CN111859147B

CN111859147B - Object recommendation method, object recommendation device and electronic equipment

Info

Publication number: CN111859147B
Application number: CN202010756547.XA
Authority: CN
Inventors: 于海燕; 施佳子; 罗涛
Original assignee: Industrial and Commercial Bank of China Ltd ICBC
Current assignee: Industrial and Commercial Bank of China Ltd ICBC
Priority date: 2020-07-31
Filing date: 2020-07-31
Publication date: 2023-08-22
Anticipated expiration: 2040-07-31
Also published as: CN111859147A

Abstract

The disclosure provides an object recommendation method, an object recommendation device and electronic equipment, which can be used in the technical field of artificial intelligence, and the method comprises the following steps: acquiring a user identifier; matching the user identifications in a knowledge graph to obtain associated layer nodes and associated object nodes of the user nodes corresponding to the user identifications, wherein the knowledge graph comprises the user nodes, the layer nodes and the association relation between every two object nodes, and each layer node comprises at least one object node; processing the association layer node, the association object node and the user node by using the object recommendation model, and determining a first association degree between each association layer node and the user node and a second association degree between the association object node and the user node; and recommending and sorting the association layer nodes and the respective association object nodes based on the first association degree and the second association degree so as to recommend the association layer and the respective association object of the association layer aiming at the user identification.

Description

Object recommendation method, object recommendation device and electronic equipment

Technical Field

The present disclosure relates to the field of artificial intelligence, and more particularly, to an object recommendation method, an object recommendation apparatus, and an electronic device and a computer-readable storage medium.

Background

With the rapid development of electronic technology, the number of information, services, and products provided by programs (such as applications and software) to users has also shown an explosive growth trend. In order to more intuitively display information, services, products and other data to users, the related art may display information, services, products and other classifications in the form of floors.

In the process of implementing the disclosed concept, the inventor finds that at least the following problems exist in the related art: the floors ranked by manual rules do not take into account the usage habits and preferences of the users, resulting in a non-compliance of the ranking with personal preferences and poor user experience. Meanwhile, if the contents displayed in floors are ordered by rules, the displayed contents lack personalization. In addition, if a model is allocated to each floor for recommendation prediction, the model training cost is too high, and when the content of the floors is predicted by multiple models at the same time, the calculation amount is large, the time consumption is long, and the online recommendation requirement cannot be met.

Disclosure of Invention

In view of the above, the disclosure provides an object recommendation method, an object recommendation device and an electronic device that combine knowledge maps to solve the problem that content recommendation for multiple floors on line cannot be satisfied.

One aspect of the present disclosure provides an object recommendation method performed by an electronic device. The method comprises the following steps: acquiring a user identifier; matching the user identifications in a knowledge graph to obtain associated layer nodes and associated object nodes of the user nodes corresponding to the user identifications, wherein the knowledge graph comprises the user nodes, the layer nodes and the association relation between every two object nodes, and each layer node comprises at least one object node; processing the association layer node, the association object node and the user node by using the object recommendation model, and determining a first association degree between each association layer node and the user node and a second association degree between the association object node and the user node; and recommending and sorting the association layer nodes and the respective association object nodes based on the first association degree and the second association degree so as to recommend the association layer and the respective association object of the association layer aiming at the user identification.

According to an embodiment of the present disclosure, constructing a knowledge-graph includes: determining user nodes, user attribute nodes, function nodes, layer nodes, product nodes and product attribute nodes; determining a first connection relation between a user node and a user attribute node, a function node, a layer node or a product node respectively, determining a second connection relation among a plurality of function nodes, determining a third connection relation between the layer node and the function node and the product node respectively, and determining a fourth connection relation between the product node and the product attribute node; and constructing a directed graph based on the user node, the user attribute node, the function node, the layer node, the product attribute node, the first connection relationship, the second connection relationship, the third connection relationship and the fourth connection relationship to generate a knowledge graph.

According to an embodiment of the present disclosure, an object recommendation model is trained by: acquiring training data, wherein the training data comprises first sub-training data, second sub-training data and third sub-training data, the first sub-training data is user history operation data before a specified time, the second sub-training data is map data determined by extending a knowledge map by using user information in the first sub-training data, and the third sub-training data is user history operation data after the specified time; and inputting the first sub-training data and/or the second sub-training data into the object recommendation model, and adjusting model parameters of the object recommendation model so that the output of the object recommendation model approaches the third sub-training data.

Another aspect of the present disclosure provides an object recommendation apparatus performed by an electronic device. The device comprises: the system comprises an identification acquisition module, a correlation layer determination module, a correlation degree determination module and a recommendation ordering module. The identification acquisition module is used for acquiring a user identification; the association layer determining module is used for matching the user identifier in a knowledge graph to obtain association layer nodes and association object nodes of the user nodes corresponding to the user identifier, the knowledge graph comprises association relations among the user nodes, the layer nodes and the object nodes, and each layer node comprises at least one object node; the association degree determining module is used for processing the association layer nodes, the association object nodes and the user nodes by using the object recommendation model, and determining a first association degree between each association layer node and the user node and a second association degree between the association object nodes and the user nodes; and the recommendation ordering module is used for recommending and ordering the association layer nodes and the respective association object nodes based on the first association degree and the second association degree so as to recommend the association layers and the respective association objects of the association layers aiming at the user identification.

Another aspect of the present disclosure provides an electronic device, comprising: one or more processors; and a memory for storing one or more instructions that, when executed by the one or more processors, cause the one or more processors to implement the object recommendation method performed by the electronic device as described above.

Another aspect of the present disclosure provides a computer-readable storage medium having stored thereon executable instructions that, when executed by a processor, cause the processor to implement an object recommendation method performed by an electronic device as above.

Another aspect of the present disclosure provides a computer program comprising computer executable instructions which, when executed, are for implementing an object recommendation method performed by an electronic device as described above.

According to the embodiment of the disclosure, the association relationship between the user and the layer node (floor) is mined based on the knowledge graph, and the association relationship between the user and the object node (such as function and product) in the floor is mined, so that the intelligent ordering of floors according to the use habit, preference and the like of the user is realized, partial objects in the floors are recommended, the user can easily find the floors which are concerned, commonly used, the objects concerned and the like, the heterogeneous network recommendation is realized, the recommendation calculation cost is reduced, the user operation cost is reduced, and the user experience is improved.

Drawings

The above and other objects, features and advantages of the present disclosure will become more apparent from the following description of embodiments thereof with reference to the accompanying drawings in which:

fig. 1 schematically illustrates an application scenario of an object recommendation method, an object recommendation apparatus, and an electronic device according to an embodiment of the present disclosure;

FIG. 2 schematically illustrates a system architecture suitable for use in an object recommendation method, an object recommendation apparatus, and an electronic device according to an embodiment of the present disclosure;

FIG. 3 schematically illustrates a flow chart of an object recommendation method performed by an electronic device according to an embodiment of the disclosure;

FIG. 4 schematically illustrates a schematic diagram of a tier node in accordance with an embodiment of the present disclosure;

FIG. 5 schematically illustrates a flow chart of a method of constructing a knowledge-graph, in accordance with an embodiment of the disclosure;

FIG. 6 schematically illustrates a schematic diagram of a knowledge-graph, in accordance with an embodiment of the disclosure;

FIG. 7 schematically illustrates a schematic diagram of a knowledge-graph, in accordance with another embodiment of the present disclosure;

FIG. 8 schematically illustrates a schematic diagram of atlas data according to an embodiment of the disclosure;

FIG. 9 schematically illustrates a schematic diagram of recommendation ordering, according to an embodiment of the present disclosure;

FIG. 10 schematically illustrates a block diagram of an object recommendation apparatus performed by an electronic device according to an embodiment of the present disclosure; and

Fig. 11 schematically illustrates a block diagram of an electronic device adapted to perform an object recommendation method performed by the electronic device, according to an embodiment of the disclosure.

Detailed Description

Hereinafter, embodiments of the present disclosure will be described with reference to the accompanying drawings. It should be understood that the description is only exemplary and is not intended to limit the scope of the present disclosure. In the following detailed description, for purposes of explanation, numerous specific details are set forth in order to provide a thorough understanding of the embodiments of the present disclosure. It may be evident, however, that one or more embodiments may be practiced without these specific details. In addition, in the following description, descriptions of well-known structures and techniques are omitted so as not to unnecessarily obscure the concepts of the present disclosure.

The terminology used herein is for the purpose of describing particular embodiments only and is not intended to be limiting of the disclosure. The terms "comprises," "comprising," and/or the like, as used herein, specify the presence of stated features, steps, operations, and/or components, but do not preclude the presence or addition of one or more other features, steps, operations, or components.

All terms (including technical and scientific terms) used herein have the same meaning as commonly understood by one of ordinary skill in the art unless otherwise defined. It should be noted that the terms used herein should be construed to have meanings consistent with the context of the present specification and should not be construed in an idealized or overly formal manner.

Where expressions like at least one of "A, B and C, etc. are used, the expressions should generally be interpreted in accordance with the meaning as commonly understood by those skilled in the art (e.g.," a system having at least one of A, B and C "shall include, but not be limited to, a system having a alone, B alone, C alone, a and B together, a and C together, B and C together, and/or A, B, C together, etc.). Where a formulation similar to at least one of "A, B or C, etc." is used, in general such a formulation should be interpreted in accordance with the ordinary understanding of one skilled in the art (e.g. "a system with at least one of A, B or C" would include but not be limited to systems with a alone, B alone, C alone, a and B together, a and C together, B and C together, and/or A, B, C together, etc.). The terms "first," "second," and the like, are used for descriptive purposes only and are not to be construed as indicating or implying relative importance or implicitly indicating the number of technical features indicated. Thus, a feature defining "a first" or "a second" may explicitly or implicitly include one or more of the described features.

With the rapid development of internet financial business, data of information, services and products provided by internet financial application also presents explosive growth trend, so as to facilitate intuitively displaying related data to users, improve the operation convenience of users, for example, facilitate finding out required functions, products and the like more quickly, and can be displayed in a classified manner in a floor form.

For example, when recommending products, all products may be ranked, if ranking is performed according to the hotspots of multiple products, then recommending the products ranked first to the user, and then dividing the ranking result into multiple rows according to the number of products that can be displayed in one row. However, the sorting mode does not consider the requirement of the user on classification display, such as various types of products, functions and the like, which are mixed together, so that the user cannot find the required object quickly and the user experience is poor.

For example, when recommending products, the floors may be ordered based on rules, such as the category of the hot product is set at the uppermost layer, and then a recommendation model is respectively allocated to each layer to determine the ordering of the products of the layer. However, the user's preferences and habits, etc., are different, resulting in a poor experience when the ranking does not correspond to the personal preferences. In addition, this results in excessive computational effort, consumes excessive computational resources, and may not meet the online recommendation requirements. The floor is a display area of a certain type of data, belongs to a row in a list, can display related content of the type, and can be customized in different types of floor layout modes.

The embodiment of the disclosure provides an object recommendation method, an object recommendation device and electronic equipment. The method includes a correlation layer determination process and a ranking process. In the process of determining the association layer, firstly, a user identifier is obtained, then the user identifier is matched in a knowledge graph to obtain association layer nodes and association object nodes of user nodes corresponding to the user identifier, the knowledge graph comprises association relations among the user nodes, the layer nodes and the object nodes, and each layer node comprises at least one object node. After the association layer determining process is completed, an ordering process is entered, first, an object recommendation model is utilized to process association layer nodes, association object nodes and user nodes, a first association degree between each association layer node and the user nodes and a second association degree between each association object node and the user nodes are determined, and then, recommendation ordering is conducted on the association layer nodes and the respective association object nodes based on the first association degree and the second association degree so as to recommend respective association objects of the association layer and the association layer identified by the user. The knowledge graph comprises the association layer and the object information, so that the ordering of the association layer and the ordering of the objects in each association layer can be obtained through the object recommendation model, the consumed computing resources are less, and the online recommendation requirement can be met.

Fig. 1 schematically illustrates an application scenario of an object recommendation method, an object recommendation apparatus, and an electronic device according to an embodiment of the present disclosure. It should be noted that, the object recommendation method, the object recommendation device and the electronic device according to the embodiments of the present disclosure may be used in the field of artificial intelligence technology, and may also be used in various fields other than the field of artificial intelligence technology, such as the field of big data technology. The application fields of the object recommendation method, the object recommendation device and the electronic equipment in the embodiment of the disclosure are not limited.

As shown in fig. 1, an Application (APP) in a mobile phone is taken as an example, and in the application, different types of functions, products and the like can be respectively displayed on multiple floors. The first layer in fig. 1 shows financial products, such as a user may find life payment, financial products, etc. in the first layer. The second layer displays products that the user can purchase online, such as various electronic products (it should be noted that products that can purchase online, such as consumer electronics products, computer office products, camera products, earphone products, clothing products, luggage products, travel products, etc.), can be subdivided. And the third layer displays information such as financial accounting, colorful fashion, etc. The floors in the related art are usually fixed, and if consumer electronics are currently the hottest, the first floor seen by all users is consumer electronics. However, the needs of each user are different, some users may be more concerned about financial products, some users may be more related to information, some users may be more related to travel products, and thus the manner of fixing floors may not be convenient for the diversified needs of users. According to the object recommending method, the object recommending device and the electronic equipment, floor recommending of thousands of people and thousands of sides can be achieved, content in floors is ranked according to the using habit, hobbies and the like of users, users can find required products, functions and the like quickly with low cost (such as time cost and the like), and user experience is improved.

Fig. 2 schematically illustrates a system architecture suitable for use in an object recommendation method, an object recommendation apparatus, and an electronic device according to an embodiment of the present disclosure. It should be noted that fig. 2 is only an example of a system architecture to which embodiments of the present disclosure may be applied to assist those skilled in the art in understanding the technical content of the present disclosure, but does not mean that embodiments of the present disclosure may not be used in other devices, systems, environments, or scenarios.

As shown in fig. 2, a system architecture 200 according to an embodiment of the present disclosure may include terminal devices 201, 202, 203, a network 204, a server 205. The terminal devices 201, 202, 203 and the server 205 may be connected through a network 204, where the network 204 may include various connection types, such as a wired, wireless communication link, or a fiber optic cable, etc.

The terminal devices 201, 202, 203 may be provided with a display screen and/or may be provided with various programs such as client applications, software, etc., including but not limited to smartphones, tablet computers, laptop portable computers, mainframe and desktop computers, etc. The terminal device 201, 202, 203 may provide multiple functional portals to the user through a client application to meet multiple needs of the user.

According to an embodiment of the present disclosure, the terminal device 201, 202, 203 may also have a processing function to process, based on the user information, floors, functions, products, etc. to be recommended to the user. And displaying the interactive components corresponding to the recommended floors, functions and products in the display interface of the client application. Referring to fig. 1, the system may include a plurality of floors including a first floor, a second floor, etc., where the first floor may include life payment products, financial products, etc., and the second floor may include electronic products such as mobile phones, cameras, etc.

The network 204 is used as a medium to provide communication links between terminal devices and servers. The network may include various connection types, such as wired and/or wireless communication links, and the like.

The server 205 may be a server providing various services. For example, the server 205 may interact with the terminal devices 201, 202, 203 via the network 204, for example, to obtain user information obtained by the terminal devices 201, 202, 203. The server 205 may be further configured to obtain a function to be recommended to a user according to user information processing, and send the function to be recommended to the user to the terminal device 201, 202, 203, so that the terminal device 201, 202, 203 preferentially displays a function entry corresponding to the function to be recommended to the user in a running client application.

It should be noted that, the object recommendation method performed by the electronic device provided in the embodiment of the present disclosure may be performed by the terminal device 201, 202, 203 or the server 205. Accordingly, the object recommendation apparatus performed by the electronic device provided by the embodiments of the present disclosure may be provided in the terminal device 201, 202, 203 or the server 205.

It should be understood that the types and numbers of terminal devices 201, 202, 203, network 204, and server 205 in fig. 2 are merely illustrative.

Fig. 3 schematically illustrates a flowchart of an object recommendation method performed by an electronic device according to an embodiment of the disclosure.

As shown in fig. 3, the method for detecting heap memory attacks performed by the server may include operations S301 to S307.

In operation S301, a user identification is acquired. The user identification may be used to uniquely identify a user or user node, such as in a knowledge-graph.

In operation S303, the user identifier is matched in the knowledge graph, and the association layer node and the association object node of the user node corresponding to the user identifier are obtained.

In this embodiment, the knowledge graph includes association relationships among user nodes, layer nodes, and object nodes, where each layer node includes at least one object node.

Specifically, the knowledge graph can comprise user nodes and layer nodes aiming at user identification, and at least one of function nodes and product nodes, user attribute nodes and product attribute nodes aiming at a plurality of functions. The system comprises a user node, a function node, a layer node, a product node, a hierarchical relationship, an association relationship, a product node and a product attribute node, wherein the association relationship exists among the user node, the function node, the layer node and the product node, the hierarchical relationship exists among the plurality of function nodes, the association relationship exists between the user node and the user attribute node, and the association relationship exists between the product node and the product attribute node. The knowledge graph can be constructed based on historical operation data of a user, functions provided by a program, products sold and the like, such as feature data of the user, the historical operation data of the user and various related relations are displayed in the form of the knowledge graph. Thus, an object required by the user can be predicted based on the knowledge-graph.

For example, after determining the user node corresponding to the user identifier in the knowledge graph, the corresponding user node may be further diffused according to the association relationship with other nodes. And obtaining the functions possibly interested by the user according to the user characteristic information diffusion, and obtaining the associated layer nodes according to the diffused result. Compared with the recommendation algorithm in the related art, the method can solve the problem of cold start of the user in the related art.

In operation S305, the association layer node, the association object node, and the user node are processed using the object recommendation model, and a first degree of association between each association layer node and the user node, and a second degree of association between the association object node and the user node are determined.

In one embodiment, the degree of association between each user and the associated tier nodes, object nodes, etc. may be determined by an artificial intelligence model to facilitate determining the floors and products, etc. required by the user based on the degree of association.

For example, artificial intelligence models may employ neural networks, which may be heterogeneous networks, to enhance the adaptability of the model and the expressive power of the features. For example, vectorization is performed on the user node, the association layer node, the object nodes included therein and the like to obtain feature vectors, then the feature vectors are input into a trained heterogeneous network model, and association degrees (such as higher association degrees when the score is higher) of a plurality of floors, products and functions for the current user node are obtained after processing the feature vectors through the heterogeneous network model. Wherein, the higher the association degree is, the higher the interest degree of the user is.

In one embodiment, the object recommendation model is trained as follows. Firstly, training data is obtained, the training data comprises first sub-training data, second sub-training data and third sub-training data, wherein the first sub-training data is user history operation data before a specified time, the second sub-training data is map data determined by extending a knowledge map by using user information in the first sub-training data, and the third sub-training data is user history operation data after the specified time. Then, the first sub-training data and/or the second sub-training data are input into the object recommendation model, and model parameters of the object recommendation model are adjusted so that the output of the object recommendation model approaches the third sub-training data.

For example, first, training data is constructed. For each user, searching a node with an association relation with the user on the knowledge graph. The association relation comprises the following steps: user node-attribute node, user node-function node, user node-layer node, user node-product node. From these nodes, the diffusion out along the edges in the knowledge-graph is continued a specified number of times, such as 2 times.

Training data is then determined, which may include: and screening user access volume (pv) data, and screening user click events of the user clicking a function menu in a floor, the user clicking the floor and the user clicking a product in the floor according to the pv data. User purchase data, a history of user purchase products is obtained through a purchase record table (wherein the user purchase data may be weighted higher than pv data, such as doubling the user purchase data, etc.). The user clicks the floor, clicks the product, clicks the function menu, and purchases the product as the forward training sample, and generates a negative sample for the purchased product by using the floor, the product and the function menu which are not clicked by the user (the screening rule of the negative sample samples according to the popularity of the floor, the product and the function menu, and the object which is popular and not operated by the user can increase the probability of sampling, namely the weight of the object).

Then, training for deep learning is performed. The deep learning model adopts a ripple network (ripple net) based on a knowledge graph, training data is used for training the model, and a prediction function is fitted to predict probability that a user may click on floors, function menus and products, and parameter adjustment is performed in the training process, so that the model is optimal. It should be noted that a similar model may be used instead of the ripple net model.

For example, the first user history data before the first moment and the data obtained by diffusing the first user history data in the knowledge graph are input, the model output result is compared with the real user history data at the first moment, and the loss value of the current model is obtained by adopting a loss function calculation. Parameters in the current model are then adjusted based on the loss value. And finally, taking the model parameter with the accuracy rate not smaller than the preset accuracy rate as the model parameter of the object recommendation model by adjusting the parameter for a plurality of times. The loss function may be, for example, a cross entropy function, a sigmoid function, or the like, and the predetermined accuracy may be, for example, any value in the range of 70% to 95%. It is to be understood that the above-mentioned types of loss functions and the range of values of the predetermined accuracy are merely examples to facilitate understanding of the present disclosure, which is not limited thereto.

After the object recommendation model is trained, predictions can be made using the object recommendation model. For example, user nodes are obtained from the graph database through user identifiers (ids), and the user nodes are continuously outwards diffused for 2 times along the edges in the graph from the user nodes, the obtained data are input into a trained model, and the model calculates the prediction score and the user vector of the user for each floor.

In operation S307, the association layer nodes and the respective association object nodes are recommended and ordered based on the first association degree and the second association degree to recommend the association layers and the respective association objects of the association layers for the user identification.

In this embodiment, the performing recommendation ordering on the association layer node and the respective association object nodes based on the first association degree and the second association degree may include the following operations.

And firstly, carrying out first recommendation ordering on the association layer nodes based on the first association degree so as to take the association layer nodes with the first designated number ordered at the front as the layer nodes to be recommended.

And then, for each of the layer nodes to be recommended, performing second recommendation ordering on the object nodes included in the layer nodes to be recommended based on the second association degree, so that the object nodes with the second designated number ordered at the front are used as the object nodes to be recommended of the layer nodes to be recommended. For example, a floor having a score higher than a predetermined score for the user node, and functions and products in the floor may be determined as functions to be recommended to the user corresponding to the user node. Alternatively, it may be that the floor of the Top 3 (Top 3) with the highest score for the user node is determined, and the functions and products in the floor are the associated floor and its objects to be recommended to the user.

For example, the floors recommended to the user are ranked, such as by floor node score. And ordering the recommended products in the floors according to the user vectors and the distances of various products in the product vector library, so that the floor sequence and the recommendation of the products in the floors are realized through one-time calculation of the model. The distance may be euclidean distance, cosine distance, etc.

In one embodiment, illustrated as a function node included in one floor, multiple functions may be ranked from large to small according to scores (which may be determined based on vector similarity) of the multiple functions for the user node. First, the scores of the functions for the user nodes are ranked, so that the scores of the functions obtained by ranking for the user nodes are sequentially reduced. Then, a predetermined number of functions with highest scores for the user nodes among the plurality of functions obtained by the ranking are determined as functions to be recommended to the user nodes. The predetermined number may be set according to a display mode of a program installed in the electronic device, for example. For example, if the program display mode is set to be capable of displaying 6 functions at a time, the predetermined number is 6, and the functions arranged on Top 6 are taken as the functions to be recommended.

According to the object recommendation method, the correlation layer nodes and the user nodes obtained by combining the knowledge graph are processed through the neural network, the scores of a plurality of correlation layer nodes, products and functions for the user can be obtained, and the problem of cold start of the user in the related technology can be solved. Meanwhile, a model is not required to be allocated to each floor to give the sequence of each object in the floor, and the floor sequence and the sequence of each object in the floor can be obtained through an object recommendation model, so that the consumption of calculation resources is reduced, and the response speed is improved.

In one embodiment, an object node comprises: at least one of the functional nodes or the product nodes, a hierarchy exists between at least two of the functional nodes, and the product nodes have associated product attribute nodes. The user nodes have associated user attribute nodes.

Accordingly, the input of the object recommendation model includes a relevance layer node and a user node, and at least one of: functional nodes, product attribute nodes, and user attribute nodes.

For example, user attribute information for a user attribute node includes, but is not limited to: at least one of age, gender, academic, professional and/or asset status, etc. The user of the user node may be any user using a predetermined program in the electronic device.

Fig. 4 schematically illustrates a schematic diagram of a tier node according to an embodiment of the present disclosure.

As shown in fig. 4, the layer node named the first layer is a layer node for a function, and the layer node includes a plurality of functions 1 to 7. Wherein function 1 is the parent node of function 2 and function 3, function 2 is the parent node of function 4, and function 6 is the parent node of function 7. Thus, there is a hierarchy between some of the functional nodes.

In one embodiment, the knowledge-graph may be constructed as follows.

Fig. 5 schematically illustrates a flowchart of a method of constructing a knowledge-graph, in accordance with an embodiment of the disclosure.

As shown in fig. 5, constructing the knowledge graph may include operations S501 to S505.

In operation S501, a user node, a user attribute node, a function node, a layer node, a product node, and a product attribute node are determined.

In operation S503, a first connection relationship between the user node and the user attribute node, the function node, the layer node, or the product node, respectively, a second connection relationship between at least two function nodes, a third connection relationship between the layer node and the function node and the product node, respectively, and a fourth connection relationship between the product node and the product attribute node are determined.

Fig. 6 schematically shows a schematic diagram of a knowledge-graph, according to an embodiment of the disclosure.

As shown in fig. 6, the knowledge graph includes a plurality of user nodes (…, m-1, m, m+1, …), a plurality of user attributes 1, 3, a plurality of layer nodes (…, n-1, n, n+1, …), a plurality of function nodes 1 to 5, etc., a plurality of product nodes 1, 2, 3, etc., and a plurality of product attribute nodes 1, etc. The first connection is a connection between the user node and the remaining nodes. The second connection relationship is a connection relationship between the function nodes. The third connection relationship is a connection relationship between the layer node and the function node or the product node. The fourth connection relationship is a connection relationship between the product node and the product attribute node. Wherein m and n are positive integers greater than zero.

When more users use the program and more functions or products are provided by the program, the storage space occupied by the knowledge graph is overlarge. In order to solve the problem, the user node related information may be stored in a database, and when the user node related information and the corresponding layer node and the object node thereof are associated through the association relationship when the user node related information is required to be used.

Specifically, the knowledge graph includes a user database and an object graph, and data of the user node, the user attribute node, and the first connection relationship is stored in the user database.

Fig. 7 schematically illustrates a schematic diagram of a knowledge-graph, according to another embodiment of the present disclosure.

As shown in fig. 7, the knowledge graph actually includes two parts: object atlases and user databases. The user database comprises user nodes, user attribute node related information and first association relation. The first association relationship may be stored in a data structure, for example, in fig. 7, there is an association relationship between the user m-1 and the layer node of the nth layer, the function 3, the layer node of the n+1th layer, and the user attribute node of the user attribute 1.

Fig. 8 schematically shows a schematic diagram of atlas data according to an embodiment of the disclosure.

As shown in fig. 8, when object recommendation is required for a user m-1, data related to the user m-1 may be extracted from a user database, and then the extracted data is associated with layer nodes, function nodes and product nodes in an object graph to obtain a complete knowledge graph for the user m-1. This can save storage space effectively. And then, based on the association relation, using the user m-1 as a starting point in the knowledge graph to diffuse outwards for 2 times, and obtaining the data of the object recommendation model to be input.

In operation S505, a directed graph is constructed based on the user node, the user attribute node, the function node, the layer node, the product attribute node, the first connection relationship, the second connection relationship, the third connection relationship, and the fourth connection relationship to generate a knowledge graph.

In one embodiment, the tier nodes are formed as follows.

First, a first inclusion relationship and a second inclusion relationship are determined, wherein the first inclusion relationship is a relationship between a layer node and a functional node which the layer node has, and the second inclusion relationship is a relationship between the layer node and a product node which the layer node displays.

Then, a layer node is generated based on the functional node and the first containment relationship, or based on the product node and the second containment relationship.

For example, first, user attributes are extracted, the user and various attribute points are put into a graph, and edges (user node-attribute node) are constructed from the user's relationship to the attributes.

Then, the function nodes are extracted, and edges (function nodes-function nodes) are constructed according to parent-child relationships between the function nodes.

Then, a floor node is generated, and an inclusion relationship is generated from the floor node and the function node included in the floor, so that an edge (floor node-function node) is created.

Then, the products and the product attributes to be displayed in the floor are put into a map, and edges (product-attributes) are constructed according to the relationship between the products and the attributes.

Next, the floor and the edges of the product (floor-product) are established based on the floor and the containment relationship before the product is displayed on the floor.

Then, the edges (user-functions) of the user and the function node are constructed according to the history of the user click functions.

Next, the edges (user-floor) of the user and the floor node are constructed from the history of the user clicking on the floor.

Then, the user's edges with the product (user-product) are constructed from the user's records of clicking and purchasing the product.

Then, according to the above rule, a directed graph is constructed as a knowledge graph.

The knowledge graph constructed by the method can embody the user attribute, the floor, the behavior of clicking the floor by the user, the relation between the displayed product and the floor in the floor, the attribute of the displayed product on the floor, the historical behavior of clicking or purchasing the product in the floor by the user and the like, namely the knowledge graph comprises a relational network among a plurality of nodes, so that the floor, the object and the like are recommended for the user conveniently based on the knowledge graph through model analysis and calculation.

In one embodiment, after constructing the knowledge graph, the above method may further include the following operations. And inputting each node of the knowledge graph into the object recommendation model to determine a node vector of each node, wherein the first association degree and the second association degree are determined based on the similarity between the node vectors.

This facilitates the formation of product libraries, functional libraries, etc. For example, after model training is completed, the product nodes, the function nodes and the layer nodes in the knowledge graph generate corresponding vectors through the trained model, and the dimension of the generated vectors can be adjusted during model training according to the number of products. The generated vectors are stored in a database to be used as a product vector library, and when the product recommendation in the floor is made, the generated vectors are ordered according to the distance between the user vector and the product vector.

In one embodiment, matching the user identifier in the knowledge graph to obtain the association layer node and the association object node of the user node corresponding to the user identifier may include the following operations. And extending the user nodes corresponding to the user identifications in the knowledge graph based on the association relation until the appointed extension depth is reached, so as to obtain association layer nodes and association object nodes of the user nodes corresponding to the user identifications.

For example, a first diffusion is performed starting from a user node, where the first diffusion may include: and determining nodes pointed by the edges taking the user node as a starting point according to the edges connected with the user node, obtaining second nodes, and taking all the second nodes as a first diffusion result. In order to make the recommended floor, product and function coverage more comprehensive, a second diffusion may be performed with the second node as a starting point to obtain a third node. And obtaining at least one diffusion node by taking the second node and the third node … … obtained through diffusion as diffusion nodes. According to embodiments of the present disclosure, the above-described diffusion may include, for example, two, three or more times in order to compromise the diversity and accuracy of the recommendation. It is to be understood that the above number of diffusions is merely an example to facilitate understanding of the present disclosure, which is not limited thereto. The diffusion times can be determined according to actual requirements.

The following illustrates a manner of determining the input of the node vector and the object recommendation model.

The user attribute node characterizes the age of the user as an example. The age information may be converted into feature vectors using One-Hot mode. For example, the ages may be sequentially divided into categories corresponding to age groups of (10, 20), (20, 30), (30, 40), (40, 50), (50, 60), (60, 70), (70, 80), (80, 90), (90, 100), and (greater than 100).

The vector of each node needing to be input into the object recommendation model can be obtained in the mode, and then the vector is spliced into a feature matrix, and the feature matrix is used as the input of the object recommendation model.

It should be noted that, considering that the characteristic information of the user and the pointing relationships of the nodes for multiple functions in the knowledge graph are relatively determined, in order to solve the problem of inaccurate recommendation caused by drift generated by the interest of the user or short-term characteristics, the method may further include the input of the recent historical operation of the user as the recommendation model of the object. For example, last 1 day, last 3 days, last 1 month, etc. In addition, when the user history operation data is recorded in each node, for example, the number of times that the user clicks on the product node recently is recorded in the product node, a recent record may be selected from the records. Therefore, the feature matrix can better represent the requirements of the current stage of the user node, and the score obtained according to the feature matrix is more consistent with the requirements of the current node of the user. Thereby, the recommendation result can be made to satisfy timeliness.

FIG. 9 schematically illustrates a schematic diagram of recommendation ordering, according to an embodiment of the present disclosure.

As shown in fig. 9, the floors recommended by the object recommendation method include: floor 1, floor 2, floor 3, etc., wherein the association of floor 1 with the user is greater than the association of floor 2 with the user, and the association of floor 2 with the user is greater than the association of floor 3 with the user. The association of object 1 with the user is greater than the association of object 2 with the user. Two functions of life payment and financial products are recommended in the floor 1, wherein the association degree of the life payment and the user is greater than that of the financial products and the user. In floor 2, two products of a mobile phone and a camera are recommended, wherein the association degree of the mobile phone and the user is larger than that of the camera and the user. Two pieces of information, namely, the financial accounting and the colorful fashion, are recommended in the floor 3, wherein the association degree of the financial accounting and the user is larger than that of the colorful fashion and the user.

According to the embodiment of the disclosure, not only the sample data of the used function but also the sample data of other functions which are not used are obtained, so that the scores of other functions for users in the output result of the neural network model can be improved by training the neural network model according to the obtained plurality of sample data. Therefore, the diversity of the finally obtained recommended results can be improved, and the scenes such as cold start and the like can be dealt with.

Fig. 10 schematically illustrates a block diagram of an object recommendation apparatus executed by an electronic device according to an embodiment of the present disclosure.

As shown in fig. 10, an object recommendation apparatus 1000 executed by an electronic device according to an embodiment of the present disclosure. The object recommendation device 1000 may include an identification acquisition module 1010, a correlation layer determination module 1020, a degree of correlation determination module 1030, and a recommendation ordering module 1040.

The identifier acquisition module 1010 is configured to acquire a user identifier.

The association layer determining module 1020 is configured to match the user identifier in a knowledge graph to obtain an association layer node and an association object node of the user node corresponding to the user identifier, where the knowledge graph includes association relationships among the user node, the layer node and the object node, and each layer node includes at least one object node.

The association determining module 1030 is configured to process the association layer node, the association object node, and the user node by using the object recommendation model, and determine a first association degree between each association layer node and the user node, and a second association degree between the association object node and the user node.

The recommendation ordering module 1040 is configured to perform recommendation ordering on the association layer node and the respective association object node based on the first association degree and the second association degree, so as to recommend the association layer and the respective association object of the association layer for the user identification.

It should be noted that, the operations that the identifier obtaining module 1010, the association layer determining module 1020, the association degree determining module 1030 and the recommendation ordering module 1040 may be respectively executable may be the same as the relevant parts of the above method, which is not described herein.

The object recommendation device disclosed by the embodiment of the disclosure is applied to the graph database and the deep neural network, so that the relation between the user and the floor is mined, and the relation between the user and the product is mined, thereby realizing intelligent sequencing of the floors according to the use habit and preference of the user, recommending the products in the floors, enabling the user to easily find the floors concerned with the user and the products concerned with the user, realizing heterogeneous network recommendation, reducing the calculation cost, reducing the operation cost and improving the user experience.

Any number of modules, sub-modules, units, sub-units, or at least some of the functionality of any number of the sub-units according to embodiments of the present disclosure may be implemented in one module. Any one or more of the modules, sub-modules, units, sub-units according to embodiments of the present disclosure may be implemented as split into multiple modules. Any one or more of the modules, sub-modules, units, sub-units according to embodiments of the present disclosure may be implemented at least in part as a hardware circuit, such as a Field Programmable Gate Array (FPGA), a Programmable Logic Array (PLA), a system-on-chip, a system-on-substrate, a system-on-package, an Application Specific Integrated Circuit (ASIC), or in any other reasonable manner of hardware or firmware that integrates or encapsulates the circuit, or in any one of or a suitable combination of three of software, hardware, and firmware. Alternatively, one or more of the modules, sub-modules, units, sub-units according to embodiments of the present disclosure may be at least partially implemented as computer program modules, which when executed, may perform the corresponding functions.

For example, any number of the identification acquisition module 1010, the association layer determination module 1020, the association degree determination module 1030, and the recommendation ordering module 1040 may be combined in one module/unit/sub-unit or any number of the modules/units/sub-units may be split into a plurality of modules/units/sub-units. Alternatively, at least some of the functionality of one or more of these modules/units/sub-units may be combined with at least some of the functionality of other modules/units/sub-units and implemented in one module/unit/sub-unit. At least one of the identity acquisition module 1010, the association layer determination module 1020, the association degree determination module 1030, and the recommendation ordering module 1040 may be implemented at least in part as hardware circuitry, such as a Field Programmable Gate Array (FPGA), a Programmable Logic Array (PLA), a system-on-chip, a system-on-substrate, a system-on-package, an application-specific integrated circuit (ASIC), or in hardware or firmware in any other reasonable manner of integrating or packaging the circuitry, or in any one of or a suitable combination of three of software, hardware, and firmware, according to embodiments of the present disclosure. Alternatively, at least one of the identity acquisition module 1010, the association layer determination module 1020, the association degree determination module 1030, and the recommendation ordering module 1040 may be implemented at least in part as a computer program module that, when executed, performs the corresponding functions.

Fig. 11 schematically illustrates a block diagram of an electronic device adapted to perform an object recommendation method according to an embodiment of the present disclosure. The electronic device shown in fig. 11 is merely an example, and should not impose any limitations on the functionality and scope of use of embodiments of the present disclosure.

As shown in fig. 11, an electronic device 1100 according to an embodiment of the present disclosure includes a processor 1101 that can perform various appropriate actions and processes according to a program stored in a Read Only Memory (ROM) 1102 or a program loaded from a storage section 1108 into a Random Access Memory (RAM) 1103. The processor 1101 may comprise, for example, a general purpose microprocessor (e.g., a CPU), an instruction set processor and/or an associated chipset and/or a special purpose microprocessor (e.g., an Application Specific Integrated Circuit (ASIC)), or the like. The processor 1101 may also include on-board memory for caching purposes. The processor 1101 may comprise a single processing unit or a plurality of processing units for performing the different actions of the method flow according to embodiments of the present disclosure.

In the RAM 1103, various programs and data necessary for the operation of the electronic device 1100 are stored. The processor 1101, ROM 1102, and RAM 1103 are connected to each other by a bus 1104. The processor 1101 performs various operations of the method flow according to the embodiments of the present disclosure by executing programs in the ROM 1102 and/or the RAM 1103. Note that the program can also be stored in one or more memories other than the ROM 1102 and the RAM 1103. The processor 1101 may also perform various operations of the method flow according to embodiments of the present disclosure by executing programs stored in one or more memories.

According to an embodiment of the disclosure, the electronic device 1100 may also include an input/output (I/O) interface 1105, the input/output (I/O) interface 1105 also being connected to the bus 1104. The electronic device 1100 may also include one or more of the following components connected to the I/O interface 1105: an input section 1106 including a keyboard, a mouse, and the like; an output portion 1107 including a Cathode Ray Tube (CRT), a Liquid Crystal Display (LCD), and the like, a speaker, and the like; a storage section 1108 including a hard disk or the like; and a communication section 1109 including a network interface card such as a LAN card, a modem, and the like. The communication section 1109 performs communication processing via a network such as the internet. The drive 1110 is also connected to the I/O interface 1105 as needed. Removable media 1111, such as a magnetic disk, an optical disk, a magneto-optical disk, a semiconductor memory, or the like, is installed as needed in drive 1110, so that a computer program read therefrom is installed as needed in storage section 1108.

According to embodiments of the present disclosure, a method flow according to embodiments of the present disclosure may be implemented as a computer program. For example, embodiments of the present disclosure include a computer program product comprising a computer program embodied on a computer readable storage medium, the computer program comprising program code for performing the method shown in the flowcharts. In such an embodiment, the computer program can be downloaded and installed from a network via the communication portion 1109, and/or installed from the removable media 1111. The above-described functions defined in the electronic device of the embodiments of the present disclosure are performed when the computer program is executed by the processor 1101. According to embodiments of the present disclosure, the above-described electronic devices, apparatuses, means, modules, units, etc. may be implemented by computer program modules.

The present disclosure also provides a computer-readable storage medium that may be embodied in the apparatus/device/system described in the above embodiments; or may exist alone without being assembled into the apparatus/device/system. The computer-readable storage medium carries one or more programs which, when executed, implement methods in accordance with embodiments of the present disclosure.

According to embodiments of the present disclosure, the computer-readable storage medium may be a non-volatile computer-readable storage medium, which may include, for example, but is not limited to: 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), 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 context of this disclosure, 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. For example, according to embodiments of the present disclosure, the computer-readable storage medium may include ROM 1102 and/or RAM 1103 described above and/or one or more memories other than ROM 1102 and RAM 1103.

The flowcharts 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 disclosure. 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.

Those skilled in the art will appreciate that the features recited in the various embodiments of the disclosure and/or in the claims may be provided in a variety of combinations and/or combinations, even if such combinations or combinations are not explicitly recited in the disclosure. In particular, the features recited in the various embodiments of the present disclosure and/or the claims may be variously combined and/or combined without departing from the spirit and teachings of the present disclosure. All such combinations and/or combinations fall within the scope of the present disclosure.

The embodiments of the present disclosure are described above. However, these examples are for illustrative purposes only and are not intended to limit the scope of the present disclosure. Although the embodiments are described above separately, this does not mean that the measures in the embodiments cannot be used advantageously in combination. The scope of the disclosure is defined by the appended claims and equivalents thereof. Various alternatives and modifications can be made by those skilled in the art without departing from the scope of the disclosure, and such alternatives and modifications are intended to fall within the scope of the disclosure.

Claims

1. An object recommendation method performed by an electronic device, comprising:

acquiring a user identifier;

constructing a knowledge graph;

matching the user identifications in the knowledge graph to obtain association layer nodes and association object nodes of user nodes corresponding to the user identifications, wherein the knowledge graph comprises association relations among the user nodes, the layer nodes and the object nodes, and each layer node comprises at least one object node;

processing the association layer node, the association object node and the user node by using an object recommendation model, and determining a first association degree between each association layer node and the user node and a second association degree between the association object node and the user node; and

The relevance layer nodes and the respective relevance object nodes are recommended and ordered based on the first relevance degree and the second relevance degree to recommend relevance layers and relevance objects of relevance layers for the user identification,

wherein the object node comprises: at least one of a functional node or a product node, wherein a hierarchical structure exists between at least two functional nodes, and the product node is provided with an associated product attribute node;

the user nodes have associated user attribute nodes; and

the input of the object recommendation model includes the association layer node and the user node, and at least one of: functional nodes, product attribute nodes and user attribute nodes,

and wherein said constructing said knowledge-graph comprises:

determining user nodes, user attribute nodes, function nodes, layer nodes, product nodes and product attribute nodes;

determining a first connection relation between the user node and the user attribute node, the function node, the layer node or the product node respectively, determining a second connection relation between at least two function nodes, determining a third connection relation between the layer node and the function node and the product node respectively, and determining a fourth connection relation between the product node and the product attribute node; and

Constructing a directed graph based on the user node, the user attribute node, the function node, the tier node, the product attribute node, the first connection relationship, the second connection relationship, the third connection relationship, and the fourth connection relationship to generate the knowledge graph,

wherein the layer node is formed by:

determining a first inclusion relationship and a second inclusion relationship, wherein the first inclusion relationship is a relationship between a layer node and a functional node of the layer node, and the second inclusion relationship is a relationship between the layer node and a product node displayed by the layer node; and

the tier node is generated based on the functional node and the first containment relationship, or based on the product node and the second containment relationship.

2. The method of claim 1, wherein the knowledge-graph comprises a user database and an object-graph, data of the user nodes, the user attribute nodes, and the first connection relationship being stored in the user database.

3. The method of claim 1, further comprising: after the construction of the knowledge-graph,

And inputting each node of the knowledge graph into the object recommendation model to determine a node vector of each node, wherein the first association degree and the second association degree are determined based on the similarity between the node vectors.

4. The method of claim 1, wherein the matching the user identifier in a knowledge graph to obtain an association layer node and an association object node of a user node corresponding to the user identifier comprises:

and extending the user node corresponding to the user identifier in the knowledge graph based on the association relation until reaching a specified extension depth to obtain an association layer node and an association object node of the user node corresponding to the user identifier.

5. The method of claim 1, wherein the object recommendation model is trained by:

acquiring training data, wherein the training data comprises first sub-training data, second sub-training data and third sub-training data, the first sub-training data is user history operation data before a specified time, the second sub-training data is map data determined by extending the knowledge maps by using user information in the first sub-training data, and the third sub-training data is user history operation data after the specified time; and

The first sub-training data and/or the second sub-training data are input to the object recommendation model, and model parameters of the object recommendation model are adjusted such that the output of the object recommendation model approaches the third sub-training data.

6. The method of claim 1, wherein the recommended ordering of the association layer nodes and the respective association object nodes based on the first degree of association and the second degree of association comprises:

performing first recommendation ordering on the associated layer nodes based on the first association degree, so as to take the associated layer nodes with the first designated number ordered at the front as layer nodes to be recommended; and

and for each to-be-recommended layer node, performing second recommendation ordering on the object nodes included in the to-be-recommended layer node based on the second association degree, so as to take the object nodes with the second designated number ordered at the front as to-be-recommended object nodes of the to-be-recommended layer node.

7. An object recommendation apparatus performed by an electronic device, comprising:

the identification acquisition module is used for acquiring the user identification;

the association layer determining module is used for constructing a knowledge graph, matching the user identifications in the knowledge graph to obtain association layer nodes and association object nodes of the user nodes corresponding to the user identifications, wherein the knowledge graph comprises association relations among the user nodes, the layer nodes and the object nodes, and each layer node comprises at least one object node;

The association degree determining module is used for processing the association layer nodes, the association object nodes and the user nodes by using an object recommendation model, and determining a first association degree between each association layer node and the user node and a second association degree between the association object nodes and the user nodes; and

a recommendation ordering module for performing recommendation ordering on the association layer nodes and the respective association object nodes based on the first association degree and the second association degree to recommend the association layer and the association object of the association layer for the user identification,

the user nodes have associated user attribute nodes; and

and wherein said constructing said knowledge-graph comprises:

wherein the layer node is formed by:

8. An electronic device, comprising:

one or more processors; and

storage means for storing one or more programs,

wherein the one or more programs, when executed by the one or more processors, cause the one or more processors to implement the method of any of claims 1-6.

9. A computer readable storage medium having stored thereon executable instructions which when executed by a processor cause the processor to implement the method of any of claims 1 to 6.