CN116976991A

CN116976991A - Advertisement recommendation data determining method, device, computer equipment and storage medium

Info

Publication number: CN116976991A
Application number: CN202310525089.2A
Authority: CN
Inventors: 邓慧挺; 陈楚丰
Original assignee: Tencent Technology Shenzhen Co Ltd
Current assignee: Tencent Technology Shenzhen Co Ltd
Priority date: 2023-05-10
Filing date: 2023-05-10
Publication date: 2023-10-31

Abstract

The application relates to an advertisement recommendation data determining method, device, equipment and storage medium. The method involves artificial intelligence, comprising: and obtaining advertisement characteristic values of the target object under different advertisement dimensions and candidate advertisement queues comprising a plurality of candidate advertisement data, wherein each candidate advertisement data corresponds to each advertisement dimension, and carrying out characteristic cross matching processing on the advertisement characteristic values and the candidate advertisement data under different advertisement dimensions to obtain advertisement matching characteristics. Determining a first advertisement feature vector stack based on advertisement features and advertisement matching features of the target use object, obtaining a first fusion feature vector stack according to the object feature vector stack and the first advertisement feature vector stack of the target use object, performing matching prediction processing based on the first fusion feature vector stack, obtaining a first matching prediction result, and determining advertisement recommendation data matched with the target use object based on the first matching prediction result. By adopting the method, the accuracy of the determined advertisement recommendation data can be improved.

Description

Advertisement recommendation data determining method, device, computer equipment and storage medium

Technical Field

The present application relates to the field of artificial intelligence technology, and in particular, to a method, an apparatus, a computer device, a storage medium, and a computer program product for determining advertisement recommendation data.

Background

With the development of artificial intelligence technology and the gradual popularization and application of Internet application, the online advertisement popularization and application method is combined with Internet application, and compared with the traditional offline advertisement delivery method, the online advertisement delivery method based on Internet has the advantages of wide path, rich content and low cost, and therefore, the online advertisement delivery method based on Internet is greatly popularized and applied.

When online advertisement promotion is carried out on the Internet, advertisement click rates or conversion rates and the like of different Internet advertisements are required to be predicted, so that a corresponding Internet advertisement delivery platform, a corresponding Internet advertisement delivery object, a corresponding Internet advertisement delivery time period, actual advertisement delivery content and the like are determined according to the obtained prediction results, and further benefits brought by advertisement delivery are improved.

Conventionally, when online advertisement promotion is performed, a mode of determining advertisement click rate or conversion rate according to historical click and access records of a use object on each internet advertisement and the like so as to determine advertisement with higher matching degree from a candidate advertisement queue for recommendation is generally adopted. However, since the clicking and access records of the used objects are disordered, and some online advertisements are missed by the used objects due to fewer delivery paths, and thus the determined advertisement clicking rate or conversion rate error is larger, the determined recommended advertisements still have the problem of lower accuracy according to the clicking rate and the like of the advertisements.

Disclosure of Invention

In view of the foregoing, it is desirable to provide an advertisement recommendation data determining method, apparatus, computer device, computer readable storage medium, and computer program product capable of improving the accuracy of recommending advertisements.

In a first aspect, the present application provides a method for determining advertisement recommendation data. The method comprises the following steps:

acquiring advertisement characteristic values of a target use object under different advertisement dimensions, and acquiring a candidate advertisement queue to be recommended, wherein the candidate advertisement queue comprises a plurality of candidate advertisement data, and each candidate advertisement data corresponds to each advertisement dimension;

respectively carrying out feature cross matching processing on advertisement feature values and candidate advertisement data under different advertisement dimensions to obtain advertisement matching features; the advertisement matching characteristics comprise advertisement dimensions and advertisement characteristic values which are matched with each other;

determining a first advertisement feature vector stack based on advertisement features corresponding to the target usage object and the advertisement matching features;

splicing to obtain a first fusion feature vector stack according to an object feature vector stack corresponding to the target use object and the first advertisement feature vector stack;

And carrying out matching prediction processing based on the first fusion feature vector stack to obtain a corresponding first matching prediction result, and determining advertisement recommendation data matched with the target use object based on the first matching prediction result.

In a second aspect, the application further provides an advertisement recommendation data determining method. The method comprises the following steps:

according to the sub-cross network matched with the advertisement dimension, carrying out cross matching processing on the advertisement characteristic value embedded vector and the advertisement dimension embedded vector under the advertisement dimension corresponding to the sub-cross network to obtain an intermediate advertisement vector;

based on an object feature vector stack corresponding to the target use object and a second advertisement feature vector stack, splicing to obtain a second fusion feature vector stack;

and carrying out matching prediction processing on the intermediate advertisement vector and the second fusion feature vector stack to obtain a corresponding second matching prediction result, and determining advertisement recommendation data matched with the target use object based on the second matching prediction result.

In a third aspect, the application further provides an advertisement recommendation data determining device. The device comprises:

the data acquisition module is used for acquiring advertisement characteristic values of the target use object under different advertisement dimensions and acquiring candidate advertisement queues to be recommended, wherein the candidate advertisement queues comprise a plurality of candidate advertisement data, and each candidate advertisement data corresponds to each advertisement dimension respectively;

the advertisement matching feature obtaining module is used for respectively carrying out feature cross matching processing on advertisement feature values and candidate advertisement data under different advertisement dimensions to obtain advertisement matching features; the advertisement matching characteristics comprise advertisement dimensions and advertisement characteristic values which are matched with each other;

a first advertisement feature vector stack determining module, configured to determine a first advertisement feature vector stack based on advertisement features corresponding to the target usage object and the advertisement matching features;

the first fusion feature vector stack obtaining module is used for splicing the object feature vector stack corresponding to the target use object and the first advertisement feature vector stack to obtain a first fusion feature vector stack;

and the advertisement recommendation data determining module is used for carrying out matching prediction processing based on the first fusion feature vector stack to obtain a corresponding first matching prediction result, and determining advertisement recommendation data matched with the target use object based on the first matching prediction result.

In a fourth aspect, the application further provides an advertisement recommendation data determining device. The device comprises:

the intermediate advertisement vector obtaining module is used for carrying out cross matching processing on the advertisement characteristic value embedded vector and the advertisement dimension embedded vector under the advertisement dimension corresponding to the sub-cross network according to the sub-cross network matched with the advertisement dimension to obtain an intermediate advertisement vector;

the second fusion feature vector stack obtaining module is used for splicing the second fusion feature vector stack based on the object feature vector stack corresponding to the target use object and the second advertisement feature vector stack;

and the advertisement recommendation data determining module is used for carrying out matching prediction processing on the intermediate advertisement vector and the second fusion feature vector stack to obtain a corresponding second matching prediction result, and determining advertisement recommendation data matched with the target use object based on the second matching prediction result.

In a fifth aspect, the present application also provides a computer device. The computer device comprises a memory storing a computer program and a processor which when executing the computer program performs the steps of:

In a sixth aspect, the present application also provides a computer device. The computer device comprises a memory storing a computer program and a processor which when executing the computer program performs the steps of:

In a seventh aspect, the present application also provides a computer-readable storage medium. The computer readable storage medium having stored thereon a computer program which when executed by a processor performs the steps of:

In an eighth aspect, the present application also provides a computer-readable storage medium. The computer readable storage medium having stored thereon a computer program which when executed by a processor performs the steps of:

In a ninth aspect, the present application also provides a computer program product. The computer program product comprises a computer program which, when executed by a processor, implements the steps of:

In a tenth aspect, the present application also provides a computer program product. The computer program product comprises a computer program which, when executed by a processor, implements the steps of:

In the method, the device, the computer equipment, the storage medium and the computer program product for determining advertisement recommendation data, advertisement characteristic values of the target usage object under different advertisement dimensions and candidate advertisement queues comprising a plurality of candidate advertisement data are obtained, wherein each candidate advertisement data corresponds to each advertisement dimension respectively, and through characteristic cross matching processing on the advertisement characteristic values under the different advertisement dimensions and the candidate advertisement data respectively, advertisement matching characteristics comprising the advertisement dimensions and the advertisement characteristic values which are matched with each other are obtained, so that comprehensive consideration is carried out on the candidate advertisement data according to the advertisement dimensions and the advertisement characteristic values, and meanwhile, the advertisement characteristic values of the target usage object under different advertisement dimensions are accurately depicted. Further, based on the advertisement feature corresponding to the target use object and the advertisement matching feature, a first advertisement feature vector stack is determined, and according to the object feature vector stack corresponding to the target use object and the first advertisement feature vector stack, a more comprehensive first fusion feature vector stack is obtained by splicing, so that a corresponding first matching prediction result can be obtained by carrying out matching prediction processing on the first fusion feature vector stack, and therefore advertisement recommendation data matched with the target use object can be determined based on the first matching prediction result, and the accuracy of the determined advertisement recommendation data is further improved.

Drawings

FIG. 1 is an application environment diagram of a method of determining advertisement recommendation data in one embodiment;

FIG. 2 is a flow diagram of a method for determining advertisement recommendation data, in one embodiment;

fig. 3 is a schematic illustration of promotion performance corresponding to an advertisement recommendation data determination method in an embodiment;

FIG. 4 is a flow diagram illustrating a general process of a method for determining advertisement recommendation data in one embodiment;

FIG. 5 is a flowchart of a method for determining advertisement recommendation data according to another embodiment;

FIG. 6 is a flow chart of a method of determining advertisement recommendation data according to yet another embodiment;

FIG. 7 is a flowchart illustrating an overall method of determining advertisement recommendation data according to another embodiment;

FIG. 8 is a block diagram of an advertisement recommendation data determination apparatus in one embodiment;

FIG. 9 is a block diagram showing a construction of an advertisement recommendation data determining apparatus according to another embodiment;

fig. 10 is an internal structural view of a computer device in one embodiment.

Detailed Description

The present application will be described in further detail with reference to the drawings and examples, in order to make the objects, technical solutions and advantages of the present application more apparent. It should be understood that the specific embodiments described herein are for purposes of illustration only and are not intended to limit the scope of the application.

The advertisement recommendation data determining method provided by the embodiment of the application relates to an artificial intelligence technology, and can be applied to various scenes such as cloud technology, artificial intelligence, intelligent traffic, network media, auxiliary driving and the like. Among these, artificial intelligence (Artificial Intelligence, AI) is the theory, method, technique and application system that uses a digital computer or a digital computer-controlled machine to simulate, extend and extend human intelligence, sense the environment, acquire knowledge and use knowledge to obtain optimal results. In other words, artificial intelligence is an integrated technology of computer science that attempts to understand the essence of intelligence and to produce a new intelligent machine that can react in a similar way to human intelligence. Artificial intelligence, i.e. research on design principles and implementation methods of various intelligent machines, enables the machines to have functions of sensing, reasoning and decision. The artificial intelligence technology is a comprehensive subject, and relates to the technology with wide fields, namely the technology with a hardware level and the technology with a software level. Artificial intelligence infrastructure technologies generally include technologies such as sensors, dedicated artificial intelligence chips, cloud computing, distributed storage, big data processing technologies, operation/interaction systems, mechatronics, and the like. The artificial intelligence software technology mainly comprises a computer vision technology, a voice processing technology, a natural language processing technology, machine learning/deep learning and other directions.

Among them, natural language processing (Nature Language processing, NLP) is an important direction in the fields of computer science and artificial intelligence. It is studying various theories and methods that enable effective communication between a person and a computer in natural language. Natural language processing is a science that integrates linguistics, computer science, and mathematics. Thus, the research in this field will involve natural language, i.e. language that people use daily, so it has a close relationship with the research in linguistics. Natural language processing techniques typically include text processing, semantic understanding, machine translation, robotic questions and answers, knowledge graph techniques, and the like.

Machine Learning (ML) is a multi-domain interdisciplinary, which relates to multiple disciplines such as probability theory, statistics, approximation theory, convex analysis, and algorithm complexity theory, and is used to specially study how a computer simulates or implements Learning behavior of a human being so as to obtain new knowledge or skill, and reorganize an existing knowledge structure to continuously improve its own performance. Machine learning is the core of artificial intelligence, a fundamental approach to letting computers have intelligence, which is applied throughout various areas of artificial intelligence. Machine learning and deep learning typically include techniques such as artificial neural networks, confidence networks, reinforcement learning, transfer learning, induction learning, and teaching learning.

With research and advancement of artificial intelligence technology, research and application of artificial intelligence technology is being developed in various fields, such as common smart home, smart wearable devices, virtual assistants, smart speakers, smart marketing, unmanned, automatic driving, unmanned aerial vehicles, robots, smart medical treatment, smart customer service, etc., and it is believed that with the development of technology, artificial intelligence technology will be applied in more fields and with increasing importance value.

The advertisement recommendation data determining method provided by the embodiment of the application particularly relates to natural language processing, machine learning and other technologies in the artificial intelligence technology, and can be applied to an application environment shown in fig. 1. Wherein the terminal 102 communicates with the server 104 via a network. The data storage system may store data that the server 104 needs to process. The data storage system may be integrated on the server 104 or may be located on a cloud or other network server. The terminal 102 may be, but not limited to, various personal computers, notebook computers, smart phones, tablet computers, internet of things devices, portable wearable devices, aircrafts, etc., and the internet of things devices may be smart speakers, smart car devices, etc. The portable wearable device may be a smart watch, smart bracelet, headset, or the like. The server 104 may be an independent physical server, or may be a server cluster formed by a plurality of physical servers, or may be a cloud server that provides cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communication, middleware services, domain name services, security services, CDNs, and basic cloud computing services such as big data and artificial intelligence platforms, where the terminal 102 and the server 104 may be directly or indirectly connected through wired or wireless communication modes, which is not limited in the embodiment of the present application.

Further, both the terminal 102 and the server 104 may be separately used to perform the advertisement recommendation data determining method provided in the embodiment of the present application, and the terminal 102 and the server 104 may also cooperatively perform the advertisement recommendation data determining method provided in the embodiment of the present application. For example, taking the terminal 102 and the server 104 cooperatively execute the method for determining advertisement recommendation data provided in the embodiment of the present application as an example, the server 104 obtains advertisement feature values of the target usage object in different advertisement dimensions, and obtains a candidate advertisement queue to be recommended including a plurality of candidate advertisement data, where each candidate advertisement data corresponds to a respective advertisement dimension. The advertisement feature values of the target usage object in different advertisement dimensions and the candidate advertisement queues may be stored in a cloud storage of the server 104, or in a data storage system, or in a local storage of the terminal 102, and may be acquired from the server 104, or the data storage system, or the terminal 102 when the advertisement recommendation data determination process is required. Further, the server 104 performs feature cross matching processing on the advertisement feature values and the candidate advertisement data in different advertisement dimensions to obtain advertisement matching features including advertisement dimensions and advertisement feature values which are matched with each other, further determines a first advertisement feature vector stack based on the advertisement features corresponding to the target usage object and the advertisement matching features, and splices the first fusion feature vector stack according to the object feature vector stack corresponding to the target usage object and the first advertisement feature vector stack. Finally, the server 104 may obtain a corresponding first matching prediction result by performing a matching prediction process based on the first fused feature vector stack, and determine advertisement recommendation data matching the target usage object based on the first matching prediction result. The server 104 may send the obtained advertisement recommendation data to the terminal 102 where the matched target usage object is located.

Likewise, taking the terminal 102 and the server 104 cooperatively execute the method for determining advertisement recommendation data provided in the embodiment of the present application as an example, specifically, the server 104 obtains advertisement feature values of the target usage object under different advertisement dimensions, and obtains a candidate advertisement queue to be recommended including candidate advertisement data belonging to a plurality of candidate advertisement data, where each candidate advertisement data corresponds to a respective advertisement dimension. Further, the server 104 performs cross matching processing on the advertisement feature value embedded vector and the advertisement dimension embedded vector in the advertisement dimension corresponding to each sub-cross network according to the sub-cross network matched with the advertisement dimension, so as to obtain an intermediate advertisement vector, and based on the object feature vector stack corresponding to the target usage object and the second advertisement feature vector stack, splices to obtain a second fusion feature vector stack. Finally, the server 104 may obtain a corresponding second matching prediction result by performing a matching prediction process on the intermediate advertisement vector and the second fusion feature vector stack, so as to determine advertisement recommendation data matching the target usage object based on the second matching prediction result. Likewise, the server 104 may send the obtained advertisement recommendation data to the terminal 102 where the matched target usage object is located.

In one embodiment, as shown in fig. 2, there is provided an advertisement recommendation data determining method, which is described by taking an example that the method is applied to the server in fig. 1, and includes the following steps:

step S202, obtaining advertisement characteristic values of a target usage object under different advertisement dimensions, and obtaining a candidate advertisement queue to be recommended, wherein the candidate advertisement queue comprises a plurality of candidate advertisement data, and each candidate advertisement data corresponds to each advertisement dimension.

The target usage object can be understood as usage objects of different application programs, platforms, web pages or the like, and personalized advertisement recommendation can be performed on each target usage object according to advertisement recommendation data matched with the target usage object in the candidate advertisement queue.

The advertisement dimension can understand dimensions of each advertisement data, such as dimensions of an advertisement industry, an advertisement recommended party, advertisement related products and the like, wherein the advertisement industry can specifically comprise a primary advertisement industry, a secondary advertisement industry and the like, the primary advertisement industry can specifically comprise a game industry, an electronic commerce industry, an entertainment industry and the like, and the secondary advertisement industry refers to a plurality of sub-industries under the primary advertisement industry, such as a game application program, a game live program and the like under the game industry, and also such as an audio and video playing program, a social application program, a short video live program and the like under the entertainment industry. The advertisement recommender may be understood as an advertiser who needs to perform advertisement promotion, and the advertisement-related product may be understood as a product that the advertiser needs to recommend, i.e. a product to be recommended, which is associated with each candidate advertisement data.

The advertisement feature value can be understood as a statistical value of each parameter of each usage object in different advertisement dimensions, and can be represented by the advertisement dimension Di-advertisement feature value Vi. The parameters in the advertisement dimension may include exposure times, click rate, conversion rate, etc., for example, the exposure times of the object a in 7 days in the primary advertisement industry may be specifically understood as the exposure times of the object a in 7 days in the game industry, the electronic commerce industry, or the entertainment industry, and if the exposure times are 50 times, the current advertisement feature value is 50 times, and when the current advertisement feature value is expressed in the Di-Vi mode, the exposure times in 7 days in different primary advertisement industries may be expressed in the game-50, the electronic commerce-50, or the entertainment-50.

The candidate advertisement queue may be understood as a plurality of candidate advertisement data to be recommended or estimated, where each candidate advertisement data corresponds to a respective advertisement dimension, for example, an advertisement dimension corresponding to a first candidate advertisement data is a first-level advertisement industry, an advertisement dimension corresponding to a second candidate advertisement data is a second-level advertisement industry, and an advertisement dimension corresponding to a third candidate advertisement data is an advertisement recommender. The advertisement dimension of the first candidate advertisement data, the second candidate advertisement data and the third candidate advertisement data are all first-level advertisement industries.

Specifically, when the advertisement feature values of the target usage object in different advertisement dimensions are obtained, specifically, the advertisement feature values of the target usage object in dimensions such as advertisement industry, advertisement recommendation party, advertisement related products and the like are obtained, for example, the exposure times (or click rate, conversion rate or the like) of the target usage object in a preset time period in the primary advertisement industry (including game industry, entertainment industry or electronic commerce industry) are obtained, if the exposure times in the preset time period (such as 7 days) are determined to be 50 times, the exposure times of the target usage object in the current advertisement dimension, namely, the primary advertisement industry, are 50 times, namely, the current advertisement feature value is 50 times, and the target usage object can be represented by 'primary advertisement industry-50'.

Similarly, the candidate advertisement queue may be understood as a plurality of candidate advertisement data to be recommended or estimated, and when the candidate advertisement queue to be recommended is obtained, the plurality of candidate advertisement data to be recommended or estimated and advertisement dimensions to which each candidate advertisement data belongs, for example, the candidate advertisement queue includes: the first candidate advertisement data-game, the second candidate advertisement data-entertainment, the third candidate advertisement data-e-commerce …, etc. can be understood that the advertisement dimension of the first candidate advertisement data is the game industry, the advertisement dimension of the second candidate advertisement data is the entertainment industry, and the advertisement dimension of the third candidate advertisement data is the e-commerce industry.

Step S204, respectively carrying out feature cross matching processing on the advertisement feature values and the candidate advertisement data under different advertisement dimensions to obtain advertisement matching features, wherein the advertisement matching features comprise advertisement dimensions and advertisement feature values which are matched with each other.

The advertisement feature values of the target usage object in different advertisement dimensions can be represented by the advertisement dimension Di-advertisement feature value Vi, for example, the exposure times of the target usage object in the first-level advertisement industry including the game industry, the entertainment industry and the e-commerce industry within 7 days are 50 times, and can be represented by the modes of game-50, e-commerce-50 and entertainment-50.

Similarly, the candidate advertisement data in the candidate advertisement queue may be represented by using advertisement number adi-advertisement dimension Di, for example, the advertisement dimension of the first candidate advertisement data (ad 1) in the candidate advertisement queue is game industry, the advertisement dimension of the second candidate advertisement data (ad 2) is entertainment industry, and the advertisement dimension of the third candidate advertisement data (ad 3) is electronic business industry, and may be represented by ad 1-game, ad 2-entertainment, and ad 3-electronic business.

Specifically, by performing feature matching processing on advertisement feature values and candidate advertisement data in different advertisement dimensions, the advertisement feature values matched with the advertisement dimensions to which the candidate advertisement data belong can be obtained. Specifically, for example, advertisement feature values of the target usage object under different advertisement dimensions, including game-50, e-commerce-50 and entertainment-50, candidate advertisement data in the candidate advertisement queue are expressed as: the ad 1-game, the ad 2-entertainment and the ad 3-e-commerce perform feature matching processing on advertisement feature values and candidate advertisement data under different advertisement dimensions, which can be understood as: for example, feature matching processing is performed on games-50 and ad 1-games, feature matching processing is performed on electronic commerce-50 and ad 3-electronic commerce, feature matching processing is performed on entertainment-50 and ad 2-entertainment, so that advertisement feature values matched with advertisement dimensions to which candidate advertisement data belong can be obtained, namely advertisement feature values expressed in the form of advertisement feature values Vi of the advertisement dimension Di-of the candidate advertisement data, such as advertisement feature values expressed in the forms of ad 1-game-50, ad 2-entertainment-50, ad 3-electronic commerce-50 and the like, namely, the exposure times of the first candidate advertisement data in the game industry are 50 times, the exposure times of the second candidate advertisement data in the entertainment industry are 50 times, and the exposure times of the third candidate advertisement data in the electronic commerce industry are 50 times, namely, the specific value of the advertisement feature values is 50 times.

Further, after the advertisement feature value matched with the advertisement dimension to which the candidate advertisement data belongs is obtained, the advertisement dimension to which the candidate advertisement data belongs and the advertisement feature value matched with the advertisement dimension are further subjected to feature cross processing.

Specifically, feature cross processing may be performed on an advertisement dimension to which the candidate advertisement data belongs, that is, a game (or an electronic commerce or entertainment), and an advertisement feature value (for example, 50 times) corresponding to the advertisement dimension, so as to obtain an advertisement matching feature. For example, if the advertisement dimension to which the candidate advertisement data belongs is a game and the advertisement feature value matched with the advertisement dimension to which the candidate advertisement data belongs is 50 times, the feature cross processing is performed on the advertisement dimension to which the candidate advertisement data belongs, that is, the game industry, and the advertisement feature value of the candidate advertisement data in the game industry for 50 times, and specifically, the advertisement matching feature corresponding to the candidate advertisement data may be game_50.

For example, according to the feature of "the exposure times of the target usage object in 7 days on the primary advertising industry", the advertisement feature value Vi of the target usage object in different advertisement dimensions Di is obtained, including: when the object number is used, namely the exposure times in the game industry (namely the game-V), the exposure times in the electronic commerce industry (namely the electronic commerce-V) and the exposure times in the entertainment industry (namely the entertainment-V) are used, if the advertisement dimension of a certain candidate advertisement data in the candidate advertisement queue is the game industry when the advertisement recommendation data is required to be determined, the obtained advertisement matching characteristic is game-V after the characteristic matching processing and the characteristic crossing processing, and if the exposure times are 50 times, the obtained advertisement matching characteristic is game-50.

Step S206, determining a first advertisement feature vector stack based on the advertisement features corresponding to the target usage object and the advertisement matching features.

Specifically, by acquiring historical log data associated with the target usage object, determining initial features corresponding to the target usage object according to the historical log data, and dividing the initial features into object features corresponding to the target usage object and advertisement features according to feature attributes, a first advertisement embedded feature vector corresponding to the advertisement features corresponding to the target usage object and a second advertisement embedded feature vector corresponding to the advertisement matching features can be determined, so that the first advertisement embedded feature vector and the second advertisement embedded feature vector are spliced, and a first advertisement feature vector stack is obtained.

The history log data associated with the target usage object may include object attribute information of the target usage object, such as an object number, a consumption record, a product purchase record, and the like, and advertisement access attribute information of the target usage object to different advertisement data, such as a click number, a conversion number, an exposure time, an access time, and the like, and further, based on the history log data associated with the target usage object, an initial feature corresponding to the target usage object may be determined, that is, an initial feature including both the object attribute information of the target usage object and the advertisement access attribute information to different advertisement data.

Further, the feature attribute may be understood as belonging to an object attribute or an advertisement access attribute, if belonging to the object attribute is determined, the corresponding initial feature is classified as an object feature, and if the object attribute is determined as an advertisement access attribute, the corresponding initial feature is classified as an advertisement feature. The first advertisement embedded feature vector corresponding to the advertisement feature and the second advertisement embedded feature vector corresponding to the advertisement matching feature may include a plurality of first advertisement embedded feature vectors and a plurality of second advertisement embedded feature vectors, which may be specifically spliced to obtain a first advertisement feature vector stack.

In one embodiment, when determining a first advertisement embedded feature vector corresponding to an advertisement feature corresponding to a target usage object and a second advertisement embedded feature vector corresponding to an advertisement matching feature, the first advertisement sub-encoded feature vector is converted into the first advertisement sub-encoded feature vector, and the advertisement matching feature is converted into the second advertisement sub-encoded feature vector, so that the first advertisement sub-encoded feature vector can be converted into the first advertisement embedded feature vector, and the second advertisement sub-encoded feature vector can be converted into the second advertisement embedded feature vector.

Specifically, when converting the advertisement feature corresponding to the target usage object into the first advertisement sub-coding feature vector and converting the advertisement matching feature into the second advertisement sub-coding feature vector, the advertisement feature and the advertisement matching feature may be specifically converted into the form of onehot vectors (i.e., unicode feature vectors), that is, the first advertisement sub-onehot vectors and the second advertisement sub-onehot vectors are obtained, and the first advertisement sub-onehot vectors are further converted into the first advertisement embedding feature vector and the second advertisement sub-onehot vectors are converted into the second advertisement embedding feature vector.

Step S208, according to the object feature vector stack corresponding to the target use object and the first advertisement feature vector stack, a first fusion feature vector stack is obtained by splicing.

Specifically, by acquiring history log data associated with the target usage object, and determining an initial feature corresponding to the target usage object from the history log data, the initial feature is further divided into an object feature corresponding to the target usage object and an advertisement feature according to the feature attribute. Further, an object feature vector stack corresponding to the object feature corresponding to the target use object is determined, and the object feature vector stack corresponding to the target use object and the first advertisement feature vector stack are spliced to obtain a first fusion feature vector stack.

When determining the object feature vector stack corresponding to the object feature corresponding to the target object, specifically, the object feature vector stack corresponding to the target object is obtained by converting the object feature corresponding to the target object into the object code feature vector, converting the object code feature vector into the object embedded feature vector, and splicing the object feature vector stack corresponding to the target object based on the object embedded feature vectors.

When the object feature corresponding to the target usage object is converted into the object encoding feature vector, specifically, the object feature is converted into a form of an onehot vector (i.e. a unihot encoding feature vector), namely, the object onehot vector is obtained, and the object onehot vector is further converted into an object embedding feature vector, so that an object feature vector stack corresponding to the target usage object can be spliced based on a plurality of object embedding feature vectors.

Step S210, carrying out matching prediction processing based on the first fusion feature vector stack to obtain a corresponding first matching prediction result, and determining advertisement recommendation data matched with the target use object based on the first matching prediction result.

Specifically, according to the trained multi-layer cross network model, matching prediction processing is carried out on the first fusion feature vector stack, a corresponding first matching prediction result is obtained, and then according to the first matching prediction result and auxiliary attribute information associated with candidate advertisement data, advertisement recommendation data matched with a target use object is determined.

The trained multi-layer cross network model may be a deep neural network (i.e. Deep Neural Networks) including a plurality of sub-cross networks and used for predicting the matching result, specifically, a convolutional neural network (Convolutional Neural Networks, i.e. CNN), a cyclic neural network (Recurrent Neural Network, i.e. RNN), and a generating countermeasure network (Generative Adversarial Networks, i.e. GAN).

Further, when the matching prediction processing is performed on the first fusion feature vector stack according to the trained multi-layer cross network model, a first matching prediction result between the target use object and each candidate advertisement data can be obtained in a prediction mode. Auxiliary attribute information associated with the candidate advertisement data, such as advertisement bid corresponding to the candidate advertisement data, advertisement calibration evaluation index (i.e., PCOC, which is generally called predict click over click, and represents a ratio of click rate after calibration to posterior click rate (i.e., approximate true probability)), and advertisement exposure policy, is also required to be acquired, so that advertisement recommendation data matched with the target usage object is determined by comprehensively considering the first matching prediction result and the auxiliary attribute information associated with the candidate advertisement data.

In one embodiment, after determining the advertisement recommendation data matched with the target usage object, the advertisement recommendation data is further sent to the corresponding target usage object, and consumption data of each target usage object in a preset time period (for example, 24 hours, 36 hours, 48 hours, 72 hours, or the like) is detected to obtain product transaction data in the preset time period, and the product transaction data is analyzed for increasing amplitude.

Further, as shown in fig. 3, a schematic popularization performance diagram corresponding to the advertisement recommendation data determining method is provided, and product transaction data in a preset time period may specifically be commodity transaction total amount (i.e. GMV, which is fully called Gross Merchandise Volume), namely, after advertisement recommendation data matched with a target usage object is determined by the advertisement recommendation data determining method provided by the embodiment of the present application, and the advertisement recommendation data is pushed to the target usage object, the obtained product transaction data (i.e. commodity transaction total amount) of the product is improved (the increase amplitude shown in fig. 3 is 4.6%), namely, by more accurately characterizing the characteristics of the target usage object in each advertisement dimension, more matched advertisement recommendation data may be determined, and the personalized recommendation is realized, and at the same time, the corresponding product transaction data is increased.

In the method for determining advertisement recommendation data, advertisement feature values of target usage objects in different advertisement dimensions and candidate advertisement queues comprising a plurality of candidate advertisement data are obtained, wherein each candidate advertisement data corresponds to each advertisement dimension respectively, feature cross matching processing is conducted on the advertisement feature values in different advertisement dimensions and the candidate advertisement data respectively, advertisement matching features comprising advertisement dimensions and advertisement feature values which are matched with each other are obtained, comprehensive consideration is conducted on the candidate advertisement data according to the advertisement dimensions and the advertisement feature values, and meanwhile advertisement feature values of the target usage objects in different advertisement dimensions are accurately depicted. Further, based on the advertisement feature corresponding to the target use object and the advertisement matching feature, a first advertisement feature vector stack is determined, and according to the object feature vector stack corresponding to the target use object and the first advertisement feature vector stack, a more comprehensive first fusion feature vector stack is obtained by splicing, so that matching prediction processing is carried out based on the first fusion feature vector stack, a corresponding first matching prediction result is obtained, and therefore advertisement recommendation data matched with the target use object can be determined based on the first matching prediction result, and accuracy of the determined advertisement recommendation data is further improved.

In one embodiment, as shown in fig. 4, there is provided an advertisement recommendation data determining method, which specifically includes:

step S401, obtaining advertisement characteristic values of a target usage object under different advertisement dimensions, and obtaining a candidate advertisement queue to be recommended, wherein the candidate advertisement queue comprises a plurality of candidate advertisement data, and each candidate advertisement data corresponds to each advertisement dimension.

Step S402, carrying out feature matching processing on the advertisement feature values and the candidate advertisement data in different advertisement dimensions to obtain advertisement feature values matched with the advertisement dimensions to which the candidate advertisement data belong.

Step S403, carrying out feature cross matching processing on the advertisement dimension to which the candidate advertisement data belongs and the advertisement feature value matched with the advertisement dimension to obtain advertisement matching features.

Step S404, acquiring history log data associated with the target usage object, and determining initial characteristics corresponding to the target usage object according to the history log data.

Step S405, the initial feature is divided into an object feature corresponding to the target usage object and an advertisement feature according to the feature attribute corresponding to the initial feature.

Step S406, converting the advertisement feature corresponding to the target usage object into a first advertisement subcode feature vector, and converting the advertisement matching feature into a second advertisement subcode feature vector.

Step S407, converting the first advertisement subcode feature vector into a first advertisement embedded feature vector and converting the second advertisement subcode feature vector into a second advertisement embedded feature vector.

Step S408, according to each first advertisement embedded feature vector and each second advertisement embedded feature vector, a first advertisement feature vector stack is obtained by splicing.

Step S409, converts the object feature corresponding to the target usage object into an object encoding feature vector, and converts the object encoding feature vector into an object embedding feature vector.

Step S410, based on the embedded feature vectors of the objects, the object feature vector stacks corresponding to the target use objects are spliced.

Step S411, according to the object feature vector stack corresponding to the target usage object and the first advertisement feature vector stack, a first fusion feature vector stack is obtained by splicing.

And step S412, carrying out matching prediction processing on the first fusion feature vector stack according to the trained multi-layer cross network model to obtain a corresponding first matching prediction result.

Step S413, determining advertisement recommendation data matching the target usage object according to the first matching prediction result and the auxiliary attribute information associated with the candidate advertisement data.

In the method for determining advertisement recommendation data, advertisement feature values of target usage objects under different advertisement dimensions and candidate advertisement queues comprising a plurality of candidate advertisement data are obtained, wherein each candidate advertisement data corresponds to each advertisement dimension respectively, feature cross matching processing is carried out on the advertisement feature values under different advertisement dimensions and the candidate advertisement data respectively, advertisement matching features comprising the advertisement dimensions and the advertisement feature values which are matched with each other are obtained, comprehensive consideration is carried out on the candidate advertisement data according to the advertisement dimensions and the advertisement feature values, and meanwhile the advertisement feature values of the target usage objects under different advertisement dimensions are accurately represented. Further, based on the advertisement feature corresponding to the target use object and the advertisement matching feature, a first advertisement feature vector stack is determined, and according to the object feature vector stack corresponding to the target use object and the first advertisement feature vector stack, a more comprehensive first fusion feature vector stack is obtained by splicing, so that matching prediction processing is carried out based on the first fusion feature vector stack, a corresponding first matching prediction result is obtained, and therefore advertisement recommendation data matched with the target use object can be determined based on the first matching prediction result, and accuracy of the determined advertisement recommendation data is further improved.

In one embodiment, as shown in fig. 5, there is provided an advertisement recommendation data determining method, which is described by taking the server in fig. 1 as an example, including the following steps:

step S502, obtaining advertisement characteristic values of a target usage object under different advertisement dimensions, and obtaining a candidate advertisement queue to be recommended, wherein the candidate advertisement queue comprises a plurality of candidate advertisement data, and each candidate advertisement data corresponds to each advertisement dimension.

The target usage object may be understood as usage objects of different application programs, platforms or web pages, etc., the advertisement dimension may be understood as a dimension to which each advertisement data belongs, such as dimensions of an advertisement industry (a primary advertisement industry, a secondary advertisement industry, etc., the primary advertisement industry includes a game industry, an electronic commerce industry, an entertainment industry, etc.), an advertisement recommender, an advertisement associated product, etc., and the advertisement feature value may be understood as a statistical value of each parameter of each usage object in different advertisement dimensions, such as obtaining exposure times of the usage object a in 7 days in the game industry, or the electronic commerce industry, or the entertainment industry, and if the exposure times are 50 times, the current advertisement feature value is 50 times, and may be represented by game-50, or electronic commerce-50, or entertainment-50. The candidate advertisement queue may be understood as a plurality of candidate advertisement data to be recommended or estimated, where each candidate advertisement data corresponds to a respective advertisement dimension, for example, an advertisement dimension corresponding to a first candidate advertisement data is a primary advertisement industry, and an advertisement dimension corresponding to a second candidate advertisement data is a secondary advertisement industry.

Specifically, when the advertisement feature values of the target usage object in different advertisement dimensions are obtained, specifically, the advertisement feature values of the target usage object in dimensions such as advertisement industry, advertisement recommendation party, advertisement related products and the like are obtained, for example, the exposure times (or click rate, conversion rate or the like) of the target usage object in a preset time period in the primary advertisement industry (including game industry, entertainment industry or electronic commerce industry) are obtained, if the exposure times in the preset time period are determined to be 50 times, the exposure times of the target usage object in the current advertisement dimension, namely the primary advertisement industry, are 50 times, namely the current advertisement feature value is 50 times, and the target usage object can be represented by 'primary advertisement industry-50'.

Similarly, the candidate advertisement queue may be understood as a plurality of candidate advertisement data to be recommended or estimated, where each candidate advertisement data corresponds to a respective advertisement dimension, and when the candidate advertisement queue to be recommended is obtained, specifically, the plurality of candidate advertisement data to be recommended or estimated and the advertisement dimension to which each candidate advertisement data belongs are obtained, where the same advertisement dimension may include a plurality of candidate advertisement data, and each candidate advertisement data corresponds to a respective advertisement dimension, for example, the candidate advertisement queue includes: the first candidate advertisement data-game, the second candidate advertisement data-entertainment, the third candidate advertisement data-e-commerce, …, etc., for example, the first candidate advertisement data and the second candidate advertisement data both belong to the game advertisement dimension, and the third candidate advertisement data belongs to the entertainment advertisement dimension, …, etc.

Step S504, according to the sub-cross network matched with the advertisement dimension, the advertisement feature value embedded vector and the advertisement dimension embedded vector under the advertisement dimension corresponding to the sub-cross network are subjected to cross matching processing to obtain an intermediate advertisement vector.

And matching sub-crossover networks are correspondingly arranged in different advertisement dimensions, and the advertisement feature values of the same advertisement dimension and the advertisement dimension embedded feature vectors are processed by adopting the same sub-crossover network. The advertisement dimension specifically comprises dimensions of advertisement industry, advertisement recommended party, advertisement related products and the like, the advertisement industry comprises a primary advertisement industry, a secondary advertisement industry and the like, the primary advertisement industry comprises a game industry, an electronic commerce industry, an entertainment industry and the like, the primary advertisement industry can comprise a plurality of secondary advertisement industries, the secondary advertisement industry can comprise a game application program, a game live broadcast program and the like in the game industry, and the secondary advertisement industry can comprise an audio and video play program, a social application program, a short video live broadcast program and the like in the entertainment industry.

It will be appreciated that a sub-crossover network is provided for primary advertising industries, such as gaming, electronics, and entertainment industries, etc., and a sub-crossover network is provided for secondary advertising industries, such as gaming applications, live game programs, etc., under the gaming industry, and a matching sub-crossover network is provided for each of the advertising recommender, the advertising associated product, etc.

Specifically, the advertisement feature value embedded vector corresponding to the advertisement feature value under each advertisement dimension and the advertisement dimension embedded vector corresponding to the advertisement dimension are determined, and cross matching processing is performed on the advertisement feature value embedded vector and the advertisement dimension embedded vector under the advertisement dimension corresponding to the sub-cross network according to the sub-cross network matched with the current advertisement dimension, so that an intermediate advertisement vector is obtained.

Wherein, the advertisement characteristic value in the advertisement dimension can be expressed in Di-Vi mode, and the determined advertisement characteristic value embedding vector can be usedRepresenting, and the advertisement dimension embedding vector can be used +.>Representing, and further embedding vector +.>Advertisement dimension embedding vector +.>Performing cross matching processing to obtain intermediate advertisement vector +.>

In one embodiment, the step of determining an advertisement feature value embedding vector corresponding to an advertisement feature value for each advertisement dimension, and an advertisement dimension embedding vector corresponding to the advertisement dimension, comprises:

converting the advertisement feature value under the advertisement dimension into an advertisement value coding feature vector, and converting the advertisement dimension into a dimension coding feature vector; the advertisement value encoding feature vector is converted into an advertisement feature value embedding vector, and the dimension encoding feature vector is converted into an advertisement dimension embedding vector.

Specifically, the method may be used to convert the advertisement feature value and the advertisement dimension into onehot vectors (i.e. unicode feature vectors), that is, obtain advertisement value onehot vectors and advertisement dimension onehot vectors, further convert the advertisement value onehot vectors into advertisement feature value embedded vectors, and convert the advertisement dimension onehot vectors into advertisement dimension embedded vectors.

Step S506, based on the object feature vector stack corresponding to the target usage object and the second advertisement feature vector stack, a second fusion feature vector stack is obtained by splicing.

Specifically, by acquiring history log data associated with a target usage object, and determining an initial feature corresponding to the target usage object from the history log data, and classifying the initial feature into an object feature corresponding to the target usage object and an advertisement feature according to a feature attribute corresponding to the initial feature.

Likewise, the feature attribute corresponding to the initial feature may be understood as whether the feature attribute of the initial feature belongs to an object attribute or an advertisement access attribute, that is, the feature attribute is used to determine whether the initial feature may be specifically classified as an object feature or an advertisement feature. If the feature attribute of the initial feature is determined to belong to the object attribute, the corresponding initial feature is classified as the object feature, and if the feature attribute is determined to be the advertisement access attribute, the corresponding initial feature is classified as the advertisement feature.

Further, by determining a second advertisement feature vector stack corresponding to the advertisement feature corresponding to the target use object and an object feature vector stack corresponding to the object feature corresponding to the target use object, the object feature vector stack corresponding to the target use object and the second advertisement feature vector stack can be spliced to obtain a second fusion feature vector stack.

Specifically, the advertisement feature corresponding to the target object is converted into a first advertisement subcode feature vector, the first advertisement subcode feature vector is converted into a first advertisement embedded feature vector, and a second advertisement feature vector stack corresponding to the target object is spliced based on each first advertisement embedded feature vector.

In one embodiment, when the advertisement feature corresponding to the target usage object is converted into a first advertisement sub-coding feature vector and the first advertisement sub-coding feature vector is converted into a first advertisement embedding feature vector, specifically, the advertisement feature corresponding to the target usage object is converted into the form of onehot vector (i.e. one-hot coding feature vector), namely, a first advertisement sub-onehot vector is obtained, and the first advertisement sub-onehot vector is further converted into a first advertisement embedding feature vector, and further, a second advertisement feature vector stack corresponding to the target usage object can be spliced based on a plurality of first advertisement embedding feature vectors.

Similarly, specifically, object feature vectors corresponding to the target use object are obtained by converting object feature vectors corresponding to the target use object into object-embedded feature vectors, and splicing the object-embedded feature vectors based on the object-embedded feature vectors.

In one embodiment, when converting the object feature corresponding to the object to be used into the object encoding feature vector, specifically, converting the object feature into a form of onehot vector (i.e. one-hot encoding feature vector), that is, obtaining the object onehot vector, and further converting the object onehot vector into the object embedding feature vector, so that the object feature vector stack corresponding to the object to be used can be spliced based on a plurality of object embedding feature vectors.

And step S508, performing matching prediction processing on the intermediate advertisement vector and the second fusion feature vector stack to obtain a corresponding second matching prediction result, and determining advertisement recommendation data matched with the target use object based on the second matching prediction result.

Specifically, according to the trained multi-layer cross network model, matching prediction processing is carried out on the intermediate advertisement vector and the second fusion feature vector stack to obtain a corresponding second matching prediction result, and auxiliary attribute information associated with candidate advertisement data is obtained to determine advertisement recommendation data matched with a target use object according to the second matching prediction result and the auxiliary attribute information associated with the candidate advertisement data.

The trained multi-layer cross network model may be a deep neural network (i.e. Deep Neural Networks) including a plurality of sub-cross networks and used for predicting the matching result, specifically, a convolutional neural network (Convolutional Neural Networks, i.e. CNN), a cyclic neural network (Recurrent Neural Network, i.e. RNN), and a generating countermeasure network (Generative Adversarial Networks, i.e. GAN). The training cross network model is also provided with sub-cross networks matched with different advertisement dimensions, and each sub-cross network is used for carrying out cross matching processing on the advertisement characteristic value embedded vector and the advertisement dimension embedded vector under the advertisement dimension matched with the sub-cross network model, so as to obtain an intermediate advertisement vector.

Further, when the matching prediction processing is performed on the intermediate advertisement vector and the second fusion feature vector stack according to the trained multi-layer cross network model, a second matching prediction result between the target use object and each candidate advertisement data can be predicted. Auxiliary attribute information associated with the candidate advertisement data, such as advertisement bid corresponding to the candidate advertisement data, advertisement calibration evaluation index (i.e., PCOC, which is generally called predict click over click, and represents a ratio of click rate after calibration to posterior click rate (i.e., approximate true probability)), and advertisement exposure policy, is also required to be obtained, so that advertisement recommendation data matched with the target usage object can be determined by comprehensively considering the intermediate advertisement vector, the second matching prediction result, and the auxiliary attribute information associated with the candidate advertisement data.

By comprehensively considering the intermediate advertisement vector and the second fusion feature vector stack, the association relationship between the advertisement dimension and the advertisement feature value of the candidate advertisement data and the association relationship between the advertisement feature and the object feature of the target object can be considered at the same time when the matching prediction processing is performed, so that the features of the target object in different advertisement dimensions and the advertisement feature value of the candidate advertisement data in the advertisement dimension can be more accurately described, and the accuracy of the obtained advertisement recommendation data matched with the target object can be improved.

In one embodiment, after determining advertisement recommendation data matched with a target usage object, the advertisement recommendation data is further sent to the corresponding target usage object, and consumption data of each target usage object in a preset time period (for example, the consumption data can be specifically 24 hours, 36 hours, 48 hours, 72 hours or the like) are detected, so that product transaction data in the preset time period are obtained, and increase amplitude analysis is performed on the product transaction data, so that actual increase amplitude of the product transaction data and benefit data brought by advertisement popularization are obtained.

According to the method for determining the advertisement recommendation data, the advertisement characteristic values of the target used object under different advertisement dimensions and the candidate advertisement queues comprising the plurality of candidate advertisement data are obtained, wherein each candidate advertisement data corresponds to each advertisement dimension respectively, the advertisement characteristic value embedded vector under the advertisement dimension corresponding to the corresponding sub-cross network and the advertisement dimension embedded vector are subjected to cross matching processing according to the sub-cross network matched with the advertisement dimension, so that the intermediate advertisement vector is obtained, further, the candidate advertisement data can be comprehensively considered in aspects of the advertisement dimension, the advertisement characteristic value and the like, and meanwhile, the characteristics of the target used object under different advertisement dimensions can be accurately depicted through the obtained intermediate advertisement vector. Further, based on the object feature vector stack corresponding to the target use object and the second advertisement feature vector stack, a second fusion feature vector stack is obtained by splicing, and then a second comprehensive and accurate matching prediction result is obtained by performing matching prediction processing on the intermediate advertisement vector and the second fusion feature vector stack, so that advertisement recommendation data matched with the target use object can be determined based on the second matching prediction result, and the accuracy of the determined advertisement recommendation data is further improved.

In one embodiment, as shown in fig. 6, there is provided an advertisement recommendation data determining method, which specifically includes the following steps:

step S601, acquiring history log data associated with the target usage object, and determining an initial feature corresponding to the target usage object according to the history log data.

Step S602, dividing the initial feature into an object feature corresponding to the target usage object and an advertisement feature according to the feature attribute corresponding to the initial feature.

Step S603 converts the object feature corresponding to the target usage object into an object encoding feature vector, and converts the object encoding feature vector into an object embedding feature vector.

Step S604, based on the embedded feature vectors of the objects, the object feature vector stacks corresponding to the target use objects are spliced.

Step S605 converts the advertisement feature corresponding to the target usage object into a first advertisement subcode feature vector, and converts the first advertisement subcode feature vector into a first advertisement embedding feature vector.

Step S606, based on the embedded feature vectors of the first advertisements, a second advertisement feature vector stack corresponding to the target use object is spliced.

Step S607, based on the object feature vector stack corresponding to the target usage object and the second advertisement feature vector stack, a second fusion feature vector stack is obtained by stitching.

In the advertisement recommendation data determining method, the initial feature corresponding to the target use object is determined according to the history log data associated with the target use object, and the initial feature is divided into the object feature corresponding to the target use object and the advertisement feature according to the feature attribute corresponding to the initial feature. Further, an object feature vector stack corresponding to the target use object is spliced based on each object embedding feature vector by converting the object feature corresponding to the target use object into an object coding feature vector, and converting the object coding feature vector into an object embedding feature vector, and similarly, an advertisement feature corresponding to the target use object is converted into a first advertisement subcode feature vector, and the first advertisement subcode feature vector is converted into a first advertisement embedding feature vector. And splicing to obtain a second advertisement feature vector stack corresponding to the target use object based on the embedded feature vectors of the first advertisements. Finally, a second fusion feature vector stack can be obtained by splicing the object feature vector stack corresponding to the target use object and the second advertisement feature vector stack. The method realizes the simultaneous comprehensive consideration of object features and advertisement features, and the second fusion feature vector stack is obtained by splicing, so that the trained multilayer cross network model can be utilized in the follow-up, the second fusion feature vector stack obtained by splicing, the advertisement feature value embedded vector under the corresponding advertisement dimension and the advertisement dimension embedded vector are directly subjected to cross matching processing to obtain the intermediate advertisement vector, and the matching prediction processing is performed, so that the second fusion vector stack generation processing and the intermediate advertisement vector generation processing can be simultaneously advanced without the mode of first performing the second fusion vector stack generation processing or first performing the intermediate advertisement vector generation processing, and the working efficiency of the matching prediction processing is improved.

In one embodiment, as shown in fig. 7, there is provided an advertisement recommendation data determining method, which specifically includes the following steps:

step S701, obtaining advertisement characteristic values of a target usage object under different advertisement dimensions, and obtaining a candidate advertisement queue to be recommended, wherein the candidate advertisement queue comprises a plurality of candidate advertisement data, and each candidate advertisement data corresponds to each advertisement dimension.

Step S702, determining an advertisement feature value embedding vector corresponding to the advertisement feature value in each advertisement dimension, and an advertisement dimension embedding vector corresponding to the advertisement dimension.

Step S703, performing cross matching processing on the advertisement feature value embedded vector and the advertisement dimension embedded vector in the advertisement dimension corresponding to the sub-cross network according to the sub-cross network matched with the advertisement dimension, to obtain an intermediate advertisement vector.

Step S704, acquiring history log data associated with the target usage object, and determining an initial feature corresponding to the target usage object according to the history log data.

Step S705, dividing the initial feature into an object feature corresponding to the target usage object and an advertisement feature according to the feature attribute corresponding to the initial feature.

Step S706 converts the object feature corresponding to the target usage object into an object encoding feature vector, and converts the object encoding feature vector into an object embedding feature vector.

Step S707, based on the embedded feature vectors of the objects, the object feature vector stack corresponding to the target usage object is spliced.

Step S708, converting the advertisement feature corresponding to the target usage object into a first advertisement subcode feature vector, and converting the first advertisement subcode feature vector into a first advertisement embedding feature vector.

Step S709, based on the embedded feature vectors of the first advertisements, a second advertisement feature vector stack corresponding to the target usage object is spliced.

Step S710, based on the object feature vector stack corresponding to the target usage object and the second advertisement feature vector stack, a second fusion feature vector stack is obtained by stitching.

Step S711, matching and predicting the intermediate advertisement vector and the second fusion feature vector stack according to the trained multi-layer cross network model to obtain a corresponding second matching and predicting result, wherein sub-cross networks matched with different advertisement dimensions are arranged in the trained cross network model.

Step S712, determining advertisement recommendation data matched with the target usage object according to the second matching prediction result and auxiliary attribute information associated with the candidate advertisement data.

According to the method for determining the advertisement recommendation data, the advertisement characteristic values of the target used object under different advertisement dimensions and the candidate advertisement queues comprising the plurality of candidate advertisement data are obtained, wherein each candidate advertisement data corresponds to each advertisement dimension respectively, the advertisement characteristic value embedded vector under the advertisement dimension corresponding to the sub-cross network and the advertisement dimension embedded vector are subjected to cross matching processing according to the sub-cross network matched with the advertisement dimension, so that the intermediate advertisement vector is obtained, further, the candidate advertisement data can be comprehensively considered in aspects of the advertisement dimension, the advertisement characteristic value and the like, and meanwhile, the characteristics of the target used object under different advertisement dimensions can be accurately depicted through the obtained intermediate advertisement vector. Further, when the intermediate advertisement vector is determined, the second fusion feature vector stack can be obtained by splicing the object feature vector stack corresponding to the target use object and the second advertisement feature vector stack, and then the second comprehensive and accurate second matching prediction result can be obtained by simultaneously carrying out matching prediction processing on the intermediate advertisement vector and the second fusion feature vector stack, so that advertisement recommendation data matched with the target use object can be determined based on the second matching prediction result, and the accuracy of the determined advertisement recommendation data is further improved.

It should be understood that, although the steps in the flowcharts related to the above embodiments are sequentially shown as indicated by arrows, these steps are not necessarily sequentially performed in the order indicated by the arrows. The steps are not strictly limited to the order of execution unless explicitly recited herein, and the steps may be executed in other orders. Moreover, at least some of the steps in the flowcharts described in the above embodiments may include a plurality of steps or a plurality of stages, which are not necessarily performed at the same time, but may be performed at different times, and the order of the steps or stages is not necessarily performed sequentially, but may be performed alternately or alternately with at least some of the other steps or stages.

Based on the same inventive concept, the embodiment of the application also provides an advertisement recommendation data determining device for realizing the above related advertisement recommendation data determining method. The implementation of the solution provided by the apparatus is similar to the implementation described in the above method, so the specific limitation in the embodiments of the apparatus for determining advertisement recommendation data provided below may refer to the limitation of the method for determining advertisement recommendation data hereinabove, and will not be described herein.

In one embodiment, as shown in fig. 8, there is provided an advertisement recommendation data determining apparatus including: a data acquisition module 802, an advertisement matching feature acquisition module 804, a first advertisement feature vector stack determination module 806, a first fused feature vector stack acquisition module 808, and an advertisement recommendation data determination module 810, wherein:

the data acquisition module 802 is configured to acquire advertisement feature values of the target usage object under different advertisement dimensions, and acquire a candidate advertisement queue to be recommended, where the candidate advertisement queue includes a plurality of candidate advertisement data, and each candidate advertisement data corresponds to a respective advertisement dimension;

the advertisement matching feature obtaining module 804 is configured to perform feature cross matching processing on advertisement feature values and candidate advertisement data in different advertisement dimensions, respectively, to obtain advertisement matching features; the advertisement matching feature comprises advertisement dimension and advertisement feature value which are matched with each other;

a first advertisement feature vector stack determination module 806 configured to determine a first advertisement feature vector stack based on advertisement features corresponding to the target usage object and advertisement matching features;

a first fused feature vector stack obtaining module 808, configured to splice the first fused feature vector stack according to the object feature vector stack corresponding to the target usage object and the first advertisement feature vector stack;

The advertisement recommendation data determining module 810 is configured to perform a matching prediction process based on the first fusion feature vector stack, obtain a corresponding first matching prediction result, and determine advertisement recommendation data matching the target usage object based on the first matching prediction result.

In the above-mentioned advertisement recommendation data determining apparatus, by acquiring advertisement feature values of the target usage object under different advertisement dimensions and candidate advertisement queues including a plurality of candidate advertisement data, wherein each candidate advertisement data corresponds to a respective advertisement dimension, and respectively performs feature cross matching processing on the advertisement feature values under different advertisement dimensions and the candidate advertisement data, so as to obtain advertisement matching features including advertisement dimensions and advertisement feature values that are matched with each other, so as to comprehensively consider the candidate advertisement data according to the advertisement dimensions and the advertisement feature values, and accurately characterize the advertisement feature values of the target usage object under different advertisement dimensions. Further, based on the advertisement feature corresponding to the target use object and the advertisement matching feature, a first advertisement feature vector stack is determined, and according to the object feature vector stack corresponding to the target use object and the first advertisement feature vector stack, a more comprehensive first fusion feature vector stack is obtained by splicing, so that matching prediction processing is carried out based on the first fusion feature vector stack, a corresponding first matching prediction result is obtained, and therefore advertisement recommendation data matched with the target use object can be determined based on the first matching prediction result, and accuracy of the determined advertisement recommendation data is further improved.

In one embodiment, the advertisement matching feature obtaining module is further configured to: performing feature matching processing on advertisement feature values and candidate advertisement data under different advertisement dimensions to obtain advertisement feature values matched with the advertisement dimensions to which the candidate advertisement data belong; and carrying out feature cross matching processing on the advertisement dimension to which the candidate advertisement data belongs and the advertisement feature value matched with the advertisement dimension to obtain advertisement matching features.

In one embodiment, the first advertisement feature vector stack determination module is further configured to: converting advertisement features corresponding to the target usage object into a first advertisement subcode feature vector, and converting advertisement matching features into a second advertisement subcode feature vector; converting the first advertisement subcode feature vector into a first advertisement embedded feature vector and converting the second advertisement subcode feature vector into a second advertisement embedded feature vector; and splicing the first advertisement embedded feature vectors and the second advertisement embedded feature vectors to obtain a first advertisement feature vector stack.

In one embodiment, an advertisement recommendation data determining apparatus is provided, further comprising:

the initial feature determining module is used for acquiring historical log data associated with the target use object and determining initial features corresponding to the target use object according to the historical log data;

The feature dividing module is used for dividing the initial feature into an object feature corresponding to the target use object and an advertisement feature according to the feature attribute corresponding to the initial feature;

an object-embedding feature vector obtaining module for converting object features corresponding to the object-use object into object-encoding feature vectors, and converting the object-encoding feature vectors into object-embedding feature vectors;

and the object feature vector stack obtaining module is used for splicing the object feature vector stacks corresponding to the target use objects based on the embedded feature vectors of the objects.

In one embodiment, the advertisement recommendation data determination module is further to: according to the trained multi-layer cross network model, carrying out matching prediction processing on the first fusion feature vector stack to obtain a corresponding first matching prediction result; and determining advertisement recommendation data matched with the target use object according to the first matching prediction result and auxiliary attribute information associated with the candidate advertisement data.

In one embodiment, as shown in fig. 9, there is provided an advertisement recommendation data determining apparatus including: a data acquisition module 902, an intermediate advertisement vector acquisition module 904, a second fused feature vector stack acquisition module 906, and an advertisement recommendation data determination module 908, wherein:

The data acquisition module 902 is configured to acquire advertisement feature values of the target usage object under different advertisement dimensions, and acquire a candidate advertisement queue to be recommended, where the candidate advertisement queue includes a plurality of candidate advertisement data, and each candidate advertisement data corresponds to a respective advertisement dimension;

the intermediate advertisement vector obtaining module 904 is configured to perform cross matching processing on the advertisement feature value embedded vector and the advertisement dimension embedded vector in the advertisement dimension corresponding to the sub-cross network according to the sub-cross network matched with the advertisement dimension, so as to obtain an intermediate advertisement vector;

a second fused feature vector stack obtaining module 906, configured to splice the second fused feature vector stack based on the object feature vector stack corresponding to the target usage object and the second advertisement feature vector stack;

the advertisement recommendation data determining module 908 is configured to perform a matching prediction process on the intermediate advertisement vector and the second fusion feature vector stack, obtain a corresponding second matching prediction result, and determine advertisement recommendation data matching the target usage object based on the second matching prediction result.

In the advertisement recommendation data determining device, the advertisement characteristic values of the target used object under different advertisement dimensions and the candidate advertisement queues comprising a plurality of candidate advertisement data are obtained, the advertisement characteristic value embedded vectors under the advertisement dimensions corresponding to the sub-cross networks and the advertisement dimension embedded vectors are subjected to cross matching processing according to the sub-cross networks matched with the advertisement dimensions, so that intermediate advertisement vectors are obtained, further, the candidate advertisement data can be comprehensively considered in the aspects of the advertisement dimensions, the advertisement characteristic values and the like, and meanwhile, the characteristics of the target used object under different advertisement dimensions can be accurately represented through the obtained intermediate advertisement vectors. Further, based on the object feature vector stack corresponding to the target use object and the second advertisement feature vector stack, a second fusion feature vector stack is obtained by splicing, and then a second comprehensive and accurate matching prediction result is obtained by performing matching prediction processing on the intermediate advertisement vector and the second fusion feature vector stack, so that advertisement recommendation data matched with the target use object can be determined based on the second matching prediction result, and the accuracy of the determined advertisement recommendation data is further improved.

the object feature vector stack obtaining module is used for obtaining an object feature vector stack corresponding to the target use object by splicing based on the embedded feature vectors of the objects;

the first advertisement embedded feature vector obtaining module is used for converting advertisement features corresponding to the target use object into first advertisement subcode feature vectors and converting the first advertisement subcode feature vectors into first advertisement embedded feature vectors;

and the second advertisement feature vector stack obtaining module is used for splicing the second advertisement feature vector stacks corresponding to the target use objects based on the embedded feature vectors of the first advertisements.

In one embodiment, the advertisement recommendation data determination module is further to: according to the trained multi-layer cross network model, matching prediction processing is carried out on the intermediate advertisement vector and the second fusion feature vector stack, and a corresponding second matching prediction result is obtained; sub-crossover networks matched with different advertisement dimensions are arranged in the trained crossover network model; and determining advertisement recommendation data matched with the target use object according to the second matching prediction result and auxiliary attribute information associated with the candidate advertisement data.

The respective modules in the above-described advertisement recommendation data determination apparatus may be implemented in whole or in part by software, hardware, and a combination thereof. The above modules may be embedded in hardware or may be independent of a processor in the computer device, or may be stored in software in a memory in the computer device, so that the processor may call and execute operations corresponding to the above modules.

In one embodiment, a computer device is provided, which may be a server, and the internal structure of which may be as shown in fig. 10. The computer device includes a processor, a memory, an Input/Output interface (I/O) and a communication interface. The processor, the memory and the input/output interface are connected through a system bus, and the communication interface is connected to the system bus through the input/output interface. Wherein the processor of the computer device is configured to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, computer programs, and a database. The internal memory provides an environment for the operation of the operating system and computer programs in the non-volatile storage media. The database of the computer equipment is used for storing advertisement feature values of target usage objects in different advertisement dimensions, a plurality of candidate advertisement data belonging to different advertisement dimensions, advertisement matching features, a first advertisement feature vector stack, an object feature vector stack corresponding to the target usage objects, a first fusion feature vector stack, a first matching prediction result, an intermediate advertisement vector, a second advertisement feature vector stack, a second fusion feature vector stack, a second matching prediction result, advertisement recommendation data and the like. The input/output interface of the computer device is used to exchange information between the processor and the external device. The communication interface of the computer device is used for communicating with an external terminal through a network connection. The computer program is executed by a processor to implement a method of advertisement recommendation data determination.

It will be appreciated by those skilled in the art that the structure shown in FIG. 10 is merely a block diagram of some of the structures associated with the present inventive arrangements and is not limiting of the computer device to which the present inventive arrangements may be applied, and that a particular computer device may include more or fewer components than shown, or may combine some of the components, or have a different arrangement of components.

In an embodiment, there is also provided a computer device comprising a memory and a processor, the memory having stored therein a computer program, the processor implementing the steps of the method embodiments described above when the computer program is executed.

In one embodiment, a computer-readable storage medium is provided, on which a computer program is stored which, when executed by a processor, carries out the steps of the method embodiments described above.

In an embodiment, a computer program product is provided, comprising a computer program which, when executed by a processor, implements the steps of the method embodiments described above.

It should be noted that, the user information (including but not limited to user equipment information, user personal information, etc.) and the data (including but not limited to data for analysis, stored data, presented data, etc.) related to the present application are information and data authorized by the user or sufficiently authorized by each party, and the collection, use and processing of the related data need to comply with the related laws and regulations and standards of the related country and region.

Those skilled in the art will appreciate that implementing all or part of the above described methods may be accomplished by way of a computer program stored on a non-transitory computer readable storage medium, which when executed, may comprise the steps of the embodiments of the methods described above. Any reference to memory, database, or other medium used in embodiments provided herein may include at least one of non-volatile and volatile memory. The nonvolatile Memory may include Read-Only Memory (ROM), magnetic tape, floppy disk, flash Memory, optical Memory, high density embedded nonvolatile Memory, resistive random access Memory (ReRAM), magnetic random access Memory (Magnetoresistive Random Access Memory, MRAM), ferroelectric Memory (Ferroelectric Random Access Memory, FRAM), phase change Memory (Phase Change Memory, PCM), graphene Memory, and the like. Volatile memory can include random access memory (Random Access Memory, RAM) or external cache memory, and the like. By way of illustration, and not limitation, RAM can be in the form of a variety of forms, such as static random access memory (Static Random Access Memory, SRAM) or dynamic random access memory (Dynamic Random Access Memory, DRAM), and the like. The databases referred to in the embodiments provided herein may include at least one of a relational database and a non-relational database. The non-relational database may include, but is not limited to, a blockchain-based distributed database, and the like. The processor referred to in the embodiments provided in the present application may be a general-purpose processor, a central processing unit, a graphics processor, a digital signal processor, a programmable logic unit, a data processing logic unit based on quantum computing, or the like, but is not limited thereto.

The technical features of the above embodiments may be arbitrarily combined, and all possible combinations of the technical features in the above embodiments are not described for brevity of description, however, as long as there is no contradiction between the combinations of the technical features, they should be considered as the scope of the description.

The foregoing examples illustrate only a few embodiments of the application and are described in detail herein without thereby limiting the scope of the application. It should be noted that it will be apparent to those skilled in the art that several variations and modifications can be made without departing from the spirit of the application, which are all within the scope of the application. Accordingly, the scope of the application should be assessed as that of the appended claims.

Claims

1. A method of determining advertisement recommendation data, the method comprising:

2. The method of claim 1, wherein the performing feature cross matching processing on the advertisement feature values and the candidate advertisement data in different advertisement dimensions to obtain advertisement matching features includes:

performing feature matching processing on advertisement feature values and candidate advertisement data under different advertisement dimensions to obtain advertisement feature values matched with the advertisement dimensions to which the candidate advertisement data belong;

and carrying out feature cross matching processing on the advertisement dimension to which the candidate advertisement data belongs and the advertisement feature value matched with the advertisement dimension to obtain advertisement matching features.

3. The method of claim 1, wherein the determining a first advertisement feature vector stack based on the advertisement features corresponding to the target usage object and the advertisement matching features comprises:

converting advertisement features corresponding to the target usage object into a first advertisement subcode feature vector, and converting the advertisement matching features into a second advertisement subcode feature vector;

converting the first advertisement subcode feature vector into a first advertisement embedded feature vector and converting the second advertisement subcode feature vector into a second advertisement embedded feature vector;

and splicing the first advertisement embedded feature vector and the second advertisement embedded feature vector to obtain a first advertisement feature vector stack.

4. A method according to any one of claims 1 to 3, further comprising, before the concatenating the first fused feature vector stack according to the object feature vector stack corresponding to the target usage object and the first advertisement feature vector stack:

acquiring history log data associated with the target use object, and determining initial characteristics corresponding to the target use object according to the history log data;

Dividing the initial feature into an object feature corresponding to the target use object and an advertisement feature according to a feature attribute corresponding to the initial feature;

converting object features corresponding to the object-use object into object-encoded feature vectors, and converting the object-encoded feature vectors into object-embedded feature vectors;

and splicing the object feature vector stacks corresponding to the target use objects based on the embedded feature vectors of the objects.

5. A method according to any one of claims 1 to 3, wherein performing a matching prediction process based on the first fused feature vector stack to obtain a corresponding first matching prediction result, and determining advertisement recommendation data matching the target usage object based on the first matching prediction result, includes:

according to the trained multi-layer cross network model, carrying out matching prediction processing on the first fusion feature vector stack to obtain a corresponding first matching prediction result;

and determining advertisement recommendation data matched with the target use object according to the first matching prediction result and auxiliary attribute information associated with the candidate advertisement data.

6. A method of determining advertisement recommendation data, the method comprising:

7. The method of claim 6, further comprising, prior to the concatenating the second fused feature vector stack based on the object feature vector stack corresponding to the target usage object and the second advertisement feature vector stack:

based on the embedded feature vectors of the objects, splicing to obtain object feature vector stacks corresponding to the target use objects;

converting advertisement features corresponding to the target usage object into a first advertisement subcode feature vector, and converting the first advertisement subcode feature vector into a first advertisement embedded feature vector;

and based on the embedded feature vectors of the first advertisements, splicing to obtain a second advertisement feature vector stack corresponding to the target use object.

8. The method of claim 6, wherein performing a matching prediction process on the intermediate advertisement vector and the second fused feature vector stack to obtain a corresponding second matching prediction result, and determining advertisement recommendation data matching the target usage object based on the second matching prediction result, comprises:

According to the trained multi-layer cross network model, matching prediction processing is carried out on the intermediate advertisement vector and the second fusion feature vector stack, and a corresponding second matching prediction result is obtained; sub-crossover networks matched with different advertisement dimensions are arranged in the trained crossover network model;

and determining advertisement recommendation data matched with the target use object according to the second matching prediction result and auxiliary attribute information associated with the candidate advertisement data.

9. An advertisement recommendation data determining apparatus, the apparatus comprising:

10. An advertisement recommendation data determining apparatus, the apparatus comprising:

11. A computer device comprising a memory and a processor, the memory storing a computer program, characterized in that the processor implements the steps of the method of any one of claims 1 to 8 when the computer program is executed.

12. A computer readable storage medium, on which a computer program is stored, characterized in that the computer program, when being executed by a processor, implements the steps of the method of any of claims 1 to 8.

13. A computer program product comprising a computer program, characterized in that the computer program, when executed by a processor, implements the steps of the method of any one of claims 1 to 8.