CN115187343A - Multi-behavior recommendation method based on attention map convolution neural network - Google Patents

Multi-behavior recommendation method based on attention map convolution neural network Download PDF

Info

Publication number
CN115187343A
CN115187343A CN202210850345.0A CN202210850345A CN115187343A CN 115187343 A CN115187343 A CN 115187343A CN 202210850345 A CN202210850345 A CN 202210850345A CN 115187343 A CN115187343 A CN 115187343A
Authority
CN
China
Prior art keywords
behavior
user
formula
matrix
item
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
CN202210850345.0A
Other languages
Chinese (zh)
Other versions
CN115187343B (en
Inventor
程志勇
韩赛
高赞
卓涛
李晓丽
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.)
Shandong Institute of Artificial Intelligence
Original Assignee
Shandong Institute of Artificial Intelligence
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 Shandong Institute of Artificial Intelligence filed Critical Shandong Institute of Artificial Intelligence
Priority to CN202210850345.0A priority Critical patent/CN115187343B/en
Publication of CN115187343A publication Critical patent/CN115187343A/en
Application granted granted Critical
Publication of CN115187343B publication Critical patent/CN115187343B/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
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/08—Learning methods
    • 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]
    • Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
    • Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
    • Y02D—CLIMATE CHANGE MITIGATION TECHNOLOGIES IN INFORMATION AND COMMUNICATION TECHNOLOGIES [ICT], I.E. INFORMATION AND COMMUNICATION TECHNOLOGIES AIMING AT THE REDUCTION OF THEIR OWN ENERGY USE
    • Y02D10/00—Energy efficient computing, e.g. low power processors, power management or thermal management

Landscapes

  • Business, Economics & Management (AREA)
  • Engineering & Computer Science (AREA)
  • Accounting & Taxation (AREA)
  • Finance (AREA)
  • Physics & Mathematics (AREA)
  • Theoretical Computer Science (AREA)
  • General Physics & Mathematics (AREA)
  • Development Economics (AREA)
  • Economics (AREA)
  • Marketing (AREA)
  • Strategic Management (AREA)
  • General Business, Economics & Management (AREA)
  • Artificial Intelligence (AREA)
  • Evolutionary Computation (AREA)
  • Health & Medical Sciences (AREA)
  • Biomedical Technology (AREA)
  • Biophysics (AREA)
  • Computational Linguistics (AREA)
  • Data Mining & Analysis (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • General Health & Medical Sciences (AREA)
  • Molecular Biology (AREA)
  • Computing Systems (AREA)
  • General Engineering & Computer Science (AREA)
  • Mathematical Physics (AREA)
  • Software Systems (AREA)
  • Information Retrieval, Db Structures And Fs Structures Therefor (AREA)
  • Management, Administration, Business Operations System, And Electronic Commerce (AREA)

Abstract

A multi-behavior recommendation method based on attention map convolution neural network utilizes a multi-channel map convolution neural network module to design a feature expression for each behavior to solve the problem of multiple embedded expressions under multiple behaviors; capturing importance degrees of different behaviors by adopting a behavior special attention mechanism module; finally, the user preference is modeled through a multi-task learning framework, the recommendation effect of the model is improved, and recommendation can be performed through modeling of the user preference.

Description

Multi-behavior recommendation method based on attention map convolution neural network
Technical Field
The invention relates to the technical field of recommendation systems and deep learning, in particular to a multi-behavior recommendation method based on an attention map convolution neural network.
Background
In the information explosion age, the recommendation system plays a great role in solving the problem of information overload. The recommendation system can predict new information which is interested by the user according to historical information of the user in the past, and is widely applied to platforms such as music, movies, television, e-commerce, online review websites, location-based recommendation services and the like. With Collaborative Filtering (CF) technology being the most widely used algorithm in the recommendation field. However, most of the traditional models and the neural network models are researched on single-type behaviors, the interest of users in real life is not only single but also on single behaviors, and therefore, the research on multi-behavior recommendation is inevitable. The current model research for multi-behavior recommendation is mainly based on the four frameworks: BPR, RNN, GCN, MTL. Sampling multi-behavior data by using a BPR frame, and sequencing the object pairs of the user by using auxiliary behaviors; the RNN network is used for capturing sequence information of a plurality of behaviors, and the importance degree between the behaviors is solved by using a behavior-specific attention mechanism. Multiple behavior information of users and item maps and item maps are captured using a GCN network to capture item attribute information. And enabling a plurality of tasks to share user embedding or model training according to the sequence relation between behaviors by utilizing the MTL framework. Although the above work has achieved great success, the association between different embedded expressions under various behaviors and the failure of deeper mining behaviors is ignored, resulting in the failure to obtain an optimal model.
Disclosure of Invention
In order to overcome the defects of the technology, the invention provides a method for modeling user preference and improving the model recommendation effect by constructing an embedded expression of various behaviors, a multi-channel graph convolution module and a behavior specific attention mechanism module.
The technical scheme adopted by the invention for overcoming the technical problems is as follows:
a multi-behavior recommendation method based on an attention map convolution neural network comprises the following steps:
a) A leave-one-out method is adopted in an e-commerce data set to obtain a training set and a test set;
b) The training set comprises N users, I articles and K types of user behaviors, and an undirected graph of the user behaviors is constructed by utilizing the training set;
c) Establishing and training a multi-task learning network model based on behavior-tailored attention multi-channel graph convolutional nerves;
d) And calculating the preference degree of a user for a certain action on the article to realize article recommendation.
Further, each user in the test set in the step a) has an article which has not interacted with the user, and the parameters of the multi-task learning network model of the multi-channel graph convolutional neural network based on behavior tailoring attention in the step c) are debugged through the test set.
Further, the K =3,3 types of user behaviors in the step b) are a behavior of purchasing an item by the user, a behavior of adding an item to the shopping cart by the user, and a behavior of clicking an item link by the user, respectively.
Further, step b) comprises the following steps:
b-1) by the formula G = (U) (k) ,I (k) ) Constructing an undirected graph G of the K-th user behavior, wherein K = {1,2, \8230 =, K }, and U in the formula (k) In order to contain the set of user nodes,
Figure BDA0003754888370000021
wherein
Figure BDA0003754888370000022
The vector matrix of the nth user under the k-th user behavior, N = {1,2, \8230;, N },
Figure BDA0003754888370000023
wherein R is a real number space, d is a dimension,
Figure BDA0003754888370000024
wherein
Figure BDA0003754888370000025
Is the vector matrix of the nth item under the k-th user behavior, m = {1,2, \8230;, I },
Figure BDA0003754888370000026
further, step c) comprises the steps of:
c-1) by the formula
Figure BDA0003754888370000027
Computing the embedding expression matrix of the u-th user at the l-th layer of the undirected graph G
Figure BDA0003754888370000028
In the formula
Figure BDA0003754888370000029
An embedded expression matrix of item i at level l-1 of the undirected graph G, l = {1,2,3}, N u For the number of items that the u-th user has interacted, N i The number of the ith item interacted by the u user;
c-2) by the formula
Figure BDA00037548883700000210
Computing an embedded expression matrix for the ith item at the l-th level of undirected graph G
Figure BDA00037548883700000211
In the formula
Figure BDA00037548883700000212
The embedded expression matrix of the user u is the l-1 layer of the undirected graph G;
c-3) by the formula
Figure BDA0003754888370000031
Calculating to obtain the final u-th user embedded expression matrix e after l-layer aggregation under each behavior u ；
c-4) by the formula
Figure BDA0003754888370000032
Calculating to obtain an embedded expression matrix e of the ith article finally after the polymerization of the layer I under each behavior i ；
c-5) by the formula
Figure BDA0003754888370000033
Calculating to obtain the kth behavior of the u userEmbedded expression matrix
Figure BDA0003754888370000034
In the formula
Figure BDA0003754888370000035
Initializing the matrix for the weights that the u-th user is affected by other behaviors under the k-th behavior, b initializing a factor for the bias,
Figure BDA0003754888370000036
initializing a matrix for the embedding expression of the u user before being influenced by the k behavior;
c-6) by the formula
Figure BDA0003754888370000037
Calculating to obtain an embedded expression matrix of the ith article under the k behavior
Figure BDA0003754888370000038
In the formula W i (k) Initializing a matrix for weights that the ith item is affected by other actions in the kth action, b a bias initialization factor,
Figure BDA0003754888370000039
initializing a matrix for embedding expression of the ith item before being influenced by the kth action, and completing establishment of a multi-task learning network model based on behavior-specific attention multi-channel graph convolutional nerves;
c-7) by the formula
Figure BDA00037548883700000310
Calculating to obtain the loss L of the kth behavior k Wherein sigma is sigmoid function, lambda is L2 norm normalization factor,
Figure BDA0003754888370000041
the possibility of performing a kth action on the ith item for the u-th user,
Figure BDA0003754888370000042
Figure BDA0003754888370000043
Figure BDA0003754888370000044
the possibility of performing a kth action on the jth item for the uth user,
Figure BDA0003754888370000045
Figure BDA0003754888370000046
c-8) by the formula
Figure BDA0003754888370000047
Calculating to obtain the total loss L MTL In the formula a k A loss weight value for the kth action;
c-9) utilization of the total loss L MTL Training a multi-task learning network model based on behavior-specific attention multi-channel graph convolutional nerves by adopting a gradient descent method, and updating an embedded expression matrix of the u-th user under the k-th behavior in the step c-5) through back propagation
Figure BDA0003754888370000048
And the weight initialization matrix of the u user under the k behavior influenced by other behaviors
Figure BDA0003754888370000049
Updating the embedded expression matrix of the ith item under the k action in the step c-6)
Figure BDA00037548883700000410
And the weight initialization matrix W of the ith item under the k action influenced by other actions i (k) . Further, in step d), the formula is used
Figure BDA00037548883700000411
Calculating the preference of the user to the articleDegree y, where T is transposed.
The invention has the beneficial effects that: designing a feature expression for each behavior by using a multi-channel graph convolution neural network module to solve the problem of multiple embedded expressions under multiple behaviors; capturing importance degrees of different behaviors by adopting a behavior special attention mechanism module; finally, the user preference is modeled through a multi-task learning framework, the recommendation effect of the model is improved, and recommendation can be performed through modeling of the user preference.
Drawings
FIG. 1 is a flow chart of the method of the present invention;
FIG. 2 is a block diagram of a behavioral specific attention mechanism of the present invention.
Detailed Description
The invention will be further explained with reference to fig. 1 and 2.
A multi-behavior recommendation method based on an attention map convolution neural network comprises the following steps:
a) A leave-one-out method is adopted in an e-commerce data set to obtain a training set and a test set;
b) The training set comprises N users, I articles and K types of user behaviors, and an undirected graph of the user behaviors is constructed by utilizing the training set;
c) Establishing and training a multi-task learning network model based on behavior-tailored attention multi-channel graph convolutional nerves;
d) And calculating the preference degree of a user for performing a certain action on the article to realize article recommendation.
Designing a feature expression for each behavior by utilizing a multi-channel graph convolutional neural network module to solve the problem of multiple embedded expressions under multiple behaviors; capturing importance degrees of different behaviors by adopting a behavior tailor-made attention mechanism module; finally, user preference modeling is performed through a multi-task learning framework, the recommendation effect of the model is improved, and recommendation can be performed through the user preference modeling.
Example 1:
each user in the test set in the step a) has an article which is not interacted with the user, and parameters of the multitask learning network model of the behavior-based special attention multichannel graph convolutional nerve in the step c) are debugged through the test set.
Example 2:
the K =3,3 types of user behaviors in the step b) are a user purchase item behavior, a user add item to shopping cart behavior and a user click item link behavior respectively. For example, a real e-commerce data set is used, which is from the largest mother-infant e-commerce platform bei in china, and comprises 3 types of user behaviors, including a user behavior of purchasing an item, a user behavior of adding an item to a shopping cart, and a user behavior of clicking on an item link.
Example 3:
the step b) comprises the following steps:
b-1) by the formula G = (U) (k) ,I (k) ) Constructing an undirected graph G of kth user behavior, K = {1,2, \8230;, K }, K =1 is a user item purchasing behavior, K =2 is a user item adding to shopping cart behavior, and K =3 is a user item clicking link behavior, wherein U is a formula (k) In order to contain the set of user nodes,
Figure BDA0003754888370000051
wherein
Figure BDA0003754888370000052
The vector matrix of the nth user under the k-th user behavior, N = {1,2, \8230;, N },
Figure BDA0003754888370000053
wherein R is a real number space, d is a dimension,
Figure BDA0003754888370000054
wherein
Figure BDA0003754888370000055
For the vector matrix of the nth item under the kth user behavior, m = {1,2, \8230;, I },
Figure BDA0003754888370000061
example 4:
the step c) comprises the following steps:
c-1) by the formula
Figure BDA0003754888370000062
Computing the embedding expression matrix of the u-th user at the l-th layer of the undirected graph G
Figure BDA0003754888370000063
In the formula
Figure BDA0003754888370000064
Is an embedded expression matrix of an item i at layer l-1 of an undirected graph G, l = {1,2,3}, N u For the number of items that the u-th user has interacted, N i Is the number of items that were interacted with by the u-th user.
c-2) by the formula
Figure BDA0003754888370000065
Computing an embedded expression matrix for the ith item at the l-th level of undirected graph G
Figure BDA0003754888370000066
In the formula
Figure BDA0003754888370000067
The embedded expression matrix of user u at layer l-1 of undirected graph G.
c-3) by the formula
Figure BDA0003754888370000068
Calculating to obtain the final u-th user embedded expression matrix e after l-layer aggregation under each behavior u 。
c-4) by the formula
Figure BDA0003754888370000069
Calculating to obtain an embedded expression matrix e of the ith article finally after the polymerization of the layer I under each behavior i 。
c-5) by the formula
Figure BDA00037548883700000610
Calculating to obtain an embedded expression matrix of the u-th user under the k-th behavior
Figure BDA00037548883700000611
In the formula
Figure BDA00037548883700000612
Initializing a matrix for the weights of the u-th user under the k-th behavior affected by the other behaviors (trainable), b biasing an initialization factor (which is a trainable real number),
Figure BDA0003754888370000071
the matrix is initialized for the embedded expression of the u-th user before being affected by the k-th behavior. c-6) by the formula
Figure BDA0003754888370000072
Calculating to obtain an embedded expression matrix of the ith article under the k behavior
Figure BDA0003754888370000073
In the formula W i (k) Initializing a matrix for weights that the ith item is affected by other actions in the kth action (trainable), b is a bias initialization factor (is a trainable real number),
Figure BDA0003754888370000074
and (5) initializing a matrix for the embedded expression of the ith item before being influenced by the kth action, and completing the establishment of a multi-task learning network model based on the behavior-specific attention multi-channel graph convolutional nerve.
c-7) by the formula
Figure BDA0003754888370000075
Calculating to obtain the loss L of the kth behavior k Wherein sigma is sigmoid function, lambda is L2 norm normalization factor,
Figure BDA0003754888370000076
the possibility of performing a kth action on the ith item for the u-th user,
Figure BDA0003754888370000077
Figure BDA0003754888370000078
Figure BDA0003754888370000079
the possibility of performing a kth action on the jth item for the uth user,
Figure BDA00037548883700000710
Figure BDA00037548883700000711
c-8) by the formula
Figure BDA00037548883700000712
Calculating to obtain the total loss L MTL In the formula a k Is the loss weight value for the kth behavior.
c-9) utilization of the total loss L MTL Training a multi-task learning network model based on behavior-specific attention multi-channel graph convolutional nerves by adopting a gradient descent method, and updating an embedded expression matrix of the u-th user under the k-th behavior in the step c-5) through back propagation
Figure BDA00037548883700000713
And the weight initialization matrix of the u user under the k behavior influenced by other behaviors
Figure BDA00037548883700000714
Updating the embedded expression matrix of the ith item under the k action in the step c-6)
Figure BDA0003754888370000081
And the weight initialization matrix W of the ith item under the k action influenced by other actions i (k) 。
Example 5:
in step d) by formula
Figure BDA0003754888370000082
And calculating to obtain the preference degree y of the user to the article, wherein T is the transposition. Finally, it should be noted that: although the present invention has been described in detail with reference to the foregoing embodiments, it will be apparent to those skilled in the art that changes may be made in the embodiments and/or equivalents thereof without departing from the spirit and scope of the invention. Any modification, equivalent replacement, or improvement made within the spirit and principle of the present invention should be included in the protection scope of the present invention.

Claims (6)

1. A multi-behavior recommendation method based on an attention map convolutional neural network is characterized by comprising the following steps:
a) A leave-one-out method is adopted in an e-commerce data set to obtain a training set and a test set;
b) The training set comprises N users, I articles and K types of user behaviors, and an undirected graph of the user behaviors is constructed by utilizing the training set;
c) Establishing and training a multi-task learning network model based on behavior-tailored attention multi-channel graph convolutional nerves;
d) And calculating the preference degree of a user for performing a certain action on the article to realize article recommendation.
2. The attention graph convolution neural network-based multi-behavior recommendation method of claim 1, wherein: each user in the test set in the step a) has an article which is not interacted with the user, and parameters of the multitask learning network model of the behavior-based special attention multichannel graph convolutional nerve in the step c) are debugged through the test set.
3. The attention graph convolution neural network-based multi-behavior recommendation method of claim 1, wherein: the K =3,3 types of user behaviors in the step b) are a user purchase item behavior, a user add item to shopping cart behavior and a user click item link behavior respectively.
4. The attention map convolutional neural network-based multi-behavior recommendation method of claim 1, wherein step b) comprises the steps of:
b-1) by the formula G = (U) (k) ,I (k) ) Constructing an undirected graph G of the K-th user behavior, wherein K = {1, 2., K }, and U in the formula (k) In order to contain the set of user nodes,
Figure FDA0003754888360000011
wherein
Figure FDA0003754888360000012
Is a vector matrix of the nth user under the k user behavior, N = {1,2,. Multidata, N },
Figure FDA0003754888360000013
wherein R is a real number space, d is a dimension,
Figure FDA0003754888360000014
wherein
Figure FDA0003754888360000015
Is a vector matrix of the nth item under the k-th user behavior, m = {1,2, ·, I },
Figure FDA0003754888360000016
5. the attention map convolution neural network-based multi-behavior recommendation method of claim 4, wherein the step c) comprises the steps of:
c-1) by the formula
Figure FDA0003754888360000021
Computing an embedded expression matrix for the u-th user at layer l of the undirected graph G
Figure FDA0003754888360000022
In the formula
Figure FDA0003754888360000023
Is an embedded expression matrix of an item i at layer l-1 of an undirected graph G, l = {1,2,3}, N u For the number of items that the u-th user has interacted, N i The number of the ith item interacted by the u user;
c-2) by the formula
Figure FDA0003754888360000024
Computing an embedded expression matrix for the ith item at the l-th level of undirected graph G
Figure FDA0003754888360000025
In the formula
Figure FDA0003754888360000026
The embedded expression matrix of the user u is the l-1 layer of the undirected graph G;
c-3) by the formula
Figure FDA0003754888360000027
Calculating to obtain the final u-th user embedded expression matrix e after the aggregation of the layer I under each behavior u ；
c-4) by the formula
Figure FDA0003754888360000028
Calculating to obtain an embedded expression matrix e of the ith article finally after the polymerization of the layer I under each behavior i ；
c-5) by the formula
Figure FDA0003754888360000029
Calculating to obtain the kth behavior of the u userEmbedded expression matrix
Figure FDA00037548883600000210
In the formula
Figure FDA00037548883600000211
Initializing the matrix for the weights of the kth user affected by other behaviors in the kth behavior, b a bias initialization factor,
Figure FDA00037548883600000212
initializing a matrix for the embedding expression of the u user before being influenced by the k behavior;
c-6) by the formula
Figure FDA0003754888360000031
Calculating to obtain an embedded expression matrix of the ith article under the k behavior
Figure FDA0003754888360000032
In the formula W i (k) Initializing a matrix for weights that the ith item is affected by other actions in the kth action, b a bias initialization factor,
Figure FDA0003754888360000033
the method comprises the steps of initializing a matrix for embedding expression of an ith article before the ith article is influenced by a kth behavior, and completing establishment of a multi-task learning network model based on behavior-specific attention multi-channel graph convolutional nerves;
c-7) by the formula
Figure FDA0003754888360000034
Calculating the loss L of the k behavior k Wherein sigma is sigmoid function, lambda is L2 norm normalization factor,
Figure FDA0003754888360000035
the possibility of performing a kth action on the ith item for the u-th user,
Figure FDA0003754888360000036
Figure FDA0003754888360000037
Figure FDA0003754888360000038
the possibility of performing a kth action on the jth item for the uth user,
Figure FDA0003754888360000039
Figure FDA00037548883600000310
c-8) by the formula
Figure FDA00037548883600000311
Calculating to obtain the total loss L MTL In the formula a k A loss weight value for the kth action;
c-9) utilization of the total loss L MTL Training a multi-task learning network model based on behavior-specific attention multi-channel graph convolutional nerves by adopting a gradient descent method, and updating an embedded expression matrix of the u-th user under the k-th behavior in the step c-5) through back propagation
Figure FDA00037548883600000312
And the weight initialization matrix of the u user under the k behavior influenced by other behaviors
Figure FDA00037548883600000313
Updating the embedded expression matrix of the ith item under the k action in the step c-6)
Figure FDA00037548883600000314
And the weight initialization matrix W of the ith item under the k behavior influenced by other behaviors i (k) 。
6. The attention graph convolution neural network-based multi-behavior recommendation method of claim 5, wherein: in step d) by formula
Figure FDA0003754888360000041
And calculating to obtain the preference degree y of the user to the article, wherein T is the transposition.
CN202210850345.0A 2022-07-20 2022-07-20 Attention graph convolution neural network-based multi-behavior recommendation method Active CN115187343B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202210850345.0A CN115187343B (en) 2022-07-20 2022-07-20 Attention graph convolution neural network-based multi-behavior recommendation method

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202210850345.0A CN115187343B (en) 2022-07-20 2022-07-20 Attention graph convolution neural network-based multi-behavior recommendation method

Publications (2)

Publication Number Publication Date
CN115187343A true CN115187343A (en) 2022-10-14
CN115187343B CN115187343B (en) 2023-08-08

Family

ID=83519039

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202210850345.0A Active CN115187343B (en) 2022-07-20 2022-07-20 Attention graph convolution neural network-based multi-behavior recommendation method

Country Status (1)

Country Link
CN (1) CN115187343B (en)

Cited By (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN116645174A (en) * 2023-07-27 2023-08-25 山东省人工智能研究院 Personalized recommendation method based on decoupling multi-behavior characterization learning
CN117171448A (en) * 2023-08-11 2023-12-05 哈尔滨工业大学 Multi-behavior socialization recommendation method and system based on graph neural network
CN117171448B (en) * 2023-08-11 2024-05-28 哈尔滨工业大学 Multi-behavior socialization recommendation method and system based on graph neural network

Citations (8)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN110580314A (en) * 2019-07-16 2019-12-17 华南师范大学 Course recommendation method and system based on graph convolution neural network and dynamic weight
CN110674407A (en) * 2019-09-30 2020-01-10 杭州电子科技大学 Hybrid recommendation method based on graph convolution neural network
CN111127142A (en) * 2019-12-16 2020-05-08 东北大学秦皇岛分校 Article recommendation method based on generalized neural attention
CN111523047A (en) * 2020-04-13 2020-08-11 中南大学 Multi-relation collaborative filtering algorithm based on graph neural network
CN111737569A (en) * 2020-06-04 2020-10-02 山东省人工智能研究院 Personalized recommendation method based on attribute perception intention-minded convolutional neural network
CN112232925A (en) * 2020-11-02 2021-01-15 哈尔滨工程大学 Method for carrying out personalized recommendation on commodities by fusing knowledge maps
US20210366024A1 (en) * 2020-05-25 2021-11-25 National University Of Defense Technology Item recommendation method based on importance of item in session and system thereof
CN114117220A (en) * 2021-11-26 2022-03-01 东北大学 Deep reinforcement learning interactive recommendation system and method based on knowledge enhancement

Patent Citations (8)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN110580314A (en) * 2019-07-16 2019-12-17 华南师范大学 Course recommendation method and system based on graph convolution neural network and dynamic weight
CN110674407A (en) * 2019-09-30 2020-01-10 杭州电子科技大学 Hybrid recommendation method based on graph convolution neural network
CN111127142A (en) * 2019-12-16 2020-05-08 东北大学秦皇岛分校 Article recommendation method based on generalized neural attention
CN111523047A (en) * 2020-04-13 2020-08-11 中南大学 Multi-relation collaborative filtering algorithm based on graph neural network
US20210366024A1 (en) * 2020-05-25 2021-11-25 National University Of Defense Technology Item recommendation method based on importance of item in session and system thereof
CN111737569A (en) * 2020-06-04 2020-10-02 山东省人工智能研究院 Personalized recommendation method based on attribute perception intention-minded convolutional neural network
CN112232925A (en) * 2020-11-02 2021-01-15 哈尔滨工程大学 Method for carrying out personalized recommendation on commodities by fusing knowledge maps
CN114117220A (en) * 2021-11-26 2022-03-01 东北大学 Deep reinforcement learning interactive recommendation system and method based on knowledge enhancement

Non-Patent Citations (1)

* Cited by examiner, † Cited by third party
Title
杨宝生;: "基于注意力机制增强图卷积神经网络的个性化新闻推荐", 兰州文理学院学报(自然科学版), no. 05, pages 52 - 56 *

Cited By (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN116645174A (en) * 2023-07-27 2023-08-25 山东省人工智能研究院 Personalized recommendation method based on decoupling multi-behavior characterization learning
CN116645174B (en) * 2023-07-27 2023-10-17 山东省人工智能研究院 Personalized recommendation method based on decoupling multi-behavior characterization learning
CN117171448A (en) * 2023-08-11 2023-12-05 哈尔滨工业大学 Multi-behavior socialization recommendation method and system based on graph neural network
CN117171448B (en) * 2023-08-11 2024-05-28 哈尔滨工业大学 Multi-behavior socialization recommendation method and system based on graph neural network

Also Published As

Publication number Publication date
CN115187343B (en) 2023-08-08

Similar Documents

Publication Publication Date Title
CN111428147B (en) Social recommendation method of heterogeneous graph volume network combining social and interest information
CN111222332B (en) Commodity recommendation method combining attention network and user emotion
CN108874914A (en) A kind of information recommendation method based on the long-pending and neural collaborative filtering of picture scroll
CN106920147A (en) A kind of commodity intelligent recommendation method that word-based vector data drives
CN111259133A (en) Personalized recommendation method integrating multiple information
CN106776928B (en) Position recommendation method based on memory computing framework and fusing social contact and space-time data
CN113407864B (en) Group recommendation method based on mixed attention network
CN109062962A (en) A kind of gating cycle neural network point of interest recommended method merging Weather information
CN107003834B (en) Pedestrian detection device and method
CN111753207B (en) Collaborative filtering method for neural map based on comments
CN114817663A (en) Service modeling and recommendation method based on class perception graph neural network
CN111737569B (en) Personalized recommendation method based on attribute perception intention-minded convolutional neural network
CN109670927A (en) The method of adjustment and its device of credit line, equipment, storage medium
CN115187343B (en) Attention graph convolution neural network-based multi-behavior recommendation method
CN113868537B (en) Recommendation method based on multi-behavior session graph fusion
CN108805291B (en) Training method of network representation learning model and server
CN104484365B (en) In a kind of multi-source heterogeneous online community network between network principal social relationships Forecasting Methodology and system
CN113409157B (en) Cross-social network user alignment method and device
CN112487305B (en) GCN-based dynamic social user alignment method
CN109657725A (en) A kind of service quality prediction technique and system based on complicated space-time context aware
CN109800424A (en) It is a kind of based on improving matrix decomposition and the recommended method across channel convolutional neural networks
CN112016000A (en) Movie group recommendation method and system based on convolution collaborative filtering
CN116258504A (en) Bank customer relationship management system and method thereof
CN111652695B (en) Collaborative filtering recommendation method based on parallel self-encoding machine
CN113158088A (en) Position recommendation method based on graph neural 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
GR01 Patent grant