CN112800207A - Commodity information recommendation method and device and storage medium - Google Patents
Commodity information recommendation method and device and storage medium Download PDFInfo
- Publication number
- CN112800207A CN112800207A CN202110041795.0A CN202110041795A CN112800207A CN 112800207 A CN112800207 A CN 112800207A CN 202110041795 A CN202110041795 A CN 202110041795A CN 112800207 A CN112800207 A CN 112800207A
- Authority
- CN
- China
- Prior art keywords
- matrix
- recommendation
- value
- commodity
- 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 36
- 239000011159 matrix material Substances 0.000 claims abstract description 214
- 238000012549 training Methods 0.000 claims abstract description 86
- 238000012360 testing method Methods 0.000 claims abstract description 15
- 239000013598 vector Substances 0.000 claims description 45
- 230000003993 interaction Effects 0.000 claims description 35
- 230000006870 function Effects 0.000 claims description 22
- 238000012545 processing Methods 0.000 claims description 19
- 230000002452 interceptive effect Effects 0.000 claims description 14
- 238000001228 spectrum Methods 0.000 claims description 9
- 238000004422 calculation algorithm Methods 0.000 claims description 6
- 230000002457 bidirectional effect Effects 0.000 claims description 5
- 238000004364 calculation method Methods 0.000 claims description 3
- 230000000694 effects Effects 0.000 abstract description 9
- 238000005065 mining Methods 0.000 description 6
- 230000009286 beneficial effect Effects 0.000 description 3
- 238000004590 computer program Methods 0.000 description 3
- 238000010586 diagram Methods 0.000 description 2
- ORILYTVJVMAKLC-UHFFFAOYSA-N Adamantane Natural products C1C(C2)CC3CC1CC2C3 ORILYTVJVMAKLC-UHFFFAOYSA-N 0.000 description 1
- 238000007418 data mining Methods 0.000 description 1
- 230000003287 optical effect Effects 0.000 description 1
- 238000012216 screening 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
- G06F16/30—Information retrieval; Database structures therefor; File system structures therefor of unstructured textual data
- G06F16/33—Querying
- G06F16/335—Filtering based on additional data, e.g. user or group profiles
-
- 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
- G06F16/33—Querying
- G06F16/3331—Query processing
- G06F16/334—Query execution
- G06F16/3346—Query execution using probabilistic model
-
- 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
- G06N3/04—Architecture, e.g. interconnection topology
-
- 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
- G06N3/084—Backpropagation, e.g. using gradient descent
-
- 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
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- Data Mining & Analysis (AREA)
- Computational Linguistics (AREA)
- Databases & Information Systems (AREA)
- Business, Economics & Management (AREA)
- Life Sciences & Earth Sciences (AREA)
- Finance (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Computing Systems (AREA)
- Molecular Biology (AREA)
- Accounting & Taxation (AREA)
- General Health & Medical Sciences (AREA)
- Health & Medical Sciences (AREA)
- Evolutionary Computation (AREA)
- Artificial Intelligence (AREA)
- Biomedical Technology (AREA)
- Biophysics (AREA)
- Strategic Management (AREA)
- Probability & Statistics with Applications (AREA)
- Marketing (AREA)
- General Business, Economics & Management (AREA)
- Economics (AREA)
- Development Economics (AREA)
- Animal Behavior & Ethology (AREA)
- Information Retrieval, Db Structures And Fs Structures Therefor (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The invention provides a commodity information recommendation method, a commodity information recommendation device and a storage medium, wherein the method comprises the following steps: importing a commodity information data set, and dividing the data set of the commodity information data set to obtain a commodity information training set and a commodity information testing set; and carrying out vectorization analysis on the commodity information training set to obtain a knowledge map matrix and a training set matrix. The method can effectively dig out implicit characteristics among data, further realize accurate recommendation, can be better distinguished mathematically, further facilitate digging more effective information in the graph, can realize quick recommendation and accurate recommendation under a specific scene, has certain generalization, has certain effect on different types of data, can effectively improve the recommendation accuracy, has good robustness, can reasonably and accurately recommend commodities to users, and realizes the commodity recommendation of the cold start to the users and the improvement of the reliability and the accuracy of the recommendation.
Description
Technical Field
The invention mainly relates to the technical field of data mining, in particular to a commodity information recommendation method, a commodity information recommendation device and a storage medium.
Background
In a recommendation system based on the knowledge graph, a plurality of researchers neglect the sufficiency of relation expression among nodes in a graph network and implicit associated information among commodity nodes, so that effective information in the knowledge graph cannot be sufficiently mined, and the recommendation accuracy rate is difficult to improve. The recommendation system based on the knowledge graph has a good effect in solving the cold start problem, but researchers of the recommendation system often ignore whether relationship vectors in the knowledge graph can accurately represent relationship information among nodes in the knowledge graph, and meanwhile, structural information among the nodes is difficult to dig out deeply.
Disclosure of Invention
The invention aims to solve the technical problem of the prior art and provides a commodity information recommendation method, a commodity information recommendation device and a storage medium.
The technical scheme for solving the technical problems is as follows: a commodity information recommendation method comprises the following steps:
importing a commodity information data set, and dividing the data set into a commodity information training set and a commodity information testing set;
vectorization analysis is carried out on the commodity information training set to obtain a knowledge map matrix and a training set matrix;
analyzing and processing the knowledge map matrix and the training set matrix to obtain a knowledge map recommendation value, an original recommendation matrix and an interaction recommendation value;
calculating loss values of the training set matrix, the knowledge graph recommendation value, the original recommendation matrix and the interaction recommendation value through a loss function to obtain a total loss value;
inputting the total loss value into a preset optimizer, and performing back propagation on the total loss value through the preset optimizer to obtain a commodity recommendation model;
and inputting the commodity information test set into the commodity information model, and obtaining a commodity recommendation result according to the commodity information model.
Another technical solution of the present invention for solving the above technical problems is as follows: an article information recommendation device comprising:
the data set dividing module is used for importing a commodity information data set, and dividing the data set into the commodity information data set to obtain a commodity information training set and a commodity information testing set;
the vectorization analysis module is used for carrying out vectorization analysis on the commodity information training set to obtain a knowledge map matrix and a training set matrix;
the analysis processing module is used for analyzing and processing the knowledge map matrix and the training set matrix to obtain a knowledge map recommended value, an original recommended matrix and an interaction recommended value;
the loss value calculation module is used for calculating the loss values of the training set matrix, the knowledge graph recommendation value, the original recommendation matrix and the interactive recommendation value through a loss function to obtain a total loss value;
the back propagation module is used for inputting the total loss value into a preset optimizer and performing back propagation on the total loss value through the preset optimizer to obtain a commodity recommendation model;
and the recommendation result obtaining module is used for inputting the commodity information test set into the commodity information model and obtaining a commodity recommendation result according to the commodity information model.
The invention has the beneficial effects that: the commodity information training set and the commodity information testing set are obtained by dividing the data set of the commodity information data set, a potential consumer group model can be constructed based on the social relation and a small amount of node purchasing information, potential consumer groups of specific commodities or commodity types are defined, and calculates the probability of the consumer purchasing the commodity, and can effectively dig out the implicit characteristics among the data, thereby realizing accurate recommendation, being capable of being better distinguished mathematically, further being beneficial to mining more effective information in the map, being capable of realizing rapid recommendation and accurate recommendation under specific scenes, and having certain generalization, has certain effect on different types of data, can effectively improve the recommendation accuracy, the commodity recommending method has good robustness, can reasonably and accurately recommend commodities to the user, realizes commodity recommendation to the user by cold start, and improves the reliability and accuracy of recommendation.
Drawings
Fig. 1 is a schematic flow chart of a commodity information recommendation method according to an embodiment of the present invention;
fig. 2 is a block diagram of a product information recommendation device according to an embodiment of the present invention.
Detailed Description
The principles and features of this invention are described below in conjunction with the following drawings, which are set forth by way of illustration only and are not intended to limit the scope of the invention.
Fig. 1 is a schematic flow chart of a commodity information recommendation method according to an embodiment of the present invention.
As shown in fig. 1, a method for recommending commodity information includes the following steps:
importing a commodity information data set, and dividing the data set into a commodity information training set and a commodity information testing set;
vectorization analysis is carried out on the commodity information training set to obtain a knowledge map matrix and a training set matrix;
analyzing and processing the knowledge map matrix and the training set matrix to obtain a knowledge map recommendation value, an original recommendation matrix and an interaction recommendation value;
calculating loss values of the training set matrix, the knowledge graph recommendation value, the original recommendation matrix and the interaction recommendation value through a loss function to obtain a total loss value;
inputting the total loss value into a preset optimizer, and performing back propagation on the total loss value through the preset optimizer to obtain a commodity recommendation model;
and inputting the commodity information test set into the commodity information model, and obtaining a commodity recommendation result according to the commodity information model.
Preferably, the preset optimizer may be an Adam optimizer.
It should be understood that the recommendation value is technically referred to as a strategy, and the nodes in the graph are described by introducing a score strategy, so that the association information between the user and the commodity is mined.
It should be understood that the commodity information training set includes a plurality of user nodes, a plurality of commodity node description nodes, relationships between a plurality of nodes, and a plurality of user purchase data.
In the above embodiment, the commodity information training set and the commodity information testing set are obtained by dividing the data set of the commodity information data set, so that the potential consumer group model can be constructed based on the social relationship and a small amount of node purchase information, the potential consumer group of a specific commodity or commodity category is defined, and calculates the probability of the consumer purchasing the commodity, and can effectively dig out the implicit characteristics among the data, thereby realizing accurate recommendation, being capable of being better distinguished mathematically, further being beneficial to mining more effective information in the map, being capable of realizing rapid recommendation and accurate recommendation under specific scenes, and having certain generalization, has certain effect on different types of data, can effectively improve the recommendation accuracy, the commodity recommending method has good robustness, can reasonably and accurately recommend commodities to the user, realizes commodity recommendation to the user by cold start, and improves the reliability and accuracy of recommendation.
Optionally, as an embodiment of the present invention, the commodity information training set includes a commodity node data set and a user node data set, and the process of performing vectorization analysis on the commodity information training set to obtain a knowledge graph matrix and a training set matrix includes:
constructing a knowledge graph of the commodity node data set and the user node data set to obtain a training set knowledge graph;
respectively carrying out data vectorization on the training set knowledge graph, the commodity node data set and the user node data set to obtain a knowledge graph matrix corresponding to the training set knowledge graph, a commodity node matrix corresponding to the commodity node data set and a user node matrix corresponding to the user node data set;
and obtaining a training set matrix according to the commodity node matrix and the user node matrix.
It should be understood that the knowledge-graph refers to triple data composed of nodes and edges, which is a form of data other than Euclidean space. The original data are described in the form of the map, so that the characteristics of the non-European data in the data can be more accurately mined. Meanwhile, the nodes in the knowledge graph can be more accurately depicted through learning of the relationship among the nodes, so that more effective associated information among data is mined, cold start recommendation is facilitated to be realized, and accuracy is improved.
It should be understood that the vectorization of the data aims to make the description information of the user, the commodity and the commodity in the recommendation data have vector space characteristics, and further, the relationship between the user, the commodity and the commodity description information and other data characteristics can be more obviously distinguished in physical space through the relationship between the vectors.
In the embodiment, the knowledge graph of the training set is constructed by constructing the knowledge graph of the commodity node data set and the knowledge graph of the user node data set, the knowledge graph matrix is obtained by vectorizing the data of the training set knowledge graph, the commodity node data set and the user node data set respectively, implicit characteristics among data can be effectively mined, accurate recommendation is achieved, better mathematical distinction can be achieved, further, more effective information in graph mining can be beneficially mined, rapid recommendation can be achieved, accurate recommendation can be achieved under a specific scene, certain generalization performance is achieved, and certain effects are achieved on different types of data.
Optionally, as an embodiment of the present invention, the process of analyzing and processing the knowledge graph matrix and the training set matrix to obtain a knowledge graph recommendation value, an original recommendation matrix, and an interaction recommendation value includes:
utilizing a Bi-Interaction bidirectional interactive learning network to perform learning processing of a training set matrix on the knowledge map matrix, the commodity node matrix and the user node matrix to obtain a user information matrix and a commodity information matrix;
calculating an original recommendation matrix for the user information matrix and the commodity information matrix through a first formula to obtain the original recommendation matrix, wherein the first formula is as follows:
wherein, e'uIs a user information matrix, e'iIs a commodity information matrix, yuiA matrix is originally recommended;
calculating an interactive recommendation value of the user information matrix and the commodity information matrix through a second formula to obtain the interactive recommendation value, wherein the second formula is as follows:
where, σ is a loss function,is dot product, e'uIs a user information matrix, e'iIs a commodity information matrix, WaAs a weight matrix, ScorecroRecommending a value for the interaction;
and calculating a knowledge graph recommendation value of the knowledge graph matrix by using a score strategy algorithm to obtain the knowledge graph recommendation value.
It should be understood that effective information between nodes can be accurately learned and extracted through interaction between users and commodities, between commodities and between users and users.
Specifically, a next user information matrix and a next commodity information matrix in the graph network are obtained through a ninth formula, where the ninth formula is:
eu+1=yui·wVV+yui Τ·wEV,
ei+1=yui·wVE+yui Τ·wEE
wherein e hereu+1,ei+1Respectively representing a next user information matrix and a next commodity information matrix in the graph network.
In the embodiment, the knowledge graph recommendation value is obtained by analyzing and processing the knowledge graph matrix, and the interaction recommendation value is obtained by analyzing and processing the training set matrix, so that the recommendation accuracy can be effectively improved, the robustness is good, the commodity can be reasonably and accurately recommended to the user, the commodity recommendation to the user by cold start is realized, and the recommendation reliability and accuracy are improved.
Optionally, as an embodiment of the present invention, the learning processing of the training set matrix on the knowledge graph matrix, the commodity node matrix, and the user node matrix by using the Bi-Interaction bidirectional Interaction learning network to obtain the user information matrix and the commodity information matrix includes:
calculating a user information matrix for the knowledge graph spectrum matrix and the user node matrix through a third formula to obtain a user information matrix, wherein the third formula is as follows:
wherein, yg=σ(W1(e+ek))+σ(W2(e·ek)),
where σ is the loss function, W1And W2Are all weight matrices, e'uAs a matrix of user information, ekIs a central matrix, k is k nodes nearest to the central node, l is the number of high-order layers, etIs a knowledge graphA tail node matrix in the spectrum matrix, e is a head node matrix in the knowledge map matrix, pi (h, r, t) is the relationship among a head node vector h, a tail node vector t and a relationship vector r among nodes,the vector is the l user node vector in the user node matrix;
calculating a commodity information matrix for the knowledge map matrix and the commodity node matrix through a fourth formula to obtain a commodity information matrix, wherein the fourth formula is as follows:
wherein, yg=σ(W1(e+ek))+σ(W2(e·ek)),
where σ is the loss function, W1And W2Are all weight matrices, e'iIs a commodity information matrix, ekIs a central matrix, k is k nodes nearest to the central node, l is the number of high-order layers, etIs a tail node matrix in the knowledge graph spectrum matrix, e is a head node matrix in the knowledge graph spectrum matrix, pi (h, r, t) is the relationship among a head node vector h, a tail node vector t and a relationship vector r among nodes,is the ith commodity node vector in the commodity node matrix.
It should be understood that the weight matrix W is a function of the user information matrix and the merchandise information matrix2And updating to obtain an updated weight matrix.
Specifically, by learning a relational expression between nodes and paying attention to an important expression, a tenth expression is expressed by:
wherein N ishIs a central node, exp is an index;
updating node information by a tenth formula, wherein the eleventh formula is as follows:
in the embodiment, the user information matrix and the commodity information matrix are obtained by learning and processing the knowledge map matrix, the commodity node matrix and the training set matrix of the user node matrix by using the Bi-Interaction bidirectional interactive learning network, a potential consumer group model can be constructed based on the community relation and a small amount of node purchase information, a potential consumer group of a specific commodity or commodity type is defined, the probability of purchasing the commodity by a consumer is calculated, the implicit characteristics among data can be effectively mined, and accurate recommendation is further achieved.
Optionally, as an embodiment of the present invention, the calculating a knowledge graph recommendation value for the knowledge graph matrix by using a score policy algorithm includes:
calculating a knowledge graph recommendation value of the knowledge graph matrix through a fifth formula to obtain the knowledge graph recommendation value, wherein the fifth formula is as follows:
Scoreg=||polar(h)W+polar(r)-polar(t)W||,
wherein gamma is variable, eposion and hidden are both preset constants, h is head node vector, t is tail node vector, r is relationship vector between nodes, e0For node vectors in a knowledge-graph matrix, knowledge-graph node vector e0Comprises a head node vector h, a tail node vector t and a relation vector r between nodes, ScoregAnd recommending values for the knowledge graph.
It should be understood that the relationship between the nodes is described through a score strategy algorithm, so that the classification result is more accurate.
In the embodiment, the knowledge graph recommended value is obtained by calculating the knowledge graph recommended value of the knowledge graph matrix by using the score strategy algorithm, so that better mathematical distinction can be realized, further more effective information in graph mining is facilitated, quick recommendation can be realized, accurate recommendation can be realized in a specific scene, certain generalization performance is realized, certain effects are realized on different types of data, meanwhile, the recommendation accuracy can be effectively improved, good robustness is realized, commodities can be reasonably and accurately recommended to a user, commodity recommendation for the user by cold start is realized, and the recommendation reliability and accuracy are improved.
Optionally, as an embodiment of the present invention, the calculating a loss value of the training set matrix, the knowledge-graph recommendation value, the original recommendation matrix, and the interaction recommendation value through a loss function to obtain a total loss value includes:
calculating a recommendation loss value of the training set matrix and the original recommendation matrix to obtain a recommendation loss value;
calculating a knowledge map loss value of the knowledge map recommendation value and the interaction recommendation value to obtain a knowledge map loss value;
calculating a total loss value of the recommended loss value and the knowledge graph loss value through a sixth formula to obtain the total loss value, wherein the sixth formula is as follows:
Loss=Lossrs+Lossgrs,
therein, LossrsTo recommend Loss value, LossgrsThe knowledge-graph Loss value and the Loss is the total Loss value.
In the embodiment, the total loss value is calculated by the loss value of the training set matrix, the knowledge graph recommended value, the original recommended matrix and the interactive recommended value through the loss function, accurate recommendation is achieved, better mathematical distinction can be achieved, further, more effective information in the graph can be beneficially mined, quick recommendation can be achieved, accurate recommendation can be achieved in a specific scene, certain generalization performance is achieved, certain effects are achieved on different types of data, meanwhile, the recommendation accuracy can be effectively improved, good robustness is achieved, commodities can be reasonably and accurately recommended for the user, commodity recommendation for the user through cold start is achieved, and reliability and accuracy of recommendation are improved.
Optionally, as an embodiment of the present invention, the calculating of the recommendation loss value on the training set matrix and the original recommendation matrix to obtain the recommendation loss value includes:
calculating a recommendation loss value of the training set matrix and the original recommendation matrix according to a seventh formula, wherein the seventh formula is as follows:
wherein, yuiIs an original recommendation matrix, O is a training set matrix, u is a user information vector, i+Is a positive sample, i-For negative examples, δ is the sigmoid function and In is the logarithmic function.
Understandably, the negative example i-The training set is obtained by screening the training set through a plurality of user purchase data.
In the above embodiment, the recommendation loss value is calculated by the seventh formula for the recommendation loss values of the training set matrix and the original recommendation matrix, so that accurate recommendation is realized, better mathematical distinction can be made, further, mining of more effective information in the graph is facilitated, quick recommendation can be realized, accurate recommendation can be realized in a specific scene, certain generalization is realized, certain effects are realized on different types of data, meanwhile, recommendation accuracy can be effectively improved, good robustness is realized, commodities can be reasonably and accurately recommended for a user, commodity recommendation for the user by cold start is realized, and reliability and accuracy of recommendation are improved.
Optionally, as an embodiment of the invention, the knowledge-graph recommendation values comprise knowledge-graph recommendation value positive sample values and knowledge-graph recommendation value negative sample values, the interaction recommendation values comprise interaction recommendation value positive sample values and interaction recommendation value negative sample values,
the process of calculating the knowledge graph loss value of the knowledge graph recommendation value and the interaction recommendation value to obtain the knowledge graph loss value comprises the following steps:
calculating the knowledge map loss value of the knowledge map recommended value positive sample value, the knowledge map recommended value negative sample value, the interaction recommended value positive sample value and the interaction recommended value negative sample value through an eighth formula, wherein the eighth formula is as follows:
therein, LossgrsPhi is the knowledge-graph loss value, phi is the softplus loss function,recommending negative sample values for the knowledge-graph,recommending positive sample values for the knowledge-graph,a positive sample value is recommended for the interaction,recommending for an interactionThe value is negative sample value.
In the above embodiment, the knowledge map loss value is calculated by the eighth formula for the knowledge map loss value of the positive sample value of the knowledge map recommended value, the negative sample value of the knowledge map recommended value, the positive sample value of the interactive recommended value and the negative sample value of the interactive recommended value, so that accurate recommendation is realized, mathematical distinction can be better performed, further, mining of more effective information in a map is facilitated, quick recommendation can be realized, accurate recommendation can be realized in a specific scene, certain generalization performance is realized, certain effects are realized on different types of data, meanwhile, recommendation accuracy can be effectively improved, good robustness is realized, commodities can be reasonably and accurately recommended for a user, cold start is realized for recommending commodities to the user, and reliability and accuracy of recommendation are improved.
Optionally, as an embodiment of the present invention, the method further includes numbering the commodity information training set to obtain a numbered commodity information training set.
Fig. 2 is a block diagram of a product information recommendation device according to an embodiment of the present invention.
Alternatively, as another embodiment of the present invention, as shown in fig. 2, a commodity information recommending apparatus includes:
the data set dividing module is used for importing a commodity information data set, and dividing the data set into the commodity information data set to obtain a commodity information training set and a commodity information testing set;
the vectorization analysis module is used for carrying out vectorization analysis on the commodity information training set to obtain a knowledge map matrix and a training set matrix;
the analysis processing module is used for analyzing and processing the knowledge map matrix and the training set matrix to obtain a knowledge map recommended value, an original recommended matrix and an interaction recommended value;
the loss value calculation module is used for calculating the loss values of the training set matrix, the knowledge graph recommendation value, the original recommendation matrix and the interactive recommendation value through a loss function to obtain a total loss value;
the back propagation module is used for inputting the total loss value into a preset optimizer and performing back propagation on the total loss value through the preset optimizer to obtain a commodity recommendation model;
and the recommendation result obtaining module is used for inputting the commodity information test set into the commodity information model and obtaining a commodity recommendation result according to the commodity information model.
Optionally, as an embodiment of the present invention, the commodity information training set includes a commodity node data set and a user node data set, and the vectorization analysis module is specifically configured to:
constructing a knowledge graph of the commodity node data set and the user node data set to obtain a training set knowledge graph;
respectively carrying out data vectorization on the training set knowledge graph, the commodity node data set and the user node data set to obtain a knowledge graph matrix corresponding to the training set knowledge graph, a commodity node matrix corresponding to the commodity node data set and a user node matrix corresponding to the user node data set;
and obtaining a training set matrix according to the commodity node matrix and the user node matrix.
Alternatively, another embodiment of the present invention provides a product information recommendation apparatus, including a memory, a processor, and a computer program stored in the memory and executable on the processor, and when the processor executes the computer program, the product information recommendation method as described above is implemented. The device may be a computer or the like.
Alternatively, another embodiment of the present invention provides a computer-readable storage medium storing a computer program which, when executed by a processor, implements the article information recommendation method as described above.
It is clear to those skilled in the art that, for convenience and brevity of description, the specific working processes of the above-described apparatuses and units may refer to the corresponding processes in the foregoing method embodiments, and are not described herein again.
In the several embodiments provided in the present application, it should be understood that the disclosed apparatus and method may be implemented in other ways. For example, the above-described apparatus embodiments are merely illustrative, and for example, a division of a unit is merely a logical division, and an actual implementation may have another division, 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.
Units described as separate parts may or may not be physically separate, and parts displayed as units may or may not be physical units, may be located in one place, or may be distributed on a plurality of network units. Some or all of the units can be selected according to actual needs to achieve the purpose of the solution of the embodiment of the present invention.
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 unit, if implemented in the form of a software functional unit and sold or used as a stand-alone product, may be stored in a computer readable storage medium. It will be understood that the technical solution of the present invention essentially contributes to the prior art, or all or part of the technical solution can be embodied in the form of a software product stored in a storage medium and including instructions for causing a computer device (which may be a personal computer, a server, or a network device) to execute all or part of the steps of the method according to the embodiments of the present invention. And the aforementioned storage medium includes: a U-disk, a removable hard disk, a Read-Only Memory (ROM), a Random Access Memory (RAM), a magnetic disk or an optical disk, and other various media capable of storing program codes.
While the invention has been described with reference to specific embodiments, it will be understood by those skilled in the art that various changes in form and details may be made therein without departing from the spirit and scope of the invention as defined by the appended claims. Therefore, the protection scope of the present invention shall be subject to the protection scope of the claims.
Claims (10)
1. A commodity information recommendation method is characterized by comprising the following steps:
importing a commodity information data set, and dividing the data set into a commodity information training set and a commodity information testing set;
vectorization analysis is carried out on the commodity information training set to obtain a knowledge map matrix and a training set matrix;
analyzing and processing the knowledge map matrix and the training set matrix to obtain a knowledge map recommendation value, an original recommendation matrix and an interaction recommendation value;
calculating loss values of the training set matrix, the knowledge graph recommendation value, the original recommendation matrix and the interaction recommendation value through a loss function to obtain a total loss value;
inputting the total loss value into a preset optimizer, and performing back propagation on the total loss value through the preset optimizer to obtain a commodity recommendation model;
and inputting the commodity information test set into the commodity information model, and obtaining a commodity recommendation result according to the commodity information model.
2. The commodity information recommendation method according to claim 1, wherein the commodity information training set includes a commodity node data set and a user node data set, and the process of performing vectorization analysis on the commodity information training set to obtain a knowledge map matrix and a training set matrix includes:
constructing a knowledge graph of the commodity node data set and the user node data set to obtain a training set knowledge graph;
respectively carrying out data vectorization on the training set knowledge graph, the commodity node data set and the user node data set to obtain a knowledge graph matrix corresponding to the training set knowledge graph, a commodity node matrix corresponding to the commodity node data set and a user node matrix corresponding to the user node data set;
and obtaining a training set matrix according to the commodity node matrix and the user node matrix.
3. The commodity information recommendation method according to claim 2, wherein the process of analyzing and processing the knowledge graph matrix and the training set matrix to obtain the knowledge graph recommendation value, the original recommendation matrix, and the interaction recommendation value includes:
utilizing a Bi-Interaction bidirectional interactive learning network to perform learning processing of a training set matrix on the knowledge map matrix, the commodity node matrix and the user node matrix to obtain a user information matrix and a commodity information matrix;
calculating an original recommendation matrix for the user information matrix and the commodity information matrix through a first formula to obtain the original recommendation matrix, wherein the first formula is as follows:
wherein, e'uIs a user information matrix, e'iIs a commodity information matrix, yuiA matrix is originally recommended;
calculating an interactive recommendation value of the user information matrix and the commodity information matrix through a second formula to obtain the interactive recommendation value, wherein the second formula is as follows:
where, σ is a loss function,is dot product, e'uIs a user information matrix, e'iIs a commodity information matrix, WaAs a weight matrix, ScorecroRecommending a value for the interaction;
and calculating a knowledge graph recommendation value of the knowledge graph matrix by using a score strategy algorithm to obtain the knowledge graph recommendation value.
4. The commodity information recommendation method according to claim 3, wherein the learning processing of the training set matrix on the knowledge map matrix, the commodity node matrix, and the user node matrix by using the Bi-Interaction bidirectional interactive learning network to obtain the user information matrix and the commodity information matrix comprises:
calculating a user information matrix for the knowledge graph spectrum matrix and the user node matrix through a third formula to obtain a user information matrix, wherein the third formula is as follows:
wherein, yg=σ(W1(e+ek))+σ(W2(e·ek)),
where σ is the loss function, W1And W2Are all weight matrices, e'uAs a matrix of user information, ekIs a central vector, k is k nodes nearest to the central node, l is the number of higher-order layers, etIs a tail node vector in the knowledge graph spectrum matrix, e is a head node vector in the knowledge graph spectrum matrix, pi (h, r, t) is a relation among the head node vector h, the tail node vector t and a relation vector r among nodes,the vector is the l user node vector in the user node matrix;
calculating a commodity information matrix for the knowledge map matrix and the commodity node matrix through a fourth formula to obtain a commodity information matrix, wherein the fourth formula is as follows:
wherein, yg=σ(W1(e+ek))+σ(W2(e·ek)),
where σ is the loss function, W1And W2Are all weight matrices, e'iIs a commodity information matrix, ekIs a central vector, k is k nodes nearest to the central node, l is the number of higher-order layers, etIs a tail node vector in the knowledge graph spectrum matrix, e is a head node vector in the knowledge graph spectrum matrix, pi (h, r, t) is a relation among the head node vector h, the tail node vector t and a relation vector r among nodes,is the ith commodity node vector in the commodity node matrix.
5. The commodity information recommendation method according to claim 3, wherein the calculating of the knowledge graph recommendation value for the knowledge graph matrix by using the score policy algorithm to obtain the knowledge graph recommendation value comprises:
calculating a knowledge graph recommendation value of the knowledge graph matrix through a fifth formula to obtain the knowledge graph recommendation value, wherein the fifth formula is as follows:
Scoreg=||polar(h)W+polar(r)-polar(t)W||,
wherein gamma is variable, eposion and hidden are both preset constants, h is head node vector, t is tail node vector, r is relationship vector between nodes, e0For node vectors in a knowledge-graph matrix, knowledge-graph node vector e0Comprises a head node vector h, a tail node vector t and a relation vector r between nodes, ScoregAnd recommending values for the knowledge graph.
6. The commodity information recommendation method according to claim 1, wherein the process of calculating the loss values of the training set matrix, the knowledge graph recommendation value, the original recommendation matrix, and the interaction recommendation value by using a loss function to obtain a total loss value comprises:
calculating a recommendation loss value of the training set matrix and the original recommendation matrix to obtain a recommendation loss value;
calculating a knowledge map loss value of the knowledge map recommendation value and the interaction recommendation value to obtain a knowledge map loss value;
calculating a total loss value of the recommended loss value and the knowledge graph loss value through a sixth formula to obtain the total loss value, wherein the sixth formula is as follows:
Loss=Lossrs+Lossgrs,
therein, LossrsTo recommend Loss value, LossgrsThe knowledge-graph Loss value and the Loss is the total Loss value.
7. The commodity information recommendation method according to claim 6, wherein the calculating of the recommendation loss value for the training set matrix and the original recommendation matrix to obtain the recommendation loss value comprises:
calculating a recommendation loss value of the training set matrix and the original recommendation matrix according to a seventh formula, wherein the seventh formula is as follows:
wherein, yuiIs an original recommendation matrix, O is a training set matrix, u is a user information vector, i+Is a positive sample, i-For negative examples, δ is the sigmoid function and In is the logarithmic function.
8. The merchandise information recommendation method of claim 6, wherein the knowledge-graph recommendation value comprises a knowledge-graph recommendation value positive sample value and a knowledge-graph recommendation value negative sample value, the interaction recommendation value comprises an interaction recommendation value positive sample value and an interaction recommendation value negative sample value,
the process of calculating the knowledge graph loss value of the knowledge graph recommendation value and the interaction recommendation value to obtain the knowledge graph loss value comprises the following steps:
calculating the knowledge map loss value of the knowledge map recommended value positive sample value, the knowledge map recommended value negative sample value, the interaction recommended value positive sample value and the interaction recommended value negative sample value through an eighth formula, wherein the eighth formula is as follows:
therein, LossgrsPhi is the knowledge-graph loss value, phi is the softplus loss function,recommending negative sample values for the knowledge-graph,recommending positive sample values for the knowledge-graph,a positive sample value is recommended for the interaction,a negative sample value is recommended for the interaction.
9. An article information recommendation device characterized by comprising:
the data set dividing module is used for importing a commodity information data set, and dividing the data set into the commodity information data set to obtain a commodity information training set and a commodity information testing set;
the vectorization analysis module is used for carrying out vectorization analysis on the commodity information training set to obtain a knowledge map matrix and a training set matrix;
the analysis processing module is used for analyzing and processing the knowledge map matrix and the training set matrix to obtain a knowledge map recommended value, an original recommended matrix and an interaction recommended value;
the loss value calculation module is used for calculating the loss values of the training set matrix, the knowledge graph recommendation value, the original recommendation matrix and the interactive recommendation value through a loss function to obtain a total loss value;
the back propagation module is used for inputting the total loss value into a preset optimizer and performing back propagation on the total loss value through the preset optimizer to obtain a commodity recommendation model;
and the recommendation result obtaining module is used for inputting the commodity information test set into the commodity information model and obtaining a commodity recommendation result according to the commodity information model.
10. The merchandise information recommendation device according to claim 9, wherein the merchandise information training set includes a merchandise node data set and a user node data set, and the vectorization analysis module is specifically configured to:
constructing a knowledge graph of the commodity node data set and the user node data set to obtain a training set knowledge graph;
respectively carrying out data vectorization on the training set knowledge graph, the commodity node data set and the user node data set to obtain a knowledge graph matrix corresponding to the training set knowledge graph, a commodity node matrix corresponding to the commodity node data set and a user node matrix corresponding to the user node data set;
and obtaining a training set matrix according to the commodity node matrix and the user node matrix.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202110041795.0A CN112800207B (en) | 2021-01-13 | 2021-01-13 | Commodity information recommendation method and device and storage medium |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202110041795.0A CN112800207B (en) | 2021-01-13 | 2021-01-13 | Commodity information recommendation method and device and storage medium |
Publications (2)
Publication Number | Publication Date |
---|---|
CN112800207A true CN112800207A (en) | 2021-05-14 |
CN112800207B CN112800207B (en) | 2022-05-13 |
Family
ID=75810421
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN202110041795.0A Active CN112800207B (en) | 2021-01-13 | 2021-01-13 | Commodity information recommendation method and device and storage medium |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN112800207B (en) |
Cited By (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN113222711A (en) * | 2021-05-28 | 2021-08-06 | 桂林电子科技大学 | Commodity information recommendation method, system and storage medium |
CN114936907A (en) * | 2022-06-15 | 2022-08-23 | 山东大学 | Commodity recommendation method and system based on node type interaction |
CN113780097B (en) * | 2021-08-17 | 2023-12-01 | 北京数慧时空信息技术有限公司 | Cultivated land extraction method based on knowledge graph and deep learning |
Citations (13)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN107862562A (en) * | 2017-09-15 | 2018-03-30 | 广州唯品会研究院有限公司 | A kind of method and device that selection progress commercial product recommending is liked according to the picture of user |
CN109447713A (en) * | 2018-10-31 | 2019-03-08 | 国家电网公司 | A kind of recommended method and device of knowledge based map |
CN110457508A (en) * | 2019-07-18 | 2019-11-15 | 西安工程大学 | Garment coordination recommended method based on matrix decomposition and knowledge mapping |
CN111199459A (en) * | 2019-12-30 | 2020-05-26 | 深圳市盟天科技有限公司 | Commodity recommendation method and device, electronic equipment and storage medium |
CN111275521A (en) * | 2020-01-16 | 2020-06-12 | 华南理工大学 | Commodity recommendation method based on user comment and satisfaction level embedding |
CN111369318A (en) * | 2020-02-28 | 2020-07-03 | 安徽农业大学 | Commodity knowledge graph feature learning-based recommendation method and system |
US20200265953A1 (en) * | 2019-02-14 | 2020-08-20 | Babylon Partners Limited | Identifying Valid Medical Data for Facilitating Accurate Medical Diagnosis |
CN111681067A (en) * | 2020-04-17 | 2020-09-18 | 清华大学 | Long-tail commodity recommendation method and system based on graph attention network |
US20200334715A1 (en) * | 2016-10-17 | 2020-10-22 | Singapore Telecommunications, Ltd. | Knowledge Model for Personalization and Location Services |
CN112085559A (en) * | 2020-08-18 | 2020-12-15 | 山东大学 | Interpretable commodity recommendation method and system based on time-sequence knowledge graph |
CN112101984A (en) * | 2020-08-16 | 2020-12-18 | 复旦大学 | Conversation recommendation model integrating user microscopic behaviors and knowledge graph |
CN112102029A (en) * | 2020-08-20 | 2020-12-18 | 浙江大学 | Knowledge graph-based long-tail recommendation calculation method |
CN112115358A (en) * | 2020-09-14 | 2020-12-22 | 中国船舶重工集团公司第七0九研究所 | Personalized recommendation method using multi-hop path features in knowledge graph |
-
2021
- 2021-01-13 CN CN202110041795.0A patent/CN112800207B/en active Active
Patent Citations (13)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US20200334715A1 (en) * | 2016-10-17 | 2020-10-22 | Singapore Telecommunications, Ltd. | Knowledge Model for Personalization and Location Services |
CN107862562A (en) * | 2017-09-15 | 2018-03-30 | 广州唯品会研究院有限公司 | A kind of method and device that selection progress commercial product recommending is liked according to the picture of user |
CN109447713A (en) * | 2018-10-31 | 2019-03-08 | 国家电网公司 | A kind of recommended method and device of knowledge based map |
US20200265953A1 (en) * | 2019-02-14 | 2020-08-20 | Babylon Partners Limited | Identifying Valid Medical Data for Facilitating Accurate Medical Diagnosis |
CN110457508A (en) * | 2019-07-18 | 2019-11-15 | 西安工程大学 | Garment coordination recommended method based on matrix decomposition and knowledge mapping |
CN111199459A (en) * | 2019-12-30 | 2020-05-26 | 深圳市盟天科技有限公司 | Commodity recommendation method and device, electronic equipment and storage medium |
CN111275521A (en) * | 2020-01-16 | 2020-06-12 | 华南理工大学 | Commodity recommendation method based on user comment and satisfaction level embedding |
CN111369318A (en) * | 2020-02-28 | 2020-07-03 | 安徽农业大学 | Commodity knowledge graph feature learning-based recommendation method and system |
CN111681067A (en) * | 2020-04-17 | 2020-09-18 | 清华大学 | Long-tail commodity recommendation method and system based on graph attention network |
CN112101984A (en) * | 2020-08-16 | 2020-12-18 | 复旦大学 | Conversation recommendation model integrating user microscopic behaviors and knowledge graph |
CN112085559A (en) * | 2020-08-18 | 2020-12-15 | 山东大学 | Interpretable commodity recommendation method and system based on time-sequence knowledge graph |
CN112102029A (en) * | 2020-08-20 | 2020-12-18 | 浙江大学 | Knowledge graph-based long-tail recommendation calculation method |
CN112115358A (en) * | 2020-09-14 | 2020-12-22 | 中国船舶重工集团公司第七0九研究所 | Personalized recommendation method using multi-hop path features in knowledge graph |
Non-Patent Citations (2)
Title |
---|
CHRISTOPHER C. YANG等: "Enriching User Experience in Online Health Communities Through Thread Recommendations and Heterogeneous Information Network Mining", 《IEEE TRANSACTIONS ON COMPUTATIONAL SOCIAL SYSTEMS 》 * |
孟祥福等: "用户-兴趣点耦合关系的兴趣点推荐方法", 《智能系统学报》 * |
Cited By (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN113222711A (en) * | 2021-05-28 | 2021-08-06 | 桂林电子科技大学 | Commodity information recommendation method, system and storage medium |
CN113222711B (en) * | 2021-05-28 | 2022-04-19 | 桂林电子科技大学 | Commodity information recommendation method, system and storage medium |
CN113780097B (en) * | 2021-08-17 | 2023-12-01 | 北京数慧时空信息技术有限公司 | Cultivated land extraction method based on knowledge graph and deep learning |
CN114936907A (en) * | 2022-06-15 | 2022-08-23 | 山东大学 | Commodity recommendation method and system based on node type interaction |
CN114936907B (en) * | 2022-06-15 | 2024-04-30 | 山东大学 | Commodity recommendation method and system based on node type interaction |
Also Published As
Publication number | Publication date |
---|---|
CN112800207B (en) | 2022-05-13 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN112800207B (en) | Commodity information recommendation method and device and storage medium | |
CN107563841B (en) | Recommendation system based on user score decomposition | |
CN107451894B (en) | Data processing method, device and computer readable storage medium | |
Bajari et al. | Demand estimation with machine learning and model combination | |
CN103678672A (en) | Method for recommending information | |
WO2017159403A1 (en) | Prediction system, method, and program | |
CN109034960B (en) | Multi-attribute inference method based on user node embedding | |
CN111461225B (en) | Customer clustering system and method thereof | |
CN106484777A (en) | A kind of multimedia data processing method and device | |
CN106157156A (en) | A kind of cooperation recommending system based on communities of users | |
CN114418035A (en) | Decision tree model generation method and data recommendation method based on decision tree model | |
CN111582443A (en) | Recommendation method based on Mask mechanism and level attention mechanism | |
CN105976070A (en) | Key-element-based matrix decomposition and fine tuning method | |
CN110348906B (en) | Improved commodity recommendation method based on multi-type implicit feedback | |
WO2017159402A1 (en) | Co-clustering system, method, and program | |
CN112102029B (en) | Knowledge graph-based long-tail recommendation calculation method | |
CN105868422B (en) | A kind of collaborative filtering recommending method based on elastic dimensional feature vector Optimizing Extraction | |
CN111460316A (en) | Knowledge system-oriented personalized recommendation method and computer storage medium | |
CN111966886A (en) | Object recommendation method, object recommendation device, electronic equipment and storage medium | |
CN113158024A (en) | Causal reasoning method for correcting popularity deviation of recommendation system | |
CN107220303A (en) | Recommendation method, device and the computer-readable medium of a kind of application | |
CN113988951A (en) | Commodity recommendation learning model construction method based on tensor decomposition and collaborative filtering | |
CN111984842B (en) | Bank customer data processing method and device | |
CN111967973B (en) | Bank customer data processing method and device | |
CN111275480B (en) | Multi-dimensional sparse sales data warehouse oriented fraud behavior mining method |
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 |