CN111967946A - Commodity recommendation method and system based on user-oriented multi-relationship information network - Google Patents

Commodity recommendation method and system based on user-oriented multi-relationship information network Download PDF

Info

Publication number
CN111967946A
CN111967946A CN202010919815.5A CN202010919815A CN111967946A CN 111967946 A CN111967946 A CN 111967946A CN 202010919815 A CN202010919815 A CN 202010919815A CN 111967946 A CN111967946 A CN 111967946A
Authority
CN
China
Prior art keywords
user
commodity
matrix
predicted
recommended
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
CN202010919815.5A
Other languages
Chinese (zh)
Other versions
CN111967946B (en
Inventor
杨博
李岸宸
于东然
张春旭
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.)
Jilin University
Original Assignee
Jilin 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 Jilin University filed Critical Jilin University
Priority to CN202010919815.5A priority Critical patent/CN111967946B/en
Publication of CN111967946A publication Critical patent/CN111967946A/en
Application granted granted Critical
Publication of CN111967946B publication Critical patent/CN111967946B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
    • G06Q30/00—Commerce
    • G06Q30/06—Buying, selling or leasing transactions
    • G06Q30/0601—Electronic shopping [e-shopping]
    • G06Q30/0631—Item recommendations
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
    • G06Q30/00—Commerce
    • G06Q30/02—Marketing; Price estimation or determination; Fundraising
    • G06Q30/0201—Market modelling; Market analysis; Collecting market data

Abstract

The invention relates to a commodity recommendation method and system based on a user-oriented multi-relationship information network. The method comprises the following steps: acquiring a recommended scene; acquiring a user to be predicted; according to the recommendation scene, acquiring the user characteristics of the user to be predicted by using a graph convolution method and an attention mechanism; acquiring commodity characteristics of a commodity to be recommended; the commodity features are a plurality of feature information of the commodities to be recommended; according to the user characteristics of the user to be predicted and the commodity characteristics of the commodity to be recommended, obtaining the score of the user to be predicted on the commodity to be recommended; judging whether the score of the user to be predicted on the commodity to be recommended is larger than a preset score threshold value or not; if so, recommending the commodity to be recommended to the user to be predicted; if not, the to-be-recommended commodity is not recommended to the to-be-predicted user. The invention can realize accurate recommendation of commodities.

Description

Commodity recommendation method and system based on user-oriented multi-relationship information network
Technical Field
The invention relates to the field of recommendation systems, in particular to a commodity recommendation method and system based on a user-oriented multi-relationship information network.
Background
The recommendation system predicts the interest preference of a user according to the historical consumption record of the user, and thereby recommends goods or services for the user. Most recommendation systems only consider users as isolated individuals when performing recommendation, but ignore the relationship between the users, resulting in low accuracy of commodity recommendation or service recommendation. In fact, there are many different types of relationships between users and users, which are considered to make the recommendation service more accurate.
Disclosure of Invention
The invention aims to provide a commodity recommendation method and system based on a user-oriented multi-relationship information network so as to realize accurate commodity recommendation.
In order to achieve the purpose, the invention provides the following scheme:
a commodity recommendation method based on a user-oriented multi-relationship information network comprises the following steps:
acquiring a recommended scene; the recommendation scene comprises a user set, a user characteristic information matrix, a user relation graph, a commodity set, a commodity characteristic information matrix and an interaction matrix of the user and the commodity; the user characteristic information matrix is a matrix formed by a plurality of pieces of characteristic information corresponding to each user; the commodity characteristic information matrix is a matrix formed by a plurality of characteristic information corresponding to each commodity; the user relationship graph is a multi-relationship information network among users;
acquiring a user to be predicted;
according to the recommendation scene, acquiring the user characteristics of the user to be predicted by using a graph convolution method and an attention mechanism; the user characteristics of the user to be predicted are fused with the characteristic information of the neighbor user of the user to be predicted, and the neighbor user of the user to be predicted is a subset of the user set;
acquiring commodity characteristics of a commodity to be recommended; the commodity features are a plurality of feature information of the commodities to be recommended;
according to the user characteristics of the user to be predicted and the commodity characteristics of the commodity to be recommended, obtaining the score of the user to be predicted on the commodity to be recommended;
judging whether the score of the user to be predicted on the commodity to be recommended is larger than a preset score threshold value or not;
when the score of the user to be predicted on the commodity to be recommended is larger than a preset score threshold value, recommending the commodity to be recommended to the user to be predicted;
and when the score of the user to be predicted on the commodity to be recommended is not larger than a preset score threshold value, the commodity to be recommended is not recommended to the user to be predicted.
Optionally, the obtaining, according to the recommendation scenario, the user feature of the user to be predicted by using a graph convolution method and an attention mechanism further includes
Acquiring a first conversion matrix;
reducing the dimension of the user characteristic information matrix by using a first conversion matrix to obtain a user characteristic information matrix after dimension reduction;
constructing a relation type matrix according to the relation type set; the ith behavior of the relationship type matrix is a one-hot vector corresponding to the ith relationship type; the value of the ith dimension in the one-hot vector corresponding to the ith relation type is 1, and the values of the other dimensions are 0;
acquiring a second conversion matrix;
reducing the dimension of the relationship type matrix by using the second conversion matrix to obtain a reduced-dimension relationship type matrix; the column number of the relation type matrix after dimension reduction is equal to the column number of the user characteristic information matrix after dimension reduction;
acquiring a third conversion matrix;
and reducing the dimension of the commodity characteristic information matrix by using the third conversion matrix to obtain a dimension-reduced commodity characteristic information matrix.
Optionally, the obtaining, according to the recommended scenario, the user characteristic of the user to be predicted by using a graph convolution method and an attention mechanism specifically includes:
for the ith iteration, the formula is used
Figure BDA0002666332790000031
Obtaining each userThe user characteristics of the current iteration, i.e. the ith iteration; wherein, F(i)(j) Is the user characteristic of the current iteration of the jth user, A is the activation function, WiIs the weight matrix of the current iteration, F(i-1)(j) The characteristics of the user in the previous iteration of the jth user, i.e. the i-1 th iteration, nei (j) is the neighbor user set of the jth user, att (j, a)j) Is the jth user and the neighbor user ajThe weight of attention in between the two,
Figure BDA0002666332790000032
is the jth user and the neighbor user ajThe type of relationship between the two or more,
Figure BDA0002666332790000033
is the jth user and the neighbor user ajFeature vector of the type of relationship between, feature vector
Figure BDA0002666332790000034
Obtaining, F, from the relationship type matrix E(i-1)(aj) For neighbor user ajUser characteristic information of the previous iteration;
judging whether the current iteration times reach the maximum iteration times or not;
if the current iteration times reach the maximum iteration times, determining the user characteristics of the user to be predicted of the current iteration as the final user characteristics of the user to be predicted;
if the current iteration times do not reach the maximum iteration times, updating the current iteration times, and returning to the utilization formula
Figure BDA0002666332790000035
And acquiring the user characteristics of the current iteration of each user, and entering the next iteration.
Optionally, the obtaining, according to the user characteristics of the user to be predicted and the commodity characteristics of the commodity to be recommended, the score of the user to be predicted on the commodity to be recommended specifically includes:
using formulas
Figure BDA0002666332790000036
Obtaining the score of the user to be predicted on the commodity to be recommended; wherein the content of the first and second substances,
Figure BDA0002666332790000037
scoring the to-be-recommended commodity for the to-be-predicted user; sigma is sigmoid function; w is apIs a transformation matrix; i is vector splicing operation; f(t)(u*) For the user u to be predicted*The user characteristics of (1); m (v)*) For a commodity v to be recommended*The characteristics of the article of commerce.
Optionally, the predefined score threshold is 0.5.
The invention also provides a commodity recommendation system based on the user-oriented multi-relationship information network, which comprises the following steps:
the recommendation scene obtaining module is used for obtaining recommendation scenes; the recommendation scene comprises a user set, a user characteristic information matrix, a user relation graph, a commodity set, a commodity characteristic information matrix and an interaction matrix of the user and the commodity; the user characteristic information matrix is a matrix formed by a plurality of pieces of characteristic information corresponding to each user; the commodity characteristic information matrix is a matrix formed by a plurality of characteristic information corresponding to each commodity; the user relationship graph is a multi-relationship information network among users;
the user to be predicted acquiring module is used for acquiring a user to be predicted;
the user feature extraction module is used for acquiring the user features of the user to be predicted according to the recommendation scene by using a graph convolution method and an attention mechanism; the user characteristics of the user to be predicted are fused with the characteristic information of the neighbor user of the user to be predicted, and the neighbor user of the user to be predicted is a subset of the user set;
the commodity feature acquisition module is used for acquiring the commodity features of the commodities to be recommended; the commodity features are a plurality of feature information of the commodities to be recommended;
the score acquisition module is used for acquiring the score of the to-be-predicted user on the to-be-recommended commodity according to the user characteristics of the to-be-predicted user and the commodity characteristics of the to-be-recommended commodity;
the judging module is used for judging whether the score of the user to be predicted on the commodity to be recommended is larger than a preset score threshold value or not;
the recommending module is used for recommending the to-be-recommended commodity to the to-be-predicted user when the score of the to-be-predicted user on the to-be-recommended commodity is larger than a preset score threshold value; and when the score of the user to be predicted on the commodity to be recommended is not larger than a preset score threshold value, the commodity to be recommended is not recommended to the user to be predicted.
Optionally, the method further includes:
the first conversion matrix obtaining module is used for obtaining a first conversion matrix before obtaining the user characteristics of the user to be predicted according to the recommendation scene by using a graph convolution method and an attention mechanism;
the first dimension reduction module is used for reducing the dimension of the user characteristic information matrix by using a first conversion matrix to obtain a user characteristic information matrix after dimension reduction;
the relationship type matrix construction module is used for constructing a relationship type matrix according to the relationship type set; the ith behavior of the relationship type matrix is a one-hot vector corresponding to the ith relationship type; the value of the ith dimension in the one-hot vector corresponding to the ith relation type is 1, and the values of the other dimensions are 0;
the second conversion matrix acquisition module is used for acquiring a second conversion matrix;
the second dimension reduction module is used for reducing the dimension of the relation type matrix by using the second conversion matrix to obtain a reduced-dimension relation type matrix; the column number of the relation type matrix after dimension reduction is equal to the column number of the user characteristic information matrix after dimension reduction;
a third transformation matrix obtaining module, configured to obtain a third transformation matrix;
and the third dimension reduction module is used for reducing the dimension of the commodity characteristic information matrix by using the third conversion matrix to obtain the commodity characteristic information matrix after dimension reduction.
Optionally, the user feature extraction module specifically includes:
a current iteration user feature calculation unit for utilizing the formula
Figure BDA0002666332790000051
Acquiring the user characteristics of the current iteration, namely the ith iteration, of each user; wherein, F(i)(j) The current iteration user characteristics of the jth user; a is an activation function; wiIs the weight matrix of the current iteration; f(i-1)(j) The characteristics of the user of the previous iteration, namely the i-1 iteration, of the jth user are obtained; nei (j) is a neighbor user set of the jth user; att (j, a)j) Is the jth user and the neighbor user ajAttention weight in between;
Figure BDA0002666332790000052
is the jth user and the neighbor user ajThe type of relationship between;
Figure BDA0002666332790000053
is the jth user and the neighbor user ajFeature vector of the type of relationship between, feature vector
Figure BDA0002666332790000054
Obtaining from the relation type matrix E; f(i-1)(aj) For neighbor user ajUser characteristic information of the previous iteration;
the judging unit is used for judging whether the current iteration times reach the maximum iteration times;
the user feature determination unit of the user to be predicted is used for determining the user feature of the user to be predicted of the current iteration as the final user feature of the user to be predicted when the current iteration number reaches the maximum iteration number;
and the returning unit is used for updating the current iteration times when the current iteration times do not reach the maximum iteration times, returning to the current iteration user characteristic calculating unit and entering the next iteration.
Optionally, the score obtaining module specifically includes:
a score calculating unit for using a formula
Figure BDA0002666332790000055
Obtaining the score of the user to be predicted on the commodity to be recommended; wherein the content of the first and second substances,
Figure BDA0002666332790000061
scoring the to-be-recommended commodity for the to-be-predicted user; sigma is sigmoid function; w is apIs a transformation matrix; i is vector splicing operation; f(t)(u*) For the user u to be predicted*The user characteristics of (1); m (v)*) For a commodity v to be recommended*The characteristics of the article of commerce.
Optionally, the predefined score threshold is 0.5.
According to the specific embodiment provided by the invention, the invention discloses the following technical effects:
the invention considers various relations among users, uses the technology of graph convolution, and more accurately extracts the user characteristics, thereby carrying out high-quality recommendation and improving the accuracy of a recommendation system.
Drawings
In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings needed to be used in the embodiments will be briefly described below, and it is obvious that the drawings in the following description are only some embodiments of the present invention, and it is obvious for those skilled in the art to obtain other drawings without inventive exercise.
FIG. 1 is a flow chart of a commodity recommendation method based on a user-oriented multi-relationship information network according to the present invention;
fig. 2 is a schematic structural diagram of a commodity recommendation system based on a user-oriented multi-relationship information network according to the present invention.
Detailed Description
The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the drawings in the embodiments of the present invention, and it is obvious that the described embodiments are only a part of the embodiments of the present invention, and not all of the embodiments. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present invention.
When recommending commodities, the relationship between users can be regarded as an abnormal picture because of the existence of various different types of relationships (friendship, relativity, classmate relationship, etc.) between users. The invention takes into account a plurality of relationships among users and uses the graph convolution technology to extract more accurate user characteristics, thereby recommending high-quality commodities.
In order to make the aforementioned objects, features and advantages of the present invention comprehensible, embodiments accompanied with figures are described in further detail below.
Fig. 1 is a schematic flow chart of a commodity recommendation method based on a user-oriented multi-relationship information network according to the present invention. As shown in fig. 1, the commodity recommendation method based on the user-oriented multi-relationship information network of the present invention includes the following steps:
step 100: and acquiring a recommended scene. The recommendation scene comprises a user set, a user characteristic information matrix, a user relation graph, a commodity set, a commodity characteristic information matrix and an interaction matrix of the user and the commodity; the user characteristic information matrix is a matrix formed by a plurality of pieces of characteristic information corresponding to each user; the commodity characteristic information matrix is a matrix formed by a plurality of characteristic information corresponding to each commodity. The recommended scenario may be denoted as Rec ═ { U, Fraw,V,Mraw,Y,G}。
Wherein U is a set of users, and U ═ U1,u2,…,uKAnd K users are contained, and each element in the set represents an individual user.
Figure BDA0002666332790000072
For the user profile information matrix, each row represents an initial profile of a user (e.g., the user's age, hobbies, gender, etc.), the profileThe feature information may be represented as a vector of D dimensions. For example, feature information of three aspects of a user is known: user ID, user gender, user age. These features are expressed as one-hot vectors and/or multi-hot vectors and then concatenated. Assume that there are only three users (K ═ 3), three age groups. Each feature information is expressed as follows:
Figure BDA0002666332790000071
then, for a user (ID 2, girl, 15 years old), the feature information vector of D dimension of the user is { [0,1,0] [0,1] [0,1,0] } { [0,1,0,0,1,0,1,0], and D { [ 8 ].
Thus, the characteristic information vectors corresponding to all the users are spliced to obtain the user characteristic information matrix FrawEach row of the matrix represents information of a user, the dimension of the matrix is K x D, and the ith row of the matrix represents a user uiCharacteristic information of (1).
V is a set of commodities, V ═ V1,v2,…,vSAnd the product contains S commodities. Each element in the set represents an individual item.
MrawIs a commodity characteristic information matrix, and the commodity characteristic information matrix,
Figure BDA0002666332790000081
each row represents an initial characteristic of an item (e.g., an item's ID, category, price, etc.), and the characteristic information may be represented as a vector in one D dimension. Assuming that there are only three items (S ═ 3), two item categories, and three price segments, each characteristic information can be expressed as:
Figure BDA0002666332790000082
then for a D-dimensional vector of a commodity (ID 2, food, 800-ary), where D8, the feature vector of the commodity is { [0,1,0] [1,0] [0,1,0] } [0,1,0,1,0 ].
Figure BDA0002666332790000083
Is the interaction matrix of the user and the commodity, if the user uaPurchased commodity viThe ith column of the a-th row of the interaction matrix Y is 1, otherwise, the ith column is 0.
For example, there are 2 users { u }a，ubAnd three commodities vi，vj，vkThere is an interaction uaPurchased viAnd the existence of an interaction ubPurchased vk. The interaction matrix Y is then:
[1,0,0]
[0,0,1]
G＝{U,Fraw,R,Erawand the user relationship graph is a multi-relationship information network among users. Since there are a plurality of relationships between users, a set including J relationships is denoted by R, and R ═ R1,r2,…rJWhere each element of R represents a type of relationship. Multiple relationships may be understood as multiple types of relationships between users. Taking 4 relations as an example (J ═ 4), 1. friend, 2. relative, 3. classmate, 4. teacher, then the set of relations R ═ friend, relativity, classmate, teacher, each relation is represented by one-hot:
friend: [1,0,0,0]
Relatives: [0,1,0,0]
The classmates: [0,0,1,0]
Teachers and students: [0,0,0,1]
The 4 relation vectors are longitudinally spliced to form a relation type matrix
Figure BDA0002666332790000091
Each row of the relationship type matrix represents one relationship in the set of relationships R. There are only 4 relationships in this example, and in reality there may be more relationships involved, so such an ErawThe dimension of the original relationship matrix is large, and the dimension of such a high-dimension matrix is reduced later for subsequent operations.
Step 200: and acquiring the user to be predicted.
After the user to be predicted is obtained, the commodity recommendation prediction is carried out by adopting the graph convolution neural network model. For parameters in the graph convolution neural network model, the invention performs training through back propagation, and defines cross entropy as a loss function. And inputting the user to be predicted, the commodity to be recommended and the recommended scene into the graph convolution neural network model, so that a commodity recommendation result can be output.
First, before prediction is performed, input data arrives at a data embedding layer, and a first conversion matrix W is usednAnd a third conversion matrix WmRespectively to user characteristic information matrix FrawAnd a characteristic information matrix M of the commodityrawTo reduce dimension, WnAnd WmDimension of D x DsWherein D issIs far less than D, thus achieving the effect of reducing dimension. The formula is as follows: f ═ Fraw·Wn；M＝Mraw·Wm。
And obtaining a user dense matrix dimension F and a commodity dense matrix M after dimensionality reduction. F has dimension K x DsDimension of M is S x Ds。
In addition, the relationship type matrix ErawIs J x J. If J is large, the length of one-hot vector is large, which is not beneficial to subsequent operation. For this purpose, a second conversion matrix W is usedrDimension reduction, W, of the relationship type matrixrDimension J x DsAfter dimensionality reduction, a relation type dense matrix E is obtained, and the dimensionality of the relation type dense matrix E is J x Ds. The formula is as follows: e ═ Eraw·Wr。
Wherein the first conversion matrix WnA second conversion matrix WrAnd a third transformation matrix WmAre trainable parameters in the graph convolution neural network model.
For each user node uiThere will be different types of relationships with each neighbor, using the set rel (i) ═ ri～a,ri～b… … } in which ri～aRepresenting user node uiWith neighbour users uaThe relationship between them.
Step 300: and acquiring the user characteristics of the user to be predicted by using a graph convolution method and an attention mechanism according to the recommended scene. The user characteristics of the user to be predicted are fused with the characteristic information of the neighbor user of the user to be predicted, and the neighbor user of the user to be predicted is a subset of the user set.
The operation is carried out in the convolutional layer, and because various relation types exist among users, the neighbor node information and the relation information with the neighbor node are considered simultaneously when the information of the user node is updated. In particular, for the jth user ujIts neighbors are represented by a set of users nei (j). In order to update the jth user ujThe feature information of (2) needs to consider the neighbor relation and the information of the neighbor users at the same time. For this purpose, an iterative process is performed using a non-linear transformation based on the attention mechanism. Specifically, update the jth user ujThe process of the characteristic information comprises the following steps:
Figure BDA0002666332790000101
wherein, F(0)F is a user characteristic information matrix after dimension reduction; f(1)(j) The user characteristics of the first iteration for the jth user; a is the activation function, where ReLU is selected as the activation function (ReLU (x) ═ max (0, x)); w1A weight matrix for a first iteration; f(0)(j) The user characteristic information of the jth user in the user characteristic information matrix F after dimension reduction is obtained; nei (j) is a neighbor user set of the jth user;
Figure BDA0002666332790000102
is the jth user and the neighbor user ajThe type of relationship between;
Figure BDA0002666332790000103
is the jth user and the neighbor user ajFeature vector of the type of relationship between, feature vector
Figure BDA0002666332790000104
Obtaining from the relation type matrix E; f(0)(aj) For the user a in the user characteristic information matrix F after dimension reductionjThe user characteristic information of (1); att (j, a)j) Is the jth user and the neighbor user ajThe attention weight in between, is a scalar.
att is a function of softmax,
Figure BDA0002666332790000105
exp(x)＝ex。
and updating the neighbor information of the jth user node once by using the formula. The network is updated simultaneously for all nodes, that is, after the first update, the user characteristic matrix of the network is updated from the initial F(0)Is changed into F(1). Updating the new feature matrix F obtained once(1)Only the characteristic information of the surrounding 1 st order neighbors is considered. In order to utilize multi-hop information, the characteristic information of 2, 3, … t order neighbors around the information can be fused by the same method, wherein t is a pre-artificially defined hyper-parameter. For each user node, updating the feature information of each node in an iterative manner, wherein for the ith iteration, the formula for updating the feature information of the node is as follows:
Figure BDA0002666332790000106
wherein, F(i)(j) The current iteration of the jth user is the user characteristic of the ith iteration; a is an activation function; wiIs the weight matrix of the current iteration; f(i-1)(j) The characteristics of the user of the previous iteration, namely the i-1 iteration, of the jth user are obtained; nei (j) is a neighbor user set of the jth user; att (j, a)j) Is the jth user and the neighbor user ajAttention weight in between;
Figure BDA0002666332790000111
is the jth user and the neighbor user ajThe type of relationship between;
Figure BDA0002666332790000112
is the jth user and the neighbor user ajFeature vector of the type of relationship between, feature vector
Figure BDA0002666332790000113
Obtaining from the relation type matrix E; f(i-1)(aj) For neighbor user ajUser characteristic information of the (i-1) th iteration.
And then judging whether the current iteration number reaches the maximum iteration number t. If the current iteration number reaches the maximum iteration number, the user characteristic F of the user j to be predicted of the current iteration is used(t)(j) And determining the user characteristics of the user to be predicted finally. And if the current iteration times do not reach the maximum iteration times, updating the current iteration times i to be i +1, updating the user characteristics of each user by using the formula, and entering the next iteration.
Step 400: and acquiring the commodity characteristics of the commodity to be recommended. The commodity features are a plurality of feature information of the commodities to be recommended, namely commodity features corresponding to the commodities to be recommended in the commodity feature information matrix.
Step 500: and obtaining the score of the to-be-recommended commodities of the to-be-predicted user according to the user characteristics of the to-be-predicted user and the commodity characteristics of the to-be-recommended commodities. In particular, using formulae
Figure BDA0002666332790000114
Obtaining the score of the user to be predicted on the commodity to be recommended; wherein the content of the first and second substances,
Figure BDA0002666332790000115
scoring the to-be-recommended commodity for the to-be-predicted user; σ is sigmoid function, and σ (x) is 1/(1+ e)-x)；wpIs a transformation matrix; i is vector splicing operation; f(t)(u*) For the user u to be predicted*The user characteristics of (1); m (v)*) For a commodity v to be recommended*The characteristics of the article of commerce.
Step 600: and judging whether the score of the to-be-recommended commodity of the to-be-predicted user is larger than a preset score threshold value. If yes, go to step 700; if not, step 800 is performed. The score threshold specified in the present invention may be set to 0.5.
Step 700: recommending the commodity to be recommended to the user to be predicted.
Step 800: and recommending the commodity to be recommended to the user to be predicted.
Fig. 2 is a schematic structural diagram of a commodity recommendation system based on a user-oriented multi-relationship information network according to the present invention. As shown in fig. 2, the commodity recommendation system based on the user-oriented multi-relationship information network of the present invention includes the following structures:
a recommended scene obtaining module 201, configured to obtain a recommended scene; the recommendation scene comprises a user set, a user characteristic information matrix, a user relation graph, a commodity set, a commodity characteristic information matrix and an interaction matrix of the user and the commodity; the user characteristic information matrix is a matrix formed by a plurality of pieces of characteristic information corresponding to each user; the commodity characteristic information matrix is a matrix formed by a plurality of characteristic information corresponding to each commodity; the user relationship graph is a multi-relationship information network among users.
A to-be-predicted user obtaining module 202, configured to obtain a to-be-predicted user.
The user feature extraction module 203 is configured to obtain the user features of the user to be predicted according to the recommended scene by using a graph convolution method and an attention mechanism; the user characteristics of the user to be predicted are fused with the characteristic information of the neighbor user of the user to be predicted, and the neighbor user of the user to be predicted is a subset of the user set.
The commodity feature acquisition module 204 is used for acquiring the commodity features of the commodities to be recommended; the commodity characteristics are a plurality of characteristic information of the commodity to be recommended.
And the score obtaining module 205 is configured to obtain a score of the to-be-predicted user for the to-be-recommended commodity according to the user characteristics of the to-be-predicted user and the commodity characteristics of the to-be-recommended commodity.
The judging module 206 is configured to judge whether the score of the to-be-predicted user for the to-be-recommended commodity is greater than a preset score threshold.
The recommending module 207 is configured to recommend the to-be-recommended commodity to the to-be-predicted user when the score of the to-be-predicted user for the to-be-recommended commodity is greater than a preset score threshold; and when the score of the user to be predicted on the commodity to be recommended is not larger than a preset score threshold value, the commodity to be recommended is not recommended to the user to be predicted.
As another embodiment, the present invention provides a commodity recommendation system based on a user-oriented multi-relationship information network, further comprising:
and the first conversion matrix acquisition module is used for acquiring a first conversion matrix before acquiring the user characteristics of the user to be predicted according to the recommended scene by using a graph convolution method and an attention mechanism.
And the first dimension reduction module is used for reducing the dimension of the user characteristic information matrix by using the first conversion matrix to obtain the user characteristic information matrix after dimension reduction.
The relationship type matrix construction module is used for constructing a relationship type matrix according to the relationship type set; the ith behavior of the relationship type matrix is a one-hot vector corresponding to the ith relationship type; the value of the ith dimension in the one-hot vector corresponding to the ith relation type is 1, and the values of the other dimensions are 0.
And the second conversion matrix acquisition module is used for acquiring a second conversion matrix.
The second dimension reduction module is used for reducing the dimension of the relation type matrix by using the second conversion matrix to obtain a reduced-dimension relation type matrix; the column number of the relation type matrix after dimension reduction is equal to the column number of the user characteristic information matrix after dimension reduction;
a third transformation matrix obtaining module, configured to obtain a third transformation matrix;
and the third dimension reduction module is used for reducing the dimension of the commodity characteristic information matrix by using the third conversion matrix to obtain the commodity characteristic information matrix after dimension reduction.
As another embodiment, in the commodity recommendation system based on the user-oriented multi-relationship information network of the present invention, the user feature extraction module 203 specifically includes:
a current iteration user feature calculation unit for utilizing the formula
Figure BDA0002666332790000131
Acquiring the user characteristics of the current iteration, namely the ith iteration, of each user; wherein, F(i)(j) The current iteration user characteristics of the jth user; a is an activation function; wiIs the weight matrix of the current iteration; f(i-1)(j) The characteristics of the user of the previous iteration, namely the i-1 iteration, of the jth user are obtained; nei (j) is a neighbor user set of the jth user; att (j, a)j) Is the jth user and the neighbor user ajAttention weight in between;
Figure BDA0002666332790000132
is the jth user and the neighbor user ajThe type of relationship between;
Figure BDA0002666332790000133
is the jth user and the neighbor user ajFeature vector of the type of relationship between, feature vector
Figure BDA0002666332790000134
Obtaining from the relation type matrix E; f(i-1)(aj) For neighbor user ajUser characteristic information of the (i-1) th iteration.
And the judging unit is used for judging whether the current iteration times reach the maximum iteration times.
And the user characteristic determining unit of the user to be predicted is used for determining the user characteristic of the user to be predicted of the current iteration as the final user characteristic of the user to be predicted when the current iteration reaches the maximum iteration.
And the returning unit is used for updating the current iteration times when the current iteration times do not reach the maximum iteration times, returning to the current iteration user characteristic calculating unit and entering the next iteration.
As another embodiment, in the present invention, based on a user-oriented multi-relationship information network-oriented product recommendation system, the score obtaining module 205 specifically includes:
a score calculating unit for using a formula
Figure BDA0002666332790000141
Obtaining the score of the user to be predicted on the commodity to be recommended; wherein the content of the first and second substances,
Figure BDA0002666332790000142
scoring the to-be-recommended commodity for the to-be-predicted user; sigma is sigmoid function; w is apIs a transformation matrix; i is vector splicing operation; f(t)(u*) For the user u to be predicted*The user characteristics of (1); m (v)*) For a commodity v to be recommended*The characteristics of the article of commerce.
The embodiments in the present description are described in a progressive manner, each embodiment focuses on differences from other embodiments, and the same and similar parts among the embodiments are referred to each other. For the system disclosed by the embodiment, the description is relatively simple because the system corresponds to the method disclosed by the embodiment, and the relevant points can be referred to the method part for description.
The principles and embodiments of the present invention have been described herein using specific examples, which are provided only to help understand the method and the core concept of the present invention; meanwhile, for a person skilled in the art, according to the idea of the present invention, the specific embodiments and the application range may be changed. In view of the above, the present disclosure should not be construed as limiting the invention.

Claims (10)

1. A commodity recommendation method based on a user-oriented multi-relationship information network is characterized by comprising the following steps:
acquiring a recommended scene; the recommendation scene comprises a user set, a user characteristic information matrix, a user relation graph, a commodity set, a commodity characteristic information matrix and an interaction matrix of the user and the commodity; the user characteristic information matrix is a matrix formed by a plurality of pieces of characteristic information corresponding to each user; the commodity characteristic information matrix is a matrix formed by a plurality of characteristic information corresponding to each commodity; the user relationship graph is a multi-relationship information network among users;
acquiring a user to be predicted;
according to the recommendation scene, acquiring the user characteristics of the user to be predicted by using a graph convolution method and an attention mechanism; the user characteristics of the user to be predicted are fused with the characteristic information of the neighbor user of the user to be predicted;
acquiring commodity characteristics of a commodity to be recommended; the commodity features are a plurality of feature information of the commodities to be recommended;
according to the user characteristics of the user to be predicted and the commodity characteristics of the commodity to be recommended, obtaining the score of the user to be predicted on the commodity to be recommended;
judging whether the score of the user to be predicted on the commodity to be recommended is larger than a preset score threshold value or not;
when the score of the user to be predicted on the commodity to be recommended is larger than a preset score threshold value, recommending the commodity to be recommended to the user to be predicted;
and when the score of the user to be predicted on the commodity to be recommended is not larger than a preset score threshold value, the commodity to be recommended is not recommended to the user to be predicted.
2. The commodity recommendation method based on the user-oriented multi-relationship information network according to claim 1, wherein the obtaining of the user characteristics of the user to be predicted according to the recommendation scenario by using a graph convolution method and an attention mechanism further comprises:
acquiring a first conversion matrix;
reducing the dimension of the user characteristic information matrix by using the first conversion matrix to obtain a user characteristic information matrix after dimension reduction;
constructing a relation type matrix according to the relation type set; the ith behavior of the relationship type matrix is a one-hot vector corresponding to the ith relationship type; the value of the ith dimension in the one-hot vector corresponding to the ith relation type is 1, and the values of the other dimensions are 0;
acquiring a second conversion matrix;
reducing the dimension of the relationship type matrix by using the second conversion matrix to obtain a reduced-dimension relationship type matrix; the column number of the relation type matrix after dimension reduction is equal to the column number of the user characteristic information matrix after dimension reduction;
acquiring a third conversion matrix;
and reducing the dimension of the commodity characteristic information matrix by using the third conversion matrix to obtain a dimension-reduced commodity characteristic information matrix.
3. The commodity recommendation method based on the user-oriented multi-relationship information network according to claim 1, wherein the obtaining of the user characteristics of the user to be predicted according to the recommendation scenario by using a graph convolution method and an attention mechanism specifically comprises:
for the ith iteration, the formula is used
Figure FDA0002666332780000021
Acquiring the user characteristics of the current iteration, namely the ith iteration, of each user; wherein, F(i)(j) The current iteration user characteristics of the jth user; a is an activation function; wiIs the weight matrix of the current iteration; f(i-1)(j) The characteristics of the user of the previous iteration, namely the i-1 iteration, of the jth user are obtained; nei (j) is a neighbor user set of the jth user; att (j, a)j) Is the jth user and the neighbor user ajAttention weight in between;
Figure FDA0002666332780000022
is the jth user and the neighbor user ajThe type of relationship between;
Figure FDA0002666332780000023
is the jth user and the neighbor user ajCharacteristic of the type of relationship betweenAn amount; f(i-1)(aj) For neighbor user ajUser characteristic information of the previous iteration;
judging whether the current iteration times reach the maximum iteration times or not;
if the current iteration times reach the maximum iteration times, determining the user characteristics of the user to be predicted of the current iteration as the final user characteristics of the user to be predicted;
if the current iteration times do not reach the maximum iteration times, updating the current iteration times, and returning to the utilization formula
Figure FDA0002666332780000031
And acquiring the user characteristics of the current iteration of each user, and entering the next iteration.
4. The user-oriented multi-relationship information network-based commodity recommendation method according to claim 2, wherein the obtaining of the score of the to-be-predicted user for the to-be-recommended commodity according to the user characteristics of the to-be-predicted user and the commodity characteristics of the to-be-recommended commodity specifically comprises:
using formulas
Figure FDA0002666332780000032
Obtaining the score of the user to be predicted on the commodity to be recommended; wherein the content of the first and second substances,
Figure FDA0002666332780000033
scoring the to-be-recommended commodity for the to-be-predicted user; sigma is sigmoid function; w is apIs a transformation matrix; i is vector splicing operation; f(t)(u*) For the user u to be predicted*The user characteristics of (1); m (v)*) For a commodity v to be recommended*The characteristics of the article of commerce.
5. The user-oriented multi-relationship information network-based commodity recommendation method according to claim 1, wherein the predetermined score threshold is 0.5.
6. A commodity recommendation system based on a user-oriented multi-relationship information network is characterized by comprising:
the recommendation scene obtaining module is used for obtaining recommendation scenes; the recommendation scene comprises a user set, a user characteristic information matrix, a user relation graph, a commodity set, a commodity characteristic information matrix and an interaction matrix of the user and the commodity; the user characteristic information matrix is a matrix formed by a plurality of pieces of characteristic information corresponding to each user; the commodity characteristic information matrix is a matrix formed by a plurality of characteristic information corresponding to each commodity; the user relationship graph is a multi-relationship information network among users;
the user to be predicted acquiring module is used for acquiring a user to be predicted;
the user feature extraction module is used for acquiring the user features of the user to be predicted according to the recommendation scene by using a graph convolution method and an attention mechanism; the user characteristics of the user to be predicted are fused with the characteristic information of the neighbor user of the user to be predicted;
the commodity feature acquisition module is used for acquiring the commodity features of the commodities to be recommended; the commodity features are a plurality of feature information of the commodities to be recommended;
the score acquisition module is used for acquiring the score of the to-be-predicted user on the to-be-recommended commodity according to the user characteristics of the to-be-predicted user and the commodity characteristics of the to-be-recommended commodity;
the judging module is used for judging whether the score of the user to be predicted on the commodity to be recommended is larger than a preset score threshold value or not;
the recommending module is used for recommending the to-be-recommended commodity to the to-be-predicted user when the score of the to-be-predicted user on the to-be-recommended commodity is larger than a preset score threshold value; and when the score of the user to be predicted on the commodity to be recommended is not larger than a preset score threshold value, the commodity to be recommended is not recommended to the user to be predicted.
7. The user-oriented multi-relationship information network-based commodity recommendation system according to claim 6, further comprising:
the first conversion matrix obtaining module is used for obtaining a first conversion matrix before obtaining the user characteristics of the user to be predicted according to the recommendation scene by using a graph convolution method and an attention mechanism;
the first dimension reduction module is used for reducing the dimension of the user characteristic information matrix by using the first conversion matrix to obtain a user characteristic information matrix after dimension reduction;
the relationship type matrix construction module is used for constructing a relationship type matrix according to the relationship type set; the ith behavior of the relationship type matrix is a one-hot vector corresponding to the ith relationship type; the value of the ith dimension in the one-hot vector corresponding to the ith relation type is 1, and the values of the other dimensions are 0;
the second conversion matrix acquisition module is used for acquiring a second conversion matrix;
the second dimension reduction module is used for reducing the dimension of the relation type matrix by using the second conversion matrix to obtain a reduced-dimension relation type matrix; the column number of the relation type matrix after dimension reduction is equal to the column number of the user characteristic information matrix after dimension reduction;
a third transformation matrix obtaining module, configured to obtain a third transformation matrix;
and the third dimension reduction module is used for reducing the dimension of the commodity characteristic information matrix by using the third conversion matrix to obtain the commodity characteristic information matrix after dimension reduction.
8. The user-oriented multi-relationship information network-based commodity recommendation system according to claim 6, wherein the user feature extraction module specifically comprises:
a current iteration user feature calculation unit for utilizing the formula
Figure FDA0002666332780000051
Acquiring the user characteristics of the current iteration, namely the ith iteration, of each user; wherein, F(i)(j) The current iteration user characteristics of the jth user; a is an activation function; wiIs the weight matrix of the current iteration; f(i-1)(j) The characteristics of the user of the previous iteration, namely the i-1 iteration, of the jth user are obtained; nei (j) is a neighbor user set of the jth user; att (j, a)j) Is the jth user and the neighbor user ajAttention weight in between;
Figure FDA0002666332780000052
is the jth user and the neighbor user ajThe type of relationship between;
Figure FDA0002666332780000053
is the jth user and the neighbor user ajA feature vector of a relationship type between; f(i-1)(aj) For neighbor user ajUser characteristic information of the previous iteration;
the judging unit is used for judging whether the current iteration times reach the maximum iteration times;
the user feature determination unit of the user to be predicted is used for determining the user feature of the user to be predicted of the current iteration as the final user feature of the user to be predicted when the current iteration number reaches the maximum iteration number;
and the returning unit is used for updating the current iteration times when the current iteration times do not reach the maximum iteration times, returning to the current iteration user characteristic calculating unit and entering the next iteration.
9. The user-oriented multi-relationship information network-based commodity recommendation system according to claim 7, wherein the score obtaining module specifically comprises:
a score calculating unit for using a formula
Figure FDA0002666332780000054
Obtaining the score of the user to be predicted on the commodity to be recommended; wherein the content of the first and second substances,
Figure FDA0002666332780000055
scoring the to-be-recommended commodity for the to-be-predicted user; sigma is sigmoid function; w is apIs a transformation matrix; i is vector splicing operation; f(t)(u*) For the user u to be predicted*The user characteristics of (1); m (v)*) For a commodity v to be recommended*The characteristics of the article of commerce.
10. The user-oriented multi-relationship information network-based commodity recommendation system according to claim 6, wherein said predetermined score threshold is 0.5.
CN202010919815.5A 2020-09-04 2020-09-04 Commodity recommendation method and system based on user-oriented multi-relation information network Active CN111967946B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202010919815.5A CN111967946B (en) 2020-09-04 2020-09-04 Commodity recommendation method and system based on user-oriented multi-relation information network

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202010919815.5A CN111967946B (en) 2020-09-04 2020-09-04 Commodity recommendation method and system based on user-oriented multi-relation information network

Publications (2)

Publication Number Publication Date
CN111967946A true CN111967946A (en) 2020-11-20
CN111967946B CN111967946B (en) 2024-04-12

Family

ID=73392095

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202010919815.5A Active CN111967946B (en) 2020-09-04 2020-09-04 Commodity recommendation method and system based on user-oriented multi-relation information network

Country Status (1)

Country Link
CN (1) CN111967946B (en)

Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN110737778A (en) * 2019-09-04 2020-01-31 北京邮电大学 Knowledge graph and Transformer based patent recommendation method
CN111160954A (en) * 2019-12-16 2020-05-15 南京理工大学 Recommendation method facing group object based on graph convolution network model
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
CN111538868A (en) * 2020-04-28 2020-08-14 中国科学技术大学 Knowledge tracking method and exercise recommendation method

Patent Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN110737778A (en) * 2019-09-04 2020-01-31 北京邮电大学 Knowledge graph and Transformer based patent recommendation method
CN111160954A (en) * 2019-12-16 2020-05-15 南京理工大学 Recommendation method facing group object based on graph convolution network model
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
CN111538868A (en) * 2020-04-28 2020-08-14 中国科学技术大学 Knowledge tracking method and exercise recommendation method

Also Published As

Publication number Publication date
CN111967946B (en) 2024-04-12

Similar Documents

Publication Publication Date Title
CN107145518B (en) Personalized recommendation system based on deep learning in social network
CN110008408A (en) A kind of session recommended method, system, equipment and medium
CN109657156A (en) A kind of personalized recommendation method generating confrontation network based on circulation
CN106779867B (en) Support vector regression recommendation method and system based on context awareness
CN109189921B (en) Comment evaluation model training method and device
CN103399858A (en) Socialization collaborative filtering recommendation method based on trust
CN109903103B (en) Method and device for recommending articles
CN103325061A (en) Community discovery method and system
CN110781407A (en) User label generation method and device and computer readable storage medium
CN112950324A (en) Knowledge graph assisted pairwise sorting personalized merchant recommendation method and system
CN110503497A (en) A kind of electric business management platform Method of Commodity Recommendation of the consumption habit based on client
CN108022116B (en) Method, system and terminal equipment for modeling user
CN107918778A (en) A kind of information matching method and relevant apparatus
CN110727855A (en) Personalized recommendation method based on improved factorization machine
CN109189922B (en) Comment evaluation model training method and device
CN114386513A (en) Interactive grading prediction method and system integrating comment and grading
CN113221019A (en) Personalized recommendation method and system based on instant learning
CN112256966A (en) Article recommendation method based on semi-supervised factorization machine
CN114298783A (en) Commodity recommendation method and system based on matrix decomposition and fusion of user social information
CN106777069A (en) Personalized recommendation method and commending system based on matrix factorisation
CN109190040A (en) Personalized recommendation method and device based on coevolution
CN116842478A (en) User attribute prediction method based on twitter content
CN111967946A (en) Commodity recommendation method and system based on user-oriented multi-relationship information network
CN116204723A (en) Social recommendation method based on dynamic hypergraph representation learning
CN112734519B (en) Commodity recommendation method based on convolution self-encoder network

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