Product recommendation method and device based on knowledge graph
By constructing customer and product knowledge graphs based on knowledge graph methods and calculating similarity using the TransR model, the diversity and cold start problems of recommendation systems are solved, personalized product recommendations are realized, and recommendation performance and customer activity are improved.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- INDUSTRIAL AND COMMERCIAL BANK OF CHINA
- Filing Date
- 2022-08-19
- Publication Date
- 2026-06-12
Smart Images
Figure CN115344783B_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of artificial intelligence technology, and more specifically, to a product recommendation method and apparatus based on knowledge graphs. Background Technology
[0002] Currently, massive amounts of data lead to information overload, prompting the development of solutions such as search and recommendation to extract information of interest to customers. In the era of big data, traditional recommendation systems are no longer the preferred approach due to their inherent limitations. The core of a recommendation system is discovering user preferences based on user item information; however, its over-reliance on user item data often results in less than ideal recommendation performance. Knowledge graphs, on the other hand, possess excellent capabilities for representing semantic relationships within knowledge. They can integrate multi-source data for semantic association, thereby improving recommendation effectiveness.
[0003] Current recommendation algorithms suffer from problems such as a lack of diversity in recommendation results, cold start, and a lack of collaborative understanding of customers or items. For example, if products are recommended based solely on customer behavior such as browsing history or recent product popularity and benefits, the recommended products may not be suitable for the current customer, resulting in a lack of personalized and persuasive recommendation services. Summary of the Invention
[0004] The main objective of this invention is to provide a product recommendation method and apparatus based on knowledge graphs, which can accurately provide product recommendations to customers, thereby offering personalized services and activating customer activity.
[0005] To achieve the above objectives, embodiments of the present invention provide a product recommendation method based on a knowledge graph, comprising:
[0006] A customer knowledge graph is constructed based on customer characteristics, and a product knowledge graph is constructed based on product characteristics.
[0007] Obtain a customer preference vector representation based on the customer knowledge graph;
[0008] The similarity between the corresponding product and the products held by the customer is defined as the customer preference similarity.
[0009] The similarity between each product and the products held by the customer is determined by the product knowledge graph and is called the customer product similarity.
[0010] The similarity of the target product is determined based on the similarity of customer preferences and customer products, and the target product is pushed based on the similarity of the target product.
[0011] In one embodiment, obtaining a customer preference vector representation based on a customer knowledge graph includes:
[0012] Obtain positive example triples from the customer's knowledge graph;
[0013] The customer preference vector representation is obtained from the positive example triples of the customer knowledge graph.
[0014] In one embodiment, obtaining a customer preference vector representation based on positive example triples from a customer knowledge graph includes:
[0015] A knowledge representation model is built based on positive example triples from the client's knowledge graph and a pre-defined TransR model;
[0016] Generate negative triples of the customer's knowledge graph based on positive triples of the customer's knowledge graph;
[0017] Train positive and negative triples of the customer knowledge graph based on the knowledge representation model to obtain a customer preference vector representation.
[0018] In one embodiment, obtaining positive example triples from the customer knowledge graph includes:
[0019] Extract resource description framework feature data from the customer knowledge graph using the resource description framework;
[0020] Obtain positive example triples of customer knowledge graph based on feature data from the resource description framework.
[0021] In one embodiment, determining the similarity between each product and the customer's held products based on the product knowledge graph includes:
[0022] Based on the product knowledge graph, determine the product density vector, the direct product relationship similarity between each product and the products held by the customer, and the indirect product relationship similarity.
[0023] Customer product similarity is determined based on product density vectors, direct product relationship similarity, and indirect product relationship similarity.
[0024] In one embodiment, determining the product dense vector based on the product knowledge graph includes:
[0025] The product knowledge graph is mapped to a continuous vector space using a pre-defined TransR model to obtain dense vectors of products.
[0026] This invention also provides a product recommendation device based on a knowledge graph, comprising:
[0027] The product knowledge graph construction module is used to construct customer knowledge graphs based on customer characteristics and product knowledge graphs based on product characteristics.
[0028] The customer preference vector representation module is used to obtain customer preference vector representations based on the customer knowledge graph.
[0029] The customer preference similarity module is used to determine the similarity between the product corresponding to the customer preference vector and the products held by the customer as the customer preference similarity.
[0030] The customer product similarity module is used to determine the similarity between each product and the customer's owned products based on the product knowledge graph.
[0031] The target product push module is used to determine the similarity of target products based on customer preference similarity and customer product similarity, and then push target products based on the similarity.
[0032] In one embodiment, the customer preference vector representation module includes:
[0033] Positive example triplet unit, used to obtain positive example triplets of customer knowledge graph based on customer knowledge graph;
[0034] The customer preference vector representation unit is used to obtain the customer preference vector representation based on the positive example triples of the customer knowledge graph.
[0035] In one embodiment, the customer preference vector representation unit includes:
[0036] The knowledge representation model building subunit is used to build a knowledge representation model based on the positive example triples of the customer's knowledge graph and the preset TransR model;
[0037] The negative example triplet subunit is used to generate negative example triplets of the customer's knowledge graph based on the positive example triplets of the customer's knowledge graph.
[0038] The customer preference vector representation subunit is used to train positive example triples and negative example triples of the customer knowledge graph based on the knowledge representation model to obtain the customer preference vector representation.
[0039] In one embodiment, the positive example triplet unit includes:
[0040] The Resource Description Framework Feature Data Subunit is used to extract Resource Description Framework feature data from the customer's knowledge graph through the Resource Description Framework.
[0041] The positive example triplet subunit is used to obtain positive example triplets of the customer knowledge graph based on the feature data of the resource description framework.
[0042] In one embodiment, the customer product similarity module includes:
[0043] The product similarity unit is used to determine the product density vector, the direct product relationship similarity, and the indirect product relationship similarity between each product and the products held by the customer, based on the product knowledge graph.
[0044] The customer product similarity unit is used to determine customer product similarity based on product density vectors, direct product relationship similarity, and indirect product relationship similarity.
[0045] In one embodiment, the product similarity unit is specifically used for:
[0046] The product knowledge graph is mapped to a continuous vector space using a pre-defined TransR model to obtain dense vectors of products.
[0047] This invention also provides a computer device, including a memory, a processor, and a computer program stored in the memory and running on the processor, wherein the processor executes the computer program to implement the steps of the knowledge graph-based product recommendation method.
[0048] This invention also provides a computer-readable storage medium storing a computer program thereon, which, when executed by a processor, implements the steps of the knowledge graph-based product recommendation method.
[0049] This invention also provides a computer program product, including a computer program / instruction, wherein when the computer program / instruction is executed by a processor, it implements the steps of the knowledge graph-based product recommendation method.
[0050] The product recommendation method and apparatus based on knowledge graphs in this invention obtains customer preference vector representations based on customer knowledge graphs constructed from customer characteristics to determine customer preference similarity, determines customer product similarity based on product knowledge graphs constructed from product characteristics, and determines target product similarity based on customer preference similarity and customer product similarity to push target products. This can provide customers with accurate product recommendations, thereby providing personalized services to customers and activating customer activity. Attached Figure Description
[0051] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.
[0052] Figure 1 This is a flowchart of the product recommendation method based on knowledge graphs in an embodiment of the present invention;
[0053] Figure 2 This is a flowchart of a product recommendation method based on knowledge graphs in another embodiment of the present invention;
[0054] Figure 3 This is a flowchart of S102 in an embodiment of the present invention;
[0055] Figure 4 This is a flowchart of S201 in an embodiment of the present invention;
[0056] Figure 5 This is a flowchart of S202 in an embodiment of the present invention;
[0057] Figure 6 This is a flowchart of S104 in an embodiment of the present invention;
[0058] Figure 7 This is a structural block diagram of the knowledge graph-based product recommendation device in an embodiment of the present invention;
[0059] Figure 8 This is a structural block diagram of the computer device in an embodiment of the present invention. Detailed Implementation
[0060] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.
[0061] Those skilled in the art will recognize that embodiments of the present invention can be implemented as a system, apparatus, device, method, or computer program product. Therefore, this disclosure can be specifically implemented in the following forms: entirely hardware, entirely software (including firmware, resident software, microcode, etc.), or a combination of hardware and software.
[0062] The acquisition, storage, use, and processing of data in the technical solution of this invention all comply with the relevant provisions of national laws and regulations.
[0063] Given the current limitations of product recommendation solutions, such as a lack of diversity in recommendation results, cold start issues, and a lack of collaborative understanding of customers or items, this invention provides a product recommendation method and system based on knowledge graphs. Building upon knowledge graphs and personalized recommendation technologies, it proposes a knowledge graph-based product recommendation scheme that incorporates product-specific features and customer preferences. This involves learning associative representations of various knowledge points within the knowledge graph, projecting the knowledge into a vector space as entity relationship vectors. This accurately represents the semantic connections between products and uncovers customer preferences. Distance metrics are used to derive the similarity between product entities, which, combined with product similarity based on customer preferences, simultaneously considers both product semantic information and customer preferences. This allows for personalized and precise product recommendation services, thereby enhancing customer engagement. The invention will be described in detail below with reference to the accompanying drawings.
[0064] Figure 1 This is a flowchart of a product recommendation method based on knowledge graphs in an embodiment of the present invention. Figure 2 This is a flowchart of a product recommendation method based on knowledge graphs in another embodiment of the present invention. For example... Figures 1-2 As shown, product recommendation methods based on knowledge graphs include:
[0065] S101: Construct a customer knowledge graph based on customer characteristics, and construct a product knowledge graph based on product characteristics.
[0066] In practice, a customer knowledge graph can be constructed based on customer social attributes (including customer age and income), customer historical transaction behavior, and customer held products, while a product knowledge graph can be constructed based on product basic information, product performance, and other product characteristics.
[0067] S102: Obtain a customer preference vector representation based on the customer knowledge graph.
[0068] Figure 3 This is a flowchart of S102 in an embodiment of the present invention. For example... Figure 3 As shown, S102 includes:
[0069] S201: Obtain positive example triples from the customer's knowledge graph.
[0070] Figure 4 This is a flowchart of S201 in an embodiment of the present invention. For example... Figure 4 As shown, S201 includes:
[0071] S301: Extract resource description framework feature data from the customer knowledge graph through the resource description framework.
[0072] RDF (Resource Description Framework) is a resource description framework that provides a unified standard for describing entities / resources, represented as SPO (Subject-Predication-Object) triples. An RDF consists of nodes and edges. Nodes represent entities / resources and attributes, while edges represent relationships between entities and between entities and attributes. Specifically, using customers in a knowledge graph as nodes, customer social attributes or relationship attributes are used as edges, connecting related nodes. These attributes or relationships include customer age, customer income, customer product transaction behavior, and customer product holdings, among others.
[0073] S302: Obtain positive example triples of customer knowledge graph based on resource description framework feature data.
[0074] In the form of knowledge graph triples, the basic units are <entity 1, relation, entity 2> and <entity, attribute 1, attribute value>. Entities are the foundation of the knowledge graph; each entity is unique within the graph. Relationships connect two different entities, such as (Customer A, Holding, Product B), where Customer A and Product B are two different entities, and Holding represents the association between these two entities. Attributes refer to the characteristics or features an entity possesses, and attribute values are the values of specific attributes of an entity, such as (Customer A, Age, 26 years old), where Customer A represents an entity, Age represents the customer attribute, and 26 years old represents the age value.
[0075] S202: Obtain the customer preference vector representation based on the positive example triples of the customer knowledge graph.
[0076] Figure 5 This is a flowchart of S202 in an embodiment of the present invention. For example... Figure 5 As shown, S202 includes:
[0077] S401: Build a knowledge representation model based on positive example triples from the customer's knowledge graph and the preset TransR model.
[0078] In practice, the TransR model can be used to efficiently learn the association representation of knowledge in positive example triples of the customer's knowledge graph, mapping entities and relations to a vector space for vectorized representation, thereby establishing a knowledge representation model.
[0079] S402: Generate negative triples of the customer knowledge graph based on positive triples of the customer knowledge graph.
[0080] S403: Train positive example triples and negative example triples of the customer knowledge graph based on the knowledge representation model to obtain the customer preference vector representation.
[0081] S103: Determine the similarity between the product corresponding to the customer preference vector and the product held by the customer as the customer preference similarity.
[0082] In practice, Euclidean distance can be used to calculate the similarity of product entities in the knowledge graph. The entity relationships in the knowledge graph are embedded in a vectorized space and represented as multi-dimensional vectors. Euclidean distance measures the absolute distance between points in this multi-dimensional space, providing a more accurate estimate of the similarity of product entities in the knowledge graph. The vectorized representation of a customer's held products is an n-dimensional vector: F h =(e 1h ,e 1h ,……e jh ...e nh ) T The vector representation of the customer preference vector corresponding to the product is an n-dimensional vector: F k =(e1k ,e 1k ,……e jk ...e nk ) T e jh For the h-th product entity (the product held by the customer), the j-th dimension attribute is e. jk For the k-th product entity (a product not held by a customer), the attribute of the j-th dimension is denoted as .
[0083] The customer preference vector represents the distance between the corresponding product and the products held by the customer:
[0084]
[0085] The result obtained from Euclidean distance is usually a number greater than zero. In order to compare similarity on the same scale, the distance result needs to be normalized:
[0086]
[0087] Where, sim c (F h ,F k This represents the similarity of customer preferences.
[0088] According to the above formula, the greater the distance between product entities, the smaller the similarity, and vice versa.
[0089] S104: Determine the similarity between each product and the customer's held products based on the product knowledge graph as the customer product similarity.
[0090] Product similarity refers to the degree of similarity between two product entities in a knowledge graph. For product entities, direct relationships alone are insufficient to distinguish the differences in similarity between two products with the same relationship type as the target product entity. Therefore, this invention considers establishing indirect relationships between product entities. That is, an indirect relationship is defined as two product entities with a direct relationship both having the same direct relationship with a third-party product entity. By mining the potential implicit relationships between two product entities, the accuracy of calculating product similarity is increased.
[0091] Figure 6 This is a flowchart of S104 in an embodiment of the present invention. For example... Figure 6 As shown, S104 includes:
[0092] S501: Determine the product density vector, the direct product relationship similarity, and the indirect product relationship similarity between each product and the products held by the customer based on the product knowledge graph.
[0093] In one embodiment, determining the dense vector of a product based on the product knowledge graph includes: mapping the product knowledge graph to a continuous vector space using a preset TransR model to obtain the dense vector of the product.
[0094] The product similarity calculation process considers the influence of both entity semantics and relationships. Before calculating product similarity, the TransR model is used to map the constructed product knowledge graph to a continuous vector space to obtain an n-dimensional dense vector of products, that is, to represent product entities and relationships as vectors.
[0095] Table 1
[0096]
[0097] Table 1 is a diagram of product similarity. As shown in Table 1, four types of relationships between product entities are established and defined during the construction of the product knowledge graph. The higher the similarity value, the higher the similarity between the two entities corresponding to the relationship.
[0098] S502: Determine customer product similarity based on product density vectors, direct product relationship similarity, and indirect product relationship similarity.
[0099] In practice, the similarity of customer products can be determined using the following formula:
[0100]
[0101] Where, sim p (F h ,F k ) represents the h-th product entity F h (Customer-held products) and the kth product entity F k The similarity between (non-customer-owned products), i.e., customer-product similarity, where i is the product ID. e jh For the h-th product entity, e is the attribute of the j-th dimension. jk For the attribute of the j-th dimension of the k-th product entity, e jh and e jk All are product-dense vectors. s g (F h ,F k Let F be the h-th product entity determined according to the product similarity table. h With the k-th product entity F k The product direct relationship similarity is given by μ, where the direct relationship between two product entities is unique. μ is the relationship coefficient, where 0 < μ < 1.
[0102] R hk ·S T For the h-th product entity F h With the k-th product entity Fk The similarity of indirect product relationships. T This represents a similarity value vector for the relationships between product entities; for example, the similarity value for the relationships between four product entities is S. T =(4,3,2,1) T .
[0103] R hk Represents the h-th product entity F h With the k-th product entity F k Is there an indirect relationship between them? hk = (1, 0, ..., 1), where 0 represents the h-th product entity F. h and the kth product entity F k It does not have the same direct relationship with any third-party product entity, where 1 represents the h-th product entity F. h and the kth product entity F k All of them have the same direct relationship with third-party product entities.
[0104] For example, R hk = (1,0,1), then the h-th product entity F h and the kth product entity F k The h-th product entity F does not have the same direct relationship with the second third-party product entity. h and the kth product entity F k All of them have the same direct relationship with the first and third third-party product entities.
[0105] S105: Determine the target product similarity based on customer preference similarity and customer product similarity, and push the target product based on the target product similarity.
[0106] In practice, the similarity of target products can be determined using the following formula:
[0107] sim(F h ,F k )=ε·sim p (F h ,F k )+(1-ε)·sim c (F h ,F k );
[0108] Where, sim(F) h ,F k ) represents the h-th product entity F h (Customer-held products) and the kth product entity F k The similarity of target products between (non-customer-owned products), where ε is the target coefficient.
[0109] After determining the similarity of the target products, the products can be arranged in descending order of similarity, and the top Z products can be selected as the target products to generate a target product recommendation list and push it to the customer.
[0110] Figure 1 The product recommendation method based on knowledge graphs shown can be implemented by a computer. Figure 1 As shown in the process, the product recommendation method based on knowledge graphs in this embodiment of the invention obtains customer preference vector representations based on customer knowledge graphs constructed from customer characteristics to determine customer preference similarity, determines customer product similarity based on product knowledge graphs constructed from product characteristics, and determines target product similarity based on customer preference similarity and customer product similarity to push target products. This can provide customers with accurate product recommendations, thereby providing personalized services to customers and activating customer activity.
[0111] The specific process of this invention embodiment is as follows:
[0112] 1. Construct a customer knowledge graph based on customer characteristics, and construct a product knowledge graph based on product characteristics.
[0113] 2. Extract resource description framework feature data from the customer knowledge graph through the resource description framework, and obtain positive example triples of the customer knowledge graph based on the resource description framework feature data.
[0114] 3. Establish a knowledge representation model based on the positive example triples of the customer's knowledge graph and the preset TransR model.
[0115] 4. Generate negative triples for the customer's knowledge graph based on the positive triples of the customer's knowledge graph.
[0116] 5. Train positive example triples and negative example triples of the customer knowledge graph based on the knowledge representation model to obtain the customer preference vector representation.
[0117] 6. Determine the similarity between the corresponding product and the product held by the customer as the customer preference similarity.
[0118] 7. Based on the product knowledge graph, determine the product density vector, the direct product relationship similarity between each product and the products held by the customer, and the indirect product relationship similarity.
[0119] 8. Determine customer product similarity based on product density vectors, direct product relationship similarity, and indirect product relationship similarity.
[0120] 9. Determine the target product similarity based on customer preference similarity and customer product similarity, and push the target product based on the target product similarity.
[0121] In summary, the knowledge graph-based product recommendation method of this invention constructs a knowledge graph to mine implicit relationships among various related information. With the help of knowledge graph construction, knowledge representation and reasoning technologies, it takes the customer as the center, associates customer information, product information and behavioral information, provides multi-perspective analysis capabilities, personalizes product recommendation services for customers, and activates customer activity.
[0122] Based on the same inventive concept, this invention also provides a product recommendation device based on a knowledge graph. Since the principle of this device in solving the problem is similar to that of the product recommendation method based on a knowledge graph, the implementation of this device can refer to the implementation of the method, and the repeated parts will not be described again.
[0123] Figure 7 This is a structural block diagram of a product recommendation device based on a knowledge graph, as described in an embodiment of the present invention. Figure 7 As shown, the knowledge graph-based product recommendation device includes:
[0124] The product knowledge graph construction module is used to construct customer knowledge graphs based on customer characteristics and product knowledge graphs based on product characteristics.
[0125] The customer preference vector representation module is used to obtain customer preference vector representations based on the customer knowledge graph.
[0126] The customer preference similarity module is used to determine the similarity between the product corresponding to the customer preference vector and the products held by the customer as the customer preference similarity.
[0127] The customer product similarity module is used to determine the similarity between each product and the customer's owned products based on the product knowledge graph.
[0128] The target product push module is used to determine the similarity of target products based on customer preference similarity and customer product similarity, and then push target products based on the similarity.
[0129] In one embodiment, the customer preference vector representation module includes:
[0130] Positive example triplet unit, used to obtain positive example triplets of customer knowledge graph based on customer knowledge graph;
[0131] The customer preference vector representation unit is used to obtain the customer preference vector representation based on the positive example triples of the customer knowledge graph.
[0132] In one embodiment, the customer preference vector representation unit includes:
[0133] The knowledge representation model building subunit is used to build a knowledge representation model based on the positive example triples of the customer's knowledge graph and the preset TransR model;
[0134] The negative example triplet subunit is used to generate negative example triplets of the customer's knowledge graph based on the positive example triplets of the customer's knowledge graph.
[0135] The customer preference vector representation subunit is used to train positive example triples and negative example triples of the customer knowledge graph based on the knowledge representation model to obtain the customer preference vector representation.
[0136] In one embodiment, the positive example triplet unit includes:
[0137] The Resource Description Framework Feature Data Subunit is used to extract Resource Description Framework feature data from the customer's knowledge graph through the Resource Description Framework.
[0138] The positive example triplet subunit is used to obtain positive example triplets of the customer knowledge graph based on the feature data of the resource description framework.
[0139] In one embodiment, the customer product similarity module includes:
[0140] The product similarity unit is used to determine the product density vector, the direct product relationship similarity, and the indirect product relationship similarity between each product and the products held by the customer, based on the product knowledge graph.
[0141] The customer product similarity unit is used to determine customer product similarity based on product density vectors, direct product relationship similarity, and indirect product relationship similarity.
[0142] In one embodiment, the product similarity unit is specifically used for:
[0143] The product knowledge graph is mapped to a continuous vector space using a pre-defined TransR model to obtain dense vectors of products.
[0144] In summary, the product recommendation device based on knowledge graphs in this invention obtains customer preference vector representations based on customer knowledge graphs constructed from customer characteristics to determine customer preference similarity, determines customer product similarity based on product knowledge graphs constructed from product characteristics, and determines target product similarity based on customer preference similarity and customer product similarity to push target products. This can provide customers with accurate product recommendations, thereby providing personalized services to customers and activating customer activity.
[0145] This invention also provides a specific implementation of a computer device capable of implementing all the steps in the knowledge graph-based product recommendation method described above. Figure 8 This is a structural block diagram of the computer device in an embodiment of the present invention, see below. Figure 8 The computer equipment specifically includes the following:
[0146] Processor 801 and memory 802.
[0147] The processor 801 is used to call the computer program in the memory 802. When the processor executes the computer program, it implements all the steps in the knowledge graph-based product recommendation method in the above embodiments. For example, when the processor executes the computer program, it implements the following steps:
[0148] A customer knowledge graph is constructed based on customer characteristics, and a product knowledge graph is constructed based on product characteristics.
[0149] Obtain a customer preference vector representation based on the customer knowledge graph;
[0150] The similarity between the corresponding product and the products held by the customer is defined as the customer preference similarity.
[0151] The similarity between each product and the products held by the customer is determined by the product knowledge graph and is called the customer product similarity.
[0152] The similarity of the target product is determined based on the similarity of customer preferences and customer products, and the target product is pushed based on the similarity of the target product.
[0153] In summary, the computer device of this invention obtains customer preference vector representations based on customer knowledge graphs constructed from customer characteristics to determine customer preference similarity, determines customer product similarity based on product knowledge graphs constructed from product characteristics, and determines target product similarity based on customer preference similarity and customer product similarity to push target products. This can provide customers with accurate product recommendations, thereby providing personalized services to customers and activating customer activity.
[0154] This invention also provides a computer-readable storage medium capable of implementing all steps of the knowledge graph-based product recommendation method in the above embodiments. The computer-readable storage medium stores a computer program that, when executed by a processor, implements all steps of the knowledge graph-based product recommendation method in the above embodiments. For example, when the processor executes the computer program, it implements the following steps:
[0155] A customer knowledge graph is constructed based on customer characteristics, and a product knowledge graph is constructed based on product characteristics.
[0156] Obtain a customer preference vector representation based on the customer knowledge graph;
[0157] The similarity between the corresponding product and the products held by the customer is defined as the customer preference similarity.
[0158] The similarity between each product and the products held by the customer is determined by the product knowledge graph and is called the customer product similarity.
[0159] The similarity of the target product is determined based on the similarity of customer preferences and customer products, and the target product is pushed based on the similarity of the target product.
[0160] In summary, the computer-readable storage medium of this invention obtains customer preference vector representations based on a customer knowledge graph constructed from customer characteristics to determine customer preference similarity, determines customer product similarity based on a product knowledge graph constructed from product characteristics, and determines target product similarity based on customer preference similarity and customer product similarity to push target products. This can provide customers with accurate product recommendations, thereby providing personalized services to customers and activating customer activity.
[0161] This invention also provides a computer program product capable of implementing all steps of the knowledge graph-based product recommendation method in the above embodiments. The computer program product includes a computer program / instruction, which, when executed by a processor, implements all steps of the knowledge graph-based product recommendation method in the above embodiments. For example, when the processor executes the computer program, it implements the following steps:
[0162] A customer knowledge graph is constructed based on customer characteristics, and a product knowledge graph is constructed based on product characteristics.
[0163] Obtain a customer preference vector representation based on the customer knowledge graph;
[0164] The similarity between the corresponding product and the products held by the customer is defined as the customer preference similarity.
[0165] The similarity between each product and the products held by the customer is determined by the product knowledge graph and is called the customer product similarity.
[0166] The similarity of the target product is determined based on the similarity of customer preferences and customer products, and the target product is pushed based on the similarity of the target product.
[0167] In summary, the computer program product of this invention obtains customer preference vector representations based on customer knowledge graphs constructed from customer characteristics to determine customer preference similarity, determines customer product similarity based on product knowledge graphs constructed from product characteristics, and determines target product similarity based on customer preference similarity and customer product similarity to push target products. This can provide customers with accurate product recommendations, thereby providing personalized services to customers and activating customer activity.
[0168] The specific embodiments described above further illustrate the purpose, technical solution, and beneficial effects of the present invention. It should be understood that the above descriptions are merely specific embodiments of the present invention and are not intended to limit the scope of protection of the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.
[0169] Those skilled in the art will also understand that the various illustrative logical blocks, units, and steps listed in the embodiments of the present invention can be implemented by electronic hardware, computer software, or a combination of both. To clearly demonstrate the interchangeability of hardware and software, the functions of the various illustrative components, units, and steps described above have been generally described. Whether such functionality is implemented through hardware or software depends on the specific application and the overall system design requirements. Those skilled in the art can implement the described functions using various methods for each specific application, but such implementation should not be construed as exceeding the scope of protection of the embodiments of the present invention.
[0170] The various illustrative logic blocks, units, or devices described in the embodiments of this invention can be implemented or operate the described functions using a general-purpose processor, digital signal processor, application-specific integrated circuit (ASIC), field-programmable gate array or other programmable logic device, discrete gate or transistor logic, discrete hardware components, or any combination thereof. The general-purpose processor can be a microprocessor; alternatively, it can be any conventional processor, controller, microcontroller, or state machine. The processor can also be implemented using a combination of computing devices, such as a digital signal processor and a microprocessor, multiple microprocessors, one or more microprocessors combined with a digital signal processor core, or any other similar configuration.
[0171] The steps of the methods or algorithms described in the embodiments of this invention can be directly embedded in hardware, a software module executed by a processor, or a combination of both. The software module can be stored in RAM, flash memory, ROM, EPROM, EEPROM, registers, hard disk, removable disk, CD-ROM, or any other form of storage medium in the art. Exemplarily, the storage medium can be connected to the processor so that the processor can read information from and write information to the storage medium. Optionally, the storage medium can also be integrated into the processor. The processor and storage medium can be housed in an ASIC, which can be housed in a user terminal. Optionally, the processor and storage medium can also be housed in different components of the user terminal.
[0172] In one or more exemplary designs, the functions described in the embodiments of the present invention can be implemented in hardware, software, firmware, or any combination of these three. If implemented in software, these functions can be stored on a computer-readable medium or transmitted on a computer-readable medium in the form of one or more instructions or code. Computer-readable media include computer storage media and communication media that facilitate the transfer of computer programs from one place to another. Storage media can be any available media that can be accessed by a general-purpose or special-purpose computer. For example, such computer-readable media can include, but is not limited to, RAM, ROM, EEPROM, CD-ROM or other optical disk storage, magnetic disk storage or other magnetic storage devices, or any other medium that can be used to carry or store program code in the form of instructions or data structures and other forms that can be read by a general-purpose or special-purpose computer, or a general-purpose or special-purpose processor. Furthermore, any connection can be suitably defined as a computer-readable medium, for example, if the software is transmitted from a website, server or other remote resource via a coaxial cable, fiber optic cable, twisted pair, digital subscriber line (DSL) or wirelessly, such as infrared, wireless and microwave, it is also included in the defined computer-readable medium. The disks and discs mentioned include compressed disks, laser discs, optical discs, DVDs, floppy disks, and Blu-ray discs. Disks typically copy data magnetically, while disks typically copy data optically using lasers. Combinations of the above can also be contained in computer-readable media.
Claims
1. A product recommendation method based on knowledge graphs, characterized in that, include: A customer knowledge graph is constructed based on customer characteristics, and a product knowledge graph is constructed based on product characteristics. A customer preference vector representation is obtained based on the customer knowledge graph; The similarity between the product represented by the customer preference vector and the products held by the customer is defined as the customer preference similarity. Based on the product knowledge graph, determine the product dense vector, the direct product relationship similarity between each product and the products held by the customer, and the indirect product relationship similarity. Based on the product dense vector, the direct product relationship similarity, and the indirect product relationship similarity, determine the customer's product similarity. The similarity of the target product is determined based on the similarity of customer preferences and the similarity of customer products, and the target product is pushed based on the similarity of the target product.
2. The product recommendation method based on knowledge graphs according to claim 1, characterized in that, The customer preference vector representation obtained from the customer knowledge graph includes: Obtain positive example triples of the customer knowledge graph based on the customer knowledge graph; The customer preference vector representation is obtained based on the positive example triples of the customer knowledge graph.
3. The product recommendation method based on knowledge graphs according to claim 2, characterized in that, The customer preference vector representation obtained from the positive example triples of the customer knowledge graph includes: A knowledge representation model is established based on the positive example triples of the customer knowledge graph and the preset TransR model; Generate negative triples of the customer knowledge graph based on the positive triples of the customer knowledge graph; Train the positive and negative triples of the customer knowledge graph based on the knowledge representation model to obtain a customer preference vector representation.
4. The product recommendation method based on knowledge graphs according to claim 2, characterized in that, The customer knowledge graph positive example triples obtained from the customer knowledge graph include: Resource description framework feature data is extracted from the customer knowledge graph using the resource description framework. The positive example triples of the customer knowledge graph are obtained based on the feature data of the resource description framework.
5. The product recommendation method based on knowledge graphs according to claim 1, characterized in that, Determining the product density vector based on the product knowledge graph includes: The product knowledge graph is mapped to a continuous vector space using a pre-defined TransR model to obtain the product dense vector.
6. A product recommendation device based on a knowledge graph, characterized in that, include: The product knowledge graph construction module is used to construct customer knowledge graphs based on customer characteristics and product knowledge graphs based on product characteristics. A customer preference vector representation module is used to obtain a customer preference vector representation based on the customer knowledge graph. The customer preference similarity module is used to determine the similarity between the product corresponding to the customer preference vector and the products held by the customer as the customer preference similarity. The customer product similarity module is used to determine the product dense vector, the direct product relationship similarity and the indirect product relationship similarity between each product and the customer's held products based on the product knowledge graph, and to determine the customer product similarity based on the product dense vector, the direct product relationship similarity and the indirect product relationship similarity. The target product push module is used to determine the target product similarity based on the customer preference similarity and the customer product similarity, and push the target product based on the target product similarity.
7. A computer device comprising a memory, a processor, and a computer program stored in the memory and running on the processor, characterized in that, When the processor executes the computer program, it implements the steps of the knowledge graph-based product recommendation method according to any one of claims 1 to 5.
8. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by a processor, it implements the steps of the knowledge graph-based product recommendation method according to any one of claims 1 to 5.
9. A computer program product comprising a computer program / instructions, characterized in that, When the computer program / instructions are executed by the processor, they implement the steps of the knowledge graph-based product recommendation method according to any one of claims 1 to 5.