CN111310056B

CN111310056B - Information recommendation method, device, equipment and storage medium based on artificial intelligence

Info

Publication number: CN111310056B
Application number: CN202010166212.2A
Authority: CN
Inventors: 卢建东
Original assignee: Shenzhen Yayue Technology Co ltd
Current assignee: Shenzhen Yayue Technology Co ltd
Priority date: 2020-03-11
Filing date: 2020-03-11
Publication date: 2023-04-14
Anticipated expiration: 2040-03-11
Also published as: CN111310056A

Abstract

The embodiment of the application discloses an information recommendation method, an information recommendation device, information recommendation equipment and a storage medium based on artificial intelligence, wherein the method comprises the following steps: acquiring target information to be recommended and historical browsing information of a target user; determining a target information characteristic vector corresponding to target information through an interest probability prediction model; determining a historical information characteristic vector corresponding to historical browsing information through the interest probability prediction model, and determining a target user characteristic vector corresponding to a target user according to the target information characteristic vector and the historical information characteristic vector; determining the interest probability of the target user for the target information according to the target information feature vector and the target user feature vector through the interest probability prediction model; whether to recommend the target information to the target user is determined based on the interest probability. The method can effectively improve the accuracy of information recommendation.

Description

Information recommendation method, device, equipment and storage medium based on artificial intelligence

Technical Field

The present disclosure relates to the field of Artificial Intelligence (AI), and in particular, to an information recommendation method, apparatus, device, and storage medium.

Background

Information which may be interested in a user is recommended to the user in a targeted manner based on the user portrait, and the service becomes one of services which are of major interest to many network platforms nowadays, for example, a news platform needs to recommend news which may be interested in the user, an audio and video resource playing platform needs to recommend audio and/or video resources which may be interested in the user, a shopping platform needs to recommend commodity information which may be interested in the user, and the like.

At present, in the related art, the degree of interest of a user in information to be recommended is predicted mainly through a neural network model such as a Deep & Wide model, a Deep fm (Factorization mechanisms) model, and the like, and then whether to recommend the information to be recommended to the user is determined based on the degree of interest. In specific implementation, the Neural network model determines a user feature vector and a feature vector of information to be recommended, and then predicts the interest degree of the user for the information to be recommended based on the two feature vectors through Deep Neural Networks (DNNs).

However, the inventor of the present application finds that the user feature vectors constructed by the neural network model are usually fixed, the characterization capability of such feature vectors for users is limited, and the degree of interest of the neural network model predicted based on such user feature vectors is usually difficult to truly and accurately reflect the degree of interest of users for the current information to be recommended, and the accuracy of network platform information recommendation is affected to a certain extent.

Disclosure of Invention

The embodiment of the application provides an information recommendation method, device, equipment and storage medium based on artificial intelligence, which can accurately predict the interest degree of a user in information to be recommended and improve the accuracy of information recommendation.

In view of this, a first aspect of the present application provides an artificial intelligence based information recommendation method, where the method includes:

acquiring target information to be recommended and historical browsing information of a target user;

determining a target information characteristic vector corresponding to the target information through an interest probability prediction model;

determining a historical information feature vector corresponding to the historical browsing information through the interest probability prediction model, and determining a target user feature vector corresponding to the target user according to the target information feature vector and the historical information feature vector;

determining the interest probability of the target user for the target information according to the target information feature vector and the target user feature vector through the interest probability prediction model;

determining whether to recommend the target information to the target user based on the probability of interest.

The second aspect of the present application provides an artificial intelligence-based information recommendation apparatus, the apparatus including:

the acquisition module is used for acquiring target information to be recommended and historical browsing information of a target user;

the information characteristic vector determining module is used for determining a target information characteristic vector corresponding to the target information through an interest probability prediction model;

the user characteristic vector determining module is used for determining a historical information characteristic vector corresponding to the historical browsing information through the interest probability prediction model, and determining a target user characteristic vector corresponding to the target user according to the target information characteristic vector and the historical information characteristic vector;

the interest probability determining module is used for determining the interest probability of the target user for the target information according to the target information feature vector and the target user feature vector through the interest probability prediction model;

and the recommending module is used for determining whether to recommend the target information to the target user based on the interest probability.

A third aspect of the application provides an apparatus comprising a processor and a memory:

the memory is used for storing a computer program;

the processor is configured to perform the steps of the artificial intelligence based information recommendation method according to the first aspect as described above according to the computer program.

A fourth aspect of the present application provides a computer-readable storage medium for storing a computer program for executing the steps of the artificial intelligence based information recommendation method according to the first aspect.

A fifth aspect of the present application provides a computer program product comprising instructions which, when run on a computer, cause the computer to perform the steps of the artificial intelligence based information recommendation method of the first aspect described above.

According to the technical scheme, the embodiment of the application has the following advantages:

the embodiment of the application provides an information recommendation method based on artificial intelligence, and an interest probability prediction model in the method can construct a corresponding user portrait along with the change of information to be recommended, so that the user portrait can dynamically reflect the interest degree of a user in the current information to be recommended. Specifically, in the method provided in the embodiment of the present application, after the target information to be recommended and the historical browsing information of the target user are obtained, a target information feature vector corresponding to the target information may be determined through an interest probability prediction model, then a historical feature vector corresponding to the historical browsing information may be determined through the interest probability prediction model, a target user feature vector corresponding to the target user is determined according to the target information feature vector and the historical feature vector, and then the interest probability of the target user with respect to the target information is predicted through the interest probability prediction model according to the target information feature vector and the target user feature vector, so as to determine whether to recommend the target information to the target user based on the interest probability. When the interest probability model determines the target user feature vector, the historical browsing information features of the target user and the current target information features to be recommended are comprehensively considered, so that the determined target user feature vector can pointedly reflect the interest degree of the target user on the current target information to be recommended, accordingly, the interest probability predicted based on the target user feature vector is more accurate, and the subsequent accurate information recommendation of a network platform is facilitated.

Drawings

Fig. 1 is a schematic view of an application scenario of an artificial intelligence based information recommendation method according to an embodiment of the present application;

FIG. 2 is a schematic flowchart of an artificial intelligence-based information recommendation method according to an embodiment of the present application;

FIG. 3 is a schematic structural diagram of an interesting probability prediction model according to an embodiment of the present application;

FIG. 4 is a schematic structural diagram of another interesting probability prediction model provided in an embodiment of the present application;

FIG. 5 is a graph of performance comparison results provided by the examples of the present application;

fig. 6 is a schematic structural diagram of an artificial intelligence-based information recommendation apparatus according to an embodiment of the present application;

FIG. 7 is a schematic structural diagram of another artificial intelligence-based information recommendation apparatus according to an embodiment of the present application;

fig. 8 is a schematic structural diagram of another artificial intelligence-based information recommendation apparatus according to an embodiment of the present application;

fig. 9 is a schematic structural diagram of a server according to an embodiment of the present application;

fig. 10 is a schematic structural diagram of a terminal device according to an embodiment of the present application.

Detailed Description

In order to make the technical solutions of the present application better understood, the technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the drawings in the embodiments of the present application, and it is obvious that the described embodiments are only a part of the embodiments of the present application, and not all of the embodiments. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present application.

The terms "first," "second," "third," "fourth," and the like in the description and in the claims of the present application and in the drawings described above, if any, are used for distinguishing between similar elements and not necessarily for describing a particular sequential or chronological order. It should be understood that the data so used may be interchanged under appropriate circumstances such that embodiments of the application described herein may be implemented in sequences other than those illustrated or described herein. Furthermore, the terms "comprises," "comprising," and "having," and any variations thereof, are intended to cover a non-exclusive inclusion, such that a process, method, system, article, or apparatus that comprises a list of steps or elements is not necessarily limited to those steps or elements expressly listed, but may include other steps or elements not expressly listed or inherent to such process, method, article, or apparatus.

Artificial intelligence is a theory, method, technique and application system that uses a digital computer or a machine controlled by a digital computer to simulate, extend and expand human intelligence, perceive the environment, acquire knowledge and use the knowledge to obtain the best results. In other words, artificial intelligence is a comprehensive technique of computer science that attempts to understand the essence of intelligence and produce a new intelligent machine that can react in a manner similar to human intelligence. Artificial intelligence is the research of the design principle and the implementation method of various intelligent machines, so that the machines have the functions of perception, reasoning and decision making.

The artificial intelligence technology is a comprehensive subject, and relates to the field of extensive technology, namely the technology of a hardware level and the technology of a software level. The artificial intelligence infrastructure generally includes 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 the like.

Machine Learning (ML) is a multi-domain cross subject, and relates to multiple subjects such as probability theory, statistics, approximation theory, convex analysis and algorithm complexity theory. The special research on how a computer simulates or realizes the learning behavior of human beings so as to acquire new knowledge or skills and reorganize the existing knowledge structure to continuously improve the performance of the computer. Machine learning is the core of artificial intelligence, is the fundamental approach for computers to have intelligence, and is applied to all fields of artificial intelligence. Machine learning and deep learning generally include techniques such as artificial neural networks, belief networks, reinforcement learning, transfer learning, inductive learning, and formula learning.

The scheme provided by the embodiment of the application relates to an artificial intelligence information recommendation technology, and is specifically explained by the following embodiment.

In the related art, when the interest degree of a user for information to be recommended is predicted through a neural network model, the neural network model usually only constructs corresponding user feature vectors based on historical browsing information of the user, the user feature vectors are usually fixed, the depicting ability of the user is poor, accordingly, the interest degree predicted based on the user feature vectors is often difficult to truly and accurately reflect the interest degree of the user for the current information to be recommended, and the accuracy of network platform information recommendation is also influenced.

In view of the technical problems in the related art, the embodiment of the application provides an information recommendation method based on artificial intelligence, and an interest probability model in the method can construct a user feature vector which changes with the change of information to be recommended, so that the constructed user feature vector can accurately reflect the interest degree of a user in the current information to be recommended.

Specifically, in the method provided by the embodiment of the application, target information to be recommended and historical browsing information of a target user are obtained first; then, determining a target information characteristic vector corresponding to the target information through an interest probability prediction model; determining a historical information characteristic vector corresponding to historical browsing information through an interest probability prediction model, and determining a target user characteristic vector corresponding to a target user according to the target information characteristic vector and the historical information characteristic vector; determining the interest degree of a target user for target information to be recommended according to the target information feature vector and the target user feature vector through an interest probability prediction model; finally, whether to recommend the target information to the target user is determined based on the interest level.

Compared with the implementation mode that the user characteristic vector is determined only according to the historical browsing information of the user in the related technology, the method provided by the embodiment of the application comprehensively considers the characteristic vector corresponding to the historical browsing information and the characteristic vector corresponding to the current target information to be recommended when the user characteristic vector is determined, so that the determined user characteristic vector can reflect the interest degree of the user in the current target information to be recommended in a more targeted manner, accordingly, the interest probability predicted based on the user characteristic vector is more accurate, and the subsequent accurate information recommendation of a network platform is facilitated.

It should be understood that, in practical applications, the information recommendation method based on artificial intelligence provided in the embodiments of the present application may be applied to devices capable of supporting operation of a neural network model, such as a terminal device, a server, and the like. The terminal device may be a computer, a smart phone, a tablet computer, a Personal Digital Assistant (PDA), or the like. The server can be an application server or a Web server; in actual deployment, the server may be an independent server or a cluster server.

In order to facilitate understanding of the technical solutions provided in the embodiments of the present application, taking an example in which the information recommendation method based on artificial intelligence provided in the embodiments of the present application is applied to a server, an application scenario of the information recommendation method based on artificial intelligence is described below.

Referring to fig. 1, fig. 1 is a schematic view of an application scenario of an artificial intelligence based information recommendation method provided in an embodiment of the present application. As shown in fig. 1, the application scenario includes a server 110 and a database 120; the server 110 is configured to execute the information recommendation method provided in the embodiment of the present application, where an interest probability prediction model 111 is operated; the database 120 stores therein historical browsing information of the user.

When the server 110 needs to determine whether to recommend the target information to the target user, the server 110 may first retrieve the historical browsing information of the target user from the database 120, for example, may retrieve the historical browsing information of the target user in the last week. Then, the interest probability prediction model 111 is called to predict the interest probability of the target user for the target information based on the target information to be recommended and the historical browsing information of the target user.

Specifically, the interest probability prediction model 111 may determine a target information feature vector corresponding to the target information and a target user feature vector corresponding to the target user, respectively. Specifically, when the target information feature vector is determined, the interest probability prediction model 111 may map the feature information of the target information to a dense vector space to obtain the target information feature vector corresponding to the target information. Specifically, when the feature vector of the target user is determined, the interest probability prediction model 111 may first determine the history information feature vector corresponding to each piece of history browsing information, and then determine the feature vector of the target user corresponding to the target user according to the feature vector of the target information and the feature vector of the history information corresponding to each piece of history browsing information.

After the interest probability prediction model 111 determines the target information feature vector and the target user feature vector, the interest probability of the target user for the target information can be determined according to the target information feature vector and the target user feature vector, and the interest probability is output.

After obtaining the interest probability output by the interest probability prediction model 111, the server 110 may determine whether to recommend the target information to the target user based on the interest probability.

It should be understood that the application scenario shown in fig. 1 is only an example, and in practical application, the terminal device may independently execute the information recommendation method provided in the embodiment of the present application, or the terminal device and the server may cooperatively execute the information recommendation method provided in the embodiment of the present application, and no limitation is imposed on the application scenario of the information recommendation method provided in the embodiment of the present application.

The information recommendation method based on artificial intelligence provided by the present application is described in detail below by way of embodiments.

Referring to fig. 2, fig. 2 is a schematic flowchart of an artificial intelligence-based information recommendation method provided in an embodiment of the present application. For convenience of description, the following embodiments are described taking a server as an execution subject as an example. As shown in fig. 2, the information recommendation method includes the following steps:

step 201: and acquiring target information to be recommended and historical browsing information of a target user.

Aiming at the current target information to be recommended, when the server determines whether the target information needs to be recommended to a target user, the server needs to acquire the target information and historical browsing information of the target user.

It should be noted that the target information to be recommended may be different types of information on different network platforms, for example, the target information may be news information, advertisement information, multimedia resources, commodity information, and the like, and the type of the target information is not specifically limited in this application. The historical browsing information of the target user refers to information browsed and/or searched by the target user in a historical time period, for example, the information browsed and/or searched by the target user in the last week or one month, and types of the historical browsing information on different network platforms are different, and the generation time period and the type corresponding to the historical browsing information are not specifically limited in this application.

In a possible implementation manner, the server may receive an information recommendation instruction issued by the upper control device, where the information recommendation instruction may carry target information to be currently recommended, and the information recommendation instruction is used to instruct the server to recommend the target information carried therein to a user of the network platform. After receiving the information recommendation instruction, the server may retrieve, from the database, historical browsing information of the user of the network platform within a preset historical time period, for example, retrieve historical browsing information of the user within the last week. Furthermore, for each user of the network platform, based on the target information in the information recommendation instruction and the historical browsing information of the user, the interest degree of the user in the target information is determined, and whether the target information is recommended to the user is determined based on the interest degree.

It should be understood that the above implementation manner is only an example, and in practical applications, the server may also obtain target information to be recommended and historical browsing information of a target user in other manners, and the obtaining manner of the target information and the historical browsing information is not limited in this application.

It should be noted that the information recommendation method provided by the embodiment of the present application may be applied to various network platforms with information recommendation services, such as a news platform, a multimedia resource playing platform, a shopping platform, and the like.

When the information recommendation method provided by the embodiment of the application is applied to a news platform, the interest probability prediction model in the embodiment of the application can be applied to a news recommendation system; the target information to be recommended can be news information or advertisement information; the historical browsing information may include at least one of the following information: historical viewed news information, historical search news information, and historical viewed advertising information. For example, assuming that the target information to be recommended currently is hot news information, the server may obtain news information browsed by the target user in the last week and news information searched for by the target user at this time as historical browsing information of the target user. For another example, assuming that the target information to be recommended currently is advertisement information provided by a certain merchant, the server may obtain advertisement information browsed by the target user in the last week as historical browsing information of the target user.

When the information recommendation method provided by the embodiment of the application is applied to a multimedia resource playing platform, the interest probability prediction model in the embodiment of the application can be applied to a multimedia resource recommendation system; the target information to be recommended can be audio resources or video resources; the historical browsing information may include at least one of the following information: audio resources played historically, video resources played historically, audio resources searched historically, and video resources searched historically. For example, assuming that the target information to be recommended currently is a certain audio resource, at this time, the server may obtain the audio resource played by the target user in the last week and the searched audio resource as the historical browsing information of the target user. For another example, assuming that the target information to be recommended currently is a certain video resource, at this time, the server may obtain the video resource played by the target user in the last week and the searched video resource as the historical browsing information of the target user.

When the information recommendation method provided by the embodiment of the application is applied to a shopping platform, the interest probability prediction model in the embodiment of the application can be applied to a commodity recommendation system; the target information to be recommended can be commodity information; the historical browsing information may include at least one of the following information: the shopping cart comprises commodity information of historical purchases, commodity information of historical searches and commodity information of historical collections, wherein the commodity information of the historical collections can comprise commodity information added into the shopping cart by a user and/or commodity information clicked and collected by the user. For example, assuming that the target information to be recommended currently is certain commodity information, the server may obtain commodity information purchased by the target user in the last week, searched commodity information, and collected commodity information as the historical browsing information of the target user.

It should be understood that, in practical application, the information recommendation method provided in the embodiment of the present application may be applied not only to the above three scenarios, but also to other scenarios that need information recommendation, and no limitation is made to the application scenario of the information recommendation method provided in the embodiment of the present application.

Step 202: and determining a target information feature vector corresponding to the target information through an interested probability prediction model.

After the server acquires the current target information to be recommended, the target information characteristic vector corresponding to the target information can be determined through a pre-trained interest probability prediction model.

Specifically, the interested probability prediction model may map feature information of the target information to a dense vector space to obtain a dense feature vector corresponding to the target information; and then, processing the dense feature vector by using a Neural network to obtain a target information feature vector corresponding to the target information, wherein the Neural network used here may be a Convolutional Neural Network (CNN) or a Long Short-Term Memory Neural network (LSTM).

The characteristic information of the target information is information capable of reflecting the characteristics of the target information; taking target information as news information as an example, the characteristic information of the news information may specifically include news headlines, news abstracts, news keywords and the like; taking the target information as an audio resource or a video resource as an example, the characteristic information of the audio resource or the video resource may specifically include a title, a creator, profile information, and the like of the audio resource or the video resource; taking the target information as the commodity information as an example, the characteristic information of the commodity information may specifically include commodity profile information and the like.

In practical application, the server can extract the characteristic information of the target information first and then input the characteristic information into the interested probability prediction model; the server can also directly input the target information into the interest probability prediction model, and the interest probability prediction model extracts the characteristic information of the target information. The present application does not limit the execution subject of the operation of extracting the feature information of the target information at all.

After the feature information of the target information is obtained, the interest probability prediction model can map the discrete feature information to a dense vector space by an embedding layer to obtain a dense feature vector of the feature information. Then, extracting features from the dense feature vector through a convolution layer and a pooling layer of the CNN, and taking the output of the pooling layer as a target information feature vector corresponding to the target information; or semantic features can be extracted from the dense feature vector through the LSTM to serve as a target information feature vector corresponding to the target information.

The following describes the above process of determining the target information feature vector by taking the interest probability prediction model to extract the target information feature vector through CNN as an example, and combining the model structure of the interest probability prediction model shown in fig. 3. As shown in fig. 3, after the feature information of the target information is input into the interest probability prediction model, the interest probability prediction model processes the feature information of the target information through a Conv emd1 module therein, that is, the feature information of the target information is mapped into a dense feature vector, and then the dense feature vector is sequentially processed through the convolution layer and the pooling layer, so as to obtain a target information feature vector corresponding to the target information.

It should be noted that the interest probability prediction model generally includes a plurality of Conv emd modules, the network structure and the processing procedure inside each Conv emd module are similar, and different Conv emd modules are used for processing different input information.

Step 203: determining a historical information feature vector corresponding to the historical browsing information through the interested probability prediction model, and determining a target user feature vector corresponding to the target user according to the target information feature vector and the historical information feature vector.

After the server obtains the historical browsing information of the target user, the server can determine the historical information feature vector corresponding to each piece of historical browsing information through the pre-trained interest probability prediction model, and further determine the target user feature vector corresponding to the target user by combining the target information feature vector corresponding to the target information determined in step 202 by the interest probability prediction model.

Specifically, the interest probability prediction model may determine, for each piece of acquired historical browsing information, a corresponding historical information feature vector thereof, and then determine a similarity between the historical information feature vector and a target information feature vector as a similarity corresponding to the piece of historical browsing information; and further, weighting the historical information characteristic vectors corresponding to the historical browsing information by utilizing the similarity corresponding to the historical browsing information, so as to obtain the target user characteristic vector corresponding to the target user.

In practical applications, the interested probability prediction model determines the implementation manner of the historical information feature vector corresponding to the historical browsing information, which is the same as the implementation manner of the target information feature vector corresponding to the target information. That is, when the interest probability prediction model processes each piece of historical browsing information, the feature information of the piece of historical browsing information is mapped to a dense vector space to obtain a dense feature vector corresponding to the piece of historical browsing information, and then the dense feature vector is processed by using the CNN or the LSTM to obtain a historical information feature vector corresponding to the piece of historical browsing information.

As shown in fig. 3, assuming that the server obtains historical browsing information of three target users, the probability prediction model of interest may process feature information of the three pieces of historical browsing information through a Conv emd2 module, a Conv emd3 module, and a Conv emd4 module, respectively, to obtain historical information feature vectors corresponding to the three pieces of historical browsing information, respectively. Here, the internal structures of the Conv emd2 module, the Conv emd3 module and the Conv emd4 module are the same as those of the Conv emd1 module described above.

Considering that the interests of users are generally diverse, when a user feature vector corresponding to the user is constructed for the user, it is required to ensure that the user feature vector can reflect the diversity of the interests of the user, so that the user feature vector has polymorphism and can change with different information to be recommended, for example, when the information to be recommended is an educational advertisement, educational advertisements historically browsed by the user can depict the interests of the user more than sports advertisements. Based on this, when determining the user feature vector, the method provided by the embodiment of the application further considers the correlation between the historical browsing information of the user and the target information to be recommended currently, and determines the user feature vector capable of specifically reflecting the interest degree of the user in the target information based on the correlation between each piece of historical browsing information and the target information.

That is, after the interested probability prediction model determines the corresponding historical information feature vector for each piece of historical browsing information, the similarity between each historical information feature vector and the target information feature vector corresponding to the target information can be sequentially determined as the similarity corresponding to each piece of historical browsing information; and further, determining a target user characteristic vector corresponding to the target user according to the similarity and the historical information characteristic vector corresponding to each piece of historical browsing information.

Specifically, when the similarity corresponding to each piece of historical browsing information is determined, the interested probability prediction model can splice the historical information feature vector corresponding to the piece of historical browsing information with the target information feature vector to obtain a first spliced feature vector; then, based on the first splicing vector, similarity between the historical information feature vector and the target information feature vector is determined through a first Deep Neural Network (DNN).

As shown in fig. 3, the interest probability prediction model may determine a similarity between each of the historical information feature vectors and the target information feature vector using an Attention-force mechanism network (Attention-net) therein. Specifically, the Conv emd1 module inputs the determined target information feature vector to the Attention-net, and the Conv emd2 module, the Conv emd3 module and the Conv emd4 module also input the respective determined historical information feature vector to the Attention-net. The Attention-net splices each historical information feature vector with a target information feature vector, and then predicts the similarity between the historical information feature vector and the target information feature vector based on the spliced vectors by using DNN in the Attention-net. Here, the network parameters of the DNN are obtained by training based on training sample data in advance.

And further, by utilizing the similarity corresponding to each piece of historical browsing information, correspondingly weighting the historical information feature vector corresponding to each piece of historical browsing information to obtain the target user feature vector corresponding to the target user. The specific calculation formula is as follows:

user_embedding＝∑＜item_of_user_embedding,item_embedding＞*item_of_user_embedding

the user _ embedding is a target user feature vector, the item _ of _ user _ embedding is a history information feature vector, the item _ embedding is a target information feature vector, and the item _ of _ user _ embedding and the item _ embedding are used for representing the similarity between the history information feature vector and the target information feature vector.

It should be noted that, in practical applications, in order to make the constructed user feature vector richer, when the server executes step 201, the server may generally obtain historical browsing information from a plurality of different data fields, so as to comprehensively measure the interest preference of the target user by using the historical browsing information of the different data fields. Taking a news platform as an example, the search behavior of the user belongs to an instant behavior, and accordingly, the data searched by the user through the news APP search engine generally reflects the short-term interest of the user, while the browsing behavior of the user based on the news APP belongs to a long-term behavior, and accordingly, the data browsed by the user through the news platform generally reflects the long-term interest of the user. The historical browsing information from different data fields is fused, so that the user portrait can be more finely depicted, and the user characteristic vector can more accurately reflect the interest preference of the user.

It should be understood that the historical browsing information of different data fields may differ in the generation manner of the historical browsing information, for example, for a news APP, the historical browsing information generated based on the search behavior is different from the historical browsing information generated based on the daily browsing behavior, that is, the historical browsing information belonging to different data fields; the difference between the historical browsing information of different data fields may also be that the type of the historical browsing information is different, for example, for a news APP, the news information historically browsed by the user is different from the advertisement information, i.e. the historical browsing information belonging to different data fields.

In addition, in the technical solution provided in this embodiment of the application, the source of the historical browsing information is not limited to the network platform that needs to perform information recommendation, and the server may also communicate with other network platforms to obtain the historical browsing information of the target user on other network platforms, for example, the server of the news platform may communicate with a certain shopping platform to obtain the historical browsing information of the target user on the shopping platform. The source of the historical browsing information is not limited in any way in the present application.

For the case that the historical browsing information comes from different data domains, the interested probability prediction model may first determine the corresponding user feature vector for each data domain. Specifically, for each piece of historical browsing information from the same data domain, the interested probability prediction model may first determine a historical information feature vector corresponding thereto, and determine a similarity between the historical information feature vector and a target information feature vector as a similarity corresponding to the piece of historical browsing information; and then, weighting the historical information characteristic vectors corresponding to the historical browsing information from the data domain by utilizing the similarity corresponding to the historical browsing information from the data domain, so as to obtain the user characteristic vector corresponding to the data domain.

It should be understood that the implementation process for determining the corresponding user feature vector for each data field is similar to the implementation process for determining the target user feature vector based on the historical browsing information from a single data field, and the detailed processing process can refer to the related description above.

Furthermore, the interested probability prediction model further needs to map the user feature vectors corresponding to the data fields to the same vector space by using a transformation matrix, and concatenate the user feature vectors mapped to the same vector space to obtain the target user feature vector corresponding to the target user.

Specifically, the interested probability prediction model may introduce a task matrix (project matrix) to implement spatial transformation between user feature vectors corresponding to respective data fields, and map the user feature vectors in different spaces to the same vector space, where the project matrix is a matrix of N × N, and N corresponds to a dimension of the user feature vector. And then, splicing the user feature vectors obtained by the project matrix conversion to obtain the target user feature vector obtained by multi-domain data characterization.

In order to facilitate understanding of the above implementation of determining the target user feature vector based on historical browsing information from multiple data fields, the above implementation is exemplarily described below with reference to the model structure of the probability prediction model of interest shown in fig. 4, taking the example that the historical browsing information includes news historical browsing information and advertisement historical browsing information.

As shown in fig. 4, for each piece of acquired news historical browsing information, the interest probability prediction model determines a historical information feature vector corresponding to each piece of news historical browsing information through a Conv emd2 module, a Conv emd3 module and a Conv emd4 module, then determines the similarity between each historical information feature vector and a target information feature vector determined by the Conv emd1 module by using Attention-net1, and further performs weighting processing on each historical information feature vector correspondingly by using the similarity corresponding to each piece of news historical browsing information to obtain a news-user-embedding corresponding to the news historical browsing information.

Similarly, for each piece of acquired advertisement historical browsing information, the interest probability prediction model determines the historical information feature vector corresponding to each piece of advertisement historical browsing information through a Conv emd5 module, a Conv emd6 module and a Conv emd7 module respectively, then determines the similarity between each historical information feature vector and the target information feature vector determined by the Conv emd1 module by using Attention-net2, and further performs weighting processing on each historical information feature vector correspondingly by using the similarity corresponding to each piece of advertisement historical browsing information to obtain the user feature vector ad-user-embedding corresponding to the advertisement historical browsing information.

Because the news historical browsing information and the advertisement historical browsing information come from different data fields, the user feature vectors corresponding to the news historical browsing information and the advertisement historical browsing information cannot be spliced directly. In order to solve the problem, a transformation matrix project is further introduced into the interest probability prediction model, the project matrix is used for mapping the news-user-embedding, so that the mapped news-user-embedding and the mapped ad-user-embedding belong to the same vector space, and the mapped news-user-embedding and the mapped ad-user-embedding are spliced to obtain a target user feature vector corresponding to a target user.

It should be understood that the project matrix introduced in the model structure shown in FIG. 4 is a matrix of N × N, where N is the dimension of ad-user-embedding; in practical applications, a project matrix of M × M may also be introduced into the model structure shown in fig. 4, where M is a dimension of news-user-embedding, and the project matrix is used for mapping the ad-user-embedding.

Step 204: and determining the interest probability of the target user for the target information according to the target information feature vector and the target user feature vector through the interest probability prediction model.

After the target information characteristic vector corresponding to the target information to be recommended and the target user characteristic vector corresponding to the target user are determined by the interest probability prediction model, the interest probability of the target user for the target information can be predicted according to the target information characteristic vector and the target user characteristic vector.

Specifically, the interest probability prediction model may splice a target information feature vector and a target user feature vector to obtain a second spliced vector; and then, determining the interest probability of the target user for the target information based on the second splicing vector through a second deep neural network in the interest probability prediction model.

As shown in fig. 3, the interest probability prediction model splices the target user feature vector and the target information feature vector, inputs the spliced feature vector into the DNN, calculates the similarity between the target user feature vector and the target information feature vector through the DNN, and uses the similarity to characterize the interest probability of the target user for the target information. Here, the network parameters of DNN are trained in advance based on training sample data.

It should be noted that the interest probability Predicted by the interest probability prediction model may be used to represent a Predicted Click probability (pctr) of the target user for the target information in practical applications, and may also be used to represent a Predicted Conversion probability (pcvr) of the target user for the target information.

Step 205: determining whether to recommend the target information to the target user based on the probability of interest.

Finally, the server can determine whether to recommend the target information to the target user according to the interest probability output by the interest probability prediction model.

In a possible implementation manner, the server may determine whether the interest probability is greater than a preset probability threshold, if so, it indicates that the target user has a high possibility of being interested in the target information, and then it is determined to recommend the target information to the target user; otherwise, if not, the target user is less likely to be interested in the target information, and the target information is determined not to be recommended to the target user.

In another possible implementation manner, in the case that multiple pieces of target information which can be recommended to the target user exist at the same time, the server may obtain interest probabilities of the target user for the pieces of target information, then perform descending order on the obtained interest probabilities, and determine the target information corresponding to the interest probability of n (n is a positive integer greater than or equal to 1) before the order as the information which needs to be recommended to the target user.

In practical application, the server may also determine whether to recommend the target information to the target user based on the interest probability in other manners, and the method for determining whether to recommend the target information to the target user by the server is not limited herein.

In order to further understand the artificial intelligence based information recommendation method provided by the embodiment of the present application, the following takes application of the method provided by the embodiment of the present application to a news platform as an example, and a whole example of the artificial intelligence based information recommendation method is described.

When the server of the news platform needs to determine whether to recommend the target advertisement information to the target user, the server can acquire the news history browsing information and the advertisement history browsing information of the target user in the last week from the database. Then, the characteristic information corresponding to the target advertisement information, the characteristic information corresponding to each news history browsing information and the characteristic information corresponding to each advertisement history browsing information are respectively determined.

And then inputting the characteristic information corresponding to the target advertisement information, the characteristic information corresponding to each news historical browsing information and the characteristic information corresponding to each advertisement historical browsing information into a pre-trained interest probability prediction model. And mapping the feature information corresponding to the target advertisement information to a dense vector space by using the interest probability prediction model, and further correspondingly processing the mapped dense vector by using the convolution layer and the pooling layer in the CNN to obtain the target information feature vector corresponding to the target advertisement information.

The interested probability prediction model correspondingly determines the historical information characteristic vector corresponding to the news historical browsing information based on the characteristic information corresponding to each piece of news historical browsing information, and then uses the Attention-net1 to splice the historical information characteristic vector and the target information characteristic vector, and predicts the similarity between the historical information characteristic vector and the target information characteristic vector through the DNN in the Attention-net1 to be used as the similarity corresponding to the news historical browsing information. And correspondingly weighting the historical information characteristic vectors corresponding to the historical browsing information by using the respective corresponding similarity of the historical browsing information to obtain the user characteristic vectors corresponding to the historical browsing information.

Similarly, the interest probability prediction model correspondingly determines the historical information feature vector corresponding to the historical advertisement browsing information based on the feature information corresponding to each piece of historical advertisement browsing information, and further, the historical information feature vector and the target information feature vector are spliced by using the Attention-net2, and the similarity between the historical information feature vector and the target information feature vector is predicted through the DNN in the Attention-net2 to serve as the similarity corresponding to the historical advertisement browsing information. And correspondingly weighting the historical information characteristic vectors corresponding to the historical browsing information of the advertisements by utilizing the respective corresponding similarity of the historical browsing information of the advertisements to obtain the user characteristic vectors corresponding to the historical browsing information of the advertisements.

And then, mapping the user characteristic vector corresponding to the news historical browsing information to a vector space of the user characteristic vector corresponding to the advertisement historical browsing information through project matrix, and splicing the user characteristic vector corresponding to the mapped news historical browsing information and the user characteristic vector corresponding to the advertisement historical browsing information to obtain a target user characteristic vector corresponding to a target user.

The interest probability prediction model splices the target information characteristic vector and the target user characteristic vector, predicts the interest probability of the target user on the target advertisement information based on the spliced characteristic vector by utilizing DNN, and outputs the interest probability. And finally, the server judges whether the interest probability is greater than a preset probability threshold value or not, and if so, the server recommends the target advertisement information to the target user.

Experimental research of the inventor of the present application proves that the performance of the probability prediction model of interest in the embodiment of the present application is significantly better than that of the model in the prior art, and fig. 5 is a schematic diagram of a performance comparison result. The Bayesian Personalized Ranking (BPR) Model, the LibFM Model, the Deep & Wide Model, the Deep fm Model, and the Deep Semantic matching Model (DSSM) Model all belong to the models commonly used in the related art at present, and the DSN Model is the interesting probability prediction Model in the embodiment of the present application. In fig. 5, a group of histograms is formed by every 6 columnar structures from left to right in fig. 5, and the columnar structures from left to right in each group of histograms respectively correspond to a BPR model, a LibFM model, a Deep & Wide model, a Deep fm model, a DSSM model, and a DSN model.

Aiming at the artificial intelligence information recommendation method, the application also provides a corresponding artificial intelligence information recommendation device, so that the information recommendation method is applied and realized in practice.

Referring to fig. 6, fig. 6 is a schematic structural diagram of an artificial intelligence based information recommendation apparatus 600 corresponding to the artificial intelligence based information recommendation method shown in fig. 2, the apparatus comprising:

an obtaining module 601, configured to obtain target information to be recommended and historical browsing information of a target user;

an information feature vector determining module 602, configured to determine, through an interest probability prediction model, a target information feature vector corresponding to the target information;

a user feature vector determining module 603, configured to determine, through the interest probability prediction model, a historical information feature vector corresponding to the historical browsing information, and determine, according to the target information feature vector and the historical information feature vector, a target user feature vector corresponding to the target user;

an interest probability determining module 604, configured to determine, according to the target information feature vector and the target user feature vector, an interest probability of the target user for the target information through the interest probability prediction model;

a recommending module 605, configured to determine whether to recommend the target information to the target user based on the interest probability.

Optionally, on the basis of the artificial intelligence based information recommendation apparatus shown in fig. 6, referring to fig. 7, fig. 7 is a schematic structural diagram of an artificial intelligence based information recommendation apparatus 700 provided in an embodiment of the present application. As shown in fig. 7, the user feature vector determination module 603 includes:

a similarity determining submodule 701, configured to determine, for each piece of acquired historical browsing information, a corresponding historical information feature vector, and determine a similarity between the historical information feature vector and the target information feature vector, where the similarity is used as the similarity corresponding to the piece of historical browsing information;

the user feature vector determining sub-module 702 is configured to perform weighting processing on the history information feature vector corresponding to each piece of history browsing information by using the similarity corresponding to each piece of history browsing information, so as to obtain the target user feature vector.

Optionally, on the basis of the artificial intelligence based information recommendation apparatus shown in fig. 6, in a case that the obtained historical browsing information is from a plurality of different data fields, referring to fig. 8, fig. 8 is a schematic structural diagram of an artificial intelligence based information recommendation apparatus 800 provided in an embodiment of the present application. As shown in fig. 8, the user feature vector determination module 603 includes:

a data domain user feature vector determining submodule 801, configured to determine, for each piece of historical browsing information from the same data domain, a corresponding historical information feature vector, and determine a similarity between the historical information feature vector and the target information feature vector, where the similarity is used as a similarity corresponding to the piece of historical browsing information; weighting the historical information characteristic vectors corresponding to the historical browsing information by using the similarity corresponding to the historical browsing information from the data field to obtain the user characteristic vectors corresponding to the data field;

the user feature vector determining submodule 802 is configured to map, by using the transformation matrix, user feature vectors corresponding to the data fields to the same vector space, so as to obtain the target user feature vector.

Optionally, on the basis of the artificial intelligence-based information recommendation apparatus shown in fig. 7 or fig. 8, the similarity determination sub-module 701 or the data domain user feature vector determination sub-module 801 is specifically configured to:

splicing the historical information characteristic vector and the target information characteristic vector to obtain a first spliced vector;

determining, by a first deep neural network, the similarity based on the first stitching vector.

Optionally, on the basis of the artificial intelligence-based information recommendation apparatus shown in fig. 6, the information feature vector determination module 602 is specifically configured to:

mapping the feature information of the target information to a dense vector space to obtain a dense feature vector corresponding to the target information;

processing the dense feature vector by using a neural network to obtain the target information feature vector; the neural network is a convolutional neural network or a long-short term memory neural network.

Optionally, on the basis of the artificial intelligence-based information recommendation apparatus shown in fig. 6, the interest probability determining module 604 is specifically configured to:

splicing the target information characteristic vector and the target user characteristic vector to obtain a second spliced vector;

determining, by a second deep neural network, the probability of interest based on the second stitching vector.

Optionally, on the basis of the artificial intelligence-based information recommendation apparatus shown in fig. 6, the interest probability prediction model is applied to a news recommendation system; the target information is news information or advertisement information; the historical browsing information comprises at least one of the following: historical viewed news information, historical search news information, and historical viewed advertising information.

Optionally, on the basis of the artificial intelligence-based information recommendation apparatus shown in fig. 6, the interest probability prediction model is applied to a multimedia resource recommendation system; the target information is audio resources or video resources; the historical browsing information comprises at least one of the following: audio resources played historically, video resources played historically, audio resources searched historically, and video resources searched historically.

Optionally, on the basis of the artificial intelligence-based information recommendation apparatus shown in fig. 6, the interest probability prediction model is applied to a commodity recommendation system; the target information is commodity information; the historical browsing information comprises at least one of the following: commodity information of historical purchases, commodity information of historical searches and commodity information of historical collections.

According to the information recommendation device provided by the embodiment of the application, when the user characteristic vector is determined, the characteristic vector corresponding to the historical browsing information and the characteristic vector corresponding to the target information to be recommended currently are comprehensively considered, so that the determined user characteristic vector can more specifically reflect the interest degree of the user on the target information to be recommended currently, accordingly, the interest probability predicted based on the user characteristic vector is more accurate, and the information recommendation can be accurately performed subsequently by a network platform.

The embodiment of the present application further provides a device for information recommendation, where the device may specifically be a server and a terminal device, and the server and the terminal device provided in the embodiment of the present application will be introduced from the perspective of hardware materialization.

Referring to fig. 9, fig. 9 is a schematic structural diagram of a server 900 according to an embodiment of the present disclosure. The server 900 may vary widely in configuration or performance and may include one or more Central Processing Units (CPUs) 922 (e.g., one or more processors) and memory 932, one or more storage media 930 (e.g., one or more mass storage devices) storing applications 942 or data 944. Memory 932 and storage media 930 can be, among other things, transient storage or persistent storage. The program stored on the storage medium 930 may include one or more modules (not shown), each of which may include a series of instruction operations on a server. Still further, a central processor 922 may be provided in communication with the storage medium 930 to execute a series of instruction operations in the storage medium 930 on the server 900.

The server 900 may also include one or more power supplies 926, one or more wired or wireless network interfaces 950, one or more input-output interfaces 958, and/or one or more operating systems 941, such as Windows Server, mac OS XTM, unixTM, linuxTM, freeBSDTM, and so forth.

The steps performed by the server in the above embodiments may be based on the server structure shown in fig. 9.

The CPU 922 is configured to execute the following steps:

determining whether to recommend the target information to the target user based on the interest probability.

Optionally, the CPU 922 can be further configured to execute the steps of any implementation manner of the artificial intelligence based information recommendation method provided in the embodiment of the present application.

Referring to fig. 10, fig. 10 is a schematic structural diagram of a terminal device according to an embodiment of the present application. For convenience of explanation, only the parts related to the embodiments of the present application are shown, and details of the specific technology are not disclosed. The terminal may be any terminal device including a smart phone, a computer, a tablet computer, a Personal digital assistant (hereinafter, referred to as "Personal digital assistant"), and the like, taking the terminal as a mobile phone as an example:

fig. 10 is a block diagram illustrating a partial structure of a mobile phone related to a terminal provided in an embodiment of the present application. Referring to fig. 10, the handset includes: radio Frequency (RF) circuit 1010, memory 1020, input unit 1030, display unit 1040, sensor 1050, audio circuit 1060, wireless fidelity (WiFi) module 1070, processor 1080, and power source 1090. Those skilled in the art will appreciate that the handset configuration shown in fig. 10 is not intended to be limiting and may include more or fewer components than those shown, or some components may be combined, or a different arrangement of components.

The memory 1020 can be used for storing software programs and modules, and the processor 1080 executes various functional applications and data processing of the mobile phone by operating the software programs and modules stored in the memory 1020. The memory 1020 may mainly include a storage program area and a storage data area, wherein the storage program area may store an operating system, an application program required by at least one function (such as a sound playing function, an image playing function, etc.), and the like; the storage data area may store data (such as audio data, a phonebook, etc.) created according to the use of the cellular phone, and the like. Further, the memory 1020 may include high speed random access memory and may also include non-volatile memory, such as at least one magnetic disk storage device, flash memory device, or other volatile solid state storage device.

The processor 1080 is the control center of the handset, connects various parts of the entire handset using various interfaces and lines, and performs various functions of the handset and processes data by running or executing software programs and/or modules stored in the memory 1020 and invoking data stored in the memory 1020. Optionally, processor 1080 may include one or more processing units; preferably, the processor 1080 may integrate an application processor, which handles primarily the operating system, user interfaces, applications, etc., and a modem processor, which handles primarily the wireless communications. It is to be appreciated that the modem processor described above may not be integrated into processor 1080.

In the embodiment of the present application, the processor 1080 included in the terminal further has the following functions:

Optionally, the processor 1080 is further configured to execute the steps of any implementation manner of the artificial intelligence based information recommendation method provided in the embodiment of the present application.

The embodiment of the present application further provides a computer-readable storage medium for storing a computer program, where the computer program is used to execute any one implementation manner of the artificial intelligence based information recommendation method described in the foregoing embodiments.

The embodiments of the present application further provide a computer program product including instructions, which when executed on a computer, cause the computer to perform any one of the implementation manners of the artificial intelligence based information recommendation method according to the foregoing embodiments.

It is clear to those skilled in the art that, for convenience and brevity of description, the specific working processes of the above-described systems, apparatuses and units may refer to the corresponding processes in the foregoing method embodiments, and are not described herein again.

In the several embodiments provided in the present application, it should be understood that the disclosed system, apparatus and method may be implemented in other manners. For example, the above-described apparatus embodiments are merely illustrative, and for example, the division of the units is only one logical division, and other divisions may be realized in practice, for example, a plurality of units or components may be combined or integrated into another system, or some features may be omitted, or not executed. In addition, the shown or discussed mutual coupling or direct coupling or communication connection may be an indirect coupling or communication connection through some interfaces, devices or units, and may be in an electrical, mechanical or other form.

The units described as separate parts may or may not be physically separate, and parts displayed as units may or may not be physical units, may be located in one position, or may be distributed on multiple network units. Some or all of the units can be selected according to actual needs to achieve the purpose of the solution of the embodiment.

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

The integrated unit, if implemented in the form of a software functional unit and sold or used as a stand-alone product, may be stored in a computer readable storage medium. Based on such understanding, the technical solution of the present application may be substantially implemented or contributed to by the prior art, or all or part of the technical solution may be embodied in a software product, which is stored in a storage medium and includes instructions for causing a computer device (which may be a personal computer, a server, or a network device) to execute all or part of the steps of the method according to the embodiments of the present application. And the aforementioned storage medium includes: a usb flash disk, a removable hard disk, a Read-Only Memory (ROM), a Random Access Memory (RAM), a magnetic disk, or an optical disk.

It should be understood that in the present application, "at least one" means one or more, "a plurality" means two or more. "and/or" for describing an association relationship of associated objects, indicating that there may be three relationships, e.g., "a and/or B" may indicate: only A, only B and both A and B are present, wherein A and B may be singular or plural. The character "/" generally indicates that the former and latter associated objects are in an "or" relationship. "at least one of the following" or similar expressions refer to any combination of these items, including any combination of single item(s) or plural items. For example, at least one (one) of a, b, or c, may represent: a, b, c, "a and b", "a and c", "b and c", or "a and b and c", wherein a, b, c may be single or plural.

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

Claims

1. An artificial intelligence based information recommendation method, characterized in that the method comprises:

determining whether to recommend the target information to the target user based on the probability of interest;

wherein the probability of interest prediction model determines the target user feature vector by:

determining a corresponding historical information feature vector for each piece of acquired historical browsing information, and determining the similarity between the historical information feature vector and the target information feature vector as the similarity corresponding to the piece of historical browsing information;

weighting the historical information characteristic vectors corresponding to the historical browsing information by using the similarity corresponding to the historical browsing information to obtain the target user characteristic vector;

alternatively, the first and second electrodes may be,

the acquired historical browsing information comes from a plurality of different data fields; the probability of interest prediction model determines the target user feature vector by:

aiming at each piece of historical browsing information from the same data field, determining a corresponding historical information feature vector, and determining the similarity between the historical information feature vector and the target information feature vector as the similarity corresponding to the piece of historical browsing information; weighting historical information characteristic vectors corresponding to the historical browsing information by utilizing the similarity corresponding to the historical browsing information from the data field to obtain user characteristic vectors corresponding to the data field;

and mapping the user characteristic vectors corresponding to the data domains to the same vector space by using the transformation matrix, and splicing the mapped user characteristic vectors to obtain the target user characteristic vector.

2. The method of claim 1, wherein determining the similarity between the historical information feature vector and the target information feature vector comprises:

3. The method of claim 1, wherein the probability of interest prediction model determines the target information feature vector by:

processing the dense feature vector by using a neural network to obtain the target information feature vector; the neural network is a convolution neural network or a long-short term memory neural network.

4. The method of claim 1, wherein the probability of interest prediction model determines the probability of interest by:

5. The method according to any one of claims 1 to 4, wherein the probabilistic predictive model of interest is applied to a news recommender system; the target information is news information or advertisement information; the historical browsing information comprises at least one of the following: historical viewed news information, historical search news information, and historical viewed advertising information.

6. The method according to any one of claims 1 to 4, wherein the probability of interest prediction model is applied to a multimedia resource recommendation system; the target information is audio resources or video resources; the historical browsing information comprises at least one of the following: audio resources played historically, video resources played historically, audio resources searched historically, and video resources searched historically.

7. The method according to any one of claims 1 to 4, wherein the probability of interest prediction model is applied to a merchandise recommendation system; the target information is commodity information; the historical browsing information comprises at least one of the following: commodity information of historical purchases, commodity information of historical searches and commodity information of historical collections.

8. An artificial intelligence-based information recommendation apparatus, characterized in that the apparatus comprises:

an interest probability determination module, configured to determine, according to the target information feature vector and the target user feature vector, an interest probability of the target user for the target information through the interest probability prediction model;

a recommendation module for determining whether to recommend the target information to the target user based on the interest probability;

the user feature vector determination module comprises:

the similarity determining submodule is used for determining a corresponding historical information characteristic vector aiming at each piece of the acquired historical browsing information, and determining the similarity between the historical information characteristic vector and the target information characteristic vector as the similarity corresponding to the piece of the historical browsing information;

the user characteristic vector determining sub-module is used for weighting the historical information characteristic vectors corresponding to the historical browsing information by utilizing the respective corresponding similarity of the historical browsing information to obtain the target user characteristic vector;

or the acquired historical browsing information comes from a plurality of different data fields; the user feature vector determination module comprises:

the data domain user feature vector determining submodule is used for determining a corresponding historical information feature vector aiming at each piece of historical browsing information from the same data domain, and determining the similarity between the historical information feature vector and the target information feature vector as the similarity corresponding to the piece of historical browsing information; weighting the historical information characteristic vectors corresponding to the historical browsing information by using the similarity corresponding to the historical browsing information from the data field to obtain the user characteristic vectors corresponding to the data field;

and the user characteristic vector determining submodule is used for mapping the user characteristic vectors corresponding to the data fields to the same vector space by using the transformation matrix to obtain the target user characteristic vector.

9. The apparatus of claim 8, wherein the information feature vector determination module is specifically configured to:

10. An apparatus, comprising a processor and a memory;

the memory is used for storing a computer program;

the processor is used for executing the artificial intelligence based information recommendation method of any one of claims 1 to 7 according to the computer program.

11. A computer-readable storage medium for storing a computer program for executing the artificial intelligence based information recommendation method according to any one of claims 1 to 7.