CN113010796A - Method for item recommendation - Google Patents

Method for item recommendation Download PDF

Info

Publication number
CN113010796A
CN113010796A CN202110404565.6A CN202110404565A CN113010796A CN 113010796 A CN113010796 A CN 113010796A CN 202110404565 A CN202110404565 A CN 202110404565A CN 113010796 A CN113010796 A CN 113010796A
Authority
CN
China
Prior art keywords
user
recommended
item
social network
attention
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.)
Granted
Application number
CN202110404565.6A
Other languages
Chinese (zh)
Other versions
CN113010796B (en
Inventor
周庆
欧娇娇
葛亮
黄智勇
仲元红
钟代笛
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.)
Chongqing University
Original Assignee
Chongqing University
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 Chongqing University filed Critical Chongqing University
Priority to CN202110404565.6A priority Critical patent/CN113010796B/en
Publication of CN113010796A publication Critical patent/CN113010796A/en
Application granted granted Critical
Publication of CN113010796B publication Critical patent/CN113010796B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Images

Classifications

    • 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
    • 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
    • G06F16/9536—Search customisation based on social or collaborative filtering
    • 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
    • G06N3/045—Combinations of networks
    • 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
    • G06N3/048—Activation functions
    • 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

Landscapes

  • Engineering & Computer Science (AREA)
  • Theoretical Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • Databases & Information Systems (AREA)
  • Data Mining & Analysis (AREA)
  • General Physics & Mathematics (AREA)
  • General Engineering & Computer Science (AREA)
  • General Health & Medical Sciences (AREA)
  • Computational Linguistics (AREA)
  • Evolutionary Computation (AREA)
  • Biophysics (AREA)
  • Molecular Biology (AREA)
  • Computing Systems (AREA)
  • Biomedical Technology (AREA)
  • Artificial Intelligence (AREA)
  • Mathematical Physics (AREA)
  • Software Systems (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Health & Medical Sciences (AREA)
  • Information Retrieval, Db Structures And Fs Structures Therefor (AREA)

Abstract

The application relates to the technical field of information recommendation, and discloses a method for item recommendation, which comprises the following steps: the method comprises the steps of obtaining preference scores of users to be recommended on items through user information and item information, obtaining social network attention aggregation data according to the user information, obtaining matching degrees of the users to be recommended on the items according to the preference scores and the social network attention aggregation data, and recommending the items corresponding to the matching degrees meeting preset conditions to the users to be recommended. In the item recommendation process, the fact that the user to be recommended is in a plurality of social networks is considered, the attention mechanism is utilized to obtain the social network attention aggregation data of the user to be recommended, the influence of other users on the user to be recommended is added into the calculation of the matching degree of the user to be recommended on the item, the obtained item to be recommended is made to be more suitable for the requirements of the user in real life, and the experience of the user in obtaining the recommended item is improved.

Description

Method for item recommendation
Technical Field
The present application relates to the field of information recommendation technology, for example, to a method for item recommendation.
Background
At present, when items need to be recommended for users, user information such as age, occupation, gender and the like is generally collected and then input into a trained item recommendation model, because users in real life do not exist independently, the users are in different social relations, the social relations are a large factor influencing recommendation performance, and the current item recommendation method does not consider the social relations, so that certain deviation exists between items recommended for the users in the prior art and requirements of the users in real life.
Disclosure of Invention
The following presents a simplified summary in order to provide a basic understanding of some aspects of the disclosed embodiments. This summary is not an extensive overview nor is intended to identify key/critical elements or to delineate the scope of such embodiments but rather as a prelude to the more detailed description that is presented later.
The embodiment of the disclosure provides a method for recommending projects, so as to better meet the requirements of users in real life.
In some embodiments, the method comprises:
acquiring user information of a user to be recommended, and acquiring a plurality of items and item information corresponding to each item; the user to be recommended is in a plurality of social networks, and users except the user to be recommended in each social network are influence users of the user to be recommended;
acquiring a user bias score corresponding to the user information, and acquiring a project bias score corresponding to the project information;
acquiring preference scores of the user to be recommended to various items according to the user bias scores and the item bias scores;
acquiring user attention values of users influencing the user to be recommended;
aggregating the attention values of the users corresponding to the social networks respectively to obtain user attention aggregation data;
acquiring the social network attention value of the user to be recommended in each social network according to the user attention aggregation data corresponding to each social network;
aggregating the attention values of the social networks to obtain aggregated data of the attention of the social networks;
acquiring the matching degree of the user to be recommended to each item according to the preference score and the social network attention aggregation data;
and recommending the item according to the matching degree.
The project recommendation method provided by the embodiment of the disclosure can achieve the following technical effects: the method comprises the steps of obtaining preference scores of users to be recommended on items through user information and item information, obtaining user attention values of all users influencing the users to be recommended according to the user information, aggregating the user attention values to obtain user attention aggregation data, obtaining social network attention values according to the user attention aggregation data, aggregating the social network attention values to obtain social network attention aggregation data, obtaining matching degrees of the users to be recommended on all items according to the preference scores and the social network attention aggregation data, and recommending items corresponding to the matching degrees meeting preset conditions to the users to be recommended. In the item recommendation process, the fact that the user to be recommended is in a plurality of social networks is considered, the attention mechanism is utilized to obtain the social network attention aggregation data of the user to be recommended, the influence of other users on the user to be recommended is added into the calculation of the matching degree of the user to be recommended on the item, the obtained item to be recommended is made to be more suitable for the requirements of the user in real life, and the experience of the user in obtaining the recommended item is improved.
The foregoing general description and the following description are exemplary and explanatory only and are not restrictive of the application.
Drawings
One or more embodiments are illustrated by way of example in the accompanying drawings, which correspond to the accompanying drawings and not in limitation thereof, in which elements having the same reference numeral designations are shown as like elements and not in limitation thereof, and wherein:
FIG. 1 is a schematic diagram of a method for item recommendation provided by embodiments of the present disclosure;
fig. 2 is a schematic diagram of a method for obtaining a user bias score according to an embodiment of the present disclosure.
Detailed Description
So that the manner in which the features and elements of the disclosed embodiments can be understood in detail, a more particular description of the disclosed embodiments, briefly summarized above, may be had by reference to the embodiments, some of which are illustrated in the appended drawings. In the following description of the technology, for purposes of explanation, numerous details are set forth in order to provide a thorough understanding of the disclosed embodiments. However, one or more embodiments may be practiced without these details. In other instances, well-known structures and devices may be shown in simplified form in order to simplify the drawing.
The terms "first," "second," and the like in the description and in the claims, and the above-described drawings of embodiments of the present disclosure, 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 present disclosure described herein may be made. Furthermore, the terms "comprising" and "having," as well as any variations thereof, are intended to cover non-exclusive inclusions.
The term "plurality" means two or more unless otherwise specified.
In the embodiment of the present disclosure, the character "/" indicates that the preceding and following objects are in an or relationship. For example, A/B represents: a or B.
The term "and/or" is an associative relationship that describes objects, meaning that three relationships may exist. For example, a and/or B, represents: a or B, or A and B.
Referring to fig. 1, an embodiment of the present disclosure provides a method for item recommendation, including:
step S101, obtaining user information of a user to be recommended, and obtaining a plurality of items and item information corresponding to each item; the user to be recommended is in a plurality of social networks, and users except the user to be recommended in each social network are influence users of the user to be recommended;
step S102, acquiring a user bias score corresponding to user information, and acquiring a project bias score corresponding to project information;
step S103, obtaining preference scores of the user to be recommended for each item according to the user bias scores and the item bias scores;
step S104, acquiring user attention values of all influencing users to treat the recommended users;
step S105, aggregating the attention values of the users corresponding to the social networks respectively to obtain user attention aggregation data;
step S106, obtaining the social network attention value of the user to be recommended in each social network according to the user attention aggregation data corresponding to each social network;
step S107, aggregating the attention values of the social networks to obtain aggregated data of the attention of the social networks;
step S108, obtaining the matching degree of the user to be recommended to each item according to the preference score and the social network attention aggregation data;
and step S109, recommending items according to the matching degree.
By adopting the method for recommending the project, in the process of recommending the project, the condition that the user to be recommended is in a plurality of social networks is considered, the social network attention aggregation data of the user to be recommended is obtained by utilizing an attention mechanism, and the influence of other users on the user to be recommended is added into the calculation of the matching degree of the user to be recommended on the project, so that the obtained project to be recommended is more suitable for the requirements of users in real life, and the experience of the user in obtaining the recommended project is improved.
In some embodiments, the user information includes numeric user information and categorical user information. Optionally, the numeric user information comprises: the age, height, etc. of the user; the category-type user information includes: gender, native place, occupation, etc. of the user.
In some embodiments, the item information includes numerical item information and categorical item information. Optionally, the category item information includes: type, attribute, etc. of the item; the numerical item information includes: the value class attribute of the item. In some embodiments, the item is a movie, and the corresponding category type item information includes a type of the movie, a director of the movie, and the like; the corresponding numerical item information includes a rating of the corresponding movie crawled from the movie rating website. The types of movies include action type, comedy type, horror type, and the like.
Optionally, the obtaining of the user bias score corresponding to the user information includes: extracting the characteristics of the user information to obtain numerical user characteristics and category user characteristics; acquiring a user sparse feature vector corresponding to the category-type user feature; converting the user sparse feature vector into user dense embedded features; splicing the dense user embedding features and the numerical user features to obtain user splicing features; and inputting the user splicing characteristics into a preset feedforward neural network model to obtain user bias scores corresponding to the user information. Therefore, the user bias scores are obtained by respectively obtaining the numerical user characteristics and the classification user characteristics, so that the user bias scores corresponding to the obtained user information are more fit with the real situation of the user.
Optionally, obtaining a user sparse feature vector corresponding to the category-type user feature includes: and generating a user sparse feature vector by using the category-type user features through one-hot or multi-hot.
Optionally, converting the user sparse feature vector into a user dense embedded feature, including: and converting the user sparse feature vector into the user dense embedded feature through the embedding layer. Optionally, the embedded layer is a fully connected network.
Optionally, the splicing the dense user embedding features and the numerical user features to obtain user splicing features includes: and splicing the dense user embedding features and the numerical user features through the connecting layer to obtain user splicing features.
Optionally, the feedforward neural network model is a three-layer perceptron, the three-layer perceptron comprising an input layer, a hidden layer and an output layer. The input layer does not perform any calculation, only transmits the user splicing characteristics to the hidden layer, the hidden layer performs calculation on the user splicing characteristics, transmits the calculation result to the output layer, and the output layer performs calculation and outputs the user bias score. The different layers of the three-layer perceptron are fully connected. In some embodiments, a feed-forward neural network model schematic as shown in FIG. 2. Stitching users together with feature x1User splice feature x2User splice feature x3And respectively entering the hidden layer through the input layer, and outputting a user bias score f (x) from the output layer. Optionally, by calculating f (x) ═ G (b)(2)+W(2)(s(b(1)+W(1)x)) to obtain a user bias score, wherein f (x) is the user bias score, W(1)Connection weight for hidden layer, b(1)To hide the bias of the layers, W(2)As connection weights of the output layers, b(2)For the bias of the output layer, the functions G, s are sigmoid activation functions, and x is a user splicing characteristic.
Optionally, the obtaining of the item bias score corresponding to the item information includes: carrying out feature extraction on the project information to obtain numerical project features and category project features; acquiring a project sparse feature vector corresponding to the category-type project feature; converting the project sparse feature vector into a project dense embedded feature through an embedded layer; splicing the dense embedding characteristics of the project and the numerical project characteristics through the connecting layer to obtain project splicing characteristics; and inputting the project splicing characteristics into a preset feedforward neural network model to obtain project bias scores corresponding to the project information. In this way, the item bias scores are obtained by respectively obtaining the numerical item features and the category type item features, so that the item bias scores corresponding to the obtained item information are more fit with the real situation of the items.
Optionally, obtaining a project sparse feature vector corresponding to the category-type project feature includes: and generating a project sparse feature vector by the one-hot or multi-hot of the category project features.
Optionally, converting the item sparse feature vector into item dense embedded features, including: and converting the project sparse feature vector into a project dense embedded feature through the embedding layer. Optionally, the embedded layer is a fully connected network.
Optionally, the item dense embedding feature and the numerical item feature are spliced to obtain an item splicing feature, including: and splicing the dense item embedding characteristics and the numerical item characteristics through the connecting layer to obtain item splicing characteristics.
Optionally, by calculating f (x') ═ G (b)(2)+W(2)(s(b(1)+W(1)x ')) to obtain a project bias score, where f (x ') is the project bias score and x ' is the project splice characteristic.
Optionally, obtaining a bias score of each item by the user to be recommended according to the user bias score and the item bias score, including:
by calculating bai＝μ+ba+biObtaining the preference score of the ith item of the ith user to be recommended;
wherein, baiScoring the preference of the ith item for the a-th user to be recommended, baA user bias score of the a-th user to be recommended, biThe item bias score of the ith item is obtained, and mu is a preset item score corresponding to the item; a. i is a positive integer.
Optionally, the obtaining of the preset item score corresponding to the item includes: and obtaining the scores of all the users on the projects in each social network, and determining the average value of the scores of all the users on the projects as the preset project score corresponding to the projects.
Optionally, the obtaining of the preset item score corresponding to the item includes: and obtaining the scores of all the users on all the projects, and determining the average value of the scores of all the users on all the projects as the preset project score corresponding to the projects.
The method comprises the steps of obtaining user deviation scores of each item according to user bias scores, item bias scores and preset item scores corresponding to the items, obtaining matching degrees of the user to be recommended to each item according to the preference scores, and recommending the items according to the matching degrees.
Optionally, obtaining user attention values of users affecting the user to treat the recommended user includes:
by calculation of
Figure BDA0003021749990000061
Obtaining the attention value of the jth influence user to the ith user to be recommended;
wherein the content of the first and second substances,
Figure BDA0003021749990000062
for the jth influence user's attention value to the ith to-be-recommended user in the tth social network, W' is a preset first weight matrix,
Figure BDA0003021749990000063
is a preset first weight vector, and is,
Figure BDA0003021749990000064
for transposing the first weight vector, LeakyReLU is a predetermined activation function,
Figure BDA0003021749990000071
a corresponding hidden vector for the a-th user to be recommended,
Figure BDA0003021749990000072
for the hidden vector corresponding to the jth influencing user,
Figure BDA0003021749990000073
for the hidden vector corresponding to the kth influencing user,
Figure BDA0003021749990000074
for the number of influencing users of the a-th to-be-recommended user in the t-th social network,
Figure BDA0003021749990000075
optionally, the preset first weight matrix W' is a trainable matrix, andlinear transformation of the vector is performed. Alternatively,
Figure BDA0003021749990000076
n is the dimension of the hidden vector corresponding to the user,
Figure BDA0003021749990000077
are real numbers.
Optionally, a preset first weight vector
Figure BDA0003021749990000078
And calculating the attention value corresponding to the t-th social network by using the trainable vector. Alternatively,
Figure BDA0003021749990000079
optionally, a unique hot code of the user information of the user to be recommended is obtained, the unique hot code corresponding to the user to be recommended is input into the embedding layer, the embedding layer executes an index operation, and a hidden vector corresponding to the user to be recommended is obtained from a preset free embedding matrix H of the user to be recommended. Alternatively,
Figure BDA00030217499900000710
wherein S is the number of all users in the social network.
Optionally, a one-hot code of user information affecting the user is obtained, the one-hot code corresponding to the user is input into the embedding layer, the embedding layer executes an index operation, and a hidden vector corresponding to the user is obtained from a preset free embedding matrix H' of the user affecting the user. Alternatively,
Figure BDA00030217499900000711
optionally, aggregating the user attention values corresponding to the social networks respectively to obtain user attention aggregated data, including:
by calculation of
Figure BDA00030217499900000712
Obtaining the tThe method comprises the steps that data of user attention aggregation of users on the a-th user to be recommended are influenced in each social network;
wherein the content of the first and second substances,
Figure BDA00030217499900000713
and aggregating data of the user attention of each influencing user to the a-th user to be recommended in the t-th social network, wherein sigma is a preset activation function.
Optionally, obtaining the social network attention value of the user to be recommended in each social network according to the user attention aggregation data corresponding to each social network includes:
by calculation of
Figure BDA0003021749990000081
Acquiring a social network attention value of a social network t to an a-th user to be recommended;
wherein eta isatThe social network attention value of the ith social network to the a-th user to be recommended is set as W, the value is a preset second weight matrix,
Figure BDA0003021749990000082
is a preset second weight vector,
Figure BDA0003021749990000083
in order to transpose the second weight vector,
Figure BDA0003021749990000084
and aggregating data of the influence users on the attention of the users to be recommended to the a-th user in the P-th social network, wherein T is the number of the social networks, and P is less than or equal to T.
Optionally, the preset second weight matrix W is a trainable matrix, for performing linear transformation of the vector,
Figure BDA0003021749990000085
optionally, a preset second weight vector
Figure BDA0003021749990000086
Is a trainable vector and is used for calculating attention values corresponding to the T social networks,
Figure BDA0003021749990000087
optionally, aggregating the social network attention values, and obtaining social network attention aggregation data includes:
by calculation of
Figure BDA0003021749990000088
Acquiring social network attention aggregation data of the a-th user to be recommended;
wherein u isaAnd T is less than or equal to T for the social network attention aggregation data of the a-th user to be recommended.
Optionally, obtaining the matching degree of the user to be recommended to each item according to the prediction score and the social network attention aggregation data includes:
by calculation of
Figure BDA0003021749990000089
Obtaining the matching degree of the items;
wherein the content of the first and second substances,
Figure BDA00030217499900000810
the matching degree of the ith item for the a-th user to be recommended, baiScoring the preference of the ith item for the a-th user to be recommended, qiHidden vector for i-th item, qi TIs pair qiAnd (5) performing transposition.
Optionally, a one-hot code of the item information corresponding to the item is obtained, the one-hot code corresponding to the item is input into the embedding layer, the embedding layer performs an index operation, and a hidden vector corresponding to the item is obtained from a preset item free embedding matrix Q. Alternatively,
Figure BDA0003021749990000091
wherein J is the number of items, and O is the dimension of the hidden vector corresponding to the item.
Optionally, the recommending items according to the matching degree includes: and recommending the items corresponding to the matching degrees meeting the preset conditions to the user to be recommended.
Optionally, the matching degree satisfying the preset condition includes: a matching degree greater than or equal to a set threshold.
The method comprises the steps of obtaining preference scores of users to be recommended on items through user information and item information, obtaining user attention values of all users influencing the users to be recommended according to the user information, aggregating the user attention values to obtain user attention aggregation data, obtaining social network attention values according to the user attention aggregation data, aggregating the social network attention values to obtain social network attention aggregation data, obtaining matching degrees of the users to be recommended on all items according to the preference scores and the social network attention aggregation data, and recommending items corresponding to the matching degrees meeting preset conditions to the users to be recommended. In the item recommendation process, the fact that the user to be recommended is in a plurality of social networks is considered, the attention mechanism is utilized to obtain the social network attention aggregation data of the user to be recommended, the influence of other users on the user to be recommended is added into the calculation of the matching degree of the user to be recommended on the item, the obtained item to be recommended is made to be more suitable for the requirements of the user in real life, and the experience of the user in obtaining the recommended item is improved.
In some embodiments, the preferences of the user to be recommended are similar to or influenced by the influencing user of the social network in which they are located. Therefore, the multiple social relationship information of the user to be recommended is a factor influencing item recommendation, and the influence of other users on the user to be recommended is represented by aggregating all the attention values influencing the user to be recommended. Because the connection strength between the user to be recommended and each influence user in the social network is different, the influence of some influence users on the user to be recommended is stronger than the influence of other influence users on the user to be recommended; secondly, various social networks exist in daily life of the user to be recommended, such as a co-worker, a classmate, a net friend and the like, and the influence degrees of different social networks on the user to be recommended are different. Therefore, the influence strength of different influencing users on the user to be recommended is represented by the user attention aggregation data; for different social networks, the influence strength of the users to be recommended in different social networks is represented by the social network attention aggregation data, and finally the influence of the users to be recommended in a plurality of social networks is obtained. The matching degree of the user to be recommended to each item is obtained according to the social network attention aggregation data, and item recommendation is carried out according to the matching degree, so that the obtained item to be recommended is more suitable for the requirements of the user in real life, and the experience of the user in obtaining the recommended item is improved.
The above description and drawings sufficiently illustrate embodiments of the disclosure to enable those skilled in the art to practice them. Other embodiments may incorporate structural, logical, electrical, process, and other changes. The examples merely typify possible variations. Individual components and functions are optional unless explicitly required, and the sequence of operations may vary. Portions and features of some embodiments may be included in or substituted for those of others. Furthermore, the words used in the specification are words of description only and are not intended to limit the claims. As used in the description of the embodiments and the claims, the singular forms "a", "an" and "the" are intended to include the plural forms as well, unless the context clearly indicates otherwise. Similarly, the term "and/or" as used in this application is meant to encompass any and all possible combinations of one or more of the associated listed. Furthermore, the terms "comprises" and/or "comprising," when used in this application, specify the presence of stated features, integers, steps, operations, elements, and/or components, but do not preclude the presence or addition of one or more other features, integers, steps, operations, elements, components, and/or groups thereof. Without further limitation, an element defined by the phrase "comprising an …" does not exclude the presence of other like elements in a process, method or apparatus that comprises the element. In this document, each embodiment may be described with emphasis on differences from other embodiments, and the same and similar parts between the respective embodiments may be referred to each other. For methods, products, etc. of the embodiment disclosures, reference may be made to the description of the method section for relevance if it corresponds to the method section of the embodiment disclosure.
Those of skill in the art would appreciate that the various illustrative elements and algorithm steps described in connection with the embodiments disclosed herein may be implemented as electronic hardware or combinations of computer software and electronic hardware. Whether such functionality is implemented as hardware or software may depend upon the particular application and design constraints imposed on the solution. Skilled artisans may implement the described functionality in varying ways for each particular application, but such implementation decisions should not be interpreted as causing a departure from the scope of the disclosed embodiments. It can be clearly understood by the skilled person that, for convenience and brevity of description, the specific working processes of the system, the apparatus and the unit described above may refer to the corresponding processes in the foregoing method embodiments, and are not described herein again.
In the embodiments disclosed herein, the disclosed methods, products (including but not limited to devices, apparatuses, etc.) may be implemented in other ways. For example, the above-described apparatus embodiments are merely illustrative, and for example, the division of the units may be merely a logical division, and in actual implementation, there may be another division, for example, multiple 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 place, or may be distributed on a plurality of network units. Some or all of the units can be selected according to actual needs to implement the present embodiment. In addition, functional units in the embodiments of the present disclosure 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 flowchart and block diagrams in the figures illustrate the architecture, functionality, and operation of possible implementations of systems, methods and computer program products according to embodiments of the present disclosure. In this regard, each block in the flowchart or block diagrams may represent a module, segment, or portion of code, which comprises one or more executable instructions for implementing the specified logical function(s). In some alternative implementations, the functions noted in the block may occur out of the order noted in the figures. For example, two blocks shown in succession may, in fact, be executed substantially concurrently, or the blocks may sometimes be executed in the reverse order, depending upon the functionality involved. In the description corresponding to the flowcharts and block diagrams in the figures, operations or steps corresponding to different blocks may also occur in different orders than disclosed in the description, and sometimes there is no specific order between the different operations or steps. For example, two sequential operations or steps may in fact be executed substantially concurrently, or they may sometimes be executed in the reverse order, depending upon the functionality involved. Each block of the block diagrams and/or flowchart illustrations, and combinations of blocks in the block diagrams and/or flowchart illustrations, can be implemented by special purpose hardware-based systems that perform the specified functions or acts, or combinations of special purpose hardware and computer instructions.

Claims (10)

1. A method for item recommendation, comprising:
acquiring user information of a user to be recommended, and acquiring a plurality of items and item information corresponding to each item; the user to be recommended is in a plurality of social networks, and users except the user to be recommended in each social network are influence users of the user to be recommended;
acquiring a user bias score corresponding to the user information, and acquiring a project bias score corresponding to the project information;
acquiring preference scores of the user to be recommended to each item according to the user bias scores and the item bias scores;
acquiring user attention values of all influencing users to the user to be recommended;
aggregating the attention values of the users corresponding to the social networks respectively to obtain user attention aggregation data;
acquiring the social network attention value of each social network to the user to be recommended according to the user attention aggregation data corresponding to each social network;
aggregating the attention values of the social networks to obtain aggregated data of the attention of the social networks;
acquiring the matching degree of the user to be recommended to each item according to the preference scores and the social network attention aggregation data;
and recommending the item according to the matching degree.
2. The method of claim 1, wherein obtaining the user bias score corresponding to the user information comprises:
extracting the characteristics of the user information to obtain numerical user characteristics and category user characteristics;
acquiring a user sparse feature vector corresponding to the category type user feature;
converting the user sparse feature vector into user dense embedded features;
splicing the user dense embedded features and the numerical user features to obtain user splicing features;
and inputting the user splicing characteristics into a preset feedforward neural network model to obtain a user bias score corresponding to the user information.
3. The method of claim 1, wherein obtaining a project bias score corresponding to the project information comprises:
extracting the characteristics of the project information to obtain numerical project characteristics and category project characteristics;
acquiring a project sparse feature vector corresponding to the category type project feature;
converting the project sparse feature vector into a project dense embedded feature through an embedded layer;
splicing the item dense embedding characteristics and the numerical item characteristics through a connecting layer to obtain item splicing characteristics;
and inputting the project splicing characteristics into a preset feedforward neural network model to obtain project bias scores corresponding to the project information.
4. The method of claim 1, wherein the obtaining the preference score of the user to be recommended for each item according to the user bias score and the item bias score comprises:
by calculating bai＝μ+ba+biObtaining the preference score of the ith item of the ith user to be recommended;
wherein, baiScoring the preference of the ith item for the a-th user to be recommended, baA user bias score of the a-th user to be recommended, biThe item bias score of the ith item is obtained, and mu is a preset item score corresponding to the item; a. i is a positive integer.
5. The method according to claim 1, wherein the obtaining user attention values of the respective influencing users to the user to be recommended comprises:
by calculation of
Figure FDA0003021749980000021
Obtaining the attention value of the jth influence user to the ith user to be recommended;
wherein the content of the first and second substances,
Figure FDA0003021749980000022
for the jth influence user's attention value to the a-th user to be recommended in the tth social network, W' is a preset first weightThe weight matrix is a matrix of the weight,
Figure FDA0003021749980000023
is a preset first weight vector, and is,
Figure FDA0003021749980000024
for transposing the first weight vector, LeakyReLU is a preset activation function,
Figure FDA0003021749980000025
a corresponding hidden vector for the a-th user to be recommended,
Figure FDA0003021749980000026
for the hidden vector corresponding to the jth influencing user,
Figure FDA0003021749980000027
for the hidden vector corresponding to the kth influencing user,
Figure FDA0003021749980000028
for the number of influencing users of the a-th to-be-recommended user in the t-th social network,
Figure FDA0003021749980000031
6. the method of claim 5, wherein the aggregating the user attention values corresponding to the social networks respectively to obtain user attention aggregated data comprises:
by calculation of
Figure FDA0003021749980000032
Acquiring user attention aggregation data of each influencing user on the a-th user to be recommended in the t-th social network;
wherein the content of the first and second substances,
Figure FDA0003021749980000033
and aggregating data of the user attention of each influencing user to the a-th user to be recommended in the t-th social network, wherein sigma is a preset activation function.
7. The method of claim 6, wherein the obtaining the social network attentiveness value of each social network to the user to be recommended according to the user attentiveness aggregation data corresponding to each social network comprises:
by calculation of
Figure FDA0003021749980000034
Acquiring a social network attention value of a social network t to an a-th user to be recommended;
wherein eta isatThe social network attention value of the ith social network to the a-th user to be recommended is set as W, the value is a preset second weight matrix,
Figure FDA0003021749980000035
is a preset second weight vector,
Figure FDA0003021749980000036
to transpose the second weight vector,
Figure FDA0003021749980000037
and aggregating data of the influence users on the attention of the users to be recommended to the a-th user in the P-th social network, wherein T is the number of the social networks, and P is less than or equal to T.
8. The method of claim 7, wherein aggregating the social network attention values to obtain social network attention aggregation data comprises:
by calculation of
Figure FDA0003021749980000038
Obtaining social network of a-th user to be recommendedCollateral attention aggregation data; wherein u isaAnd T is less than or equal to T for the social network attention aggregation data of the a-th user to be recommended.
9. The method of claim 8, wherein the obtaining the matching degree of the user to be recommended to each item according to the preference score and the social network attention aggregation data comprises:
by calculation of
Figure FDA0003021749980000041
Obtaining the matching degree of the items;
wherein the content of the first and second substances,
Figure FDA0003021749980000042
the matching degree of the ith item for the a-th user to be recommended, baiScoring the preference of the ith item for the a-th user to be recommended, qiHidden vector for i-th item, qi TTo said q isiAnd (5) performing transposition.
10. The method according to any one of claims 1 to 9, wherein the recommending items according to the matching degree comprises:
and recommending the items corresponding to the matching degrees meeting the preset conditions to the user to be recommended.
CN202110404565.6A 2021-04-15 2021-04-15 Method for item recommendation Active CN113010796B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202110404565.6A CN113010796B (en) 2021-04-15 2021-04-15 Method for item recommendation

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202110404565.6A CN113010796B (en) 2021-04-15 2021-04-15 Method for item recommendation

Publications (2)

Publication Number Publication Date
CN113010796A true CN113010796A (en) 2021-06-22
CN113010796B CN113010796B (en) 2023-04-07

Family

ID=76389390

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202110404565.6A Active CN113010796B (en) 2021-04-15 2021-04-15 Method for item recommendation

Country Status (1)

Country Link
CN (1) CN113010796B (en)

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN114021811A (en) * 2021-11-03 2022-02-08 重庆大学 Attention-based improved traffic prediction method and computer medium

Citations (10)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20090164400A1 (en) * 2007-12-20 2009-06-25 Yahoo! Inc. Social Behavior Analysis and Inferring Social Networks for a Recommendation System
CN106649657A (en) * 2016-12-13 2017-05-10 重庆邮电大学 Recommended system and method with facing social network for context awareness based on tensor decomposition
CN109410080A (en) * 2018-10-16 2019-03-01 合肥工业大学 A kind of social image recommended method based on level attention mechanism
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
CN111428147A (en) * 2020-03-25 2020-07-17 合肥工业大学 Social recommendation method of heterogeneous graph volume network combining social and interest information
CN111523051A (en) * 2020-04-24 2020-08-11 山东师范大学 Social interest recommendation method and system based on graph volume matrix decomposition
CN111681084A (en) * 2020-06-09 2020-09-18 南京邮电大学 E-commerce platform recommendation method based on social relationship influence factors
CN112084428A (en) * 2020-09-17 2020-12-15 辽宁工程技术大学 Collaborative filtering recommendation method based on coupling network embedding and knowledge graph
CN112256980A (en) * 2020-10-23 2021-01-22 辽宁工程技术大学 Dynamic graph attention network-based multi-relation collaborative filtering recommendation
CN112559905A (en) * 2020-12-24 2021-03-26 北京理工大学 Conversation recommendation method based on dual-mode attention mechanism and social similarity

Patent Citations (10)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20090164400A1 (en) * 2007-12-20 2009-06-25 Yahoo! Inc. Social Behavior Analysis and Inferring Social Networks for a Recommendation System
CN106649657A (en) * 2016-12-13 2017-05-10 重庆邮电大学 Recommended system and method with facing social network for context awareness based on tensor decomposition
CN109410080A (en) * 2018-10-16 2019-03-01 合肥工业大学 A kind of social image recommended method based on level attention mechanism
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
CN111428147A (en) * 2020-03-25 2020-07-17 合肥工业大学 Social recommendation method of heterogeneous graph volume network combining social and interest information
CN111523051A (en) * 2020-04-24 2020-08-11 山东师范大学 Social interest recommendation method and system based on graph volume matrix decomposition
CN111681084A (en) * 2020-06-09 2020-09-18 南京邮电大学 E-commerce platform recommendation method based on social relationship influence factors
CN112084428A (en) * 2020-09-17 2020-12-15 辽宁工程技术大学 Collaborative filtering recommendation method based on coupling network embedding and knowledge graph
CN112256980A (en) * 2020-10-23 2021-01-22 辽宁工程技术大学 Dynamic graph attention network-based multi-relation collaborative filtering recommendation
CN112559905A (en) * 2020-12-24 2021-03-26 北京理工大学 Conversation recommendation method based on dual-mode attention mechanism and social similarity

Non-Patent Citations (2)

* Cited by examiner, † Cited by third party
Title
MUNAN LI ET AL.: "An Efficient Adaptive Attention Neural Network for Social Recommendation", 《IEEE ACCESS》 *
潘可: "面向多源社交网络的用户兴趣爱好特征分析与推荐技术研究", 《中国优秀硕士学位论文全文数据库 信息科技辑》 *

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN114021811A (en) * 2021-11-03 2022-02-08 重庆大学 Attention-based improved traffic prediction method and computer medium

Also Published As

Publication number Publication date
CN113010796B (en) 2023-04-07

Similar Documents

Publication Publication Date Title
Pena et al. Bias in multimodal AI: Testbed for fair automatic recruitment
CN108537624B (en) Deep learning-based travel service recommendation method
CN106599022B (en) User portrait forming method based on user access data
Rutkowski et al. A content-based recommendation system using neuro-fuzzy approach
WO2014160282A1 (en) Classifying resources using a deep network
Ernawati et al. Implementation of the Naïve Bayes algorithm with feature selection using genetic algorithm for sentiment review analysis of fashion online companies
CN112138403B (en) Interactive behavior recognition method and device, storage medium and electronic equipment
CN112580352B (en) Keyword extraction method, device and equipment and computer storage medium
CN110427480B (en) Intelligent personalized text recommendation method and device and computer readable storage medium
CN112257841A (en) Data processing method, device and equipment in graph neural network and storage medium
CN114201516B (en) User portrait construction method, information recommendation method and related devices
CN111625715A (en) Information extraction method and device, electronic equipment and storage medium
CN114298783A (en) Commodity recommendation method and system based on matrix decomposition and fusion of user social information
CN112085171A (en) Recommendation method based on clustering multi-entity graph neural network
Tsinganos et al. Utilizing convolutional neural networks and word embeddings for early-stage recognition of persuasion in chat-based social engineering attacks
Asgarnezhad et al. An Effective Algorithm to Improve Recommender Systems using Evolutionary Computation Algorithms and Neural Network
Fang Making recommendations using transfer learning
Hoang et al. Academic event recommendation based on research similarity and exploring interaction between authors
KR20190130410A (en) Apparatus and method for recommending customized content for smart senior
CN117557331A (en) Product recommendation method and device, computer equipment and storage medium
CN113010796B (en) Method for item recommendation
CN110795640A (en) Adaptive group recommendation method for compensating group member difference
Zeng et al. RACMF: robust attention convolutional matrix factorization for rating prediction
Nguyen et al. Collaborative multi-key learning with an anonymization dataset for a recommender system
Bagkar et al. Sentiment analysis on Netflix

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
GR01 Patent grant