sample user according to the records such as the search record, the comment record and the attention record.

And S220, generating a feature matrix of the sample user group based on the user features.

In some embodiments, generating the feature matrix of the sample user group based on the user features may be to perform feature processing on the user features to obtain a feature vector of each sample user. And forming a feature matrix based on the feature vector of each sample user.

Optionally, the feature processing on the user feature may include at least one of discretization processing, regularization processing and normalization processing, or a combination thereof.

The discretization process may be to divide continuous user features into intervals, where each interval is a new feature. Specifically, the features may be sorted first and then discretized into N intervals according to the equal frequency. Because the importance of the continuous values of the features in different intervals is different, discretization of the continuous features can better reflect that the features have different weights in different intervals, and the discrete features are easy to increase and decrease, so that the model can be more stable due to rapid iteration of the model during model training. The risk of model overfitting can also be reduced after feature discretization.

The standardization treatment and the normalization treatment are used for eliminating the influence of different characteristics on different index dimensions, and the comparability between indexes can be facilitated after the standardization treatment.

In the present embodiment, discretization, normalization, and normalization processing of user features can be performed to facilitate better use of the processed user features by the machine learning model.

And S230, training according to the feature matrix to obtain a pre-trained classification model.

In some embodiments, as shown in fig. 3, S230 may include:

and S231, extracting a positive sample from the sample user group based on the characteristic matrix, and determining the sample user group with the positive sample extracted as a residual sample.

In some embodiments, the positive sample is extracted from the sample user group based on the feature matrix, and the positive sample is extracted from the sample user group according to the feature matrix and a preset positive sample feature identifier, and the sample user group from which the positive sample is extracted is determined as the remaining sample.

The user characteristics of the users can be marked, so that a preset positive sample characteristic mark can be obtained, and the positive sample can be extracted from the sample user group through the preset positive sample characteristic mark and the characteristic matrix because the characteristic matrix describes the whole sample user group. In the embodiment, the positive sample can be quickly and accurately extracted from the sample user group through the feature matrix and the preset positive sample feature identification.

And S232, extracting a negative sample from the residual samples.

The remaining samples may refer to users in the sample user group who do not have the preset positive sample feature identifier.

In some embodiments, the negative sample may be extracted from the remaining samples by selecting the negative sample from the remaining samples based on a positive unlabeled learning (PU learning) machine learning algorithm and user characteristics of the positive sample.

The classification learning is usually supervised learning, that is, learning rules from positive and negative samples determined to predict new data. However, in an actual application scenario, data samples may not be abundant due to problems of insufficient data accumulation, high cost for labeling data, and the like. While the positive sample label-free learning is a semi-supervised learning algorithm, which can be mainly used for solving the problem that the positive sample can be clearly determined but the negative sample cannot be determined, and is generally applied to text classification, material sample classification and the like. In the embodiment, the positive sample label-free learning machine learning algorithm can be used for finding out a reliable negative sample set in an unlabeled sample set according to the labeled positive sample, the algorithm is applied to sample classification of users, and a large number of unknown users are classified, so that effective negative samples can be obtained, and the effect of a subsequent classification model is improved.

In other embodiments, the negative examples may also be extracted from the remaining examples by labeling the negative example feature identifiers in advance and then extracting the negative examples based on the negative example feature identifiers and the feature matrix. Specifically, the user needing to be used as the negative sample can be determined in advance, wherein the user characteristics with the larger user characteristic difference of the positive sample can be selected to be used as the user characteristics of the negative sample, and then the user characteristics of the negative sample are marked, so that the preset negative sample characteristic identification can be obtained. In the embodiment, the negative sample can be quickly and accurately extracted from the sample user group through the feature matrix and the preset negative sample feature identification.

And S233, inputting the positive sample and the negative sample into the machine learning model for training to obtain a pre-trained classification model.

In some embodiments, the positive sample and the negative sample are input into the machine learning model for training, and the pre-trained classification model is obtained by processing the positive sample and the negative sample through a Logistic Regression (LR) machine learning classification algorithm or a Random Forest (RF) machine learning classification algorithm.

In the embodiment, the pre-trained classification model is obtained by processing the positive sample and the negative sample through a logistic regression machine learning classification algorithm or processing the positive sample and the negative sample through a random forest machine learning classification algorithm, so that the prediction precision of the classification model can be improved on the premise that the calculation amount is not obviously improved, and the preset result is relatively stable even for missing and unbalanced data.

S240, obtaining user characteristics of a plurality of users.

The specific implementation of S240 may refer to S110, and therefore is not described herein.

And S250, inputting the user characteristics of a plurality of users into a pre-trained classification model to obtain the pushing probability corresponding to each user, wherein the classification model is constructed based on the user characteristics, and the pushing probability is used for representing the probability that the user is a pushing object.

The specific implementation of S250 can refer to S120, and therefore is not described herein.

And S260, determining a pushing object from the plurality of users according to the pushing probability.

The specific implementation of S240 may refer to S130, and therefore is not described herein.

Referring to fig. 4, fig. 4 is a flowchart illustrating a method for determining a pushed object according to another embodiment of the present application, where the method may include the following steps:

s310, user characteristics of a plurality of users are obtained.

The specific implementation of S310 can refer to S110, and therefore is not described herein.

And S320, inputting the user characteristics of a plurality of users into a pre-trained classification model to obtain the pushing probability corresponding to each user, wherein the classification model is constructed based on the user characteristics, and the pushing probability is used for representing the probability that the user is a pushing object.

The specific implementation of S320 may refer to S120, and therefore is not described herein.

S330, determining a pushing object from a plurality of users according to the pushing probability.

The specific implementation of S230 can refer to S130, and therefore is not described herein.

S340, taking the plurality of pushing objects as a target pushing group, and acquiring the type of the target pushing group according to the times of the target pushing group accessing the specified webpage.

When there are a plurality of push objects obtained in S330, the plurality of push objects may be regarded as one target push group, so that the push can be performed in units of one group in the following. Since the multiple push objects obtained through S310 to S330 have a certain similarity, then the type of the target push group is obtained according to the number of times that the target push group accesses the specified web page, and the common push requirement of the multiple push objects in the target push group can be further determined.

The designated web pages or websites can be a plurality of web pages or websites, and the plurality of web pages or websites can be classified in advance, for example, sports websites, news websites, shopping websites, game websites, and the like. Each type of web page or website corresponds to a type of target push group, for example, the type of target push group corresponding to a sports type website is a sports fan, the type of target push group corresponding to a game type website is a game fan, and the like.

As an example, for example, the target push group may access a sports website 1000 times in a period of time, a news website 700 times, and a shopping website 500 times, and since the target user group accesses the sports website the most times in the same period of time, the type of the target push group may be determined as a sports enthusiast.

In some embodiments, as shown in fig. 5, S340 may include:

s341, extracting an effective push object from the plurality of push objects, and setting the effective push object as a target push group.

Because some push objects may not visit the website or the webpage, or visit the designated webpage or the website less frequently, if information related to push is pushed to the type of push according to the type of the whole target push group, a push error may be caused, for example, some push objects are not sports enthusiasts but belong to the target push group of which the type is sports enthusiasts, and the push objects push useless information. Therefore, the invalid push objects can be filtered out from the target push group, and the valid push objects are reserved for pushing, so that the pushing accuracy can be ensured.

In some embodiments, as shown in fig. 6, S341 may include:

s3411, respectively obtaining access times of each pushed object accessing the designated web page, and obtaining a plurality of second access times, where the plurality of second access times correspond to the plurality of pushed objects one to one.

As an example, the push objects include a first push object, a second push object, and a third push object, where the second access number corresponding to the first push object is 80 times, the second access number corresponding to the second push object is 54 times, and the second access number corresponding to the third push object is 20 times.

S3412, extracting a target access count from the plurality of second access counts, the target access count being greater than or equal to the second count threshold.

As an example, assuming that the second time threshold is 50 times, the second access times corresponding to the first push object and the second access times corresponding to the second push object may be determined as the target access times.

S3413, determining the push objects corresponding to the target access times as valid push objects, and using the valid push objects as a target push group.

As an example, the first push object and the second push object may be determined to be valid push objects, and the first push object and the second push object may be taken as target push groups.

In the embodiment, the second number of times that each pushing object accesses the designated webpage is obtained, whether the second number is greater than or equal to the second number threshold is judged, the pushing object corresponding to the second number threshold or greater is taken as the effective pushing object, and the effective pushing object is updated to the target pushing group for pushing, so that the pushing error is avoided, and the pushing accuracy is ensured.

S342, acquiring the first access times of the effective push object to access the specified webpage.

As an example, if the specified web page includes a sports web page, a game web page, and a news web page, the first access time for the active push object to access the sports web page is 500 times, the first access time for the game web page is 100 times, and the first access time for the news web page is 300 times.

S343, it is determined whether the first access time is greater than or equal to the first time threshold.

Alternatively, when the first number of accesses is less than the first number threshold, S342 is performed.

And S344, when the first access times is larger than or equal to the first time threshold value, acquiring the type label of the specified webpage.

As an example, if the first access number of times that the effective push object accesses the sports webpage is 500 times and the first access number of times that the effective push object accesses the news webpage is 300 times, the sports webpage and the news webpage satisfy a condition that the first access number is greater than or equal to the first number threshold, and a genre tag of the sports webpage, such as "sports", and a genre tag of the news webpage, such as "news", may be acquired.

And S345, determining the type of the target push group according to the type label.

As an example, the type of the target push group is determined to be a sports enthusiast according to the type tag "sports", while the type of the target push group may be determined to be a news follower according to the type tag "news".

In consideration that one target push group may be interested in multiple types of specified webpages, in the embodiment, by judging whether the first access times of the target push group consisting of effective push objects to different specified webpages is greater than or equal to the first time threshold value or not, and according to the type tags of the specified webpages meeting the conditions, one or more types of the target push group can be effectively determined, so that push contents are enriched.

And S350, sending push information corresponding to the type to the target push group. Alternatively, the push information may be user information, content information, commodity information, or the like.

As an example, if it is determined in S340 that the type of the target push group is a sports enthusiast, some information about sports, such as sports news, sports goods, sports event announcements, etc., may be pushed to the target push group.

In this embodiment, a plurality of pushing objects with certain similarity are used as a target pushing group, and pushing is performed in units of the group, so that the complexity of pushing one by one can be avoided. In addition, the type of the target push group is obtained according to the number of times that the target push group accesses the specified webpage, the type of the target push group can be simply and effectively determined, the push requirement of the target push group can be accurately obtained based on the type, and therefore the push accuracy is effectively improved.

Referring to fig. 7, a device 400 for determining a pushed object according to an embodiment of the present application is shown, where the device 400 includes: a user characteristic obtaining module 410, a push probability obtaining module 420 and a push object selecting module. The user characteristic obtaining module 410 is configured to obtain user characteristics of a plurality of users. The push probability obtaining module 420 is configured to input user characteristics of multiple users into a pre-trained classification model to obtain a push probability corresponding to each user, where the classification model is constructed based on a user portrait, and the push probability is used to represent a probability that the user is a push object. The push object selection module is used for determining a push object from a plurality of users according to the push probability.

Further, the push object determination further includes: a sample user feature obtaining module 410, a feature matrix generating module and a classification model generating module.

A sample user characteristic obtaining module 410, configured to obtain user characteristics of multiple sample users in a sample user group;

the characteristic matrix generation module is used for generating a characteristic matrix of a sample user group based on the user characteristics;

and the classification model generation module is used for obtaining a pre-trained classification model according to the characteristic matrix training.

And the push object determining module 430 determines a push object from the plurality of users according to the push probability.

Further, the classification model generation module comprises:

and the positive sample extraction unit is used for extracting a positive sample from the sample user group based on the characteristic matrix and determining the sample user group with the extracted positive sample as a residual sample.

And a negative sample extraction unit for extracting a negative sample from the remaining samples.

And the classification model generation unit is used for inputting the positive sample and the negative sample into the machine learning model for training to obtain a pre-trained classification model.

Further, the positive sample extraction unit is specifically configured to extract a positive sample from the sample user group according to the feature matrix and a preset positive sample feature identifier, and determine the sample user group from which the positive sample is extracted as a remaining sample.

Further, the negative sample extraction unit is used for selecting a negative sample from the remaining samples based on the positive sample label-free learning machine learning algorithm and the user characteristics of the positive sample.

Furthermore, the sample user group comprises a plurality of sample users, and the feature matrix generation module is further used for performing feature processing on the user features to obtain a feature vector of each sample user; a feature matrix is composed based on the feature vectors of each sample user.

Further, the user characteristics include at least one of user attribute characteristics, user behavior characteristics, and user interest characteristics.

Further, the sample user characteristic obtaining module 410 further includes:

the user behavior characteristic acquisition unit is used for acquiring an operation log of a sample user within a preset time length, wherein the operation log records operation behaviors generated by the user when the user uses a network, and the operation behaviors comprise at least one of purchasing behaviors, accessing behaviors and browsing behaviors; and determining the user behavior characteristics of the sample user according to the operation log.

Further, the sample user characteristic obtaining module 410 further includes:

the user attribute feature acquisition unit is used for acquiring attribute information of the sample user, wherein the attribute information comprises at least one of gender, birth date, occupation and education background; and determining the user attribute characteristics of the sample user according to the attribute information.

Further, the sample user characteristic obtaining module 410 further includes:

the user interest characteristic acquisition unit is used for acquiring social records of the sample user, wherein the social records comprise at least one of search records, comment records and attention records; and determining user interest characteristics according to the social records.

Further, the feature processing includes at least one of discretization processing, regularization processing and normalization processing or a combination of the discretization processing, the regularization processing and the normalization processing.

Further, the classification model generation unit is used for processing the positive samples and the negative samples through an LR machine learning classification algorithm or a random forest machine learning classification algorithm to obtain a pre-trained classification model.

Further, the pushed object determining module 430 is further configured to select a preset number of users from the multiple users according to a sequence from a large pushed probability to a small pushed probability; and determining a preset number of users as push objects.

Further, the pushed object determining module 430 is further configured to select at least one target pushing probability from the multiple pushing probabilities, where the target pushing probability is greater than or equal to the probability threshold; and determining a pushing object for the user corresponding to the target pushing probability.

Further, the number of the push objects is multiple, and the push object determining module 430 is further configured to use the multiple push objects as a target push group, and obtain the type of the target push group according to the number of times that the target push group accesses the specified webpage; and sending push information corresponding to the type to a target push group.

Further, the push object determining module 430 is further configured to extract an effective push object from the multiple push objects, and use the effective push object as a target push group.

And acquiring the first access times of the effective push object to access the specified webpage.

And when the first access times are larger than or equal to the first time threshold value, acquiring the type label of the specified webpage.

And determining the type of the target push group according to the type tag.

Further, the pushed object determining module 430 is further configured to obtain the number of times that each pushed object accesses the specified webpage, respectively, to obtain a plurality of second access times, where the plurality of second access times correspond to the plurality of pushed objects one to one.

And extracting a target access frequency from the plurality of second access frequencies, wherein the target access frequency is greater than or equal to a second frequency threshold.

And determining the push object corresponding to the target access times as an effective push object.

It can be clearly understood by those skilled in the art that, for convenience and brevity of description, the specific working processes of the apparatus 400 and the modules described above may refer to corresponding processes in the foregoing method embodiments, and are not described herein again.

In the several embodiments provided in this application, the coupling or direct coupling or communication connection between the modules shown or discussed may be through some interfaces, the indirect coupling or communication connection between the device 400400 or the modules, and may be in an electrical, mechanical or other form.

In addition, functional modules in the embodiments of the present application may be integrated into one processing module, or each of the modules may exist alone physically, or two or more modules are integrated into one module. The integrated module can be realized in a hardware mode, and can also be realized in a software functional module mode.

Referring to fig. 8, a block diagram of a device terminal according to an embodiment of the present disclosure is shown. The device terminal 500 may be the device terminal 500 capable of running the application program in the foregoing embodiment. The device terminal 500 in the present application may include one or more of the following components: a processor 510, a memory 520, and one or more applications, wherein the one or more applications may be stored in the memory 520 and configured to be executed by the one or more processors 510, the one or more programs configured to perform a method as described in the aforementioned method embodiments.

Processor 510 may include one or more processing cores. The processor 510 connects various parts within the overall terminal device using various interfaces and lines, and performs various functions of the terminal device and processes data by executing or executing instructions, programs, code sets, or instruction sets stored in the memory 520 and calling data stored in the memory 520. Alternatively, the processor 510 may be implemented in hardware using at least one of Digital Signal Processing (DSP), Field-Programmable Gate Array (FPGA), and Programmable Logic Array (PLA). The processor 510 may integrate one or a combination of a Central Processing Unit (CPU) 510, a Graphics Processing Unit (GPU) 510, a modem, and the like. Wherein, the CPU mainly processes an operating system, a user interface, an application program and the like; the GPU is used for rendering and drawing display content; the modem is used to handle wireless communications. It is understood that the modem may not be integrated into the processor 510, but may be implemented by a communication chip.

The Memory 520 may include a Random Access Memory (RAM) 520 and may also include a Read-Only Memory (Read-Only Memory) 520. The memory 520 may be used to store instructions, programs, code sets, or instruction sets. The memory 520 may include a stored program area and a stored data area, wherein the stored program area may store instructions for implementing an operating system, instructions for implementing at least one function (such as a touch function, a sound playing function, an image playing function, etc.), instructions for implementing various method embodiments described below, and the like. The storage data area may also store data created by the terminal in use, such as a phonebook, audio-video data, chat log data, and the like.

Referring to fig. 9, a block diagram of a computer-readable storage medium according to an embodiment of the present application is shown. The computer readable medium 600 has stored therein a program code 610, the program code 610 being capable of being invoked by a processor to perform the method described in the method embodiments above.

The computer-readable storage medium 600 may be an electronic memory such as a flash memory, an EEPROM (electrically erasable programmable read only memory), an EPROM, a hard disk, or a ROM. Alternatively, the computer-readable storage medium includes a non-transitory computer-readable storage medium. The computer readable storage medium has a storage space for program code for performing any of the method steps of the above-described method. The program code can be read from or written to one or more computer program products. The program code may be compressed, for example, in a suitable form.

In summary, according to the method, the apparatus, the terminal device, and the storage medium for determining a push object provided in the embodiments of the present application, a classification model is constructed based on user characteristics, when user characteristics of a user are input into the classification model, a probability of whether the user is a user to be pushed can be quickly and accurately output, and a certain number of users to be pushed can be effectively determined from a large number of user groups to be pushed as a push target user group based on the probability of the user to be pushed, so that the number of users to be pushed to the target user group is better expanded, a coverage of the user to be effectively pushed in the user group is increased, and the push efficiency is improved.

Finally, it should be noted that: the above embodiments are only used to illustrate the technical solutions of the present application, and not to limit the same; although the present application has been described in detail with reference to the foregoing embodiments, it will be understood by those of ordinary skill in the art that: the technical solutions described in the foregoing embodiments may still be modified, or some technical features may be equivalently replaced; such modifications and substitutions do not necessarily depart from the spirit and scope of the corresponding technical solutions in the embodiments of the present application.

Claims

A method for push object determination, the method comprising:

acquiring user characteristics of a plurality of users;

inputting the user characteristics of the users into a pre-trained classification model to obtain a pushing probability corresponding to each user, wherein the classification model is constructed based on the user characteristics, and the pushing probability is used for representing the probability that the user is a pushing object;

and determining a pushing object from the plurality of users according to the pushing probability.
The method of claim 1, further comprising, prior to said inputting user characteristics of said plurality of users into a pre-trained classification model:

acquiring user characteristics of a plurality of sample users in a sample user group;

generating a feature matrix of the sample user population based on the user features;

and training according to the characteristic matrix to obtain a pre-trained classification model.
The method of claim 2, wherein the training from the feature matrix results in a pre-trained classification model comprising:

extracting a positive sample from the sample user group based on the characteristic matrix, and determining the sample user group with the positive sample extracted as a residual sample;

extracting a negative sample from the residual sample;

and inputting the positive sample and the negative sample into a machine learning model for training to obtain a pre-trained classification model.
The method according to claim 3, wherein the extracting a positive sample from the sample user group based on the feature matrix and determining the sample user group after the extraction of the positive sample as a remaining sample comprises:

and taking the positive sample from the sample user group according to the feature matrix and a preset positive sample feature identifier, and determining the sample user group with the positive sample taken as a residual sample.
The method of claim 3, wherein the extracting negative samples from the remaining samples comprises:

and selecting the negative sample from the rest samples based on a positive sample label-free learning machine learning algorithm and the user characteristics of the positive sample.
The method of any one of claims 2-5, wherein the sample user population includes a plurality of sample users, and wherein generating the feature matrix for the sample user population based on the user features comprises:

performing feature processing on the user features to obtain a feature vector of each sample user;

and forming the feature matrix based on the feature vector of each sample user.
The method of claim 6, wherein the user characteristics comprise at least one of user attribute characteristics, user behavior characteristics, and user interest characteristics.
The method of claim 7, wherein the user characteristics comprise user behavior characteristics, and wherein the obtaining the user characteristics of the sample user comprises:

obtaining an operation log of the sample user within a preset time length, wherein the operation log records operation behaviors generated by the user when the user uses a network, and the operation behaviors comprise at least one of purchasing behaviors, accessing behaviors and browsing behaviors;

and determining the user behavior characteristics of the sample user according to the operation log.
The method of claim 7, wherein the user characteristics comprise user attribute characteristics, and wherein the obtaining the user characteristics of the sample user comprises:

acquiring attribute information of the sample user, wherein the attribute information comprises at least one of gender, birth date, occupation and educational background;

and determining the user attribute characteristics of the sample user according to the attribute information.
The method of claim 7, wherein the user characteristics comprise user interest characteristics, and wherein the obtaining the user characteristics of the sample user comprises:

obtaining social records of the sample user, wherein the social records comprise at least one of search records, comment records and attention records;

and determining the user interest characteristics according to the social records.
The method according to any one of claims 6 to 10, wherein the feature processing comprises at least one of discretization processing, regularization processing and normalization processing or a combination thereof.
The method according to any one of claims 3-5, wherein the inputting the positive samples and the negative samples into a machine learning model for training, and obtaining a pre-trained classification model comprises:

and processing the positive sample and the negative sample through an LR machine learning classification algorithm or a random forest machine learning classification algorithm to obtain the pre-trained classification model.
The method according to any one of claims 1-12, wherein said determining a push object from said plurality of users according to said push probability comprises:

selecting a preset number of users from the plurality of users according to the sequence of the pushing probability from large to small;

and determining the preset number of users as the push objects.
The method according to any one of claims 1-12, wherein said determining a push object from said plurality of users according to said push probability comprises:

selecting at least one target push probability from the plurality of push probabilities, wherein the target push probability is greater than or equal to a probability threshold;

and determining the push object for the user corresponding to the target push probability.
The method according to any of claims 1-14, wherein the number of the push objects is multiple, and after determining the push object from the multiple users according to the push probability, the method further comprises:

taking a plurality of pushing objects as a target pushing group, and acquiring the type of the target pushing group according to the times of accessing specified webpages by the target pushing group;

and sending push information corresponding to the type to the target push group.
The method of claim 15, wherein the taking a plurality of the push objects as a target push group and obtaining the type of the target push group according to the number of times that the target push group accesses a specified webpage comprises:

extracting effective push objects from the plurality of push objects, and taking the effective push objects as the target push group;

acquiring a first access frequency of the effective push object to access an appointed webpage;

when the first access times are larger than or equal to a first time threshold value, acquiring a type label of the specified webpage;

and determining the type of the target push group according to the type tag.
The method of claim 16, wherein extracting a valid push object from the plurality of push objects and regarding the valid push object as the target push group comprises:

respectively obtaining the access times of each pushed object to the specified webpage to obtain a plurality of second access times, wherein the second access times are in one-to-one correspondence with the pushed objects;

extracting a target access frequency from the plurality of second access frequencies, wherein the target access frequency is greater than or equal to a second frequency threshold;

determining the push objects corresponding to the target access times as the effective push objects, and taking the effective push objects as the target push groups.
A push object determination apparatus, the apparatus comprising:

the user characteristic acquisition module is used for acquiring user characteristics of a plurality of users;

the pushing probability obtaining module is used for inputting the user characteristics of the users into a pre-trained classification model to obtain a pushing probability corresponding to each user, the classification model is constructed based on the user portrait, and the pushing probability is used for representing the probability that the user is a pushing object;

and the push object determining module is used for determining a push object from the plurality of users according to the push probability.
A terminal device, comprising:

one or more processors;

a memory;

one or more applications, wherein the one or more applications are stored in the memory and configured to be executed by the one or more processors, the one or more applications configured to perform the method of any of claims 1-17.
A computer-readable storage medium having program code stored therein, the program code being invoked by a processor to perform the method of any one of claims 1 to 17.