CN112016961A - Push method, apparatus, electronic device, and computer-readable storage medium - Google Patents

Abstract

The present application provides a push method, device, electronic device and computer-readable storage medium, the method includes: extracting user characteristics and context characteristics of unknown users, and industry characteristics of the game to be pushed; Game category, a first prediction model is determined from multiple prediction models, wherein the multiple prediction models are prediction models based on different game categories, the game to be pushed belongs to the first game category, and the first The estimation model is the estimation model of the first game category; the user characteristics and context characteristics of the unknown user and the industry characteristics of the game to be pushed are input into the first estimation model, and the unknown user is output. target user; push the to-be-pushed game to the target user.

Description

Push method, apparatus, electronic device, and computer-readable storage medium

technical field

The present application relates to the field of advertising technology, and more particularly, to a push method, apparatus, electronic device, and computer-readable storage medium.

Background technique

When advertising, for some shallow goals (such as clicks, downloads, etc.), the advertising platform itself can obtain user data, but for some deep conversion goals (such as second retention, payment, etc.), many large game manufacturers are Those who are unwilling to return data to help the advertising platform make in-depth conversion estimates are mainly because this part of the data is the core secret of game manufacturers. In this case, how to obtain user data of such games to improve the advertising effect is an urgent need. solved problem.

SUMMARY OF THE INVENTION

The present application provides a push method, device, electronic device and computer-readable storage medium. By constructing models of different game categories, when recommending a game, you only need to input the game to be recommended into the corresponding category model to get the game. precise users.

In a first aspect, an embodiment of the present application provides a push method, including:

Extract user characteristics and contextual characteristics of unknown users, as well as industry characteristics of games to be pushed;

According to the game category to which the game to be pushed belongs, a first prediction model is determined among multiple prediction models, wherein the multiple prediction models are prediction models based on different game categories, and the game to be pushed belongs to the first prediction model. A game category, the first prediction model is an estimation model of the first game category;

Inputting the user characteristics and context characteristics of the unknown user and the industry characteristics of the game to be pushed into the first estimation model, and outputting the target user among the unknown users;

Push the to-be-pushed game to the target user.

In some possible implementations, the method further includes:

Input the target distinguishing characteristics of the users of the first game category, the industry characteristics of the games belonging to the first game category, and the positive and negative samples of the first game category into the first prediction model for training, and obtain parameters of the first prediction model.

In some possible implementations, the method further includes:

Perform feature engineering on the data of existing users of games of different game categories to obtain the target distinguishing features of users of different game categories.

In some possible implementations, the feature engineering is performed on the data of existing users of games of different game categories, and the target distinguishing characteristics of users of different game categories are obtained, including:

constructing positive and negative samples of the first game category according to data of existing users of at least one game belonging to the first game category;

Analyze the contribution of user characteristics corresponding to positive and negative samples of different game categories, and determine the basic characteristics of users of different game categories.

In some possible implementations, performing feature engineering on the data of existing users of games of different game categories to obtain target distinguishing features of users of different game categories further includes:

According to the basic characteristics of positive and negative samples of different game categories, analyze the crowd portraits of users of different game categories, and determine the distinguishing characteristics of users of different game categories;

Feature association is performed on the distinguishing features to obtain target distinguishing features for distinguishing users of different game categories.

In some possible implementations, the feature association is performed on the distinguishing features to obtain target distinguishing features for distinguishing users of different game categories, including:

Obtain the historical behavior data of game users of the different game categories through big data;

Perform feature association on the distinguishing features according to the historical behavior data of game users of the different game categories to obtain the target distinguishing features.

For example, the historical behavior data of game users of the different game categories is obtained through big data.

Optionally, the big data includes at least one of the following: user behavior data on a search platform, user behavior data on other advertising platforms, and user behavior data obtained by a data management platform.

In some possible implementations, constructing positive and negative samples of the first game category according to data of existing users of at least one game belonging to the first game category, including:

For each game in the at least one game, the user data and game data that are activated by the user and have deep conversion events within a period of time are taken as positive samples of the first game category, The user data and game data of the deep conversion event are used as negative samples of the first game category.

In some possible implementations, the method further includes:

If the same user data and game data exist in both the positive samples of the first game category and the negative samples of the first game category, the user data and game data are excluded from the negative samples of the first game category data.

In some possible implementations, the method further includes:

removing game data in the positive and negative samples of the first game category;

Input the positive and negative samples of the first game category with game data removed into the first prediction model for training.

In a second aspect, an embodiment of the present application provides a push method, including: acquiring data of existing users of games of different game categories;

According to the data of existing users of at least one game belonging to the same game category, construct positive and negative samples of the same game category;

Perform feature engineering on the data of existing users of games of different game categories to obtain the target distinguishing characteristics of users of different game categories;

Input the industry characteristics of games belonging to the same game category, the target distinguishing characteristics of users of games of the same game category, and the positive and negative samples of the same game category into the prediction model corresponding to the same game category for training , to obtain the prediction model corresponding to the same game category.

In some possible implementations, constructing positive and negative samples of the same game category according to data of existing users of at least one game belonging to the same game category, including:

For each game in the at least one game, the user data and game data that have been activated by the user for a period of time and have deep conversion events are taken as positive samples of the same game category, The user data and game data of the conversion event are used as negative samples of the same game category.

In some possible implementations, the method further includes:

If the same user data and game data exist in both the positive samples of the same game category and the negative samples of the same game category, the user data and game data are excluded from the negative samples of the same game category.

In some possible implementations, the method further includes:

removing the game data in the positive and negative samples of the same game category;

Input the positive and negative samples of the same game category with the game data removed into the prediction model corresponding to the same game category for training.

In some possible implementations, the feature engineering is performed on the data of existing users of games of different game categories, and the target distinguishing characteristics of users of different game categories are obtained, including:

Analyze the contribution of user characteristics corresponding to positive and negative samples of different game categories, and determine the basic characteristics of users of different game categories;

According to the basic characteristics of positive and negative samples of different game categories, analyze the crowd portraits of users of different game categories, and determine the distinguishing characteristics of users of different game categories;

Feature association is performed on the distinguishing features to obtain target distinguishing features for distinguishing users of different game categories.

In some possible implementations, the feature association is performed on the distinguishing features to obtain target distinguishing features for distinguishing users of different game categories, including:

Obtain the historical behavior data of game users of the different game categories;

Perform feature association on the distinguishing features according to the historical behavior data of game users of the different game categories to obtain the target distinguishing features.

For example, the historical behavior data of game users of the different game categories is obtained through big data.

Optionally, the big data includes at least one of the following: user behavior data on a search platform, user behavior data on other advertising platforms, and user behavior data obtained by a data management platform.

In a third aspect, an embodiment of the present application provides a push device, which is configured to execute the steps of the method in the first aspect or any possible implementation manner of the first aspect. Specifically, the pushing device includes a unit for executing the method in the first aspect or any possible implementation manner of the first aspect.

In a fourth aspect, an embodiment of the present application provides another push apparatus, which is configured to execute the steps of the method in the second aspect or any possible implementation manner of the second aspect. Specifically, the pushing device includes a unit for executing the method in the second aspect or any possible implementation manner of the second aspect.

In a fifth aspect, an embodiment of the present application provides an electronic device, including: a processor, a storage medium, and a bus, where the storage medium stores machine-readable instructions executable by the processor, and when the electronic device runs, all Communication between the processor and the storage medium is through a bus, and the processor executes the machine-readable instructions to perform the steps of any optional push method in the first aspect.

In a sixth aspect, an embodiment of the present application provides an electronic device, including: a processor, a storage medium, and a bus, where the storage medium stores machine-readable instructions executable by the processor, and when the electronic device runs, all Communication between the processor and the storage medium is through a bus, and the processor executes the machine-readable instructions to perform the steps of any optional push method in the second aspect.

In a seventh aspect, an embodiment of the present application provides a computer-readable storage medium, where a computer program is stored on the computer-readable storage medium, and the computer program is executed by a processor when it is executed as optional in the first aspect The steps of the push method.

In an eighth aspect, an embodiment of the present application provides a computer-readable storage medium, where a computer program is stored on the computer-readable storage medium, and the computer program is executed by a processor when it is executed as optional in the second aspect The steps of the push method.

Based on the above technical solutions, the present application can establish an estimation model for accurate users of the game theme based on the data of existing users of the same game theme, combined with the industry characteristics of games of different game categories, and can accurately mine the subsequent unknown users. It can identify potential accurate users and filter out some noisy users, so that advertisers have more accurate goals when advertising, and improve the cost performance of advertising. For advertising platforms, it improves the traffic efficiency of the advertising platform and can ultimately improve the advertising platform. 's earnings.

Description of drawings

1 is a schematic flowchart of a method for pushing provided by an embodiment of the present application;

2 is a schematic flowchart of another method for pushing provided by an embodiment of the present application;

3 is a schematic flowchart of another push method provided by an embodiment of the present application;

FIG. 4 is a schematic structural diagram of a push device according to an embodiment of the present application;

FIG. 5 is a schematic structural diagram of a push device according to an embodiment of the present application;

FIG. 6 is a schematic structural diagram of an electronic device according to an embodiment of the present application.

Detailed ways

The technical solutions in the embodiments of the present application will be described below with reference to the accompanying drawings in the embodiments of the present application. Obviously, the described embodiments are part of the embodiments of the present application, not all of the embodiments. With regard to the embodiments in the present application, all other embodiments obtained by persons of ordinary skill in the art without creative efforts shall fall within the protection scope of the present application.

FIG. 1 is a schematic flowchart of a push method 100 provided by an embodiment of the present application. As shown in FIG. 1 , the method 100 includes at least some of the following contents:

S101, extracting user characteristics and contextual characteristics of unknown users, and industry characteristics of the game to be pushed;

S102, according to the game category to which the game to be pushed belongs, determine a first prediction model from multiple prediction models, wherein the multiple prediction models are prediction models based on different game categories, and the game to be pushed It belongs to the first game category, and the first estimation model is an estimation model of the first game category;

S103, input the user characteristics and context characteristics of the unknown user and the industry characteristics of the game to be pushed into the first estimation model, and output the target user among the unknown users;

S104. Push the to-be-pushed game to the target user.

It should be noted that the embodiments of the present application are applicable to the recommendation of game-type applications, or may also be applied to the recommendation of other types of applications, for example, clue-type applications. In the specific implementation, different types of clue applications can be distinguished, such as education, examination, financial management, etc., and the estimation model corresponding to each type of clue application can be further constructed, and the implementation methods are similar. In the following, the prediction model corresponding to the game construction of different game categories is described as an example, but the present application is not limited to this.

In the embodiments of this application, games can be divided into different themes, or categories, such as but not limited to game themes such as Three Kingdoms, Two-dimensional, Legendary, Women's Orientation, War, Official Fight, Business War, and Science Fiction. Game themes, respectively train prediction models based on game themes, when performing model prediction, you can determine the theme of the game to be pushed, and further estimate the unknown user based on the prediction model of the theme to determine the content of the game to be pushed. The target user can then push the to-be-pushed game to the target user.

In order to facilitate a better understanding of the embodiments of the present application, the training process of the prediction model based on the game theme is first described.

1. Construct positive and negative samples of different game themes

In the embodiments of the present application, positive and negative samples of the different game themes can be constructed by using existing users of games with different game themes, and the existing users may include high-quality users of the game, such as users with deep conversion events. Using the existing high-quality users of existing games as samples to construct a model is beneficial to filter out high-quality users for the game theme, thereby improving the effect of advertising.

In the embodiment of the present application, positive and negative samples of the game theme can be constructed according to the data of existing users of all games belonging to the same game theme, that is, games of the same game theme have common positive and negative samples.

Specifically, for each game in all games belonging to the same game theme, the positive and negative samples of each game can be counted. For example, the user activates the game within a period of time, such as 6 hours, or 12 hours, and there are data of deep conversion events (such as user identification (ID) and game ID) as a positive sample. Take the data (such as user identification (ID) and game ID) of the event that the user activates the category within a period of time, such as 6 hours, or 12 hours, but has not been deeply converted as a negative sample. The positive samples of all games belonging to the same game theme are regarded as the positive samples of the game theme, and the negative samples of all games are regarded as the negative samples of the game theme.

Wherein, the deep conversion event may be, for example, second stay, payment, and the like.

It should be understood that the above construction methods of positive and negative samples are only examples, and may be adjusted according to actual needs, and the present application is not limited thereto.

In the embodiment of the present application, the positive and negative samples may include user IDs, game IDs, and labels identifying the sample types. When inputting training data into the estimation model corresponding to the game theme, the game IDs in the positive and negative samples may be removed. That is, the binding relationship between the user and a single game is released, so that the user and the game theme are bound.

Optionally, in some embodiments, if the same user data and game data exist in both the positive samples of the game theme and the negative samples of the game theme, the user is excluded from the negative samples of the game theme. data and game data.

In a specific implementation, if the positive and negative samples of the same game theme are not balanced, oversampling or undersampling can be used to solve the above problem. For example, if the number of users of legendary themes is usually large, the number of positive samples is large, and the number of positive samples can be reduced by means of undersampling. For another example, if the number of users of a game with a niche theme is small, the number of positive samples is small, and oversampling can be used to increase the number of positive samples.

2. Build a feature pool

After constructing positive and negative samples of different game themes, the user characteristics corresponding to the positive and negative samples of different game themes can be further obtained, such as but not limited to any one or more of the user's age, gender and occupation. Describe the industry characteristics of games with different game themes, such as industry description information, such as industry, sub-industry and category information, and further add the above characteristics to the feature pool.

It can be understood that age, gender and occupation are only a preferred way of limiting user characteristics in this application. In practical applications, according to the specific type of the game to be pushed, the user characteristics may also include other limiting factors. For example, income level, educational background, etc., will not be repeated here.

Further, feature engineering can be performed on the user characteristics and industry characteristics of the obtained positive and negative samples of different game themes, so as to determine the distinguishing characteristics of different game themes.

Specifically, as shown in FIG. 2 , in S110 , the contribution of existing features of different game categories may be analyzed first, and the feature with a larger contribution may be determined as the basic feature. (InformationValue, IV).

For example, the features with a void value rate greater than a certain threshold in the existing features are eliminated, and the features with an IV value greater than a certain threshold are used as one of the basic features.

Then in S120, based on the basic features of the users in the positive and negative samples of different game themes, analyze the user's portraits of each game theme, and select features that are significantly different for different game themes as distinguishing features from the basic features.

Further, in S130, the distinguishing features are associated, and the feature range is expanded to obtain the target distinguishing features. For example, if the distinguishing feature includes feature A, and a user with feature A also has feature B, but feature B is not included in the distinguishing feature, feature B can be added to the distinguishing feature.

For example, for game users with legendary themes, they often have one thing in common: they like to watch Hong Kong movies. For the distinguishing features of games with legendary themes, a feature similar to "Browsing Hong Kong movies in the past period of time" can be added.

In some implementations, the behavior data of the users of the different game themes can be obtained through big data, and the obtained user behavior data can be added to the feature pool, based on the existing features in the feature pool and obtained through big data The user behavior data is used for feature association, which can improve the generalization ability of the model.

Optionally, the big data includes: user behavior data on the search platform, user behavior data on other advertising platforms, and user behavior data obtained by a data management platform.

For example, user data who have searched for game titles, game competitions, and game team names may be input into the feature pool.

For another example, user behavior data may be obtained from other game platform manufacturers. For example, user behavior data may be acquired through a software development kit (Software Development Kit, SDK), and the acquired user behavior data may be further added to the feature pool.

For another example, user behavior data may be acquired from a data management platform (Data Management Platform, DMP), and the acquired user behavior data may be further added to the feature pool.

It can be understood that user behavior data can include any behavior of users on the Internet, such as but not limited to users watching a certain game advertisement, clicking a certain game advertisement, installing and activating a certain game, and paying for a certain game. Or other likes or comments, etc.

Based on these user behavior data, the user's preference for game themes can be determined, and further targeted recommendations can be made when making game recommendations, which can improve the effect of advertising.

Optionally, after constructing the distinguishing features that distinguish different game themes, before training, the above features can also be subjected to data preprocessing, for example, data normalization, unified coding format, etc. The coded value of the feature corresponds to the coded value of the existing feature.

The constructed features of the same game theme (including user features and industry features of the game theme) and the positive and negative samples of the same game theme are used as training data, and the prediction model of the game theme is obtained by training according to a preset algorithm. Optionally, the prediction model may be a ranking model, such as a Factorization Machines (FM) model, a Field-aware Factorization Machine (FFM) model, an LR model, and the like.

When estimating according to the trained prediction model, industry features of the game to be pushed can be extracted through the feature pool, and user features of unknown users (specifically, the distinguishing features constructed in the foregoing embodiment) and context can be further extracted. feature, input into the estimation model of the subject matter of the game to be recommended, estimate the target user of the game to be recommended among the unknown users through the estimation model, and deliver the game to be pushed to the target user. .

Optionally, the context information is context information of the advertisement, such as opening the screen, opening the first screen of the screen, refreshing the advertisement upward, refreshing the advertisement downward, and the like.

To sum up, according to the data of existing users of the same game theme, the embodiment of the present application can establish an estimation model of accurate users of the game theme, and for subsequent unknown users, potential accurate users can be accurately mined, and some noise can be filtered out. Users can make advertisers more accurate when advertising, improve the cost-effectiveness of advertising, and for advertising platforms, improve the traffic efficiency of the advertising platform, and ultimately increase the revenue of the advertising platform.

FIG. 3 is a schematic flowchart of a push method according to another embodiment of the present application. As shown in FIG. 3 , the method 200 includes at least some of the following contents:

S201, acquiring data of existing users of games of different game categories;

S202, construct positive and negative samples of the same game category according to the data of existing users of at least one game belonging to the same game category;

S203, perform feature engineering on data of existing users of games of different game categories, and obtain target distinguishing features of users of different game categories;

S204, input the industry characteristics of the games belonging to the same game category, the target distinguishing characteristics of the users of the games of the same game category, and the positive and negative samples of the same game category into the prediction model corresponding to the same game category Perform training to obtain the estimated model corresponding to the same game category.

It should be understood that for the specific implementation of the method 200, reference may be made to the relevant implementation in the method 100, which is not repeated here for brevity.

Optionally, in some embodiments, constructing positive and negative samples of the same game category according to data of existing users of at least one game belonging to the same game category includes:

For each game in the at least one game, the user data and game data that have been activated by the user for a period of time and have deep conversion events are taken as positive samples of the same game category, The user data and game data of the conversion event are used as negative samples of the same game category.

Optionally, in some embodiments, the method 200 further includes:

If the same user data and game data exist in both the positive samples of the same game category and the negative samples of the same game category, the user data and game data are excluded from the negative samples of the same game category.

Optionally, in some embodiments, the method 200 further includes:

removing the game data in the positive and negative samples of the same game category;

Input the positive and negative samples of the same game category with the game data removed into the prediction model corresponding to the same game category for training.

Optionally, in some embodiments, the feature engineering is performed on the data of existing users of games of different game categories to obtain target distinguishing features of users of different game categories, including:

Analyze the contribution of user characteristics corresponding to positive and negative samples of different game categories, and determine the basic characteristics of users of different game categories;

According to the basic characteristics of positive and negative samples of different game categories, analyze the crowd portraits of users of different game categories, and determine the distinguishing characteristics of users of different game categories;

Feature association is performed on the distinguishing features to obtain target distinguishing features for distinguishing users of different game categories.

Optionally, in some embodiments, performing feature association on the distinguishing features to obtain target distinguishing features for distinguishing users of different game categories, including:

Obtain the historical behavior data of game users of the different game categories;

Perform feature association on the distinguishing features according to the historical behavior data of game users of the different game categories to obtain the target distinguishing features.

Optionally, in some embodiments, the big data includes at least one of the following: user behavior data on a search platform, user behavior data on other advertising platforms, and user behavior data obtained by a data management platform.

Therefore, the push method of the embodiment of the present application can establish an estimation model of accurate users of the game theme according to the data of existing users of the same game theme, combined with the industry characteristics of games of different game categories, and for subsequent unknown users, it can be Accurately dig out potential accurate users, and filter out some noisy users, so that advertisers have more accurate goals when advertising, and improve the cost performance of advertising. For advertising platforms, it improves the traffic efficiency of advertising platforms, and can ultimately improve Advertising platform revenue.

The method embodiments of the present application are described in detail above with reference to FIGS. 1 to 3 , and the apparatus embodiments of the present application are described in detail below with reference to FIGS. 4 to 6 . It should be understood that the device embodiments and the method embodiments correspond to each other, and are similar to For the description, refer to the method embodiment.

FIG. 4 shows a schematic block diagram of a push apparatus 300 according to an embodiment of the present application. As shown in Figure 4, the device 300 includes:

An extraction unit 301, configured to extract user characteristics and contextual characteristics of unknown users, as well as industry characteristics of the game to be pushed;

The determining unit 302 is configured to determine a first estimation model among multiple estimation models according to the game category to which the game to be pushed belongs, wherein the multiple estimation models are estimation models based on different game categories, and the The game to be pushed belongs to a first game category, and the first estimation model is an estimation model of the first game category;

Estimating unit 303, configured to input the user characteristics and context characteristics of the unknown user and the industry characteristics of the game to be pushed into the first estimation model, and output the target user among the unknown users;

The pushing unit 304 is configured to push the to-be-pushed game to the target user.

Optionally, in some embodiments, the extracting unit 301 is further configured to: perform feature engineering on data of existing users of games of different game categories to obtain target distinguishing features of users of different game categories.

Optionally, in some embodiments, the extraction unit 301 is further configured to:

constructing positive and negative samples of the first game category according to data of existing users of at least one game belonging to the first game category;

Analyze the contribution of user characteristics corresponding to positive and negative samples of different game categories, and determine the basic characteristics of users of different game categories.

Optionally, in some embodiments, the extraction unit 301 is further configured to: analyze the crowd portraits of users of different game categories according to the basic characteristics of positive and negative samples of different game categories, and determine the distinguishing characteristics of users of different game categories. ;

Feature association is performed on the distinguishing features to obtain target distinguishing features for distinguishing users of different game categories.

Optionally, in some embodiments, the extracting unit 301 is further configured to: acquire historical behavior data of game users of the different game categories;

Perform feature association on the distinguishing features according to the historical behavior data of game users of the different game categories to obtain the target distinguishing features.

Optionally, in some embodiments, the big data includes at least one of the following: user behavior data on a search platform, user behavior data on other advertising platforms, and user behavior data obtained by a data management platform.

Optionally, in some embodiments, the apparatus 300 further includes:

A training unit for inputting the industry characteristics of games belonging to the first game category, the target distinguishing characteristics of users of the first game category, and the positive and negative samples of the first game category into the first estimation The model is trained to obtain parameters of the first estimated model.

Optionally, in some embodiments, the training unit is further configured to: for each game in the at least one game, use the user data and game data that are activated by the user and have deep conversion events within a period of time as all the game data. For the positive samples of the first game category, the user data and game data that are activated by the user but have no deep conversion events within a period of time are used as negative samples of the first game category.

Optionally, in some embodiments, the training unit is further used for:

If the same user data and game data exist in both the positive samples of the first game category and the negative samples of the first game category, the user data and game data are excluded from the negative samples of the first game category data.

Optionally, in some embodiments, the training unit is further configured to: remove game data in the positive and negative samples of the first game category;

Input the positive and negative samples of the first game category with game data removed into the first prediction model for training.

Therefore, according to the push device of the embodiment of the present application, according to the data of existing users of the same game theme, combined with the industry characteristics of games of different game categories, an estimation model of accurate users of the game theme can be established. For subsequent unknown users, It can accurately dig out potential accurate users and filter out some noisy users, so that advertisers can target more accurate advertisements and improve the cost performance of advertisements. For advertising platforms, it improves the traffic efficiency of advertising platforms, and can ultimately Increase the revenue of the advertising platform.

FIG. 5 shows a schematic block diagram of a push apparatus 400 according to an embodiment of the present application. As shown in Figure 5, the device 400 includes:

an acquisition unit 401, configured to acquire data of existing users of games of different game categories;

A construction unit 402, configured to construct positive and negative samples of the same game category according to the data of existing users of at least one game belonging to the same game category;

The processing unit 403 is configured to perform feature engineering on the data of existing users of games of different game categories, and obtain target distinguishing features of users of different game categories;

The training unit 404 is configured to input the industry characteristics of the games belonging to the same game category, the target distinguishing characteristics of the users of the games of the same game category, and the positive and negative samples of the same game category into the corresponding game category. The prediction model is trained to obtain the prediction model corresponding to the same game category.

Optionally, in some embodiments, the construction unit 402 is specifically configured to:

For each game in the at least one game, the user data and game data that have been activated by the user for a period of time and have deep conversion events are taken as positive samples of the same game category, The user data and game data of the conversion event are used as negative samples of the same game category.

Optionally, in some embodiments, the constructing unit 402 is further configured to: if the same user data and game data exist both in the positive samples of the same game category and in the negative samples of the same game category, from The user data and game data are excluded from the negative samples of the same game category.

Optionally, in some embodiments, the construction unit 402 is specifically configured to: remove the game data in the positive and negative samples of the same game category;

The training unit 404 is further configured to: input the positive and negative samples of the same game category from which the game data is removed into the prediction model corresponding to the same game category for training.

Optionally, in some embodiments, the processing unit 403 is further configured to:

Analyze the contribution of user characteristics corresponding to positive and negative samples of different game categories, and determine the basic characteristics of users of different game categories;

According to the basic characteristics of positive and negative samples of different game categories, analyze the crowd portraits of users of different game categories, and determine the distinguishing characteristics of users of different game categories;

Feature association is performed on the distinguishing features to obtain target distinguishing features for distinguishing users of different game categories.

Optionally, in some embodiments, the processing unit 403 is further configured to:

Obtain the historical behavior data of game users of the different game categories;

Perform feature association on the distinguishing features according to the historical behavior data of game users of the different game categories to obtain the target distinguishing features.

Optionally, in some embodiments, the big data includes at least one of the following: user behavior data on a search platform, user behavior data on other advertising platforms, and user behavior data obtained by a data management platform.

Therefore, the push device of the embodiment of the present application can establish an estimation model of accurate users of the game theme according to the data of existing users of the same game theme, and can accurately mine potential accurate users for subsequent unknown users, and Filtering out some noisy users makes advertisers more accurate when advertising, and improves the cost-effectiveness of advertising. For advertising platforms, it improves the traffic efficiency of the advertising platform, and can ultimately increase the revenue of the advertising platform.

6 is a schematic structural diagram of an electronic device provided by an embodiment of the present application, including: a processor 501, a memory 502, and a bus 503, the memory 502 stores machine-readable instructions executable by the processor 501, and the The processor 501 communicates with the memory 502 through the bus 1103, and the processor 501 executes the machine-readable instructions to execute the steps in the method embodiments shown in FIG. 1 to FIG. 3 .

An embodiment of the present application further provides a chip, the chip includes an input and output interface, at least one processor, at least one memory, and a bus, the at least one memory is used to store instructions, and the at least one processor is used to call the at least one memory. , to execute the steps in the method embodiment shown in FIG. 1 to FIG. 3 .

Embodiments of the present application further provide a computer-readable storage medium, where a computer program is stored on the computer-readable storage medium, and when the computer program is run by a processor, the steps in the method embodiments shown in FIG. 1 to FIG. 3 are executed.

It should be understood that the specific examples in the embodiments of the present application are only for helping those skilled in the art to better understand the embodiments of the present application, rather than limiting the scope of the embodiments of the present application.

It should be understood that the terms used in the embodiments of the present application and the appended claims are only for the purpose of describing specific embodiments, and are not intended to limit the embodiments of the present application. For example, as used in the embodiments of this application and the appended claims, the singular forms "a," "above," and "the" are intended to include the plural forms as well, unless the context clearly dictates otherwise.

It should be understood that the processor or processing unit in this embodiment of the present application may be an integrated circuit chip, which has the capability of processing signals. In the implementation process, each step of the above method embodiments may be completed by a hardware integrated logic circuit in a processor or an instruction in the form of software. The above-mentioned processor may be a general-purpose processor, a digital signal processor (Digital Signal Processor, DSP), an application specific integrated circuit (Application Specific Integrated Circuit, ASIC), an off-the-shelf programmable gate array (Field Programmable Gate Array, FPGA) or other programmable Logic devices, discrete gate or transistor logic devices, discrete hardware components. The methods, steps, and logic block diagrams disclosed in the embodiments of this application can be implemented or executed. A general purpose processor may be a microprocessor or the processor may be any conventional processor or the like. The steps of the method disclosed in conjunction with the embodiments of the present application may be directly embodied as executed by a hardware decoding processor, or executed by a combination of hardware and software modules in the decoding processor. The software modules may be located in random access memory, flash memory, read-only memory, programmable read-only memory or electrically erasable programmable memory, registers and other storage media mature in the art. The storage medium is located in the memory, and the processor reads the information in the memory, and completes the steps of the above method in combination with its hardware.

It can be understood that the memory of the embodiment of the present application may be a volatile memory or a non-volatile memory, or may include both volatile and non-volatile memory. Wherein, the non-volatile memory may be Read-Only Memory (ROM), Programmable Read-Only Memory (Programmable ROM, PROM), Erasable Programmable Read-Only Memory (Erasable PROM, EPROM), Electrically Erasable Memory Except programmable read-only memory (Electrically EPROM, EEPROM) or flash memory. The volatile memory may be random access memory (RAM), which is used as an external cache. By way of example and not limitation, many forms of RAM are available, such as Static RAM (SRAM), Dynamic RAM (DRAM), Synchronous DRAM, SDRAM), double data rate synchronous dynamic random access memory (Double Data Rate SDRAM, DDR SDRAM), enhanced synchronous dynamic random access memory (Enhanced SDRAM, ESDRAM), synchronous link dynamic random access memory (Synchlink DRAM, SLDRAM) And direct memory bus random access memory (DirectRambus RAM, DR RAM). It should be noted that the memory of the systems and methods described herein is intended to include, but not be limited to, these and any other suitable types of memory.

The embodiments of the present application also provide a computer program, where the computer program includes instructions, when the computer program is executed by a computer, so that the computer can execute the content of the method embodiments.

It should be noted that, on the premise of no conflict, each embodiment described in this application and/or the technical features in each embodiment can be arbitrarily combined with each other, and the technical solution obtained after the combination should also fall within the protection scope of this application .

It should be understood that the specific examples in the embodiments of the present application are only to help those skilled in the art to better understand the embodiments of the present application, rather than limiting the scope of the embodiments of the present application, and those skilled in the art can Various improvements and modifications can be made, and these improvements or modifications all fall within the protection scope of the present application.

The above are only specific embodiments of the present application, but the protection scope of the present application is not limited to this. should be covered within the scope of protection of this application. Therefore, the protection scope of the present application should be subject to the protection scope of the claims.

Claims (21)

1. A push method, comprising:

extracting user characteristics and context characteristics of unknown users and industry characteristics of games to be pushed;

determining a first estimation model in a plurality of estimation models according to the game category to which the game to be pushed belongs, wherein the estimation models are estimation models based on different game categories, the game to be pushed belongs to a first game category, and the first estimation model is an estimation model of the first game category;

inputting the user characteristics and the context characteristics of the unknown users and the industry characteristics of the games to be pushed into the first estimation model, and outputting target users of the unknown users;

and pushing the game to be pushed to the target user.

2. The method of claim 1, further comprising:

inputting the target distinguishing characteristics of the users of the first game category, the industry characteristics of the games belonging to the first game category and the positive and negative samples of the first game category into the first estimation model for training to obtain the parameters of the first estimation model.

3. The method of claim 2, further comprising:

and performing characteristic engineering on the data of the existing users of the games of different game categories to acquire the target distinguishing characteristics of the users of different game categories.

4. The method of claim 3, wherein said performing feature engineering on data of existing users of games of different game categories to obtain target distinguishing features of users of different game categories comprises:

constructing positive and negative samples of the first game item according to data of existing users of at least one game belonging to the first game item;

analyzing the contribution degrees of the user characteristics corresponding to the positive and negative samples of different game categories, and determining the basic characteristics of the users of different game categories;

analyzing the crowd portraits of the users of different game categories according to the basic characteristics of the positive and negative samples of the different game categories, and determining the distinguishing characteristics of the users of the different game categories;

and performing feature association on the distinguishing features to obtain target distinguishing features of users distinguishing different game categories.

5. The method of claim 4, wherein said feature associating said distinctive features to obtain a target distinctive feature for distinguishing users of different categories of games comprises:

acquiring historical behavior data of game users of different game categories;

and performing feature association on the distinguishing features according to the historical behavior data of the game users of different game categories to obtain the target distinguishing features.

6. The method according to any one of claims 2-5, wherein the extracting the user features of the unknown user comprises:

and extracting the target distinguishing features of the unknown user.

7. The method of claim 4, wherein said constructing positive and negative examples of said first category of games from data of existing users of at least one game belonging to said first category of games comprises:

for each game of the at least one game, user data and game data that have been user activated for a period of time and have a depth conversion event are taken as positive examples of the first game category, and user data and game data that have been user activated for a period of time and have not had a depth conversion event are taken as negative examples of the first game category.

8. The method of claim 7, further comprising:

and if the same user data and game data exist in both the positive sample and the negative sample of the first game item, removing the user data and the game data from the negative sample of the first game item.

9. The method according to claim 7 or 8, characterized in that the method further comprises:

removing game data in the positive and negative samples of the first game item;

and inputting the positive and negative samples of the first game item with the game data removed into the first estimation model for training.

10. A push method, comprising:

acquiring data of existing users of games of different game categories;

constructing a positive and negative sample of the same game item according to the data of the existing users of at least one game belonging to the same game item;

performing characteristic engineering on the data of the existing users of the games of different game categories to acquire target distinguishing characteristics of the users of different game categories;

inputting the industry characteristics of games belonging to the same game category, the target distinguishing characteristics of users of the games belonging to the same game category and the positive and negative samples of the same game category into the estimation model corresponding to the same game category for training to obtain the estimation model corresponding to the same game category.

11. The method of claim 10, wherein said constructing positive and negative examples of the same game item based on data from existing users of at least one game belonging to the same game item comprises:

for each game of the at least one game, the user data and game data that have been user activated for a period of time and have a depth conversion event are taken as positive examples of the same game category, and the user data and game data that have been user activated for a period of time and have no depth conversion event are taken as negative examples of the same game category.

12. The method of claim 11, further comprising:

and if the same user data and game data exist in the positive sample of the same game item and the negative sample of the same game item, removing the user data and the game data from the negative sample of the same game item.

13. The method according to any one of claims 10-12, further comprising:

removing game data in the positive and negative samples of the same game item class;

and inputting the positive and negative samples of the same game item without the game data into the estimation model corresponding to the same game item for training.

14. The method of any one of claims 10 to 12, wherein said performing feature engineering on data of existing users of games of different game categories to obtain target distinguishing features of users of different game categories comprises:

analyzing the contribution degrees of the user characteristics corresponding to the positive and negative samples of different game categories, and determining the basic characteristics of the users of different game categories;

analyzing the crowd portraits of the users of different game categories according to the basic characteristics of the positive and negative samples of the different game categories, and determining the distinguishing characteristics of the users of the different game categories;

and performing feature association on the distinguishing features to obtain target distinguishing features of users distinguishing different game categories.

15. The method of claim 14, wherein said feature associating said distinctive features to obtain a target distinctive feature for distinguishing users of different categories of games comprises:

acquiring historical behavior data of game users of different game categories;

and performing feature association on the distinguishing features according to the historical behavior data of the game users of different game categories to obtain the target distinguishing features.

16. A pushing device, comprising:

the extracting unit is used for extracting the user characteristics and the context characteristics of the unknown user and the industry characteristics of the game to be pushed;

the determining unit is used for determining a first pre-estimation model in a plurality of pre-estimation models according to the game category to which the game to be pushed belongs, wherein the plurality of pre-estimation models are pre-estimation models based on different game categories, the game to be pushed belongs to a first game category, and the first pre-estimation model is a pre-estimation model of the first game category;

the estimation unit is used for inputting the user characteristics and the context characteristics of the unknown user and the industry characteristics of the game to be pushed into the first estimation model and outputting a target user in the unknown user;

and the pushing unit is used for pushing the game to be pushed to the target user.

17. A pushing device, comprising:

the system comprises an acquisition unit, a storage unit and a processing unit, wherein the acquisition unit is used for acquiring data of existing users of games of different game categories;

the building unit is used for building a positive sample and a negative sample of the same game item according to the data of the existing users of at least one game belonging to the same game item;

the processing unit is used for performing characteristic engineering on the data of the existing users of the games of different game categories to acquire target distinguishing characteristics of the users of different game categories;

and the training unit is used for inputting the industrial characteristics of the games belonging to the same game category, the target distinguishing characteristics of the users of the games belonging to the same game category and the positive and negative samples of the same game category into the estimation model corresponding to the same game category for training to obtain the estimation model corresponding to the same game category.

18. An electronic device, comprising: a processor, a storage medium and a bus, the storage medium storing machine-readable instructions executable by the processor, the processor and the storage medium communicating via the bus when the electronic device is operating, the processor executing the machine-readable instructions to perform the steps of the push method according to any one of claims 1 to 9.

19. A computer-readable storage medium, characterized in that the computer-readable storage medium is used for storing a computer program which, when executed by a processor, performs the push method according to any one of claims 1 to 9.

20. An electronic device, comprising: a processor, a storage medium and a bus, the storage medium storing machine-readable instructions executable by the processor, the processor and the storage medium communicating via the bus when the electronic device is operating, the processor executing the machine-readable instructions to perform the steps of the push method according to any one of claims 10 to 15.

21. A computer-readable storage medium, characterized in that the computer-readable storage medium is used for storing a computer program which, when executed by a processor, performs the push method according to any one of claims 10 to 15.

Images