CN114048826B - Recommendation model training method, device, equipment and medium - Google Patents

Recommendation model training method, device, equipment and medium Download PDF

Info

Publication number
CN114048826B
CN114048826B CN202111437512.0A CN202111437512A CN114048826B CN 114048826 B CN114048826 B CN 114048826B CN 202111437512 A CN202111437512 A CN 202111437512A CN 114048826 B CN114048826 B CN 114048826B
Authority
CN
China
Prior art keywords
vector
user
target user
item
target
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Active
Application number
CN202111437512.0A
Other languages
Chinese (zh)
Other versions
CN114048826A (en
Inventor
赵錾
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
China Construction Bank Corp
Original Assignee
China Construction Bank Corp
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by China Construction Bank Corp filed Critical China Construction Bank Corp
Priority to CN202111437512.0A priority Critical patent/CN114048826B/en
Publication of CN114048826A publication Critical patent/CN114048826A/en
Application granted granted Critical
Publication of CN114048826B publication Critical patent/CN114048826B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
    • G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
    • G06F16/90—Details of database functions independent of the retrieved data types
    • G06F16/95—Retrieval from the web
    • G06F16/953—Querying, e.g. by the use of web search engines
    • G06F16/9535—Search customisation based on user profiles and personalisation
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/25—Fusion techniques
    • G06F18/253—Fusion techniques of extracted features
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/04—Architecture, e.g. interconnection topology
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/08—Learning methods

Abstract

The application relates to the technical field of information recommendation, and provides a recommendation model training method, device, equipment and medium, which are used for improving the accuracy of information recommendation. The method comprises the following steps: fusing a user vector of each user and a first evaluation vector of each user on a target item to obtain a first interaction vector of each user on the target item, inputting the item vector of the target item and the first interaction vector into a first neural network to fuse, obtaining a first attention weight of each user, weighting and fusing a plurality of first interaction vectors of the plurality of users on the target item based on the first attention weights of the plurality of users to obtain a first vector of the target item, fusing the first vector of the target item and a second vector of the target user to obtain a prediction rating of the target item by the target user, adjusting model parameters of a recommendation model based on the actual rating and the prediction rating to obtain a trained recommendation model.

Description

Recommendation model training method, device, equipment and medium
Technical Field
The application relates to the technical field of information recommendation, in particular to a recommendation model training method, device, equipment and medium.
Background
With the rapid development of networks and the continuous enrichment of network resources, the amount of information on the networks is also increased, and information overload has become an important challenge. Because of the large amount of redundant data on the network, the search engine can help users to filter information, but cannot help users with ambiguous demands, which severely interfere with the user's acquisition of valid data.
The recommendation system does not require the user to provide specific requirements, models the user and the articles by analyzing and processing the information of the user and the articles, further digs the relation of the user and the articles, and then actively recommends the related articles for the user. However, conventional recommendation systems often lack the ability to distinguish between user trustworthiness, and users often trust friends' feedback rather than other common users when making decisions, resulting in a recommendation system with lower accuracy.
Disclosure of Invention
The embodiment of the application provides a recommendation method, a recommendation device, recommendation equipment and recommendation media, which are used for improving the accuracy of information recommendation.
In a first aspect, the present application provides a recommendation model training method, including:
Fusing the user vector of each user and the first evaluation vector of each user on the target item to obtain a first interaction vector of each user on the target item;
Inputting the item vector of the target item and the first interaction vector into a first neural network for fusion to obtain a first attention weight of each user, wherein the first attention weight is used for indicating the contribution of each user to the target item;
Based on first attention weights of a plurality of users, carrying out weighted fusion on a plurality of first interaction vectors of the target item by the plurality of users to obtain a first vector of the target item, wherein the plurality of users are users who evaluate the target item, and the first vector is used for indicating the characteristics of the target item;
Obtaining a second vector of a target user based on item vectors of a plurality of items, a second evaluation vector of the target user on the plurality of items, and a third evaluation vector of a plurality of neighbor users of the target user on the plurality of items, wherein the plurality of neighbor users are friends with interactive relation with the target user, the plurality of items are items evaluated by the target user, and the second vector is used for indicating characteristics of the target user;
fusing the first vector and the second vector to obtain a prediction rating of the target user on the target item;
Based on the actual rating of the target user on the target item and the prediction rating, adjusting model parameters of the recommendation model to obtain a trained recommendation model;
and generating a recommended item list of the target user based on the trained recommended model.
In one possible embodiment, obtaining the second vector of the target user based on the item vector of the plurality of items, the second rating vector of the target user for the plurality of items, and the third rating vector of the plurality of items by the plurality of neighbor users of the target user comprises:
obtaining a first sub-vector of a target user based on item vectors of a plurality of items and second evaluation vectors of the target user on the plurality of items, wherein the first sub-vector is used for indicating the evaluation condition of the target user on the plurality of items;
obtaining a second sub-vector of the target user based on item vectors of a plurality of items and third evaluation vectors of a plurality of neighbor users of the target user on the plurality of items, wherein the second sub-vector is used for indicating evaluation conditions of the plurality of neighbor users on the plurality of items;
and fusing the first sub-vector and the second sub-vector to obtain a second vector of the target user.
In one possible embodiment, obtaining a first sub-vector of a target user based on an item vector of a plurality of items, a second evaluation vector of the plurality of items by the target user, includes:
Fusing the project vector of each project and the second evaluation vector of each project by the target user to obtain a second interaction vector of each project by the target user;
Inputting the user vector of the target user and the second interaction vector into a second neural network for fusion to obtain a second attention weight of each item, wherein the second attention weight is used for indicating the contribution of the target user to each item;
And based on the second attention weights of the plurality of items, carrying out weighted fusion on a plurality of second interaction vectors of the plurality of items by the target user to obtain a first sub-vector of the target user.
In one possible embodiment, obtaining the second sub-vector of the target user based on the item vector of the plurality of items, the third evaluation vector of the plurality of items by the plurality of neighbor users of the target user, includes:
fusing the project vector of each project and the third evaluation vector of each project by each neighbor user to obtain a third interaction vector of each neighbor user on each project;
Based on the second attention weights of the plurality of items, weighting and fusing a plurality of third interaction vectors of the plurality of items by each neighbor user to obtain a third sub-vector of each neighbor user, wherein the third sub-vector is used for indicating the evaluation condition of each neighbor user on the plurality of items;
Fusing the user vector of the target user and the third interaction vector to obtain a third attention weight of each neighbor user, wherein the third attention weight is used for indicating the importance of each neighbor user to the target user;
And carrying out weighted fusion on a plurality of third sub-vectors of the plurality of neighbor users based on the third attention weights of the plurality of neighbor users to obtain a second sub-vector of the target user.
In a second aspect, the present application provides a recommendation model training method, including:
The acquisition module is used for fusing the user vector of each user and the first evaluation vector of each user on the target item to acquire the first interaction vector of each user on the target item;
the obtaining module is further configured to input the item vector of the target item and the first interaction vector into a first neural network to be fused, so as to obtain a first attention weight of each user, where the first attention weight is used to indicate a contribution of each user to the target item;
the obtaining module is further configured to weight and fuse a plurality of first interaction vectors of the target item by a plurality of users based on first attention weights of the plurality of users, to obtain a first vector of the target item, where the plurality of users are users who have evaluated the target item, and the first vector is used to indicate characteristics of the target item;
the obtaining module is further configured to obtain a second vector of the target user based on an item vector of a plurality of items, a second evaluation vector of the target user on the plurality of items, and a third evaluation vector of a plurality of neighbor users of the target user on the plurality of items, where the plurality of neighbor users are friends having an interaction relationship with the target user, the plurality of items are items evaluated by the target user, and the second vector is used to indicate characteristics of the target user;
the obtaining module is further configured to fuse the first vector and the second vector to obtain a prediction rating of the target user on the target item;
The adjustment module is used for adjusting model parameters of the recommendation model based on the actual rating of the target user on the target item and the prediction rating to obtain a trained recommendation model;
And the generation module is used for generating a recommended item list of the target user based on the trained recommended model.
In a possible embodiment, the obtaining module is specifically configured to:
obtaining a first sub-vector of a target user based on item vectors of a plurality of items and second evaluation vectors of the target user on the plurality of items, wherein the first sub-vector is used for indicating the evaluation condition of the target user on the plurality of items;
obtaining a second sub-vector of the target user based on item vectors of a plurality of items and third evaluation vectors of a plurality of neighbor users of the target user on the plurality of items, wherein the second sub-vector is used for indicating evaluation conditions of the plurality of neighbor users on the plurality of items;
and fusing the first sub-vector and the second sub-vector to obtain a second vector of the target user.
In a possible embodiment, the obtaining module is specifically configured to:
Fusing the project vector of each project and the second evaluation vector of each project by the target user to obtain a second interaction vector of each project by the target user;
Inputting the user vector of the target user and the second interaction vector into a second neural network for fusion to obtain a second attention weight of each item, wherein the second attention weight is used for indicating the contribution of the target user to each item;
And based on the second attention weights of the plurality of items, carrying out weighted fusion on a plurality of second interaction vectors of the plurality of items by the target user to obtain a first sub-vector of the target user.
In a possible embodiment, the obtaining module is specifically configured to:
fusing the project vector of each project and the third evaluation vector of each project by each neighbor user to obtain a third interaction vector of each neighbor user on each project;
Based on the second attention weights of the plurality of items, weighting and fusing a plurality of third interaction vectors of the plurality of items by each neighbor user to obtain a third sub-vector of each neighbor user, wherein the third sub-vector is used for indicating the evaluation condition of each neighbor user on the plurality of items;
Fusing the user vector of the target user and the third interaction vector to obtain a third attention weight of each neighbor user, wherein the third attention weight is used for indicating the importance of each neighbor user to the target user;
And carrying out weighted fusion on a plurality of third sub-vectors of the plurality of neighbor users based on the third attention weights of the plurality of neighbor users to obtain a second sub-vector of the target user.
In a third aspect, the present application provides an electronic device comprising:
a memory for storing program instructions;
A processor for invoking program instructions stored in the memory and executing the method of any of the first aspects in accordance with the obtained program instructions, comprising.
In a fourth aspect, the present application provides a computer readable storage medium storing a computer program comprising program instructions which, when executed by a computer, cause the computer to perform the method of any one of the first aspects.
In a fifth aspect, the present application provides a computer program product comprising: computer program code which, when run on a computer, causes the computer to perform the method of any of the first aspects.
In the embodiment of the application, the attention weight is distributed to the evaluation of each user on the target item through the attention mechanism, and the first interaction vectors of the target item are weighted and fused by the users based on the first attention weights of the users to obtain the first vector of the target item, so that the characteristics more in accordance with the actual target item are mined. Based on item vectors of a plurality of items, second evaluation vectors of a target user on the plurality of items and third evaluation vectors of a plurality of neighbor users of the target user on the plurality of items, the second vectors of the target user are obtained, evaluation of the plurality of neighbor users is combined, so that more characteristics of the target user are mined, the first vectors and the second vectors are finally fused, the obtained prediction rating is closer to the actual rating, and further the accuracy of a recommendation result of the trained recommendation model is higher. Compared with a common self-attention model, the neural network is faster in operation speed, a plurality of attention weights can be obtained rapidly, the training efficiency of the whole recommendation model is further improved, and the recommendation efficiency of the recommendation model is also improved.
Drawings
Fig. 1 is a schematic diagram of an application scenario of a recommendation model training method according to an embodiment of the present application;
FIG. 2 is a diagram of graphical data in social recommendation provided by an embodiment of the present application;
FIG. 3 is a schematic structural diagram of a recommendation model according to an embodiment of the present application;
FIG. 4 is a flowchart of a recommendation model training method according to an embodiment of the present application;
FIG. 5 is a schematic structural diagram of an item module according to an embodiment of the present application;
fig. 6 is a schematic structural diagram of a user module according to an embodiment of the present application;
Fig. 7 is a schematic structural diagram of a first obtaining module according to an embodiment of the present application;
fig. 8 is a schematic structural diagram of a second obtaining module according to an embodiment of the present application;
FIG. 9 is a block diagram of a recommendation model training apparatus according to an embodiment of the present application;
fig. 10 is a block diagram of an electronic device according to an embodiment of the present application.
Detailed Description
For the purpose of making the objects, technical solutions and advantages of the present application more apparent, the technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present application, and it is apparent that the described embodiments are only some embodiments of the present application, not all embodiments of the present application. All other embodiments, which can be made by those skilled in the art based on the embodiments of the application without making any inventive effort, are intended to be within the scope of the application. Embodiments of the application and features of the embodiments may be combined with one another arbitrarily without conflict. Also, while a logical order of illustration is depicted in the flowchart, in some cases the steps shown or described may be performed in a different order than presented.
The terms first and second in the description and claims of the application and in the above-mentioned figures are used for distinguishing between different objects and not for describing a particular sequential order. Furthermore, the term "include" and any variations thereof is intended to cover non-exclusive protection. For example, a process, method, system, article, or apparatus that comprises a list of steps or elements is not limited to only those listed steps or elements but may include other steps or elements not listed or inherent to such process, method, article, or apparatus.
In the embodiment of the present application, the "plurality" may mean at least two, for example, two, three or more, and the embodiment of the present application is not limited.
In order to improve accuracy of information recommendation, the embodiment of the application provides a recommendation model training method which can be executed by recommendation model training equipment. For the sake of simplifying the description, the recommended model training apparatus will be simply referred to as training apparatus hereinafter. The training device may be implemented by a terminal, such as a personal computer, a cell phone, or an embedded device, such as a camera, or a server, such as a physical service or a virtual server.
An application scenario schematic diagram of the recommendation model training method is described below. Referring to fig. 1, an application scenario diagram of a recommendation model training method according to an embodiment of the present application is shown. The application scenario diagram includes sample data 110 and training device 120.
After the training device 120 obtains the sample data 110, the recommendation model is trained according to the sample data 110, and a trained recommendation model is obtained, and is used for generating a recommendation item list for the user. The process of how the recommendation model is trained will be described below.
Referring to fig. 2, a schematic diagram of graphical data in social recommendation provided by an embodiment of the present application is shown, where the graphical data includes a user-project diagram 201 and a user-user diagram 202, the user-project diagram 201 includes user ratings for different projects, the user ratings for project 1 are 3 stars, the user ratings for project 2 are 4 stars, and the user ratings for project 3 are 5 stars. The user-user graph 202 may also be referred to as a social relationship graph, including connections of users to different neighbor users.
Since the graphic data in the social recommendation includes a user-item graph and a user-user graph, the recommendation model needs to extract relevant information from the two graphs, respectively. Referring to fig. 3, a schematic structural diagram of a recommendation model according to an embodiment of the present application is shown. The recommendation model in the embodiment of the application comprises an item module 301, a user module 302 and a rating prediction module 303, wherein the item module 301 is used for extracting item characteristics, namely potential influencing factors of the item. The user module 302 is used to extract user features, i.e., user space potential influencing factors. The rating prediction module 303 is configured to fuse the item features output by the item module 301 and the user features output by the user module 302 to obtain a rating prediction result.
Based on the foregoing discussion of fig. 1-3, a recommendation model training method is described below as an example of the training apparatus of fig. 1. Fig. 4 is a flowchart of a recommendation model training method according to an embodiment of the application.
S401, fusing the user vector of each user and the first evaluation vector of each user on the target item to obtain the first interaction vector of each user on the target item.
The training device may represent each user with a vector, referred to as a user vector, e.g., user u t has a user vector of P t. And vector representation of each item, referred to as an item vector, e.g., item vector q j for item v j. After a user purchases an item or receives a service for an item, the user typically evaluates the item, and the training device may vector the user's evaluation of the item, referred to as an evaluation vector. For example, in a five-star rating system, the star rating is denoted as r, r.epsilon.1, 2,3,4,5, and the rating vector is e r.
Specifically, the training device may splice the user vector of each user and the first evaluation vector of each user for the target item, input a Multi-layer Perceptron (MLP), and output the first interaction vector of each user for the target item.
For example:
Where p t represents the user vector for user u t, e r represents the first rating vector for user u t for target item v j, Representing vector stitching, f jt represents the first interaction vector of user u t to target item v j, and g u () represents the processing of the MLP model.
S402, inputting the item vector of the target item and the first interaction vector into a first neural network for fusion, and obtaining a first attention weight of each user.
Specifically, the training device may input the project vector of the target project and the first interaction vector of each user on the target project into the first neural network to be fused, output the attention score of each user, normalize the attention score of each user, and obtain the first attention weight of each user. The first attention weight is used to indicate the contribution of each user to the target item, and can also be understood as the impact weight of each user's evaluation of the target item on the item module.
For example, the first neural network is a two-layer neural network, which may also be referred to as a first attention network, defined as:
Where q j represents the item vector of the target item v j, f jt represents the first interaction vector of the user u t on the target item v j, W 1 and W 2 represent the weights of the first layer and the second layer, b 1 and b 2 represent the deviations of the first layer and the second layer, respectively, and σ represents the nonlinear activation function.
The normalization formula is as follows:
Wherein, Representing the attention score of user u t to target item v j, B (j) representing the set of users who have rated target item v j, μ jt representing the first attention weight of user u t to target item v j in user set B (j).
S403, based on the first attention weights of the users, the first interaction vectors of the target items are weighted and fused by the users to obtain the first vectors of the target items.
After the training device obtains the first attention weight of each user, the training device may perform weighted fusion on a plurality of first interaction vectors of the target item by the plurality of users based on the first attention weights of the plurality of users, and use the weighted fused vector as the first vector of the target item. Wherein the plurality of users are users evaluating the target item and the first vector is used to indicate the characteristics of the target item.
For example, the weighted fusion formula is as follows:
Where μ jt represents the first attention weight of user u t to target item v j, f jt represents the first interaction vector of user u t to target item v j, B (j) represents the set of users who evaluated target item v j, and z j represents the first vector of target item v j.
Or the training device may perform weighted fusion on the first attention weights of the plurality of users and the plurality of first interaction vectors of the plurality of users on the target item to obtain a first vector of the target item.
For example, the formula for a neural network is as follows:
Wherein σ represents a nonlinear activation function, W and B represent weights and deviations of the neural network, respectively, μ jt、fjt,zj and B (j) refer to the content discussed above, and are not repeated here.
Considering that the contributions of different users to the same target item are different, the attention mechanism is introduced in the embodiment of the application, a first attention weight is allocated to the evaluation of each user to the target item, and the first attention weight can capture heterogeneous influences from the user-item graph so as to reflect the different contributions of different users to the target item. And the attention mechanism is realized through the neural network, so that the operation efficiency can be improved, and a plurality of first attention weights of a plurality of users to the target item can be obtained rapidly.
Referring to fig. 5, a schematic structural diagram of an item module according to an embodiment of the present application is provided, the item module includes three first fusion modules 501, a first attention network 502, and a second fusion module 503, and fig. 5 is an example of three first fusion modules 501, where in fact, the number of first fusion modules 501 is the same as the number of users evaluating the target item. Dashed arrows represent inputs and outputs of the first attention network 502.
The first fusion module 501 is configured to fuse a user vector of a user and a first evaluation vector of the user on a target item, so as to obtain a first interaction vector of the user on the target item. For example, the user vector p 1 of the first user and the first evaluation vector e 1 of the first user on the target item are input into one of the first fusion modules 501, so as to obtain a first interaction vector f 1 of the first user on the target item, and so on, the other two first fusion modules 501 output a first interaction vector f 2 of the second user on the target item and a first interaction vector f 3 of the third user on the target item respectively.
The first attention network 502, that is, the first neural network, is configured to fuse the item vector of the target item and the first interaction vector of each user on the target item to obtain a first attention weight of each user, for example, the first attention weights of three users are μ 1、μ2、μ3 respectively. The second fusion module 503 is configured to perform weighted fusion on the first attention weights of the three users and the three first interaction vectors of the target item by the three users, so as to obtain a first vector of the target item.
S404, obtaining a second vector of the target user based on the project vectors of the projects, the second evaluation vectors of the target user on the projects and the third evaluation vectors of the neighbor users on the projects.
The recommendation model can be combined with the evaluation condition of the target user on the plurality of items and the evaluation condition of the plurality of neighbor users on the plurality of items to obtain a second vector of the target user. The plurality of neighbor users are friends with interactive relation with the target user, and the neighbor users are users who pay attention to each other with the target user, or send messages to each other with the target user, or transfer money to each other with the target user. The plurality of items are items evaluated by the target user, and the second vector is used for indicating the characteristics of the target user, namely the space potential factors of the target user.
Specifically, the training device may obtain a first sub-vector of the target user based on the project vectors of the plurality of projects and the second evaluation vectors of the target user on the plurality of projects, where the first sub-vector is used to indicate the evaluation condition of the target user on the plurality of projects, that is, the project space influence factor of the target user, and may be understood as the influence of the project factor on the target user. And obtaining a second sub-vector of the target user based on the project vector of the plurality of projects and the third evaluation vector of the plurality of neighbor users on the plurality of projects, wherein the second sub-vector is used for indicating the evaluation condition of the plurality of neighbor users on the plurality of projects, namely social influence factors of the target user, and can be understood as influence of the social influence factors on the target user. And then fusing the first sub-vector and the second sub-vector to obtain the second vector of the target user, namely the space potential influence factor of the target user.
In one possible embodiment, the training device obtains the first sub-vector of the target user as follows:
s1.1, fusing the project vector of each project and the second evaluation vector of the target user on each project to obtain a second interaction vector of the target user on each project.
Specifically, the training device may splice the item vector of each item and the second evaluation vector of the target user for each item, and then input the second interaction vector of the target user for each item into the MLP model.
For example:
Where q a represents the item vector of item v a, e r represents the second evaluation vector of item v a by the target user u i, Representing vector stitching, x ia represents the second interaction vector of user u i to item v a, and g v () represents the output of the MLP model.
S1.2, inputting the user vector of the target user and the second interaction vector into a second neural network for fusion, and outputting the second attention weight of each item.
Specifically, the training device inputs the user vector of the target user and a plurality of second interaction vectors of the target user for each item into the first neural network to be fused, outputs the attention score of each item, normalizes the attention score of each item, and obtains a second attention weight of each item, wherein the second attention weight is used for indicating the contribution of the target user to each item.
For example, the second neural network is a two-layer neural network, defined as:
Where p i represents the user vector of the target user u i, x ia represents the second interaction vector of the target user u i with the item v a, Representing vector stitching, W 1 and W 2 represent weights of the first and second layers, respectively, b 1 and b 2 represent deviations of the second layer, respectively, σ represents a nonlinear activation function.
The normalization formula is as follows:
Wherein, Representing the attention score of the target user u i to the item v a, C (j) representing the number of items evaluated by the target user u i, and α ia representing the second attention weight of the target user u i to the item v a.
Considering that the contribution of the target user to different projects is different, the attention mechanism is introduced in the embodiment of the application, and a second attention weight is allocated to the evaluation of the target user to each project so as to reflect the different contributions of the target user to the different projects. And the attention mechanism is realized through the neural network, so that the operation efficiency can be improved, and a plurality of second attention weights of the target user to a plurality of projects can be obtained rapidly.
S1.3, based on the second attention weights of the plurality of items, carrying out weighted fusion on a plurality of second interaction vectors of the plurality of items by the target user, and obtaining a first sub-vector of the target user.
Specifically, the training device may perform weighted fusion on the second attention weights of the plurality of items and the plurality of second interaction vectors input by the target user to the neural network of the plurality of items, to obtain a first sub-vector of the target user.
For example, the formula for a neural network is:
Where a ia represents the second attention weight of the target user u i to item v a, x ia represents the second interaction vector of the target user u i to item v a, The first sub-vector representing the target user u i, σ is the nonlinear activation function, and W and b represent the weight and bias, respectively.
In one possible embodiment, the training device obtains the second sub-vector of the target user as follows:
S2.1, fusing the project vector of each project and the third evaluation vector of each neighbor user on each project to obtain the third interaction vector of each neighbor user on each project.
Specifically, the training device may splice the project vector of each project and the third evaluation vector of each neighboring user for each project, and then input the third interaction vector of each neighboring user for each project into the MLP model. The process of obtaining the third interaction vector is referred to the process of obtaining the second interaction vector, which is discussed above, and will not be described herein.
S2.2, based on the second attention weights of the plurality of items, weighting and fusing the plurality of third interaction vectors of the plurality of items by each neighbor user to obtain a third sub-vector of each neighbor user.
Specifically, the training device may perform weighted fusion on the second attention weights of the plurality of items and the plurality of third interaction vectors input by each neighbor user to the neural network of the plurality of items, to obtain a third sub-vector of each neighbor user. The third sub-vector is used to indicate the evaluation of multiple items by each neighbor user. The process of obtaining the third sub-vector is referred to the process of obtaining the first sub-vector of the target user, which is not described herein.
S2.3, fusing the user vector of the target user and the third interaction vector to obtain the third attention weight of each neighbor user.
The third attention weight is used to indicate the importance of each neighbor user to the target user. There are various ways of obtaining the third attention weight, and the description will be made below.
Mode one obtains a third attention weight through a self-attention model.
The training device may integrate the user vector of the target user and the third interaction vector of each neighboring user for each item with the self-attention model, output the attention score of each neighboring user, normalize the attention score of each neighboring user, and obtain the third attention weight of each neighboring user.
The self-attention model is based on scaling dot product, and comprises three inputs of query, key and value. For example, the formula for the self-attention model is as follows:
Wherein, A third sub-vector representing neighbor user u o, p i representing the user vector of target user u i,/>Representing vector concatenation, reLU representing a linear rectification function, W qi and W ki representing the weight matrices of the query and key, respectively, S i representing the correlation matrices of W qi and W ki, softmax representing a normalization function,/>Representing the attention score of neighbor user u o, d is the embedded length of user vector p i, β io represents the third attention weight of neighbor user u o, and N (i) represents the neighbor user set of target user u i.
And secondly, realizing an attention mechanism through a neural network.
The training device may input the user vector of the target user and the third interaction vector of each neighboring user for each item into the third neural network to be fused, output the attention score of each neighboring user, normalize the attention score of each neighboring user, and obtain the third attention weight of each neighboring user. The process of obtaining the third attention weight through the third neural network refers to the content of obtaining the first attention weight or the second attention weight discussed above, and will not be described herein.
In consideration of different importance of different neighbor users to the target user, the attention mechanism is introduced in the embodiment of the application, and a third attention weight is distributed to each neighbor user of the target user so as to embody the importance of different neighbor users to the target user. And the attention mechanism is realized through the neural network, so that the operation efficiency can be improved, and a plurality of third attention weights of a plurality of neighbor users to the target user can be obtained rapidly.
S2.4, based on the third attention weights of the plurality of neighbor users, carrying out weighted fusion on the plurality of third sub-vectors of the plurality of neighbor users to obtain a second sub-vector of the target user.
Specifically, the training device may input the third attention weights of the plurality of neighbor users and the plurality of third sub-vectors of the plurality of neighbor users into the neural network to perform weighted fusion, so as to obtain the second sub-vector of the target user.
For example, the formula for a neural network is as follows:
Where beta io denotes the third attention weight of the neighbor user u o of the target user u i, A third sub-vector representing neighbor user u o, N (i) representing the neighbor user set of target user u i, σ being the nonlinear activation function, W and b representing weights and biases, respectively,/>A second sub-vector representing the target user u i.
In one possible embodiment, the training device may input the first sub-vector and the second sub-vector into the MLP model, outputting the second vector of the target user.
For example:
c2＝σ(W2·c1+b2)
……
hi＝σ(Wl·cl-1+bl)
Wherein, First sub-vector representing target user u i,/>A second sub-vector representing the target user u i,/>Representing vector concatenation, i is the index of the hidden layer, c l-1 represents the output of the hidden layer of layer 1, W l and b l represent the weight and bias of the hidden layer of layer 1, respectively, σ is a nonlinear activation function, and h i represents the second vector of the target user u i.
Referring to fig. 6, a schematic structural diagram of a user module according to an embodiment of the present application is provided, where the user module includes a first obtaining module 601, a second obtaining module 602, and a fusion module 603. Wherein, the first obtaining module 601 is configured to learn to obtain a first sub-vector of the target user from the user-project graph. The second obtaining module 602 is configured to learn to obtain a second sub-vector of the target user from the social relationship diagram. The fusion module 603 is configured to fuse the first sub-vector and the second sub-vector to obtain a second vector of the target user. The meaning of the first sub-vector, the second sub-vector and the second vector is referred to in the foregoing discussion, and will not be repeated here.
Referring to fig. 7, a schematic structural diagram of a first obtaining module according to an embodiment of the present application is provided, where the first obtaining module includes three fourth fusion modules 701, a second attention network 702 and a fifth fusion module 703, and fig. 7 is an example of three fourth fusion modules 701, and in fact, the number of fourth fusion modules 701 is the same as the number of items evaluated by a target user. The dashed arrows represent the input and output of the second attention network 702.
The fourth fusion module 701 is configured to fuse the item vector of each item and the second evaluation vector of each item by the target user, so as to obtain a second interaction vector of each item by the target user. For example, the item vector q 1 of the first item and the second evaluation vector e 1 of the target user on the first item are input into one fourth fusion module 701 to obtain the second interaction vector x 1 of the target user on the first item, and so on, the other two fourth fusion modules 701 output the second interaction vector x 2 of the target user on the second item and the second interaction vector x 3 of the target user on the third item respectively.
The second attention network 702, that is, the second neural network, is configured to fuse the user vector of the target user and the second interaction vector of the target user for each item, and output a second attention weight of each item, for example, the second attention weights of the three items are α 1、α2、α3, respectively. The fifth fusion module 703 is configured to perform weighted fusion on the second attention weights of the three items and the three second interaction vectors of the three items by the target user, so as to obtain a first sub-vector of the target user.
Referring to fig. 8, a schematic structural diagram of a second obtaining module according to an embodiment of the present application is provided, where the second obtaining module includes three sixth fusion modules 801, a third attention network 802, and a seventh fusion module 803. Fig. 8 illustrates three sixth fusion modules 801 as an example, where in practice the number of sixth fusion modules 801 is the same as the number of neighbor users of the target user. Dashed arrows represent the input and output of the third attention network 802.
The third fusing module 801 is configured to obtain third sub-vectors of three neighboring users, where the meaning and the obtaining process of the third sub-vectors refer to the content discussed above, and are not described herein again. The third attention network 802, that is, the third neural network, is configured to fuse the user vector of the target user and the third interaction vector of each item by each neighboring user, and output the third attention weight of each neighboring user, for example, the third attention weights of the three neighboring users are β 1、β2、β3, respectively. The seventh fusion module 803 is configured to perform weighted fusion on the third attention weights of the three neighbor users and the three third sub-vectors of the three neighbor users, so as to obtain a second sub-vector of the target user.
S405, fusing the first vector and the second vector to obtain the prediction rating of the target user on the target item.
Specifically, the training device may input the first vector and the second vector into the MLP model, outputting a prediction rating of the target item by the target user.
For example:
g2＝σ(W2·g1+b2)
……
gl-1＝σ(Wl·gl-1+bl)
r′ij＝wT·gl-1
Where z j denotes a first vector, h i denotes a second vector, Representing the concatenation of two vectors, i being the index of the hidden layer, g l-1 representing the output of the l-1 hidden layer, W l and b l representing the weights and deviations of the l hidden layer, respectively, σ being the nonlinear activation function, r' ij representing the prediction rating of the target user u i on the target item v j. /(I)
S406, based on the actual rating and the prediction rating of the target user on the target item, adjusting the model parameters of the recommendation model to obtain the trained recommendation model.
Specifically, the training device may define a loss function according to the actual rating and the prediction rating, and continuously optimize the loss function to update the model parameters of the recommendation model according to the negative gradient direction until the loss function reaches the minimum value, so as to consider that the recommendation model converges and obtain the recommendation model after training.
For example, the loss function is formulated as follows:
where |o| is the number of ratings, r' ij denotes the predicted rating of the target user u i on the target item v j, r ij denotes the actual rating of the target user u i on the target item v j, and Loss denotes the Loss value.
In one possible embodiment, to alleviate the over-fitting problem in the neural network, the training device may apply Dropout to the recommendation model during training, randomly discard some neurons of the recommendation model during training, and update only some parameters when updating parameters of the recommendation model.
S407, generating a recommended item list of the target user based on the trained recommended model.
After the training device obtains the trained recommendation model, a recommendation item list of the target user can be generated based on the trained recommendation model, wherein the recommendation item list comprises a plurality of items which are arranged according to the high-low order of the prediction rating, and the higher the prediction rating is, the more interested the target user is in the items.
It should be noted that even if Dropout is employed by the training device during training of the recommendation model, since Dropout is disabled during testing, the training device does not discard neurons of the recommendation model during testing, and uses the entire recommendation model to generate a recommendation item list for the target user.
For example, a user purchases some financial products or financial products at a bank, and the recommendation model generates a recommended product list of the user according to a historical purchase record of the user and a social relationship of the user, wherein the historical purchase record comprises products purchased by the user and ratings of the products, and the social relationship comprises a neighbor user of the user and the historical purchase record of the neighbor user.
Based on the same inventive concept, an embodiment of the present application provides a recommended model training device, which is disposed in the training apparatus discussed above, please refer to fig. 9, and includes:
the obtaining module 901 is configured to fuse a user vector of each user with a first evaluation vector of each user on a target item, so as to obtain a first interaction vector of each user on the target item;
The obtaining module 901 is further configured to input an item vector of a target item and a first interaction vector into a first neural network to be fused, so as to obtain a first attention weight of each user, where the first attention weight is used to indicate a contribution of each user to the target item;
The obtaining module 901 is further configured to weight and fuse a plurality of first interaction vectors of the target item by a plurality of users based on first attention weights of the plurality of users, to obtain a first vector of the target item, where the plurality of users are users who evaluate the target item, and the first vector is used to indicate a feature of the target item;
The obtaining module 901 is further configured to obtain a second vector of the target user based on an item vector of a plurality of items, a second evaluation vector of the target user on the plurality of items, and a third evaluation vector of a plurality of neighbor users of the target user on the plurality of items, where the plurality of neighbor users are friends having an interaction relationship with the target user, the plurality of items are items evaluated by the target user, and the second vector is used to indicate characteristics of the target user;
the obtaining module 901 is further configured to fuse the first vector and the second vector to obtain a prediction rating of the target user on the target item;
The adjusting module 902 is configured to adjust model parameters of the recommendation model based on an actual rating and a prediction rating of the target user on the target item, and obtain a trained recommendation model;
the generating module 903 is configured to generate a recommended item list of the target user based on the trained recommendation model.
In one possible embodiment, the obtaining module 901 is specifically configured to:
Based on item vectors of the plurality of items and second evaluation vectors of the target user on the plurality of items, a first sub-vector of the target user is obtained, wherein the first sub-vector is used for indicating the evaluation condition of the target user on the plurality of items;
Obtaining a second sub-vector of the target user based on the project vector of the plurality of projects and a third evaluation vector of the plurality of neighbor users of the target user on the plurality of projects, wherein the second sub-vector is used for indicating the evaluation condition of the plurality of neighbor users on the plurality of projects;
And fusing the first sub-vector and the second sub-vector to obtain a second vector of the target user.
In one possible embodiment, the obtaining module 901 is specifically configured to:
Fusing the project vector of each project and the second evaluation vector of each project by the target user to obtain a second interaction vector of each project by the target user;
Inputting the user vector of the target user and the second interaction vector into a second neural network for fusion to obtain a second attention weight of each item, wherein the second attention weight is used for indicating the contribution of the target user to each item;
and based on the second attention weights of the plurality of items, carrying out weighted fusion on the plurality of second interaction vectors of the plurality of items by the target user to obtain a first sub-vector of the target user.
In one possible embodiment, the obtaining module 901 is specifically configured to:
fusing the project vector of each project and the third evaluation vector of each project by each neighbor user to obtain a third interaction vector of each neighbor user on each project;
Based on the second attention weights of the plurality of items, weighting and fusing a plurality of third interaction vectors of the plurality of items by each neighbor user to obtain a third sub-vector of each neighbor user, wherein the third sub-vector is used for indicating the evaluation condition of the plurality of items by each neighbor user;
fusing the user vector of the target user and the third interaction vector to obtain a third attention weight of each neighbor user, wherein the third attention weight is used for indicating the importance of each neighbor user to the target user;
And carrying out weighted fusion on a plurality of third sub-vectors of the plurality of neighbor users based on the third attention weights of the plurality of neighbor users to obtain a second sub-vector of the target user.
As an embodiment, the apparatus discussed in fig. 9 may implement any of the recommended model training methods discussed above, and will not be described here.
Based on the same inventive concept, an embodiment of the present application provides an electronic device, which may implement the functions of the training device discussed above, referring to fig. 10, the device includes a processor 1001 and a memory 1002.
A memory 1002 for storing program instructions;
The processor 1001 is configured to invoke the program instructions stored in the memory 1002, and execute the steps included in any of the recommended model training methods discussed above according to the obtained program instructions. Because the principle of solving the problem of the electronic device is similar to that of the recommended model training method, the implementation of the electronic device can be referred to the implementation of the method, and the repetition is omitted.
The processor 1001 may be a central processing unit (central processing unit, CPU), or be a digital processing unit, or be a combination of one or more of an image processor, etc. The memory 1002 may be a volatile memory (RAM) such as a random-access memory (RAM); the memory 1002 may also be a non-volatile memory (non-volatile memory), such as a read-only memory, a flash memory (flash memory), a hard disk (HARD DISK DRIVE, HDD) or a Solid State Disk (SSD), or any other medium that can be used to carry or store desired program code in the form of instructions or data structures and that can be accessed by a computer, but is not limited thereto. The memory 1002 may be a combination of the above.
As an example, the processor 1001 in fig. 10 may implement any of the recommended model training methods discussed above, and the processor 1001 may also implement the functions of the recommended model training apparatus discussed above in fig. 9.
Based on the same inventive concept, embodiments of the present application provide a computer-readable storage medium storing a computer program comprising program instructions that, when executed by a computer, cause the computer to perform a recommended model training method as any one of the preceding discussion. Since the principle of solving the problem by the computer readable storage medium is similar to that of the recommended model training method, implementation of the computer readable storage medium can refer to implementation of the method, and repeated parts are omitted.
Based on the same inventive concept, embodiments of the present application also provide a computer program product comprising: computer program code which, when run on a computer, causes the computer to perform the recommended model training method as any of the previous discussions. Since the principle of the solution of the problem of the computer program product is similar to that of the recommended model training method, the implementation of the computer program product can refer to the implementation of the method, and the repetition is omitted.
It will be appreciated by those skilled in the art that embodiments of the present application may be provided as a method, system, or computer program product. Accordingly, the present application may take the form of an entirely hardware embodiment, an entirely software embodiment or an embodiment combining software and hardware aspects. Furthermore, the present application may take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, and the like) having computer-usable program code embodied therein.
The present application is described with reference to flowchart illustrations and/or block diagrams of methods, apparatus (systems) and computer program products according to the application. It will be understood that each flow and/or block of the flowchart illustrations and/or block diagrams, and combinations of flows and/or blocks in the flowchart illustrations and/or block diagrams, can be implemented by computer program instructions. These computer program instructions may be provided to a processor of a general purpose computer, special purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, create means for implementing the functions specified in the flowchart flow or flows and/or block diagram block or blocks.
These computer program instructions may also be stored in a computer-readable memory that can direct a computer or other programmable data processing apparatus to function in a particular manner, such that the instructions stored in the computer-readable memory produce an article of manufacture including instruction means which implement the function specified in the flowchart flow or flows and/or block diagram block or blocks.
These computer program instructions may also be loaded onto a computer or other programmable data processing apparatus to cause a series of operational steps to be performed on the computer or other programmable apparatus to produce a computer implemented process such that the instructions which execute on the computer or other programmable apparatus provide steps for implementing the functions specified in the flowchart flow or flows and/or block diagram block or blocks.
It will be apparent to those skilled in the art that various modifications and variations can be made to the present application without departing from the spirit or scope of the application. Thus, it is intended that the present application also include such modifications and alterations insofar as they come within the scope of the appended claims or the equivalents thereof.

Claims (11)

1. A recommendation model training method, comprising:
Fusing the user vector of each user and the first evaluation vector of each user on the target item to obtain a first interaction vector of each user on the target item;
Inputting the item vector of the target item and the first interaction vector into a first neural network for fusion to obtain a first attention weight of each user, wherein the first attention weight is used for indicating the contribution of each user to the target item;
Based on first attention weights of a plurality of users, carrying out weighted fusion on a plurality of first interaction vectors of the target item by the plurality of users to obtain a first vector of the target item, wherein the plurality of users are users who evaluate the target item, and the first vector is used for indicating the characteristics of the target item;
Obtaining a second vector of a target user based on item vectors of a plurality of items, a second evaluation vector of the target user on the plurality of items, and a third evaluation vector of a plurality of neighbor users on the plurality of items, wherein the plurality of neighbor users are friends with interactive relation with the target user, the plurality of items are items evaluated by the target user, and the second vector is used for indicating characteristics of the target user;
fusing the first vector and the second vector to obtain a prediction rating of the target user on the target item;
Based on the actual rating of the target user on the target item and the prediction rating, adjusting model parameters of the recommendation model to obtain a trained recommendation model;
and generating a recommended item list of the target user based on the trained recommended model.
2. The method of claim 1, wherein obtaining the second vector for the target user based on the item vector for the plurality of items, the second vector for the target user for the plurality of items, and the third vector for the plurality of items for the plurality of neighbor users of the target user comprises:
obtaining a first sub-vector of a target user based on item vectors of a plurality of items and second evaluation vectors of the target user on the plurality of items, wherein the first sub-vector is used for indicating the evaluation condition of the target user on the plurality of items;
obtaining a second sub-vector of the target user based on item vectors of a plurality of items and third evaluation vectors of a plurality of neighbor users of the target user on the plurality of items, wherein the second sub-vector is used for indicating evaluation conditions of the plurality of neighbor users on the plurality of items;
and fusing the first sub-vector and the second sub-vector to obtain a second vector of the target user.
3. The method of claim 2, wherein obtaining the first sub-vector of the target user based on the item vector of the plurality of items, the second rating vector of the plurality of items by the target user, comprises:
Fusing the project vector of each project and the second evaluation vector of each project by the target user to obtain a second interaction vector of each project by the target user;
Inputting the user vector of the target user and the second interaction vector into a second neural network for fusion to obtain a second attention weight of each item, wherein the second attention weight is used for indicating the contribution of the target user to each item;
And based on the second attention weights of the plurality of items, carrying out weighted fusion on a plurality of second interaction vectors of the plurality of items by the target user to obtain a first sub-vector of the target user.
4. The method of claim 3, wherein obtaining the second sub-vector of the target user based on the item vector of the plurality of items, a third evaluation vector of the plurality of items by a plurality of neighbor users of the target user, comprises:
fusing the project vector of each project and the third evaluation vector of each project by each neighbor user to obtain a third interaction vector of each neighbor user on each project;
Based on the second attention weights of the plurality of items, weighting and fusing a plurality of third interaction vectors of the plurality of items by each neighbor user to obtain a third sub-vector of each neighbor user, wherein the third sub-vector is used for indicating the evaluation condition of each neighbor user on the plurality of items;
Fusing the user vector of the target user and the third interaction vector to obtain a third attention weight of each neighbor user, wherein the third attention weight is used for indicating the importance of each neighbor user to the target user;
And carrying out weighted fusion on a plurality of third sub-vectors of the plurality of neighbor users based on the third attention weights of the plurality of neighbor users to obtain a second sub-vector of the target user.
5. A recommendation model training device, comprising:
The acquisition module is used for fusing the user vector of each user and the first evaluation vector of each user on the target item to acquire the first interaction vector of each user on the target item;
The obtaining module is used for inputting the item vector of the target item and the first interaction vector into a first neural network to be fused, so as to obtain a first attention weight of each user, wherein the first attention weight is used for indicating the contribution of each user to the target item;
the obtaining module is further configured to weight and fuse a plurality of first interaction vectors of the target item by a plurality of users based on first attention weights of the plurality of users, to obtain a first vector of the target item, where the plurality of users are users who have evaluated the target item, and the first vector is used to indicate characteristics of the target item;
The obtaining module is further configured to obtain a second vector of a target user based on an item vector of a plurality of items, a second evaluation vector of the target user on the plurality of items, and a third evaluation vector of a plurality of neighbor users on the plurality of items, where the plurality of neighbor users are friends having an interaction relationship with the target user, the plurality of items are items evaluated by the target user, and the second vector is used to indicate characteristics of the target user;
the obtaining module is further configured to fuse the first vector and the second vector to obtain a prediction rating of the target user on the target item;
The adjustment module is used for adjusting model parameters of the recommendation model based on the actual rating of the target user on the target item and the prediction rating to obtain a trained recommendation model;
And the generation module is used for generating a recommended item list of the target user based on the trained recommended model.
6. The apparatus of claim 5, wherein the obtaining module is specifically configured to:
obtaining a first sub-vector of a target user based on item vectors of a plurality of items and second evaluation vectors of the target user on the plurality of items, wherein the first sub-vector is used for indicating the evaluation condition of the target user on the plurality of items;
obtaining a second sub-vector of the target user based on item vectors of a plurality of items and third evaluation vectors of a plurality of neighbor users of the target user on the plurality of items, wherein the second sub-vector is used for indicating evaluation conditions of the plurality of neighbor users on the plurality of items;
and fusing the first sub-vector and the second sub-vector to obtain a second vector of the target user.
7. The apparatus of claim 6, wherein the obtaining module is specifically configured to:
Fusing the project vector of each project and the second evaluation vector of each project by the target user to obtain a second interaction vector of each project by the target user;
Inputting the user vector of the target user and the second interaction vector into a second neural network for fusion to obtain a second attention weight of each item, wherein the second attention weight is used for indicating the contribution of the target user to each item;
And based on the second attention weights of the plurality of items, carrying out weighted fusion on a plurality of second interaction vectors of the plurality of items by the target user to obtain a first sub-vector of the target user.
8. The apparatus of claim 7, wherein the obtaining module is specifically configured to:
fusing the project vector of each project and the third evaluation vector of each project by each neighbor user to obtain a third interaction vector of each neighbor user on each project;
Based on the second attention weights of the plurality of items, weighting and fusing a plurality of third interaction vectors of the plurality of items by each neighbor user to obtain a third sub-vector of each neighbor user, wherein the third sub-vector is used for indicating the evaluation condition of each neighbor user on the plurality of items;
Fusing the user vector of the target user and the third interaction vector to obtain a third attention weight of each neighbor user, wherein the third attention weight is used for indicating the importance of each neighbor user to the target user;
And carrying out weighted fusion on a plurality of third sub-vectors of the plurality of neighbor users based on the third attention weights of the plurality of neighbor users to obtain a second sub-vector of the target user.
9. An electronic device, comprising:
a memory for storing program instructions;
a processor for invoking program instructions stored in the memory and for performing the method of any of claims 1-4 in accordance with the obtained program instructions.
10. A computer readable storage medium, characterized in that the computer readable storage medium stores a computer program comprising program instructions which, when executed by a computer, cause the computer to perform the method of any of claims 1-4.
11. A computer program product, the computer program product comprising: computer program code which, when run on a computer, causes the computer to perform the method of any of the preceding claims 1-4.
CN202111437512.0A 2021-11-30 2021-11-30 Recommendation model training method, device, equipment and medium Active CN114048826B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202111437512.0A CN114048826B (en) 2021-11-30 2021-11-30 Recommendation model training method, device, equipment and medium

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202111437512.0A CN114048826B (en) 2021-11-30 2021-11-30 Recommendation model training method, device, equipment and medium

Publications (2)

Publication Number Publication Date
CN114048826A CN114048826A (en) 2022-02-15
CN114048826B true CN114048826B (en) 2024-04-30

Family

ID=80211726

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202111437512.0A Active CN114048826B (en) 2021-11-30 2021-11-30 Recommendation model training method, device, equipment and medium

Country Status (1)

Country Link
CN (1) CN114048826B (en)

Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN110059262A (en) * 2019-04-19 2019-07-26 武汉大学 A kind of construction method and device, item recommendation method of the project recommendation model based on hybrid neural networks
CN110162706A (en) * 2019-05-22 2019-08-23 南京邮电大学 A kind of personalized recommendation method and system based on interaction data cluster
CN110765353A (en) * 2019-10-16 2020-02-07 腾讯科技（深圳）有限公司 Processing method and device of project recommendation model, computer equipment and storage medium
CN111241423A (en) * 2020-01-17 2020-06-05 江西财经大学 Deep recommendation method and system integrating trust distrust relation and attention mechanism
CN113312644A (en) * 2021-06-15 2021-08-27 杭州金智塔科技有限公司 Cross-domain recommendation model training method and system based on privacy protection

Patent Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN110059262A (en) * 2019-04-19 2019-07-26 武汉大学 A kind of construction method and device, item recommendation method of the project recommendation model based on hybrid neural networks
CN110162706A (en) * 2019-05-22 2019-08-23 南京邮电大学 A kind of personalized recommendation method and system based on interaction data cluster
CN110765353A (en) * 2019-10-16 2020-02-07 腾讯科技（深圳）有限公司 Processing method and device of project recommendation model, computer equipment and storage medium
CN111241423A (en) * 2020-01-17 2020-06-05 江西财经大学 Deep recommendation method and system integrating trust distrust relation and attention mechanism
CN113312644A (en) * 2021-06-15 2021-08-27 杭州金智塔科技有限公司 Cross-domain recommendation model training method and system based on privacy protection

Also Published As

Publication number Publication date
CN114048826A (en) 2022-02-15

Similar Documents

Publication Publication Date Title
US10762422B2 (en) Wide and deep machine learning models
US11550871B1 (en) Processing structured documents using convolutional neural networks
CN109766557B (en) Emotion analysis method and device, storage medium and terminal equipment
Salant et al. Contextualized word representations for reading comprehension
EP3574453A1 (en) Optimizing neural network architectures
CN111506820B (en) Recommendation model, recommendation method, recommendation device, recommendation equipment and recommendation storage medium
CN110598118A (en) Resource object recommendation method and device and computer readable medium
CN112925892B (en) Dialogue recommendation method and device, electronic equipment and storage medium
Mohanty et al. Predicting fluctuations in cryptocurrencies' price using users' comments and real-time prices
Li et al. An efficient co-attention neural network for social recommendation
Eide et al. Deep neural network marketplace recommenders in online experiments
Renström et al. Fraud Detection on Unlabeled Data with Unsupervised Machine Learning
CN114048826B (en) Recommendation model training method, device, equipment and medium
Bach et al. Cross-domain intention detection in discussion forums
CN115238188A (en) Object recommendation method and system and object recommendation model system
CN115049458A (en) Commodity pushing method and device based on user crowd modeling, medium and equipment
Ranjbar et al. Explaining recommendation system using counterfactual textual explanations
Li et al. Adaptive gating in mixture-of-experts based language models
EP4081949A1 (en) Machine learning algorithm search
CN112559905A (en) Conversation recommendation method based on dual-mode attention mechanism and social similarity
de Oliveira Monteiro et al. Market prediction in criptocurrency: A systematic literature mapping
CN110795232A (en) Data processing method, data processing device, computer readable storage medium and computer equipment
Ding et al. Deep Feature Fusion over Multi-Field Categorical Data for Rating Prediction
Liu et al. Mixing-RNN: a recommendation algorithm based on recurrent neural network
He et al. Capturing Temporal Dynamics of Implicit Feedbacks for Collaborative Filtering by Using Deep Recurrent Neural Networks

Legal Events

Date Code Title Description
PB01 Publication
PB01 Publication
SE01 Entry into force of request for substantive examination
SE01 Entry into force of request for substantive examination
GR01 Patent grant