CN114912033A - Knowledge graph-based recommendation popularity deviation adaptive buffering method - Google Patents
Knowledge graph-based recommendation popularity deviation adaptive buffering method Download PDFInfo
- Publication number
- CN114912033A CN114912033A CN202210531980.2A CN202210531980A CN114912033A CN 114912033 A CN114912033 A CN 114912033A CN 202210531980 A CN202210531980 A CN 202210531980A CN 114912033 A CN114912033 A CN 114912033A
- Authority
- CN
- China
- Prior art keywords
- popularity
- graph
- user
- item
- knowledge
- 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
Links
- 238000000034 method Methods 0.000 title claims abstract description 44
- 230000003044 adaptive effect Effects 0.000 title claims description 15
- 230000003139 buffering effect Effects 0.000 title abstract description 5
- 238000003062 neural network model Methods 0.000 claims abstract description 14
- 230000002159 abnormal effect Effects 0.000 claims abstract description 11
- 230000000116 mitigating effect Effects 0.000 claims description 15
- 230000002776 aggregation Effects 0.000 claims description 9
- 238000004220 aggregation Methods 0.000 claims description 9
- 239000011159 matrix material Substances 0.000 claims description 7
- 239000013598 vector Substances 0.000 claims description 7
- 238000004364 calculation method Methods 0.000 claims description 5
- 230000003993 interaction Effects 0.000 claims description 4
- 238000012935 Averaging Methods 0.000 claims description 2
- 230000008569 process Effects 0.000 abstract description 5
- 238000004590 computer program Methods 0.000 description 9
- 230000006870 function Effects 0.000 description 6
- 230000008878 coupling Effects 0.000 description 3
- 238000010168 coupling process Methods 0.000 description 3
- 238000005859 coupling reaction Methods 0.000 description 3
- 230000000694 effects Effects 0.000 description 3
- 238000012545 processing Methods 0.000 description 3
- 102100038591 Endothelial cell-selective adhesion molecule Human genes 0.000 description 2
- 101000882622 Homo sapiens Endothelial cell-selective adhesion molecule Proteins 0.000 description 2
- 230000006399 behavior Effects 0.000 description 2
- 238000004422 calculation algorithm Methods 0.000 description 2
- 238000004891 communication Methods 0.000 description 2
- 230000001186 cumulative effect Effects 0.000 description 2
- 238000001914 filtration Methods 0.000 description 2
- 238000012986 modification Methods 0.000 description 2
- 230000004048 modification Effects 0.000 description 2
- 238000003672 processing method Methods 0.000 description 2
- 238000012549 training Methods 0.000 description 2
- 238000013473 artificial intelligence Methods 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 230000008859 change Effects 0.000 description 1
- 238000006243 chemical reaction Methods 0.000 description 1
- 238000002790 cross-validation Methods 0.000 description 1
- 230000003247 decreasing effect Effects 0.000 description 1
- 230000007812 deficiency Effects 0.000 description 1
- 238000013461 design Methods 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 238000010586 diagram Methods 0.000 description 1
- 238000005516 engineering process Methods 0.000 description 1
- 230000002708 enhancing effect Effects 0.000 description 1
- 238000002474 experimental method Methods 0.000 description 1
- 238000000605 extraction Methods 0.000 description 1
- 230000006872 improvement Effects 0.000 description 1
- 230000003287 optical effect Effects 0.000 description 1
- 238000011160 research Methods 0.000 description 1
- 230000000717 retained effect Effects 0.000 description 1
- 238000012360 testing method Methods 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/90—Details of database functions independent of the retrieved data types
- G06F16/95—Retrieval from the web
- G06F16/953—Querying, e.g. by the use of web search engines
- G06F16/9536—Search customisation based on social or collaborative filtering
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/30—Information retrieval; Database structures therefor; File system structures therefor of unstructured textual data
- G06F16/36—Creation of semantic tools, e.g. ontology or thesauri
- G06F16/367—Ontology
-
- 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
-
- 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
- Y02D30/00—Reducing energy consumption in communication networks
- Y02D30/70—Reducing energy consumption in communication networks in wireless communication networks
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Databases & Information Systems (AREA)
- Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Computational Linguistics (AREA)
- Life Sciences & Earth Sciences (AREA)
- Health & Medical Sciences (AREA)
- Molecular Biology (AREA)
- Artificial Intelligence (AREA)
- Biomedical Technology (AREA)
- Biophysics (AREA)
- Evolutionary Computation (AREA)
- General Health & Medical Sciences (AREA)
- Animal Behavior & Ethology (AREA)
- Computing Systems (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)
- Information Transfer Between Computers (AREA)
Abstract
The invention provides a recommendation popularity deviation self-adaptive buffering method based on a knowledge graph, which comprises the following steps: grouping the items according to the popularity of the items; creating popularity nodes with the same quantity as the article groups; embedding popularity nodes into a knowledge graph with the relation of (entity s, relation of s and t, entity t) to obtain the relation including (i, popularity, p) n ) Wherein i represents an item and p represents a popularity knowledge graph of n Representing a user's preference for an item; creating a user preference graph; constructing an abnormal graph based on the popularity knowledge graph and the user preference graph; predicting the scoring of the user to the item in the abnormal image based on the graph neural network model; and obtaining a final score through the score of the user on the item and the popularity score, and recommending the item to the user through the final score. The invention is in the process of pushingWhen the popularity deviation problem in the recommendation system is solved, the knowledge graph is introduced for the first time, so that the recommendation is facilitated, the popularity problem can be effectively relieved, and the use experience of the user is considered by utilizing the preference of the user to the popularity.
Description
Technical Field
The invention belongs to the technical field of artificial intelligence, and particularly relates to a recommendation popularity deviation self-adaptive buffering method based on a knowledge graph.
Background
In the internet era, users can enjoy various services on various electronic platforms. However, as the number of users continues to increase, the problem of information overload becomes more serious, which makes it impossible for users to effectively search for their desired content. The Recommendation System (RS) is a powerful tool for people to solve information overload, and helps users to filter information and recommend information, products and the like which the users are interested in to the users according to the information needs, interests and the like of the users. With the rapid development of recommendation technology, personalized recommendation systems are continuously perfected and successfully applied to various fields, which are closely related to our lives, such as e-commerce, video and audio, news information and other fields, the personal images of the recommendation systems exist. The collaborative filtering is one of the most successful methods in the recommendation system, the rating of a certain user on a certain article can be predicted, and the preference of a certain user group is utilized to generate the recommendation. So accurately capturing the user's interests is the core of an effective recommendation system, which faces a major challenge of the Popularity bias problem (Popularity bias): collaborative filtering is generally biased towards recommending more popular items to users. The popularity bias can lead to the Long tail effect (Long tail phenomenon) in the well-known LastFM: the cumulative rating of the top 20% of the items with popularity is much higher than the remaining 80% of the long-tailed items. After training data with long tail effects, the model inherits the bias and expands the popularity bias by recommending popular items, further resulting in popular items becoming more popular while less popular items are not recommended.
Although popular items are often recommended well, such recommendations are not meaningful since popular items are well known. It is most valuable to the user to find out the appropriate recommendations from the non-popular items. The recommendation system should seek a balance between popular and unpopular items. Previous research has focused primarily on increasing the number of recommended unpopular (long tailed) items. For example, Jones proposes a re-weighting method to improve the extraction performance for small items. Chen Zhihong et al propose a method to correct the popularity bias by introducing regularization terms. In addition, there are methods that use scoring to reduce the proportion of popular items. These existing methods all increase the exposure of non-popular items. They largely overlook the user's interest in popular items, which can degrade the user experience.
Disclosure of Invention
In view of the above-mentioned shortcomings of the prior art, the present invention provides a knowledge-graph based recommendation popularity bias adaptive mitigation method to solve at least one of the deficiencies in the prior art.
In order to achieve the above objects and other related objects, the present invention provides a method for adaptively mitigating a deviation of a recommendation popularity based on a knowledge graph, comprising:
grouping the articles according to the popularity of the articles to obtain a plurality of article groups; wherein the popularity represents a popularity of the item;
creating popularity nodes with the same quantity as the article groups, and associating the article groups with the popularity nodes, wherein each article group corresponds to one popularity node;
embedding the popularity nodes into a knowledge graph with the relation of (entity s, relation of s and t, entity t) to obtain the relation including (i, popularity, p) n ) Wherein i represents an item, p n Representing a user's preference for an item;
respectively taking a user and an article as entity nodes, and taking the preference of the user to the article as an edge to create a user preference graph;
constructing an abnormal graph based on the popularity knowledge graph and the user preference graph;
predicting the scoring of the user to the item in the abnormal image based on the graph neural network model;
and obtaining a final score through the score of the user on the item and the popularity score, and recommending the item to the user through the final score.
Optionally, the information of the neighbor nodes from the target node i in the aggregation graph neural network model includes: message calculation and message aggregation;
the message computation includes:
and (3) calculating the Message (u, e, i) of the source node u passing through the edge e:
the message aggregation comprises:
computing heterogeneous mutual Attention (u, e, i) between a source node u and a target node i:
the corresponding messages from the source node are averaged using the attention vector as a weight and the output is updated.
Optionally, the Message (u, e, i) of the source node u passing through the edge e is calculated by the following formula:
wherein ,Mτ(u) Is a linear projection of the node type tau (u),
representing a matrix of messages, H (l-1) The output of layer l-1 of the heterogeneous graph neural network model is represented.
Optionally, the heterogeneous mutual Attention (u, e, i) between the source node u and the target node i is calculated by:
averaging corresponding messages from the source node u by using the attention vector as a weight, and updating output;
wherein N (t) represents a one-hop neighbor node of the source node u, Q τ(i) 、K τ(u) Is a linear projection, phi (e) is a type of edge,
is an edge-based attention matrix.
Optionally, the method further comprises:
the euclidean norm is used to align the relational embedding in the popularity knowledgegraph with the preference embedding in the preference graph.
Optionally, the loss function of the graph neural network model is:
wherein ,LHGT Loss function, L, representing a neural network model of the graph BPR Represents a pairwise BPR loss, L align Representing alignment loss, theta represents a model parameter set, alpha and lambda are used to constrain alignment loss and euclidean norm L, respectively 2 。
Optionally, the scoring the item by the user and the scoring the popularity, and the obtaining the final score includes:
the final score is calculated by the following function:
wherein ,
in order to be the final score,
indicating the user's rating of the item,
representing popularity score, popularity represents popularity, w represents weight, w ═ 1-rp (u), rp (u) is the proportion of popular items in the interaction data of user u.
Optionally, the user scores the item as:
as described above, the adaptive recommended popularity deviation mitigation method based on the knowledge graph of the invention has the following beneficial effects:
the invention discloses a recommendation popularity deviation self-adaptive buffering method based on a knowledge graph, which comprises the following steps: grouping the articles according to the popularity of the articles to obtain a plurality of article groups; wherein the popularity represents a popularity of the item; creating popularity nodes with the same quantity as the article groups, and associating the article groups with the popularity nodes, wherein each article group corresponds to one popularity node; embedding the popularity nodes into a knowledge graph with the relation of (entity s, relation of s and t, entity t) to obtain the relation including (i, popularity, p) n ) Wherein i represents an item and p represents a popularity knowledge graph of n Representing a user's preference for an item; respectively taking a user and an article as entity nodes, and taking the preference of the user to the article as an edge to create a user preference graph; constructing an abnormal graph based on the popularity knowledge graph and the user preference graph; predicting the scoring of the user to the item in the abnormal image based on the graph neural network model; and obtaining a final score through the score of the user on the item and the popularity score, and recommending the item to the user through the final score. When the popularity deviation problem in the recommendation system is processed, the knowledge graph is introduced for the first time, so that the recommendation is facilitated, the popularity problem can be effectively relieved, and the use experience of the user is considered by utilizing the preference of the user to the popularity.
Drawings
FIG. 1 is a schematic block diagram of a method for adaptive mitigation of recommended popularity bias based on knowledge-graph in an embodiment of the present invention;
fig. 2 is a flowchart of a method for adaptive mitigation of recommended popularity bias based on a knowledge-graph in an embodiment of the present invention.
Detailed Description
The embodiments of the present invention are described below with reference to specific embodiments, and other advantages and effects of the present invention will be easily understood by those skilled in the art from the disclosure of the present specification. The invention is capable of other and different embodiments and of being practiced or of being carried out in various ways, and its several details are capable of modification in various respects, all without departing from the spirit and scope of the present invention. It should be noted that the features in the following embodiments and examples may be combined with each other without conflict.
It should be noted that the drawings provided in the following embodiments are only for illustrating the basic idea of the present invention, and the drawings only show the components related to the present invention rather than being drawn according to the number, shape and size of the components in actual implementation, and the type, amount and proportion of each component in actual implementation can be changed freely, and the layout of the components can be more complicated.
When the two problems that the knowledge graph lacks popularity deviation knowledge and different users have different preferences on popular articles are solved, the invention provides a recommendation popularity deviation self-adaptive buffer method based on the knowledge graph, the knowledge graph is used for recommending the preference degree of the users on the popular articles, so that the popularity deviation problem in a recommendation system is reduced from the perspective of the users, and the user experience can be taken care of. The invention solves the main problems of popularity deviation in a recommendation system: that is, each user's preference for popular items is not uniform, and recommendations for decreasing popular items or increasing non-popular items cannot be indiscriminately.
In solving the problem of popularity deviation in a recommendation system, the adaptive mitigation method for recommendation popularity deviation based on knowledge graph provided by the invention, as shown in fig. 1, roughly comprises the following steps: 1) establishing a knowledge graph embedded with popular nodes; the popular nodes are used for enhancing entity relations in the knowledge graph; 2) constructing a user preference graph for characterizing user-item relationships using multiple preferences; it should be noted that, in the user preference modeling process, each preference is aligned with the relationship embedded in the knowledge graph of popularity nodes; 3) establishing a heterogeneous graph conversion model for learning user, article and fine-grained preference representation; 4) establishing a prediction model; and the prediction model generates a recommendation list with the popularity deviation removed by each user in an individualized way according to the preference of the users to the popular goods.
As shown in fig. 2, a method for adaptively relieving a deviation of a recommendation popularity based on a knowledge graph includes:
s100, grouping the articles according to the popularity of the articles to obtain a plurality of article groups; wherein the popularity represents a popularity of the item;
s200, creating popularity nodes with the same quantity as the article groups, and associating the article groups with the popularity nodes, wherein each article group corresponds to one popularity node;
s300, embedding the popularity nodes into a knowledge graph with relation (entity S, relation of S and t, entity t) to obtain a popularity knowledge graph with relation (i, popularity, pn), wherein i represents an article, and pn represents preference of a user on the article;
s400, respectively taking a user and an article as entity nodes, and taking the preference of the user to the article as a side to create a user preference graph;
s500, constructing an abnormal graph based on the popularity knowledge graph and the user preference graph;
s600, predicting the scoring of the user to the article in the abnormal image based on the image neural network model;
s700, obtaining a final score through the score of the user on the item and the popularity score, and recommending the item to the user through the final score.
In steps S100 to S300, the articles are first classified into K groups according to their popularity. Recreating K pieces called p i The popularity nodes of i e { 1.,. K } represent these groups and connect items to their corresponding nodes. Before creating the popularity nodes, the relationships in the knowledge-graph are (the relationship of the entities s, s and t, entity t). After the popular nodes are created, a new relationship (i, popularity, p) is introduced to the knowledge-graph n ) Thereby integrating popularity information into the knowledge-graph, as shown in (2) of FIG. 1Shown in the figure. The knowledge graph integrated with the Popularity information is expressed as a Popularity knowledge graph KGEPN (KG Embedded with Popularity nodes).
In step S400, a user preference graph is created with the user and the item as entity nodes and the preference of the user for the item as an edge. The purpose of the user preference graph is to capture multi-preference intuitive knowledge that affects the user's behavior, reflecting the commonality of all the user's behaviors, with possible preferences being different considerations of music attributes, such as artist, genre, or popularity, as exemplified by music recommendations; preference may therefore be given as a reason for the user to select an item. This embodiment models the user u-item i relationship with a granularity of preference, assuming that P is a set of preferences shared by all users, cuts the unified user u-item i relationship into | P | preferences, and decomposes each (u, interaction _ with, i) triplet into { (u, P, i) | P ∈ P } form, as shown in (1) of fig. 1, called a Preference Graph (PG).
In one embodiment, the euclidean norm is used to align the relationship embedding in the popularity knowledge graph with the preference embedding in the preference graph.
Because the preferences are expressed as hidden vectors, which is ambiguous for deeper understanding, the number of preferences in the user preference graph is set as the number of relationships in the popularity knowledge graph KGEPN, and the relationship information in KGEPN is converted into preferences. Specifically, the euclidean norm is utilized to align the preference embedding and the relationship embedding in KGEPN.
In the step S500-step S600, constructing an abnormal graph based on the popularity knowledge graph and the user preference graph;
in this embodiment, KGEPN and PG are combined into a new Heterogeneous Graph (HG).
On the basis of the Heterogeneous map, a Heterogeneous map converter (HGT) was developed on the basis of HG. The HGT is intended to aggregate information from neighbor nodes of a target node t in a graph neural network model, this process comprising: message calculation and message aggregation;
the message computation includes:
and (3) calculating the Message (u, e, i) of the source node u passing through the edge e:
the message aggregation comprises:
computing heterogeneous mutual Attention (u, e, i) between a source node u and a target node i:
the corresponding messages from the source node are averaged using the attention vector as a weight and the output is updated.
Specifically, the output of the (L) -th HGT layer is represented by H (L), and the depth of the HGT layer is represented by L. The message computation part incorporates a message matrix
To mitigate the distribution differences of different types of nodes and edges. Given a source node u and an edge e, the HGT can calculate the information that node u passes edge e by the following equation:
wherein ,Mτ(u) Is a unique linear projection of node type τ (u).
In the message aggregation part, heterogeneous mutual attention between the source node u and the target node i is firstly calculated by utilizing the HGT so as to control the influence of the source node u on the target node i. HGT uses a specific linear projection Q for each type of node τ(i) Or K τ(u) And using a different edge-based matrix for each edge type phi (e)
To model the distribution for maximum variance. And (c) representing a one-hop neighbor node of the source node u by using N (t), and further calculating the heterogeneous mutual Attention (s, e, t) of the heterogeneous graph:
the HGT re-uses the attention vector as a weight to average the corresponding messages from the source node and updates the output:
the paired BPR losses are selected in this embodiment to train the HGT by combining the alignment and BPR losses, by minimizing
Learning model parameters, wherein theta represents a model parameter set, and respectively constraining alignment loss and L by utilizing alpha and lambda 2 And (4) norm.
In step S700, a final score is obtained by scoring the item by the user and scoring the popularity, and the item is recommended to the user by the final score.
In particular, with the independent edge-based matrix for each edge type, user preferences may be more finely quantified. Given the final representation of user u and item i, a corresponding score between u and i is calculated for user u's preferences to determine the likelihood that u accepts i, as shown in (4) of FIG. 1. In the calculation, the influence of popularity on the recommendation result is controlled through a weight w, and a certain proportion of popular item recommendations are removed from the user u in a targeted manner according to the proportion of popular items in the personal profile of the user u.
In particular, recommendations for popular items are suppressed by the following function
Wherein w is 1-RP (u)
Rp (u) is the proportion of popular items (popular items-if the popularity of a commodity is 20% of all items, it is a popular item) in the interaction data of the user u, and if the user is very interested in the popular items, the popular item preference of the user is rarely deleted, so as to ensure the user experience, and vice versa. Therefore, the invention achieves the purposes of pertinently reducing the popularity deviation problem and giving consideration to the user experience according to different preference degrees of the users to popular articles.
Compared with recommendation algorithms BPRMF, LightGCN, KTUP, IPS-CN and ESAM, the recommendation popularity deviation self-adaptive mitigation method based on the knowledge graph (the method for short) uses the public data sets LastFM and DBbook-2014 in experiments. To ensure the quality of the data sets, a 5-core setup is used, i.e. at least 5 interacting users and items are retained and an item knowledge is constructed for each data set. The 80% of the items associated with each user were randomly selected to make up the training set and all remaining 20% were used as the test set and the cross-validation method was employed. The following are the attributes of the data set and experimental results for each method.
Where AWING-APS is a variant of AWING that removes a proportion of popular item recommendations when recommended. From the experimental results, it can be seen that the BPRMF model performs the worst on the two data sets, and although LightGCN performs better than BPRMF, the Average occupancy of the long-tailed Articles (APT) is low, indicating that a popular deviation is prevalent in the recommendation system. Secondly, the performance of KTUP is superior to the LightGCN and BPRMF models in all indicators, which indicates that introducing knowledge-maps not only facilitates recommendations, but also helps mitigate popularity bias. Furthermore, IPS-CN and ESAM provide a large improvement in APT compared to the three models described above, but do not maintain the ratio of popular to non-popular items, which may affect the user experience. The method, AWING, outperforms all comparison methods on both datasets in terms of Recall (Recall) and Normalized Discounted Cumulative Gain (NDCG), indicating that identifying fine-grained preferences facilitates recommendations. In addition, the AWING-APS of the method has the best performance in the Popularity preference consistency (delta GAP) of the user, the overall Diversity (AD) and the Average Percentage of Tail items (APT), verifies the meaning of removing a certain proportion of popular item recommendation to relieve the Popularity deviation, maintains the similar proportion of popular and non-popular items and ensures good user experience. Therefore, the method has certain advantages in reasonably relieving the popularity deviation problem.
The present invention also provides a storage medium storing a computer program which, when executed by a processor, performs the aforementioned video data processing method.
The present invention also provides an apparatus comprising:
a memory for storing a computer program;
a processor for executing the computer program stored by the memory to cause the apparatus to perform the aforementioned video data processing method.
The Processor may be a Central Processing Unit (CPU), other general purpose Processor, a Digital Signal Processor (DSP), an Application Specific Integrated Circuit (ASIC), a Field Programmable Gate Array (FPGA) or other Programmable logic device, discrete Gate or transistor logic device, discrete hardware component, etc. A general purpose processor may be a microprocessor or the processor may be any conventional processor or the like.
The memory may be an internal storage unit or an external storage device, such as a plug-in hard disk, a Smart Media Card (SMC), a Secure Digital Card (SD), a Flash memory Card (Flash Card), and the like. Further, the memory may also include both an internal storage unit and an external storage device. The memory is used for storing the computer program and other programs and data. The memory may also be used to temporarily store data that has been output or is to be output.
It will be apparent to those skilled in the art that, for convenience and brevity of description, only the above-mentioned division of the functional units and modules is illustrated, and in practical applications, the above-mentioned function distribution may be performed by different functional units and modules according to needs, that is, the internal structure of the apparatus is divided into different functional units or modules to perform all or part of the above-mentioned functions. Each functional unit and module in the embodiments may be integrated in one processing unit, or each unit may exist alone physically, or two or more units are integrated in one unit, and the integrated unit may be implemented in a form of hardware, or in a form of software functional unit. In addition, specific names of the functional units and modules are only for convenience of distinguishing from each other, and are not used for limiting the protection scope of the present application. The specific working processes of the units and modules in the system may refer to the corresponding processes in the foregoing method embodiments, and are not described herein again.
In the above embodiments, the descriptions of the respective embodiments have respective emphasis, and reference may be made to the related descriptions of other embodiments for parts that are not described or illustrated in a certain embodiment.
Those of ordinary skill in the art will appreciate that the various illustrative elements and algorithm steps described in connection with the embodiments disclosed herein may be implemented as electronic hardware, or combinations of computer software and electronic hardware. Whether such functionality is implemented as hardware or software depends upon the particular application and design constraints imposed on the implementation. Skilled artisans may implement the described functionality in varying ways for each particular application, but such implementation decisions should not be interpreted as causing a departure from the scope of the present invention.
In the embodiments provided in the present invention, it should be understood that the disclosed apparatus/terminal device and method may be implemented in other ways. For example, the above-described embodiments of the apparatus/terminal device are merely illustrative, and for example, the division of the modules or units is only one logical division, and there may be other divisions when actually implemented, for example, a plurality of units or components may be combined or integrated into another system, or some features may be omitted, or not executed. In addition, the shown or discussed mutual coupling or direct coupling or communication connection may be an indirect coupling or communication connection through some interfaces, devices or units, and may be in an electrical, mechanical or other form.
The units described as separate parts may or may not be physically separate, and parts displayed as units may or may not be physical units, may be located in one position, or may be distributed on multiple network units. Some or all of the units can be selected according to actual needs to achieve the purpose of the solution of the embodiment.
In addition, functional units in the embodiments of the present invention may be integrated into one processing unit, or each unit may exist alone physically, or two or more units are integrated into one unit. The integrated unit can be realized in a form of hardware, and can also be realized in a form of a software functional unit.
The integrated modules/units, if implemented in the form of software functional units and sold or used as separate products, may be stored in a computer readable storage medium. Based on such understanding, all or part of the flow of the method according to the embodiments of the present invention may also be implemented by a computer program, which may be stored in a computer-readable storage medium, and when the computer program is executed by a processor, the steps of the method embodiments may be implemented. Wherein the computer program comprises computer program code, which may be in the form of source code, object code, an executable file or some intermediate form, etc. The computer readable medium may comprise any entity or device capable of carrying the computer program code, a recording medium, a U disk, a removable hard disk, a magnetic disk, an optical disk, a computer Memory, a Read-Only Memory (ROM), a Random Access Memory (RAM), an electrical carrier signal, a telecommunications signal, a software distribution medium, etc.
The foregoing embodiments are merely illustrative of the principles and utilities of the present invention and are not intended to limit the invention. Any person skilled in the art can modify or change the above-mentioned embodiments without departing from the spirit and scope of the present invention. Accordingly, it is intended that all equivalent modifications or changes which may be made by those skilled in the art without departing from the spirit and scope of the present invention as defined in the appended claims.
Claims (8)
1. A recommendation popularity deviation adaptive mitigation method based on knowledge graph is characterized by comprising the following steps:
grouping the articles according to the popularity of the articles to obtain a plurality of article groups; wherein the popularity represents a popularity of the item;
creating popularity nodes with the same quantity as the article groups, and associating the article groups with the popularity nodes, wherein each article group corresponds to one popularity node;
embedding the popularity nodes into a knowledge graph with the relation of (entity s, relation of s and t, entity t) to obtain the relation including (i, popularity, p) n ) Wherein i represents an item and p represents a popularity knowledge graph of n Representing a user's preference for an item;
respectively taking a user and an article as entity nodes, and taking the preference of the user to the article as an edge to create a user preference graph;
constructing an abnormal graph based on the popularity knowledge graph and the user preference graph;
predicting the scoring of the user to the item in the abnormal image based on the graph neural network model;
and obtaining a final score through the scoring of the user on the item and the popularity score, and recommending the item to the user through the final score.
2. The adaptive knowledge-graph-based recommended popularity bias mitigation method of claim 1,
the information of the neighbor nodes from the target node i in the aggregation graph neural network model comprises the following steps: message calculation and message aggregation;
the message calculation includes:
and (3) calculating the Message (u, e, i) of the source node u passing through the edge e:
the message aggregation comprises:
computing heterogeneous mutual Attention (u, e, i) between the source node u and the target node i:
the corresponding messages from the source node are averaged using the attention vector as a weight and the output is updated.
3. The adaptive knowledge-graph-based recommended popularity bias mitigation method of claim 2, wherein,
and calculating the Message (u, e, i) of the source node u passing through the edge e by the following formula:
wherein ,Mτ(u) Is a linear projection of the node type tau (u),
representing a matrix of messages, H (l-1) The output of layer l-1 of the heterogeneous graph neural network model is represented.
4. The knowledge-graph based recommended popularity bias adaptive mitigation method of claim 3, wherein,
the heterogeneous mutual Attention (u, e, i) between the source node u and the target node i is calculated by:
averaging corresponding messages from the source node u by using the attention vector as a weight, and updating output;
wherein N (t) represents a one-hop neighbor node of the source node u, Q τ(i) 、K τ(u) Is a linear projection, phi (e) is a type of edge,
is an edge-based attention matrix.
5. The knowledge-graph-based recommended popularity bias adaptive mitigation method of claim 4, further comprising:
the euclidean norm is used to align the relational embedding in the popularity knowledgegraph with the preference embedding in the preference graph.
6. The knowledge-graph-based recommended popularity bias adaptive mitigation method of claim 5, wherein a penalty function of the graph neural network model is:
wherein ,LHGT Loss function, L, representing a neural network model of the graph BPR Represents a pairwise BPR loss, L align Representing alignment loss, theta represents a model parameter set, alpha and lambda are used to constrain alignment loss and euclidean norm L, respectively 2 。
7. The adaptive knowledge-graph-based recommendation popularity bias mitigation method of claim 6, wherein obtaining a final score through the user's scoring of items and popularity scoring comprises:
the final score is calculated by the following function:
wherein ,
in order to be the final score,
indicating the user's rating of the item,
representing popularity score, popularity represents popularity, w represents weight, w ═ 1-rp (u), rp (u) is the proportion of popular items in the interaction data of user u.
8. The adaptive knowledge-graph-based recommendation popularity bias mitigation method of claim 7, wherein the user scores items as:
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210531980.2A CN114912033B (en) | 2022-05-16 | 2022-05-16 | Recommendation popularity deviation self-adaptive relieving method based on knowledge graph |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210531980.2A CN114912033B (en) | 2022-05-16 | 2022-05-16 | Recommendation popularity deviation self-adaptive relieving method based on knowledge graph |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN114912033A true CN114912033A (en) | 2022-08-16 |
| CN114912033B CN114912033B (en) | 2023-04-21 |
Family
ID=82765843
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202210531980.2A Active CN114912033B (en) | 2022-05-16 | 2022-05-16 | Recommendation popularity deviation self-adaptive relieving method based on knowledge graph |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN114912033B (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN115809374A (en) * | 2023-02-13 | 2023-03-17 | 四川大学 | Method, system, device and storage medium for correcting mainstream deviation of recommendation system |
Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111339404A (en) * | 2020-02-14 | 2020-06-26 | 腾讯科技（深圳）有限公司 | Content popularity prediction method and device based on artificial intelligence and computer equipment |
| CN112989064A (en) * | 2021-03-16 | 2021-06-18 | 重庆理工大学 | Recommendation method for aggregating knowledge graph neural network and self-adaptive attention |
| CN113158033A (en) * | 2021-03-19 | 2021-07-23 | 浙江工业大学 | Collaborative recommendation model construction method based on knowledge graph preference propagation |
| CN113239181A (en) * | 2021-05-14 | 2021-08-10 | 廖伟智 | Scientific and technological literature citation recommendation method based on deep learning |
| CN113609306A (en) * | 2021-08-04 | 2021-11-05 | 北京邮电大学 | Social network link prediction method and system for resisting residual image variation self-encoder |
| CN114049930A (en) * | 2021-11-12 | 2022-02-15 | 东南大学 | Traditional Chinese medicine prescription relocation method based on heterogeneous network representation learning |
| CN114282122A (en) * | 2021-12-21 | 2022-04-05 | 郑州大学 | Efficient non-sampling graph convolution network recommendation method |
-
2022
- 2022-05-16 CN CN202210531980.2A patent/CN114912033B/en active Active
Patent Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111339404A (en) * | 2020-02-14 | 2020-06-26 | 腾讯科技（深圳）有限公司 | Content popularity prediction method and device based on artificial intelligence and computer equipment |
| CN112989064A (en) * | 2021-03-16 | 2021-06-18 | 重庆理工大学 | Recommendation method for aggregating knowledge graph neural network and self-adaptive attention |
| CN113158033A (en) * | 2021-03-19 | 2021-07-23 | 浙江工业大学 | Collaborative recommendation model construction method based on knowledge graph preference propagation |
| CN113239181A (en) * | 2021-05-14 | 2021-08-10 | 廖伟智 | Scientific and technological literature citation recommendation method based on deep learning |
| CN113609306A (en) * | 2021-08-04 | 2021-11-05 | 北京邮电大学 | Social network link prediction method and system for resisting residual image variation self-encoder |
| CN114049930A (en) * | 2021-11-12 | 2022-02-15 | 东南大学 | Traditional Chinese medicine prescription relocation method based on heterogeneous network representation learning |
| CN114282122A (en) * | 2021-12-21 | 2022-04-05 | 郑州大学 | Efficient non-sampling graph convolution network recommendation method |
Non-Patent Citations (4)
| Title |
|---|
| CHENGFENG XU 等: "Long- and short-term self-attention network for sequential recommendation" * |
| 周炫余: "联合知识图谱和时间特性的数学知识自动推荐方法" * |
| 张雪茹 等: "基于知识图谱的用户偏好推荐算法" * |
| 汪菁瑶 等: "用户行为序列个性化推荐研究综述" * |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN115809374A (en) * | 2023-02-13 | 2023-03-17 | 四川大学 | Method, system, device and storage medium for correcting mainstream deviation of recommendation system |
| CN115809374B (en) * | 2023-02-13 | 2023-04-18 | 四川大学 | Method, system, device and storage medium for correcting mainstream deviation of recommendation system |
Also Published As
| Publication number | Publication date |
|---|---|
| CN114912033B (en) | 2023-04-21 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Ning et al. | A comprehensive survey of neighborhood-based recommendation methods | |
| US11715132B2 (en) | Adaptive and recursive system and method | |
| US7493294B2 (en) | Mutually adaptive systems | |
| US8027943B2 (en) | Systems and methods for observing responses to invitations by users in a web-based social network | |
| US7526458B2 (en) | Adaptive recommendations systems | |
| US8850325B2 (en) | Scaling social networks | |
| Zhang et al. | Being “red” on the internet | |
| Shang et al. | Wisdom of the crowd: Incorporating social influence in recommendation models | |
| CN111611472A (en) | Binding recommendation method and system based on graph convolution neural network | |
| CN103119620A (en) | Action suggestions based on inferred social relationships | |
| CN107786943A (en) | A kind of tenant group method and computing device | |
| WO2022142001A1 (en) | Target object evaluation method based on multi-score card fusion, and related device therefor | |
| Alaimo et al. | Encoding the everyday: The infrastructural apparatus of social data | |
| CN115270001B (en) | Privacy protection recommendation method and system based on cloud collaborative learning | |
| Peng et al. | N-dimensional Markov random field prior for cold-start recommendation | |
| Wibawa et al. | Virtual influencers: is the persona trustworthy? | |
| CN108536753A (en) | The determination method and relevant apparatus of duplicate message | |
| Yang et al. | A Time‐Aware CNN‐Based Personalized Recommender System | |
| US20240046311A1 (en) | Adaptive and recursive system and method | |
| JP2013164704A (en) | Information processing apparatus, information processing method, and program | |
| CN114912033A (en) | Knowledge graph-based recommendation popularity deviation adaptive buffering method | |
| CN111047410B (en) | Recommendation method, recommendation device, terminal equipment and storage medium | |
| Chen et al. | Graph enhanced neural interaction model for recommendation | |
| CN117033754A (en) | Model processing method, device, equipment and storage medium for pushing resources | |
| CN113487389A (en) | Information recommendation method and device |
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 |