CN116664240B - Shopping behavior prediction method, system, storage medium and terminal based on graphic neural network - Google Patents
Shopping behavior prediction method, system, storage medium and terminal based on graphic neural network Download PDFInfo
- Publication number
- CN116664240B CN116664240B CN202310656430.8A CN202310656430A CN116664240B CN 116664240 B CN116664240 B CN 116664240B CN 202310656430 A CN202310656430 A CN 202310656430A CN 116664240 B CN116664240 B CN 116664240B
- Authority
- CN
- China
- Prior art keywords
- customer
- commodity
- graph
- characteristic information
- behavior prediction
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
- 238000000034 method Methods 0.000 title claims abstract description 35
- 238000013528 artificial neural network Methods 0.000 title claims abstract description 33
- 239000013598 vector Substances 0.000 claims abstract description 46
- 238000010586 diagram Methods 0.000 claims abstract description 14
- 230000002776 aggregation Effects 0.000 claims abstract description 10
- 238000004220 aggregation Methods 0.000 claims abstract description 10
- 230000006399 behavior Effects 0.000 claims description 36
- 239000011159 matrix material Substances 0.000 claims description 18
- 230000006870 function Effects 0.000 claims description 12
- 230000015654 memory Effects 0.000 claims description 12
- 230000004913 activation Effects 0.000 claims description 9
- 238000004590 computer program Methods 0.000 claims description 9
- 238000006116 polymerization reaction Methods 0.000 claims description 4
- 230000004931 aggregating effect Effects 0.000 claims description 3
- 230000005540 biological transmission Effects 0.000 claims description 3
- 238000004364 calculation method Methods 0.000 claims description 3
- 238000012545 processing Methods 0.000 claims description 3
- 238000010276 construction Methods 0.000 claims description 2
- 238000000605 extraction Methods 0.000 claims description 2
- 238000004458 analytical method Methods 0.000 description 6
- 238000011161 development Methods 0.000 description 5
- 238000013135 deep learning Methods 0.000 description 4
- 230000000694 effects Effects 0.000 description 4
- 238000005516 engineering process Methods 0.000 description 4
- 230000008901 benefit Effects 0.000 description 3
- 230000008569 process Effects 0.000 description 3
- 238000013459 approach Methods 0.000 description 2
- 238000001914 filtration Methods 0.000 description 2
- 230000009286 beneficial effect Effects 0.000 description 1
- 238000007405 data analysis Methods 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 238000002474 experimental method Methods 0.000 description 1
- 230000007246 mechanism Effects 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000003058 natural language processing Methods 0.000 description 1
- 238000003062 neural network model Methods 0.000 description 1
- 230000003287 optical effect Effects 0.000 description 1
- 238000011160 research Methods 0.000 description 1
- 238000012549 training Methods 0.000 description 1
- 238000012546 transfer Methods 0.000 description 1
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
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/90—Details of database functions independent of the retrieved data types
- G06F16/95—Retrieval from the web
- G06F16/953—Querying, e.g. by the use of web search engines
- G06F16/9535—Search customisation based on user profiles and personalisation
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/90—Details of database functions independent of the retrieved data types
- G06F16/95—Retrieval from the web
- G06F16/953—Querying, e.g. by the use of web search engines
- G06F16/9537—Spatial or temporal dependent retrieval, e.g. spatiotemporal queries
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/0464—Convolutional networks [CNN, ConvNet]
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/047—Probabilistic or stochastic networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/048—Activation functions
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
-
- 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
- G06Q10/00—Administration; Management
- G06Q10/04—Forecasting or optimisation specially adapted for administrative or management purposes, e.g. linear programming or "cutting stock problem"
-
- 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
Abstract
The invention relates to the field of recommendation systems, in particular to a shopping behavior prediction method, a shopping behavior prediction system, a shopping behavior prediction storage medium and a shopping behavior prediction terminal based on a graphic neural network. The method comprises the following steps: constructing a customer diagram and a commodity diagram: the method comprises the steps of initially describing characteristics of customers and commodities, converting customer data into a form of ternary vectors, and for the generated ternary vectors, taking a first element and a third element in the ternary vectors as nodes in a graph structure, wherein a second element serves as an edge connecting the two nodes, and respectively constructing a customer graph structure and a commodity graph structure through connection of the nodes and the edges; collecting graph characteristic information; customer and merchandise information aggregation. The invention can better utilize the complex relationship between the user and the commodity, fully mine the potential user interests and commodity correlation, and provide more accurate and personalized shopping recommendation experience for the user.
Description
Technical Field
The invention relates to the field of recommendation systems, in particular to a shopping behavior prediction method, a shopping behavior prediction system, a shopping behavior prediction storage medium and a shopping behavior prediction terminal based on a graphic neural network.
Background
First, early research focused mainly on traditional recommender approaches, such as collaborative filtering and content filtering. The method ignores the complex relationship between the user and the commodity in shopping behavior analysis, and potential user interests and commodity correlations cannot be fully mined. Subsequently, with the rise of graphic neural networks, researchers began to introduce them into shopping behavior analysis. The graph neural network provides a new approach that better models and analyzes the relationship network between the user and the commodity. Through the representation of the learning nodes and edges, the graph neural network can capture complex relations and potential modes in the graph, so that the accuracy and individuation degree of shopping recommendation are improved. In recent years, with the rapid development of deep learning technology, graphic neural networks have been widely used in shopping behavior analysis. Deep learning has the advantage that it can automatically learn feature representations and can process large-scale and complex data. The deep learning training graph neural network model can be utilized to effectively learn the representations of the user and the commodity, so that the shopping behaviors of the user can be better understood and predicted. Currently, researchers are further exploring the use of graph neural networks in shopping behavior analysis. They are constantly improving model structures and algorithms, introducing techniques of attention mechanisms, graph convolution, etc., to improve the performance and efficiency of the model. Meanwhile, shopping behavior analysis is combined with technologies in other fields, such as natural language processing, time sequence data analysis and the like, so that more comprehensive and accurate shopping recommendation is realized.
In summary, the technical development of analyzing shopping behavior by using the graph neural network has undergone the introduction from the conventional recommendation system to the graph neural network, and the continuous development of technologies such as deep learning and the like has been combined. This development of technology provides a more accurate and personalized solution to shopping recommendations and is expected to further advance the development of the shopping behavior analysis field in the future.
Disclosure of Invention
Aiming at the defects of the prior art, the invention develops a shopping behavior prediction method, a shopping behavior prediction system, a storage medium and a terminal based on a graphic neural network, and the graphic neural network is utilized to analyze shopping information, so that operators can analyze the shopping information conveniently, and store can be operated better.
The technical scheme for solving the technical problems is as follows: in one aspect, the invention provides a shopping behavior prediction method based on a graph neural network.
Shopping behavior prediction method based on graph neural network comprises the following steps of
S1, constructing a customer diagram and a commodity diagram:
the method comprises the steps of initially describing characteristics of customers and commodities, converting customer data into a form of ternary vectors, and for the generated ternary vectors, taking a first element and a third element in the ternary vectors as nodes in a graph structure, wherein a second element serves as an edge connecting the two nodes, and respectively constructing a customer graph structure and a commodity graph structure through connection of the nodes and the edges;
s2, collecting characteristic information of the graph:
extracting detail information from the characteristic information of customers and commodities by using a graph structure, and exploring key information in the characteristic information;
s3, aggregating customer and commodity information:
the updated graph structure is aggregated, and then the purchase probability of the commodity by the customer is predicted.
Based on the shopping behavior prediction method based on the graph neural network, S2 is specifically as follows:
1) The original eigenvector calculation formula of the ab side in the graph structure is as follows:n a is the original representation vector of node a in the graph structure, e ab Is the original representation vector of the ab-side, con (n a ，e ab ) The original representation vectors of the nodes and the edges are subjected to the same-scale stitching, and the scale of the stitched vector is d c ，d c Is defined as n a And e ab The values of the respective scales are added to each other to obtain W c Is a transposed matrix of dimension d x d c The value of d is the same as the scale of the vector to be output by the final model, and sigma represents an activation function Relu;
2) The formula of the characteristic information received by the ab side is as follows:
is the characteristic information received by edge ab in the t+1st cycle, { N (a) \b } indicates that the current set is the set of all nodes adjacent to node a excluding node b, { N (a) \b }, respectively>Is the characteristic information of the edge ka obtained by the t-th cycle; />The three vectors in brackets are spliced in the same scale, and the vector scale obtained after splicing is d o The value of which is n a 、n c 、/>Is obtained by vector scale addition; w (W) o Is a transposed matrix with a dimension d x d o The matrix scale d is adjusted by the model output requirement; sigma represents the activation function Relu;
3) The obtained characteristic information of the edge ab in the last cycleUpdating to the characteristic information obtained in the current cycle +.>W n Is a learning matrix, the matrix scale is d multiplied by d, the specific value of d is adjusted according to the output setting of the model, and sigma represents an activation function Relu;
4) After T times of circulation, the side ab characteristic information is obtained
Based on the shopping behavior prediction method based on the graph neural network, S3 is specifically as follows:
1) Customer graph structure G using customer-commodity relation matrix W L And commodity graph structure G L Processing is performed to know that the customer b purchased the commodity 4. Selecting node information of a customer b, node information of a commodity 4 and side information connected with the two nodes respectively to participate in aggregation operation;
2) Polymerization operation:
a. the customer relationship score and commodity relationship score are calculated as follows:
representing normalized customer score,/->Represents normalized commodity fraction, N (V) represents commodity graph G V The function h ()' is the inner product of the calculated vectors;
c. calculating final representative vector L for each customer and each commodity V And V L The method comprises the steps of carrying out a first treatment on the surface of the The specific formula is as follows:
x n representing initial representation information of the corresponding node;
3) The purchase probability of the customer L on the commodity V is calculated by the following specific formula:
V LV ＝(L V ，V L )。
in another aspect, the present invention provides a storage medium having stored thereon a computer program which, when executed by a processor, implements a method for predicting shopping behavior based on a graph neural network.
In yet another aspect, the present invention further provides a terminal, which is characterized by comprising: a processor and a memory; the memory is used for storing a computer program, and the processor is used for executing the computer program stored in the memory so as to enable the terminal to execute the shopping behavior prediction method based on the graph neural network.
The effects provided in the summary of the invention are merely effects of embodiments, not all effects of the invention, and the above technical solution has the following advantages or beneficial effects:
according to the shopping behavior prediction method, the shopping behavior prediction system, the storage medium, the construction of the terminal customer graph and the commodity graph, the graph characteristic information acquisition and the customer and commodity information aggregation based on the graph neural network, the purchasing probability of the customer to the commodity can be analyzed and predicted, the complex relationship between the user and the commodity can be better utilized, the potential user interests and commodity correlation can be fully mined, and therefore more accurate and personalized shopping recommendation experience is provided for the user.
Drawings
The accompanying drawings are included to provide a further understanding of the invention and are incorporated in and constitute a part of this specification, illustrate the invention and together with the embodiments of the invention, serve to explain the invention.
FIG. 1 is a schematic diagram of the structure of the present invention;
FIG. 2 is a flow chart of information collection according to the present invention;
FIG. 3 is a flow chart of the polymerization operation of the present invention.
Detailed Description
Other advantages and effects of the present invention will become apparent to those skilled in the art from the following disclosure, which describes the embodiments of the present invention with reference to specific examples. The invention may be practiced or carried out in other embodiments that depart from the specific details, and the details of the present description may be modified or varied from the spirit and scope of the present invention. It should be noted that the following embodiments and features in the embodiments may be combined with each other without conflict.
It should be noted that the illustrations provided in the following embodiments merely illustrate the basic concept of the present invention by way of illustration, and only the components related to the present invention are shown in the drawings and are not drawn according to the number, shape and size of the components in actual implementation, and the form, number and proportion of the components in actual implementation may be arbitrarily changed, and the layout of the components may be more complicated.
Example 1
A shopping behavior prediction method based on a graph neural network comprises the following steps of
S1, constructing a customer diagram and a commodity diagram:
a. the experiment is carried out by adopting a public information data set of a food supermarket, and the operation is carried out by adopting python and Neo4j map library, the user map structure is constructed by utilizing supermarket customer information, and the characteristics of 7 aspects of customers are selected to carry out initial description on the characteristics of the customers in the embodiment, wherein the 7 aspects are respectively: the city scale of the customer living, the annual income range of the customer, the age range of the customer, the academic level of the customer, the age of the customer, the monthly consumption level of the customer and the number of family personnel of the customer;
b. converting customer data into a form of a ternary vector, such as a customer number: 1023, city scale, two-line city, assuming 8000 customer data, 56000 ternary vectors are finally obtained;
c. 5 attributes of the commodity are selected to initially describe the characteristics of the commodity, and the 6 aspects are as follows: the price range of the commodity, whether the commodity can be sold for a second time, the specific name of the commodity, the class of the commodity and the brand to which the commodity belongs;
d. the data of commodity is also converted into a form of a ternary vector, such as commodity number: 653, commodity category, sports goods, assuming 900 commodity data, 4500 ternary vectors are finally obtained;
c. for the generated ternary vector, taking a first element and a third element in the ternary vector as nodes in the graph structure, and taking a second element as an edge connecting the two nodes, and respectively constructing a customer graph structure and a commodity graph structure shown in figure 1 through the connection of the nodes and the edges;
s2, collecting characteristic information of the graph:
extracting detail information from the characteristic information of customers and commodities by using a graph structure, exploring key information in the characteristic information, improving the judging capability of a model, and adopting a unidirectional circulation transfer graph characteristic information acquisition mode in order to acquire richer characteristic information and reduce redundant information in the information acquisition process;
s3, aggregating customer and commodity information:
the updated graph structure is aggregated, and then the purchase probability of the commodity by the customer is predicted.
Referring to fig. 2, fig. 2 illustrates a diagram information collection process of a customer diagram and a commodity diagram by taking a side ab as an example, an upper half of fig. 2 is an overall flow description, and a lower half of fig. 2 is an description of a specific structure of three modules, so that the detailed description S2 is as follows:
1) The original eigenvector calculation formula of the ab side in the graph structure is as follows:n a is the original representation vector of node a in the graph structure, e ab Is the original representation vector of the ab-side, con (n a ，e ab ) The original representation vectors of the nodes and the edges are subjected to the same-scale stitching, and the scale of the stitched vector is d c ，d c Is defined as n a And e ab The values of the respective scales are added to each other to obtain W c Is a transposed matrix of dimension d x d c ，d c The value of d can be adjusted according to the requirement, the value of d is the same as the scale of the vector to be output by the final model, and sigma represents an activation function Relu;
2) Taking the t+1 cycle as an example, the characteristic information received by the edge ab is the characteristic information of the edge adjacent to the node a and directed to the node a in the previous cycle, in order to reduce redundancy information, the adjacent nodes do not include the node b, and therefore the characteristic information of the edge ba is not involved in transmission, so in the figure, only the edge ca and the edge ea are involved in transmission of the characteristic information, and the formula of the characteristic information received by the edge ab is as follows:
is the characteristic information received by edge ab in the t+1st cycle, { N (a) \b } indicates that the current set is the set of all nodes adjacent to node a excluding node b, { N (a) \b }, respectively>Is the characteristic information of the edge ka obtained by the t-th cycle; />The three vectors in brackets are spliced in the same scale, and the vector scale obtained after splicing is d o The value of which is n a 、n c 、/>Is obtained by vector scale addition; w (W) o Is a transposed matrix with a dimension d x d o The matrix scale d is adjusted by the model output requirement; sigma represents the activation function Relu;
3) The obtained characteristic information of the edge ab in the last cycleUpdating to the characteristic information obtained in the current cycle +.>W n Is a learning matrix, the matrix scale is d multiplied by d, the specific value of d is adjusted according to the output setting of the model, and sigma represents an activation function Relu;
4) After T times of circulation, the side ab characteristic information is obtainedAfter the related operation of the customer graph and the commodity graph, the initial information of the nodes is kept unchanged, and the characteristic information of the edges already contains rich detail information transmitted by other edges.
The whole flow is shown in fig. 3, and s3 is specifically as follows:
1) Customer graph structure G using customer-commodity relation matrix W L And commodity graph structure G L Processing to acquire that the customer b purchases the commodity 4, selecting node information of the customer b, node information of the commodity 4 and side information connected with the two nodes respectively, and participating in aggregation operation;
2) Polymerization operation:
a. the customer relationship score and commodity relationship score are calculated as follows:
representing normalized advisorPassenger score, ->Represents normalized commodity fraction, N (V) represents commodity graph G V The function h ()' is the inner product of the calculated vectors;
c. calculating final representative vector L for each customer and each commodity V And V L The method comprises the steps of carrying out a first treatment on the surface of the The specific formula is as follows:
x n representing initial representation information of the corresponding node;
3) The purchase probability of the customer L on the commodity V is calculated by the following specific formula:
Y LV ＝(L V ，V L ) The trained network can predict the purchase probability of the commodity which is not purchased by the customer, and the merchant can determine the commodity incoming condition according to the number of registered customers and daily mobile customers so as to improve sales.
Example 2
A shopping behavior prediction system based on a graph neural network, the system comprising:
the building module is used for building a customer graph structure and a commodity graph structure;
the extraction module is used for extracting detailed information of the characteristic information of the customers and the commodities by using a graph structure;
and the aggregation module is used for conducting aggregation operation on the updated graph structure.
Example 3
A storage medium having stored thereon a computer program which when executed by a processor implements a method of predicting shopping behavior based on a graph neural network. The storage medium may be nonvolatile or volatile. Such as floppy disks, read-Only memories (ROMs), random access memories (Random Access Memory, RAMs), FLASH memories (FLASH), hard disks, optical disks, etc. of computers.
Example 4
A terminal, comprising: a processor and a memory; the memory is used for storing a computer program, and the processor is used for executing the computer program stored in the memory so as to enable the terminal to execute the shopping behavior prediction method based on the graph neural network. The processor may be a general-purpose processor including a central processor, a network processor, etc.; but may also be a digital signal processor, an application specific integrated circuit, a Field-programmable gate array (FPGA) or other programmable logic device, a discrete gate or transistor logic device, a discrete hardware component.
While the foregoing description of the embodiments of the present invention has been presented with reference to the drawings, it is not intended to limit the scope of the invention, but rather, it is apparent that various modifications or variations can be made by those skilled in the art without the need for inventive work on the basis of the technical solutions of the present invention.
Claims (4)
1. The shopping behavior prediction method based on the graph neural network is characterized by comprising the following steps of:
s1, constructing a customer diagram and a commodity diagram:
the method comprises the steps of initially describing characteristics of customers and commodities, converting customer data into a form of ternary vectors, and for the generated ternary vectors, taking a first element and a third element in the ternary vectors as nodes in a graph structure, wherein a second element serves as an edge connecting the two nodes, and respectively constructing a customer graph structure and a commodity graph structure through connection of the nodes and the edges;
s2, collecting characteristic information of the graph:
extracting detail information of the characteristic information of the customers and the commodities by using a graph structure, and exploring key information in the characteristic information;
s3, aggregating customer and commodity information:
the updated graph structure is subjected to aggregation operation, and then the purchasing probability of the commodity by a customer is predicted;
s2, specifically, the method comprises the following steps:
1) The original feature vector of the ab side in the graph structure is the side which is connected with the a node and the b node and points to the node a, and the calculation formula is as follows:n a is the original representation vector of node a in the graph structure, e ab Is the original representation vector of the ab-side, con (n a ，e ab ) The original representation vectors of the nodes and the edges are subjected to the same-scale stitching, and the scale of the stitched vector is d c ，d c Is defined as n a And e ab The values of the respective scales are added to each other to obtain W c Is a transposed matrix of dimension d x d c The value of d is the same as the scale of the vector to be output by the final model, and sigma represents an activation function Relu;
2) In the t+1 cycle, the characteristic information received by the ab side is the characteristic information of the side adjacent to the node a and directed to the node a in the t cycle, the adjacent nodes do not comprise the node b, the characteristic information of the ba side does not participate in the transmission, and the formula of the characteristic information received by the ab side is as follows:
is the characteristic information received by edge ab in the t+1 cycle, where { N (a) \b } indicates that the current set is a set of all nodes adjacent to node a excluding node b, { N (a) \b }, a combination of all nodes adjacent to node a, and }>Is the characteristic information of the edge ka obtained by the t-th cycle; />The three vectors in brackets are spliced in the same scale, and the vector scale obtained after splicing is d o The value of which is n a 、n k 、/>Is obtained by vector scale addition; wo is the transposed matrix with dimensions d x d o The matrix scale d is adjusted by the model output requirement; sigma represents the activation function Relu;
3) The obtained characteristic information of the edge ab in the last cycleUpdated to the characteristic information obtained in the current cycle W n Is a learning matrix, the matrix scale is d multiplied by d, the specific value of d is adjusted according to the output setting of the model, and sigma represents an activation function Relu;
4) After T times of circulation, the side ab characteristic information is obtainedAfter the customer diagram and the commodity diagram are operated, the initial information of the nodes is kept unchanged, and the characteristic information of the edges already contains rich detail information transmitted by other edges;
wherein, S3 is specifically as follows:
1) Processing the customer graph structure GL and the commodity graph structure GV by utilizing a customer-commodity relation matrix W to acquire that a certain commodity is purchased by a customer, selecting node information of the customer, node information of the commodity and characteristic information of edges connected with the two nodes respectively to participate in aggregation operation;
2) Polymerization operation:
a. calculating customer relationship scoresScore->Representing a normalized score of the customer,represents normalized commodity fraction, N (V) represents commodity graph G V N (L) represents a customer graph G L The function h ()' is the inner product of the calculated vectors;
b. calculating final representative vector L for each customer and each commodity V And V L The method comprises the steps of carrying out a first treatment on the surface of the The specific formula is as follows:x n representing initial representation information of the corresponding node;
3) The purchase probability of the customer L on the commodity V is calculated by the following specific formula:
Y LV ＝(L V ，V L )。
2. a shopping behavior prediction system based on a graph neural network, which is characterized by comprising:
the construction module is used for constructing a customer graph structure and a commodity graph structure and realizing the S1 of the shopping behavior prediction method based on the graph neural network of claim 1;
the extraction module is used for extracting detailed information of characteristic information of customers and commodities by using a graph structure and is used for realizing S2 of the shopping behavior prediction method based on the graph neural network of claim 1;
the aggregation module is used for realizing the S3 of the shopping behavior prediction method based on the graph neural network of claim 1 by carrying out aggregation operation on the updated graph structure;
the shopping behavior prediction system based on the graphic neural network is integrally used for realizing the shopping behavior prediction method based on the graphic neural network of claim 1.
3. A storage medium having stored thereon a computer program, which when executed by a processor implements the neural network-based shopping behavior prediction method of claim 1.
4. A terminal, comprising: a processor and a memory; the memory is used for storing a computer program, and the processor is used for executing the computer program stored in the memory, so that the terminal executes the shopping behavior prediction method based on the graph neural network as claimed in claim 1.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202310656430.8A CN116664240B (en) | 2023-06-05 | 2023-06-05 | Shopping behavior prediction method, system, storage medium and terminal based on graphic neural network |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202310656430.8A CN116664240B (en) | 2023-06-05 | 2023-06-05 | Shopping behavior prediction method, system, storage medium and terminal based on graphic neural network |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN116664240A CN116664240A (en) | 2023-08-29 |
| CN116664240B true CN116664240B (en) | 2024-02-02 |
Family
ID=87720325
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202310656430.8A Active CN116664240B (en) | 2023-06-05 | 2023-06-05 | Shopping behavior prediction method, system, storage medium and terminal based on graphic neural network |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN116664240B (en) |
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112613602A (en) * | 2020-12-25 | 2021-04-06 | 神行太保智能科技(苏州)有限公司 | Recommendation method and system based on knowledge-aware hypergraph neural network |
| CN113313164A (en) * | 2021-05-27 | 2021-08-27 | 复旦大学附属肿瘤医院 | Digital pathological image classification method and system based on superpixel segmentation and image convolution |
| CN114637863A (en) * | 2022-03-31 | 2022-06-17 | 东北大学秦皇岛分校 | Knowledge graph recommendation method based on propagation |
-
2023
- 2023-06-05 CN CN202310656430.8A patent/CN116664240B/en active Active
Patent Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112613602A (en) * | 2020-12-25 | 2021-04-06 | 神行太保智能科技(苏州)有限公司 | Recommendation method and system based on knowledge-aware hypergraph neural network |
| CN113313164A (en) * | 2021-05-27 | 2021-08-27 | 复旦大学附属肿瘤医院 | Digital pathological image classification method and system based on superpixel segmentation and image convolution |
| CN114637863A (en) * | 2022-03-31 | 2022-06-17 | 东北大学秦皇岛分校 | Knowledge graph recommendation method based on propagation |
Non-Patent Citations (1)
| Title |
|---|
| 基于双知识图谱和图卷积神经网络的商品推荐研究;李贤宗;《中国优秀硕士学位论文全文数据库信息科技辑》(第01期);正文第2-4章 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN116664240A (en) | 2023-08-29 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| US11586880B2 (en) | System and method for multi-horizon time series forecasting with dynamic temporal context learning | |
| Pan et al. | Study on convolutional neural network and its application in data mining and sales forecasting for E-commerce | |
| CN112819023B (en) | Sample set acquisition method, device, computer equipment and storage medium | |
| CN111784455A (en) | Article recommendation method and recommendation equipment | |
| CN114372573B (en) | User portrait information recognition method and device, computer equipment and storage medium | |
| CN112487199B (en) | User characteristic prediction method based on user purchasing behavior | |
| CN112148973B (en) | Data processing method and device for information push | |
| WO2022252458A1 (en) | Classification model training method and apparatus, device, and medium | |
| CN112085615A (en) | Method and device for training graph neural network | |
| CN113761250A (en) | Model training method, merchant classification method and device | |
| Liu et al. | Research of animals image semantic segmentation based on deep learning | |
| Zhang et al. | Learning to detect salient object with multi-source weak supervision | |
| CN115860880B (en) | Personalized commodity recommendation method and system based on multi-layer heterogeneous graph convolution model | |
| CN115797781A (en) | Crop identification method and device, computer equipment and storage medium | |
| CN112131261A (en) | Community query method and device based on community network and computer equipment | |
| US20240037133A1 (en) | Method and apparatus for recommending cold start object, computer device, and storage medium | |
| US20220027722A1 (en) | Deep Relational Factorization Machine Techniques for Content Usage Prediction via Multiple Interaction Types | |
| CN115309997B (en) | Commodity recommendation method and device based on multi-view self-coding features | |
| CN116664240B (en) | Shopping behavior prediction method, system, storage medium and terminal based on graphic neural network | |
| CN116910357A (en) | Data processing method and related device | |
| CN108229572B (en) | Parameter optimization method and computing equipment | |
| CN115631008B (en) | Commodity recommendation method, device, equipment and medium | |
| CN115345291A (en) | Graph processing method and related device | |
| CN115994331A (en) | Message sorting method and device based on decision tree | |
| CN115344794A (en) | Scenic spot recommendation method based on knowledge map semantic embedding |
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 | ||
| TA01 | Transfer of patent application right | ||
| TA01 | Transfer of patent application right |
Effective date of registration: 20240112 Address after: No. 101, Building 20, Zone 3, Zhonghai International Community, Shizhong District, Jinan City, Shandong Province, 250000 Applicant after: Xie Bicong Address before: 250101 1101h, 11th floor, building A4 - (3), Hanyu financial and business center, No. 7000, Jingshi Road, high tech Zone, Jinan, Shandong Province Applicant before: Shandong RuiZheng Information Technology Co.,Ltd. |
|
| GR01 | Patent grant | ||
| GR01 | Patent grant |