CN112749323A - Method and device for constructing user portrait - Google Patents
Method and device for constructing user portrait Download PDFInfo
- Publication number
- CN112749323A CN112749323A CN201911053500.0A CN201911053500A CN112749323A CN 112749323 A CN112749323 A CN 112749323A CN 201911053500 A CN201911053500 A CN 201911053500A CN 112749323 A CN112749323 A CN 112749323A
- Authority
- CN
- China
- Prior art keywords
- user
- information
- commodity
- node
- representation
- 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.)
- Pending
Links
- 238000000034 method Methods 0.000 title claims abstract description 40
- 238000010276 construction Methods 0.000 claims abstract description 59
- 230000002159 abnormal effect Effects 0.000 claims abstract description 44
- 238000013528 artificial neural network Methods 0.000 claims abstract description 31
- 238000012549 training Methods 0.000 claims abstract description 9
- 239000013598 vector Substances 0.000 claims description 85
- 238000004590 computer program Methods 0.000 claims description 9
- 239000000126 substance Substances 0.000 claims description 6
- 238000013461 design Methods 0.000 abstract description 9
- 238000010586 diagram Methods 0.000 description 21
- 230000006870 function Effects 0.000 description 8
- 238000004891 communication Methods 0.000 description 7
- 230000008569 process Effects 0.000 description 7
- 230000006399 behavior Effects 0.000 description 5
- 230000007246 mechanism Effects 0.000 description 5
- 230000003287 optical effect Effects 0.000 description 4
- 238000012545 processing Methods 0.000 description 4
- 230000004048 modification Effects 0.000 description 3
- 238000012986 modification Methods 0.000 description 3
- 230000009286 beneficial effect Effects 0.000 description 2
- 239000000835 fiber Substances 0.000 description 2
- 230000000644 propagated effect Effects 0.000 description 2
- 239000004065 semiconductor Substances 0.000 description 2
- 230000005856 abnormality Effects 0.000 description 1
- 238000013145 classification model Methods 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- 230000006872 improvement Effects 0.000 description 1
- 239000004973 liquid crystal related substance Substances 0.000 description 1
- 238000007620 mathematical function Methods 0.000 description 1
- 239000013307 optical fiber Substances 0.000 description 1
- 230000003068 static effect Effects 0.000 description 1
- 238000006467 substitution reaction Methods 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/90—Details of database functions independent of the retrieved data types
- G06F16/95—Retrieval from the web
- G06F16/953—Querying, e.g. by the use of web search engines
- G06F16/9535—Search customisation based on user profiles and personalisation
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q30/00—Commerce
- G06Q30/02—Marketing; Price estimation or determination; Fundraising
- G06Q30/0201—Market modelling; Market analysis; Collecting market data
Abstract
The invention discloses a method and a device for constructing a user portrait, and relates to the technical field of computers. One embodiment of the method comprises: acquiring a user information abnormal picture; training to obtain a user portrait construction model for constructing a user portrait based on a graph neural network according to the user nodes, the commodity information nodes, the first edges and the second edges of the user information abnormal composition; and adding commodity information related to the user to be constructed to the user information abnormal composition, and constructing a model by using the user portrait to construct the user portrait of the user to be constructed. According to the embodiment, the user portrait prediction model can be directly constructed on the basis of various data in the user information abnormal composition under the condition that a manual design mixing method is not carried out, a large amount of label data is not needed, and the method is high in efficiency and low in cost.
Description
Technical Field
The invention relates to the technical field of computers, in particular to a method and a device for constructing a user portrait.
Background
User portrayal, namely user information tagging (such as gender, age and the like), is a basic mode for abstracting a user overall for supporting big data applications such as personalized recommendation and the like by collecting and analyzing data of main information such as user static attributes, social attributes, behavior attributes and the like. The user portrait has a wide application prospect, especially in the field of electronic commerce, and can quickly and accurately position the user group, the value information such as user demands and the like based on the user portrait, so that the user portrait has a vital role in the aspects of commodity search, commodity recommendation, advertisement, accurate marketing and the like.
At present, a common method for constructing a user portrait is to construct a user portrait as a supervised classification task, regard each user as an independent data instance, regard a known user portrait as a label for supervised learning, and construct a user portrait classification model based on artificial design features such as user historical behaviors.
In the process of implementing the invention, the inventor finds that at least the following problems exist in the prior art: the existing method for constructing the user portrait needs a large amount of user logos for supervised learning, and is long in time consumption and high in cost; in the process of supervised learning, only data such as historical behaviors of the user (such as commodity purchasing and commodity clicking) and the like are adopted, other data which can be used for constructing a user portrait (such as similarity between the user and other users or similarity of purchased commodities and the like) are not considered, and when multiple data need to be considered simultaneously, a hybrid method needs to be designed manually to model the multiple data.
Disclosure of Invention
In view of the above, the present invention provides a method and an apparatus for constructing a user portrait, which can construct a user portrait prediction model based on various data in a user information heterogeneous composition directly without manually designing a hybrid method, and do not require a large amount of tag data, and are efficient and low in cost.
To achieve the above object, according to one aspect of the present invention, there is provided a method of constructing a user representation, comprising: acquiring a user information abnormal composition, wherein nodes of the user information abnormal composition comprise user nodes and commodity information nodes related to the users, a first edge of the user information abnormal composition indicates the association between different users, and a second edge of the user information abnormal composition indicates the association between the users and the commodity information; training to obtain a user portrait construction model for constructing a user portrait based on a graph neural network according to the user nodes, the commodity information nodes, the first edge and the second edge; and adding commodity information related to the user to be constructed to the user information abnormal composition, and constructing a model by using the user portrait to construct the user portrait of the user to be constructed.
Optionally, the graph network comprises: the system comprises a user information input layer, a commodity information representation layer, a first user representation layer, a second user representation layer and a user portrait construction layer; the user information input layer is used for acquiring the user information abnormal picture so as to acquire the user node, the commodity information node, the first edge and the second edge; the user information representation layer is used for generating a commodity information vector representing the characteristics of the commodity information node; the first user representation layer is used for generating a first user vector representing the characteristics of the user node according to the commodity information vector and the second edge; the second user representation layer is used for generating a second user vector representing the characteristics of the user node according to the first user vector and the first edge; and the user portrait construction layer is used for constructing the user portrait of the user node according to the second user vector.
Optionally, the node of the user information heteromorphic graph further includes: a commodity attribute information node for describing the commodity information; the third side of the user information heteromorphic graph indicates an association between the commodity attribute information and the commodity information.
Optionally, the graph neural network further comprises: a commodity attribute information presentation layer; the commodity attribute information representation layer is used for generating a commodity attribute information vector representing the characteristics of the commodity attribute information node; and the user information representation layer is used for generating a commodity information vector representing the characteristics of the commodity information node according to the commodity attribute information vector and the third edge.
Optionally, a first user node in the user nodes in the user information heterogeneous composition graph has a user portrait label, a second user node in the user nodes in the user information heterogeneous composition graph does not have a user portrait label, and a user portrait construction model for constructing a user portrait is obtained through training based on a graph neural network according to the first user node, the second user node, the commodity information node, the first edge and the second edge.
To achieve the above object, according to another aspect of the present invention, there is provided an apparatus for constructing a user representation, comprising: the system comprises an abnormal composition acquisition module, a construction model acquisition module and a user portrait construction module; the system comprises an abnormal composition acquisition module, a product information acquisition module and a product information acquisition module, wherein the abnormal composition acquisition module is used for acquiring a user information abnormal composition, nodes of the user information abnormal composition comprise user nodes and product information nodes related to the users, a first edge of the user information abnormal composition indicates the association between different users, and a second edge of the user information abnormal composition indicates the association between the users and the product information; the model construction model acquisition module is used for training to obtain a user portrait construction model used for constructing a user portrait based on a graph neural network according to the user nodes, the commodity information nodes, the first edges and the second edges; the user portrait construction module is used for adding commodity information related to a user to be constructed to the user information abnormal composition, constructing a model by using the user portrait, and constructing the user portrait of the user to be constructed.
Optionally, the graph neural network comprises: the system comprises a user information input layer, a commodity information representation layer, a first user representation layer, a second user representation layer and a user portrait construction layer; the user information input layer is used for acquiring the user information abnormal picture so as to acquire the user node, the commodity information node, the first edge and the second edge; the user information representation layer is used for generating a commodity information vector representing the characteristics of the commodity information node; the first user representation layer is used for generating a first user vector representing the characteristics of the user node according to the commodity information vector and the second edge; the second user representation layer is used for generating a characteristic second user vector representing the user node according to the first user vector and the first edge; and the user portrait construction layer is used for constructing the user portrait of the user node according to the second user vector.
Optionally, the node of the user information heteromorphic graph further includes: a commodity attribute information node for describing the commodity information; the third side of the user information heteromorphic graph indicates an association between the commodity attribute information and the commodity information.
Optionally, the graph neural network further comprises: a commodity attribute information presentation layer; the commodity attribute information representation layer is used for generating a characteristic commodity attribute information vector representing the commodity attribute information node; and the user information representation layer is used for generating a commodity information vector representing the characteristics of the commodity information node according to the commodity attribute information vector and the third edge.
Optionally, a first user node of the user nodes in the user information heteromorphic image has a user portrait label, and a second user node of the user nodes in the user information heteromorphic image does not have a user portrait label; the construction model obtaining module is further configured to train to obtain a user portrait construction model for constructing a user portrait based on a graph neural network according to the first user node, the second user node, the commodity information node, the first edge and the second edge.
To achieve the above object, according to still another aspect of the present invention, there is provided a server for predicting a user representation, comprising: one or more processors; a storage device to store one or more programs that, when executed by the one or more processors, cause the one or more processors to implement
To achieve the above object, according to still another aspect of the present invention, there is provided a computer readable medium having a computer program stored thereon, wherein the program is executed by a processor to implement any one of the methods of constructing a user representation as described above.
The invention has the following advantages or beneficial effects: the nodes and edges of the user information heteromorphic graph with various nodes are adopted to represent various data such as users, commodity information, similarity among different users, corresponding relation between the users and the commodity information and the like, and the nodes and edges of the user information heteromorphic graph are learned through a graph neural network so as to obtain a user portrait construction model; the method not only realizes simultaneous modeling of various data, but also avoids the problem of manual design of various data mixing methods or manual design characteristics, does not need a large number of data labels, and realizes quick, efficient and low-cost construction of user images.
Further effects of the above-mentioned non-conventional alternatives will be described below in connection with the embodiments.
Drawings
The drawings are included to provide a better understanding of the invention and are not to be construed as unduly limiting the invention. Wherein:
FIG. 1 is a schematic diagram of a main flow of a method of constructing a user representation according to an embodiment of the invention;
FIG. 2a is a schematic diagram of a user information exception diagram according to an embodiment of the present invention;
FIG. 2b is a diagram of another user information exception diagram according to an embodiment of the present invention;
FIG. 2c is a schematic diagram of the structure of a neural network according to an embodiment of the present invention;
FIG. 2d is a schematic diagram of another embodiment of a neural network according to the present invention;
FIG. 3 is a schematic diagram of yet another user information exception diagram in accordance with an embodiment of the present invention;
FIG. 4 is a schematic diagram of the major modules of an apparatus for constructing a user representation in accordance with an embodiment of the present invention;
FIG. 5 is an exemplary system architecture diagram in which embodiments of the present invention may be employed;
fig. 6 is a schematic block diagram of a computer system suitable for use in implementing a terminal device or server of an embodiment of the invention.
Detailed Description
Exemplary embodiments of the present invention are described below with reference to the accompanying drawings, in which various details of embodiments of the invention are included to assist understanding, and which are to be considered as merely exemplary. Accordingly, those of ordinary skill in the art will recognize that various changes and modifications of the embodiments described herein can be made without departing from the scope and spirit of the invention. Also, descriptions of well-known functions and constructions are omitted in the following description for clarity and conciseness.
As shown in fig. 1, an embodiment of the present invention provides a method for constructing a user portrait, which may specifically include the following steps:
step S101, obtaining a user information heteromorphic graph, wherein nodes of the user information heteromorphic graph comprise user nodes and commodity information nodes related to the users, a first edge of the user information heteromorphic graph indicates the association between different users, and a second edge of the user information heteromorphic graph indicates the association between the users and the commodity information.
The graph is composed of nodes and edges connected among the nodes, and can be divided into a homogeneous graph and a heterogeneous graph according to the number of the types of the nodes; the isomorphic graph refers to a graph with only one type of nodes, and the heterogeneous graph refers to a graph with two or more types of nodes. The user information heteromorphic graph referred to in the embodiment of the application refers to a graph at least including two nodes of user and commodity information, and the commodity information refers to information capable of distinguishing commodities such as a commodity ID, a commodity identification, a commodity name and the like. Correspondingly, the association indicated by the first edge between different users refers to a similarity relationship between different users, and in the field of electronic commerce, the similarity relationship between users may be calculated based on the similarity of historical behaviors (such as clicking, collecting, purchasing commodities, and the like) of two users. The association of the user and the commodity information indicated by the second edge means that the historical behavior of the user relates to the commodity, for example, if the user clicks or purchases some commodities, the second edge is established between the user and the corresponding commodity information to represent the tendency of the user to purchase the commodity, that is, the user is the user who purchases or consumes the commodity.
As shown in fig. 2a, in a preferred embodiment, a user information abnormal graph is provided, which has two types of nodes, i.e. user u and commodity information i; the user nodes comprise a plurality of u1, u2, u3, u4, u5 and the like, and a plurality of commodity information used for identifying or representing that the user purchases, clicks or collects and the like are respectively i1, i2, i3, i4 and the like; the first sides u1u2, u2u5, u3u6, u4u5 and the like indicate that the number of commodities purchased or having a purchasing tendency by a user among different users reaches a set threshold (such as 1 or 2); the second sides u1i1, u2i1, u3i2, etc. represent items purchased or having a tendency to be purchased by the user. Specifically, taking user node u2 as an example, user u2 purchases or tends to purchase commodity i1, so that user u2 and commodity i1 have second side u2i1, and user u2 and user u1 both purchase or tend to purchase commodity i1, so that user u2 and user u1 have similar relationship, and thus have first side u1u 2.
It can be understood that, since the commodity information (such as a mobile phone) has commodity attribute information which can be used for describing the commodity information, such as commodity attribute information of color, brand, model and the like, in the process of constructing the user information heteromorphic image, the user information heteromorphic image can be constructed according to various information such as the commodity information, the commodity attribute information and the like of a user, so that the establishment of a user portrait construction model can be realized based on more information, and the user portrait can be constructed more comprehensively, stereoscopically and accurately.
In an optional implementation manner, the node of the user information heteromorphic graph further includes: a commodity attribute information node for describing the commodity information; the third side of the user information heteromorphic graph indicates an association between the commodity attribute information and the commodity information.
Specifically, referring to fig. 2b, the nodes of the user information heteromorphic graph constructed in a preferred embodiment include: three types of user nodes u, commodity information nodes i and commodity attribute information nodes t; the user nodes are provided with a plurality of user nodes, including u1, u2, u3, u4, u5 and the like, the user nodes are also provided with a plurality of commodity information used for identifying or representing that the user purchases, clicks or collects and the like, the commodity information is i1, i2, i3, i4 and the like, and the commodity attribute information used for describing the commodity information includes t1, t2, t3, t4 and the like; the first sides u1u2, u2u5, u3u6, u4u5 and the like indicate that the number of commodities purchased or having a purchasing tendency by a user among different users reaches a set threshold (such as 1 or 2); the second sides u1i1, u2i1, u3i2, etc. represent items purchased or having a tendency to be purchased by the user; the third sides i1t1, i1t2, i2t2, and the like indicate the product attribute information that the product information has. Specifically, taking user node u2 as an example, since user u2 purchases or tends to purchase commodity i1, user u2 and commodity i1 have a second edge u2i1 therebetween; meanwhile, since both user u2 and user u1 purchase or prefer to purchase commodity i1, user u2 and user u1 have a similar relationship, and thus have a first side u1u 2; since the commodity information i1 has two kinds of commodity attribute information, t1 and t2, the commodity information i1 has third sides i1t1 and i1t2 at the same time.
And S102, training to obtain a user portrait construction model for constructing the user portrait based on a graph neural network according to the user nodes, the commodity information nodes, the first edge and the second edge.
In an alternative embodiment, the graph network comprises: the system comprises a user information input layer, a commodity information representation layer, a first user representation layer, a second user representation layer and a user portrait construction layer; the user information input layer is used for acquiring the user information abnormal picture so as to acquire the user node, the commodity information node, the first edge and the second edge; the user information representation layer is used for generating a commodity information vector representing the characteristics of the commodity information node; the first user representation layer is used for generating a first user vector representing the characteristics of the user node according to the commodity information vector and the second edge; the second user representation layer is used for generating a second user vector representing the characteristics of the user node according to the first user vector and the first edge; and the user portrait construction layer is used for constructing the user portrait of the user node according to the second user vector.
Specifically, referring to fig. 2a and 2c, after receiving the user information heterogeneous diagram shown in fig. 2a, the user information input layer of the neural network (shown in fig. 2 c) obtains user nodes, such as u1, u2, u3, u4, and u5, commodity information nodes, such as i1, i2, i3, and i4, commodity attribute information nodes, such as t1, t2, t3, and t4, and third edges, such as first edges u1u2, u2u5, u3u6, u4u5, second edges u1i1, u2i1, and u3i2, indicated by the user information heterogeneous diagram. The product information presentation layer is configured to present each product information as a low-dimensional vector composed of a plurality of real numbers, and for example, the product information vector corresponding to the product information i1 may be a low-dimensional vector having a length T: s ═ s1,s2,…,sT]Wherein s isiAre real numbers. After the first user representation layer receives the commodity information vector, the second edges u1i1, u2i1, u3i2 and the like transmitted by the commodity information representation layer, each user may purchase or tends to purchase a plurality of commodities, and therefore according to the second edges of each user node and the corresponding commodity information vector, through an attention mechanics learning mechanism, the corresponding weight of the commodity information vector representing the first user is determined, and the corresponding first user vector is obtained through weighted summation of the weights of the commodity information vectors, for example, u ═ v1,v2,…,vT]Wherein v isiAre real numbers. After receiving the first user vector, the first edges u1u2, u2u5, u3u6, u4u5 and the like transmitted by the first user, the second user representation layer determines the weight of the first user vector according to all the second edges of each user and the corresponding first user representation through an attention mechanics learning mechanism and obtains the corresponding second user vector through weighted summation of the first user vector weights, for example, the first user vector, the first edges u1u2, u2u5, u3u6, u4u5 and the likeu＝[u1,u2,…,uT]Wherein u isiAre real numbers. On the basis, the user portrait construction layer constructs the user portrait corresponding to each user based on the second user vector of the user.
In an optional implementation manner, the node of the user information heteromorphic graph further includes: a commodity attribute information node for describing the commodity information; in a case where the third side of the user information abnormality map indicates an association between the commodity attribute information and the commodity information, the map neural network further includes: a commodity attribute information presentation layer; the commodity attribute information representation layer is used for generating a commodity attribute information vector representing the characteristics of the commodity attribute information node; and the user information representation layer is used for generating a commodity information vector representing the characteristics of the commodity information node according to the commodity attribute information vector and the third edge.
Referring to fig. 2b and 2d, after receiving the user information heterogeneous diagram shown in fig. 2b, the user information input layer of the neural network (shown in fig. 2 d) acquires user nodes such as u1, u2, u3 and u4, commodity information nodes such as i1, i2, i3 and i4, first sides u1u2, u2u5, u3u6 and u4u5, second sides u1i1, u2i1 and u3i2, and third sides i1t1, i1t2 and i2t2, which are indicated by the user information heterogeneous diagram. The product attribute information representation layer is configured to represent each piece of product attribute information as a low-dimensional vector composed of a plurality of real numbers, and for example, the product attribute information vector corresponding to the product attribute information T1 may be a low-dimensional vector having a length T: e ═ e1,e2,…,eT]Wherein e isiAre real numbers. After the commodity information representation layer receives the commodity attribute information vector, the third edge i1t1, i1t2, i2t2 and the like, each piece of commodity information may correspond to a plurality of pieces of commodity attribute information, so that the weight of each commodity attribute information vector is determined through an attention mechanics learning mechanism according to all the third edges of each piece of commodity information and the corresponding commodity attribute information, and the commodity information vector corresponding to the commodity information is determined through weighted summation of the weights of the commodity attribute information vectors, for example, s ═ s [ ([ s ])1,s2,…,sT]Which isMiddle SiAre real numbers. After the first user representation layer receives the commodity information vector, the second edges u1i1, u2i1, u3i2 and the like transmitted by the commodity information representation layer, each user may purchase or tends to purchase a plurality of commodities, and therefore according to the second edges of each user node and the corresponding commodity information vectors, weights corresponding to the commodity information vectors representing the first user are determined through an attention mechanics learning mechanism, and corresponding first user vectors are obtained through weighted summation of the weights of the commodity information vectors, wherein u is [ v ═ v { (v) } for example1,v2,…,vT]Wherein v isiAre real numbers. After the second user representation layer receives the first user vector, the first edges u1u2, u2u5, u3u6, u4u5 and the like transmitted by the first user, since each user may have a plurality of similar users, according to all the second edges of each user and the corresponding first user representation, the weights of the first user vector are determined through an attention mechanics learning mechanism, and the corresponding second user vector is obtained through weighted summation of the first user vector weights, such as u [ [ u ] ]1,u2,…,uT]Wherein u isiAre real numbers. On the basis, the user portrait construction layer constructs the user portrait corresponding to each user based on the second user vector of the user.
Furthermore, a first user node in the user nodes in the user information abnormal composition has a user portrait label, a second user node in the user nodes in the user information abnormal composition does not have a user portrait label, and a user portrait construction model for constructing a user portrait is obtained through training based on a graph neural network according to the first user node, the second user node, the commodity information node, the first edge and the second edge. Specifically, parameters of a user information input layer, a commodity information representation layer, a first user representation layer, a second user representation layer, a user portrait construction layer and the like in the graph neural network are trained and optimized by comparing the constructed user portrait with consistent user portrait labels and calculating a loss function through cross entropy. Therefore, the model can be built based on the trained user portrait, and the user portrait of each user in the user information abnormal picture can be built.
Referring to fig. 3, in a preferred embodiment, a user information exception pattern containing a new user is provided. Specifically, the user images of the users u1, u2, u5, and u7 are illustrated as male, female, and female, respectively. During the user representation construction process, we consider the representations of users u2 and u7 as unknown. For both users u1 and u2, the commodities i1 and i3 are purchased, the purchasing habits are similar, and the user representation of the user u2 constructed based on the model is male. For both users u5 and u7, only one identical commodity i5 was purchased. Since both merchandise items i5 and i6 were purchased by female user u7, they are likely to be more female merchandise items, and thus the corresponding merchandise attributes t6 and t7 are also more likely to be female. Since the article i4 includes the attribute t6, it is judged that the article i4 is more female, and the user representation of the user u5 is constructed as a female based on the characteristics that both the articles i5 and i4 are more female. At the same time, if the user images corresponding to u2 and u7 of the model construction do not match the known user images, then the user image construction model continues to be optimized based on the mathematical function.
And step S103, adding commodity information related to the user to be constructed to the user information abnormal composition, and constructing a user image of the user to be constructed by using the user image construction model.
It can be understood that the user portrait corresponding to the user in the user information heterogeneous graph can be constructed based on the user portrait construction model obtained after the user information heterogeneous graph learning. Therefore, when a user portrait of a new user needs to be constructed, the new user, corresponding commodity information, commodity attribute information and the like are added to the user information heterogeneous graph as nodes, corresponding first sides, second sides, third sides and the like are constructed at the same time for construction, and then the user portrait of the new user needs to be constructed based on the user information heterogeneous graph containing the new user.
The invention has the following advantages or beneficial effects: the nodes and edges of the user information heteromorphic graph with various nodes are adopted to represent various data such as users, commodity information, similarity among different users, corresponding relation between the users and the commodity information and the like, and the nodes and edges of the user information heteromorphic graph are learned through a graph neural network so as to obtain a user portrait construction model; the method not only realizes simultaneous modeling of various data, but also avoids the problem of manual design of various data mixing methods or manual design characteristics, does not need a large number of data labels, and realizes quick, efficient and low-cost construction of user images.
Referring to fig. 4, on the basis of the above embodiment, there is provided an apparatus 400 for constructing a user representation, comprising: a heterogeneous graph acquisition module 401, a construction model acquisition module 402 and a user portrait construction module 403; wherein the content of the first and second substances,
the heterogeneous graph obtaining module 401 is configured to obtain a user information heterogeneous graph, where nodes of the user information heterogeneous graph include user nodes and commodity information nodes related to the user, where a first edge of the user information heterogeneous graph indicates an association between different users, and a second edge of the user information heterogeneous graph indicates an association between the user and the commodity information; the model construction model acquisition module 402 is configured to train to obtain a user portrait construction model for constructing a user portrait based on a graph neural network according to the user node, the commodity information node, the first edge, and the second edge; the user portrait creating module 403 is configured to add commodity information related to a user to be created to the user information heterogeneous composition, and create a user portrait of the user to be created by using the user portrait creating model.
In an alternative embodiment, the graph neural network comprises: the system comprises a user information input layer, a commodity information representation layer, a first user representation layer, a second user representation layer and a user portrait construction layer; the user information input layer is used for acquiring the user information abnormal picture so as to acquire the user node, the commodity information node, the first edge and the second edge; the user information representation layer is used for generating a commodity information vector representing the characteristics of the commodity information node; the first user representation layer is used for generating a first user vector representing the characteristics of the user node according to the commodity information vector and the second edge; the second user representation layer is used for generating a second user vector representing the characteristics of the user node according to the first user vector and the first edge; and the user portrait construction layer is used for constructing the user portrait of the user node according to the second user vector.
In an optional implementation manner, the node of the user information heteromorphic graph further includes: a commodity attribute information node for describing the commodity information; the third side of the user information heteromorphic graph indicates an association between the commodity attribute information and the commodity information.
In an optional embodiment, the graph neural network further comprises: a commodity attribute information presentation layer; the commodity attribute information representation layer is used for generating a commodity attribute information vector representing the characteristics of the commodity attribute information node; and the user information representation layer is used for generating a commodity information vector representing the characteristics of the commodity information node according to the commodity attribute information vector and the third edge.
In an alternative embodiment, a first user node of the user nodes in the user information composition has a user portrait label and a second user node of the user nodes in the user information composition does not have a user portrait label; the construction model obtaining module 402 is further configured to train to obtain a user portrait construction model for constructing a user portrait based on a graph neural network according to the first user node, the second user node, the commodity information node, the first edge, and the second edge.
FIG. 5 illustrates an exemplary system architecture 500 for a method of building a user representation or a device for building a user representation to which embodiments of the present invention may be applied.
As shown in fig. 5, the system architecture 500 may include terminal devices 501, 502, 503, a network 504, and a server 505. The network 504 serves to provide a medium for communication links between the terminal devices 501, 502, 503 and the server 505. Network 504 may include various connection types, such as wired, wireless communication links, or fiber optic cables, to name a few.
The user may use the terminal devices 501, 502, 503 to interact with a server 505 over a network 504 to receive or send messages or the like. The terminal devices 501, 502, 503 may have various communication client applications installed thereon, such as a shopping application, a web browser application, a search application, an instant messaging tool, a mailbox client, social platform software, and the like.
The terminal devices 501, 502, 503 may be various electronic devices having a display screen and supporting web browsing, including but not limited to smart phones, tablet computers, laptop portable computers, desktop computers, and the like.
The server 505 may be a server that provides various services, such as a background management server that supports shopping websites browsed by users using the terminal devices 501, 502, 503. The background management server can analyze and process the received data such as the product information inquiry request and feed back the processing result (the constructed user portrait) to the terminal equipment.
It should be noted that the method for constructing a user representation provided by the embodiment of the present invention is generally executed by the server 505, and accordingly, the apparatus for constructing a user representation is generally disposed in the server 505.
It should be understood that the number of terminal devices, networks, and servers in fig. 5 is merely illustrative. There may be any number of terminal devices, networks, and servers, as desired for implementation.
Referring now to FIG. 6, a block diagram of a computer system 600 suitable for use with a terminal device implementing an embodiment of the invention is shown. The terminal device shown in fig. 6 is only an example, and should not bring any limitation to the functions and the scope of use of the embodiments of the present invention.
As shown in fig. 6, the computer system 600 includes a Central Processing Unit (CPU)601 that can perform various appropriate actions and processes according to a program stored in a Read Only Memory (ROM)602 or a program loaded from a storage section 608 into a Random Access Memory (RAM) 603. In the RAM 603, various programs and data necessary for the operation of the system 600 are also stored. The CPU 601, ROM 602, and RAM 603 are connected to each other via a bus 604. An input/output (I/O) interface 605 is also connected to bus 604.
The following components are connected to the I/O interface 605: an input portion 606 including a keyboard, a mouse, and the like; an output portion 607 including a display such as a Cathode Ray Tube (CRT), a Liquid Crystal Display (LCD), and the like, and a speaker; a storage section 608 including a hard disk and the like; and a communication section 609 including a network interface card such as a LAN card, a modem, or the like. The communication section 609 performs communication processing via a network such as the internet. The driver 610 is also connected to the I/O interface 605 as needed. A removable medium 611 such as a magnetic disk, an optical disk, a magneto-optical disk, a semiconductor memory, or the like is mounted on the drive 610 as necessary, so that a computer program read out therefrom is mounted in the storage section 608 as necessary.
In particular, according to the embodiments of the present disclosure, the processes described above with reference to the flowcharts may be implemented as computer software programs. For example, embodiments of the present disclosure include a computer program product comprising a computer program embodied on a computer readable medium, the computer program comprising program code for performing the method illustrated in the flow chart. In such an embodiment, the computer program may be downloaded and installed from a network through the communication section 609, and/or installed from the removable medium 611. The computer program performs the above-described functions defined in the system of the present invention when executed by the Central Processing Unit (CPU) 601.
It should be noted that the computer readable medium shown in the present invention can be a computer readable signal medium or a computer readable storage medium or any combination of the two. A computer readable storage medium may be, for example, but not limited to, an electronic, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any combination of the foregoing. More specific examples of the computer readable storage medium may include, but are not limited to: an electrical connection having one or more wires, a portable computer diskette, a hard disk, a Random Access Memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disc read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the foregoing. In the present invention, a computer readable storage medium may be any tangible medium that can contain, or store a program for use by or in connection with an instruction execution system, apparatus, or device. In the present invention, however, a computer readable signal medium may include a propagated data signal with computer readable program code embodied therein, for example, in baseband or as part of a carrier wave. Such a propagated data signal may take many forms, including, but not limited to, electro-magnetic, optical, or any suitable combination thereof. A computer readable signal medium may also be any computer readable medium that is not a computer readable storage medium and that can communicate, propagate, or transport a program for use by or in connection with an instruction execution system, apparatus, or device. Program code embodied on a computer readable medium may be transmitted using any appropriate medium, including but not limited to: wireless, wire, fiber optic cable, RF, etc., or any suitable combination of the foregoing.
The flowchart and block diagrams in the figures illustrate the architecture, functionality, and operation of possible implementations of systems, methods and computer program products according to various embodiments of the present invention. In this regard, each block in the flowchart or block diagrams may represent a module, segment, or portion of code, which comprises one or more executable instructions for implementing the specified logical function(s). It should also be noted that, in some alternative implementations, the functions noted in the block may occur out of the order noted in the figures. For example, two blocks shown in succession may, in fact, be executed substantially concurrently, or the blocks may sometimes be executed in the reverse order, depending upon the functionality involved. It will also be noted that each block of the block diagrams or flowchart illustration, and combinations of blocks in the block diagrams or flowchart illustration, can be implemented by special purpose hardware-based systems which perform the specified functions or acts, or combinations of special purpose hardware and computer instructions.
The modules described in the embodiments of the present invention may be implemented by software or hardware. The described modules may also be provided in a processor, which may be described as: a processor comprises a heterogeneous image acquisition module, a construction model acquisition module and a user portrait construction module. The names of these modules do not in some cases form a limitation on the modules themselves, and for example, the heterogeneous map acquisition module may also be described as a "module that acquires a user information heterogeneous map".
As another aspect, the present invention also provides a computer-readable medium that may be contained in the apparatus described in the above embodiments; or may be separate and not incorporated into the device. The computer readable medium carries one or more programs, and when the one or more programs are executed by a device, the device comprises a function of acquiring a user information heteromorphic graph, wherein nodes of the user information heteromorphic graph comprise user nodes and commodity information nodes related to the user, a first edge of the user information heteromorphic graph indicates an association between different users, and a second edge of the user information heteromorphic graph indicates an association between the user and the commodity information; training to obtain a user portrait construction model for constructing a user portrait based on a graph neural network according to the user nodes, the commodity information nodes, the first edge and the second edge; and adding commodity information related to the user to be constructed to the user information abnormal composition, and constructing a model by using the user portrait to construct the user portrait of the user to be constructed.
According to the technical scheme of the embodiment of the invention, the nodes and edges of the user information abnormal composition graph with various nodes are adopted to represent various data such as users, commodity information, similarity among different users, corresponding relation between the users and the commodity information and the like, and the nodes and edges of the user information abnormal composition graph are learned through a graph neural network so as to obtain a user portrait construction model; the method not only realizes simultaneous modeling of various data, but also avoids the problem of manual design of various data mixing methods or manual design characteristics, does not need a large number of data labels, and realizes quick, efficient and low-cost construction of user images.
The above-described embodiments should not be construed as limiting the scope of the invention. Those skilled in the art will appreciate that various modifications, combinations, sub-combinations, and substitutions can occur, depending on design requirements and other factors. Any modification, equivalent replacement, and improvement made within the spirit and principle of the present invention should be included in the protection scope of the present invention.
Claims (12)
1. A method of constructing a user representation, comprising:
acquiring a user information abnormal composition, wherein nodes of the user information abnormal composition comprise user nodes and commodity information nodes related to the users, a first edge of the user information abnormal composition indicates the association between different users, and a second edge of the user information abnormal composition indicates the association between the users and the commodity information;
training to obtain a user portrait construction model for constructing a user portrait based on a graph neural network according to the user nodes, the commodity information nodes, the first edge and the second edge;
and adding commodity information related to the user to be constructed to the user information abnormal composition, and constructing a model by using the user portrait to construct the user portrait of the user to be constructed.
2. A method of constructing a user representation as claimed in claim 2, wherein the picture network comprises: the system comprises a user information input layer, a commodity information representation layer, a first user representation layer, a second user representation layer and a user portrait construction layer; wherein the content of the first and second substances,
the user information input layer is used for acquiring the user information abnormal picture so as to acquire the user node, the commodity information node, the first edge and the second edge;
the user information representation layer is used for generating a commodity information vector representing the characteristics of the commodity information node;
the first user representation layer is used for generating a first user vector representing the characteristics of the user node according to the commodity information vector and the second edge;
the second user representation layer is used for generating a second user vector representing the characteristics of the user node according to the first user vector and the first edge;
and the user portrait construction layer is used for constructing the user portrait of the user node according to the second user vector.
3. The method of constructing a user representation according to claim 2, wherein said nodes of the user information composition further comprise: a commodity attribute information node for describing the commodity information; the third side of the user information heteromorphic graph indicates an association between the commodity attribute information and the commodity information.
4. The method of constructing a user representation of claim 3, wherein the graph neural network further comprises: a commodity attribute information presentation layer; wherein the content of the first and second substances,
the commodity attribute information representation layer is used for generating a commodity attribute information vector representing the characteristics of the commodity attribute information node;
and the user information representation layer is used for generating a commodity information vector representing the characteristics of the commodity information node according to the commodity attribute information vector and the third edge.
5. The method of claim 2, wherein a first user node of the user nodes in the user information composition has a user representation label, a second user node of the user nodes in the user information composition has no user representation label, and a user representation construction model for constructing a user representation is trained based on a graph neural network from the first user node, the second user node, the merchandise information node, the first edge, and the second edge.
6. An apparatus for constructing a representation of a user, comprising: the system comprises an abnormal composition acquisition module, a construction model acquisition module and a user portrait construction module; wherein the content of the first and second substances,
the system comprises an abnormal composition acquisition module, a product information acquisition module and a product information acquisition module, wherein the abnormal composition acquisition module is used for acquiring a user information abnormal composition, nodes of the user information abnormal composition comprise user nodes and product information nodes related to the users, a first edge of the user information abnormal composition indicates the association between different users, and a second edge of the user information abnormal composition indicates the association between the users and the product information;
the model construction model acquisition module is used for training to obtain a user portrait construction model used for constructing a user portrait based on a graph neural network according to the user nodes, the commodity information nodes, the first edges and the second edges;
the user portrait construction module is used for adding commodity information related to a user to be constructed to the user information abnormal composition, constructing a model by using the user portrait, and constructing the user portrait of the user to be constructed.
7. An apparatus for constructing a user representation as claimed in claim 6, wherein said graph neural network comprises: the system comprises a user information input layer, a commodity information representation layer, a first user representation layer, a second user representation layer and a user portrait construction layer; wherein the content of the first and second substances,
the user information input layer is used for acquiring the user information abnormal picture so as to acquire the user node, the commodity information node, the first edge and the second edge;
the user information representation layer is used for generating a commodity information vector representing the characteristics of the commodity information node;
the first user representation layer is used for generating a first user vector representing the characteristics of the user node according to the commodity information vector and the second edge;
the second user representation layer is used for generating a second user vector representing the characteristics of the user node according to the first user vector and the first edge;
and the user portrait construction layer is used for constructing the user portrait of the user node according to the second user vector.
8. An apparatus for constructing a user representation as claimed in claim 7, wherein said nodes of said user information profile further comprise: a commodity attribute information node for describing the commodity information; the third side of the user information heteromorphic graph indicates an association between the commodity attribute information and the commodity information.
9. The apparatus for constructing a user representation according to claim 8, wherein said graph neural network further comprises: a commodity attribute information presentation layer; wherein the content of the first and second substances,
the commodity attribute information representation layer is used for generating a commodity attribute information vector representing the characteristics of the commodity attribute information node;
and the user information representation layer is used for generating a commodity information vector representing the characteristics of the commodity information node according to the commodity attribute information vector and the third edge.
10. An apparatus for constructing a user representation as claimed in claim 7,
a first user node of the user nodes in the user information heteromorphic image has a user portrait label, and a second user node of the user nodes in the user information heteromorphic image does not have a user portrait label;
the construction model obtaining module is further configured to train to obtain a user portrait construction model for constructing a user portrait based on a graph neural network according to the first user node, the second user node, the commodity information node, the first edge and the second edge.
11. A server for building a user representation, comprising:
one or more processors;
a storage device for storing one or more programs,
when executed by the one or more processors, cause the one or more processors to implement the method of any one of claims 1-5.
12. A computer-readable medium, on which a computer program is stored, which, when being executed by a processor, carries out the method according to any one of claims 1-5.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201911053500.0A CN112749323A (en) | 2019-10-31 | 2019-10-31 | Method and device for constructing user portrait |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201911053500.0A CN112749323A (en) | 2019-10-31 | 2019-10-31 | Method and device for constructing user portrait |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN112749323A true CN112749323A (en) | 2021-05-04 |
Family
ID=75645069
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201911053500.0A Pending CN112749323A (en) | 2019-10-31 | 2019-10-31 | Method and device for constructing user portrait |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN112749323A (en) |
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112465565A (en) * | 2020-12-11 | 2021-03-09 | 加和（北京）信息科技有限公司 | User portrait prediction method and device based on machine learning |
| CN113378051A (en) * | 2021-06-16 | 2021-09-10 | 南京大学 | Crowd-sourced task recommendation method based on user-task association of graph neural network |
| CN115883147A (en) * | 2022-11-22 | 2023-03-31 | 浙江御安信息技术有限公司 | Attacker portrait drawing method based on graph neural network |
Citations (9)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20160239892A1 (en) * | 2013-07-11 | 2016-08-18 | Odd Concepts Inc. | User interest-based product information recommendation system |
| CN106874435A (en) * | 2017-01-25 | 2017-06-20 | 北京航空航天大学 | User portrait construction method and device |
| WO2017157146A1 (en) * | 2016-03-15 | 2017-09-21 | 平安科技（深圳）有限公司 | User portrait-based personalized recommendation method and apparatus, server, and storage medium |
| CN107464141A (en) * | 2017-08-07 | 2017-12-12 | 北京京东尚科信息技术有限公司 | For the method, apparatus of information popularization, electronic equipment and computer-readable medium |
| CN109242633A (en) * | 2018-09-20 | 2019-01-18 | 阿里巴巴集团控股有限公司 | A kind of commodity method for pushing and device based on bigraph (bipartite graph) network |
| CN109903117A (en) * | 2019-01-04 | 2019-06-18 | 苏宁易购集团股份有限公司 | A kind of knowledge mapping processing method and processing device for commercial product recommending |
| CN109993966A (en) * | 2018-01-02 | 2019-07-09 | 中国移动通信有限公司研究院 | A kind of method and device of building user portrait |
| WO2019157928A1 (en) * | 2018-02-13 | 2019-08-22 | 阿里巴巴集团控股有限公司 | Method and apparatus for acquiring multi-tag user portrait |
| CN110309405A (en) * | 2018-03-08 | 2019-10-08 | 腾讯科技（深圳）有限公司 | A kind of item recommendation method, device and storage medium |
-
2019
- 2019-10-31 CN CN201911053500.0A patent/CN112749323A/en active Pending
Patent Citations (9)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20160239892A1 (en) * | 2013-07-11 | 2016-08-18 | Odd Concepts Inc. | User interest-based product information recommendation system |
| WO2017157146A1 (en) * | 2016-03-15 | 2017-09-21 | 平安科技（深圳）有限公司 | User portrait-based personalized recommendation method and apparatus, server, and storage medium |
| CN106874435A (en) * | 2017-01-25 | 2017-06-20 | 北京航空航天大学 | User portrait construction method and device |
| CN107464141A (en) * | 2017-08-07 | 2017-12-12 | 北京京东尚科信息技术有限公司 | For the method, apparatus of information popularization, electronic equipment and computer-readable medium |
| CN109993966A (en) * | 2018-01-02 | 2019-07-09 | 中国移动通信有限公司研究院 | A kind of method and device of building user portrait |
| WO2019157928A1 (en) * | 2018-02-13 | 2019-08-22 | 阿里巴巴集团控股有限公司 | Method and apparatus for acquiring multi-tag user portrait |
| CN110309405A (en) * | 2018-03-08 | 2019-10-08 | 腾讯科技（深圳）有限公司 | A kind of item recommendation method, device and storage medium |
| CN109242633A (en) * | 2018-09-20 | 2019-01-18 | 阿里巴巴集团控股有限公司 | A kind of commodity method for pushing and device based on bigraph (bipartite graph) network |
| CN109903117A (en) * | 2019-01-04 | 2019-06-18 | 苏宁易购集团股份有限公司 | A kind of knowledge mapping processing method and processing device for commercial product recommending |
Non-Patent Citations (1)
| Title |
|---|
| 温昂展: "基于多源异构大数据的学者用户画像关键技术研究", 《中国硕士学位论文全文数据库》 * |
Cited By (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112465565A (en) * | 2020-12-11 | 2021-03-09 | 加和（北京）信息科技有限公司 | User portrait prediction method and device based on machine learning |
| CN112465565B (en) * | 2020-12-11 | 2023-09-26 | 加和（北京）信息科技有限公司 | User portrait prediction method and device based on machine learning |
| CN113378051A (en) * | 2021-06-16 | 2021-09-10 | 南京大学 | Crowd-sourced task recommendation method based on user-task association of graph neural network |
| CN113378051B (en) * | 2021-06-16 | 2024-03-22 | 南京大学 | User-task association crowdsourcing task recommendation method based on graph neural network |
| CN115883147A (en) * | 2022-11-22 | 2023-03-31 | 浙江御安信息技术有限公司 | Attacker portrait drawing method based on graph neural network |
| CN115883147B (en) * | 2022-11-22 | 2023-10-13 | 浙江御安信息技术有限公司 | Attacker portrait method based on graphic neural network |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN107729937B (en) | Method and device for determining user interest tag | |
| CN107944481B (en) | Method and apparatus for generating information | |
| CN110020162B (en) | User identification method and device | |
| CN112016796B (en) | Comprehensive risk score request processing method and device and electronic equipment | |
| CN112749323A (en) | Method and device for constructing user portrait | |
| CN110866040A (en) | User portrait generation method, device and system | |
| CN112925973A (en) | Data processing method and device | |
| CN108512674B (en) | Method, device and equipment for outputting information | |
| CN114493786A (en) | Information recommendation method and device | |
| CN112950321A (en) | Article recommendation method and device | |
| CN112449217B (en) | Method and device for pushing video, electronic equipment and computer readable medium | |
| US20210049665A1 (en) | Deep cognitive constrained filtering for product recommendation | |
| US20120265588A1 (en) | System and method for recommending new connections in an advertising exchange | |
| CN114780847A (en) | Object information processing and information pushing method, device and system | |
| CN110838019A (en) | Method and device for determining trial supply distribution crowd | |
| CN111125502A (en) | Method and apparatus for generating information | |
| CN114677174A (en) | Method and device for calculating sales volume of unladen articles | |
| CN114996579A (en) | Information pushing method and device, electronic equipment and computer readable medium | |
| CN114663015A (en) | Replenishment method and device | |
| CN113450172A (en) | Commodity recommendation method and device | |
| CN113627454A (en) | Article information clustering method, pushing method and pushing device | |
| CN113159877A (en) | Data processing method, device, system and computer readable storage medium | |
| CN113762994A (en) | Method and device for user operation management | |
| CN113450167A (en) | Commodity recommendation method and device | |
| CN112860858A (en) | Method and device for answering questions |
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 |