KR102518389B1 - Method, Apparatus and System for Item Recommendation Using Consumer Bigdata - Google Patents

Abstract

A product recommendation method, device, and system using customer big data are disclosed. The product recommendation method receives a plurality of individual shopping mall customer identifiers and a plurality of personal information element groups corresponding to the plurality of individual shopping mall identifiers from a plurality of shopping malls, and generates an integrated customer identifier for each customer based on the plurality of personal information element groups. and collects order history data of a plurality of customers identified as a plurality of integrated customer identifiers from a plurality of shopping malls, converts the collected order history data into graph-type data consisting of at least three node hierarchies, and graph-type data is stored in the database, and the similarity is determined based on the number of connection paths between at least two nodes among nodes included in at least three node hierarchies in response to queries about similar products based on the stored graph-type data and returned. can do. Accordingly, it is possible to recommend products customized for each customer.

Description

Product recommendation method, apparatus and system using customer big data {Method, Apparatus and System for Item Recommendation Using Consumer Bigdata}

The present invention relates to a product recommendation method, apparatus, and system using customer big data, which builds big data from a plurality of shopping malls and uses big data to recommend products optimized for customers based on it. It is about product recommendation method.

In general, marketing is an indispensable success factor in online-based e-commerce. As a conventional marketing method using the Internet, a banner advertisement or image advertisement is exposed on a web page such as a portal site or an e-commerce site that provides e-mail service, search service, and e-commerce service, etc., and the user clicks the advertisement. If so, it is connected to the server of the linked advertiser.

However, since these conventional marketing methods collectively provide advertisement contents to an unspecified number of Internet users regardless of their region of residence, age, taste, etc., customers perceive advertisements as rather noise information, causing irritation and advertising effect. There are also minor problems.

Accordingly, target advertising technology is being actively developed. For example, Korea Patent Publication No. 2013-0157475 discloses a standard information collection unit that provides an advertisement environment setting interface to users and collects advertisement environment setting information from users and collected advertisements.

A technology such as a target advertisement providing device including an advertisement management unit that detects candidate advertisements based on environment setting information and selects and organizes exposure advertisements for each advertisement slot among the detected candidate advertisements is disclosed.

However, such a conventional technology for providing targeted advertisements is cumbersome because the customer has to set each category of advertisements to be exposed or advertisements to be blocked through the environment setting interface, and since it is based on setting by category, the accuracy of target advertisements is also low. It is true that the actual utilization of targeted advertising is not high.

In addition, the conventional technology for providing targeted advertisements uses a method of connecting a user terminal and a user information collection server by inserting a link for collecting customer information into an internal frame (I-Frame) in each web page, but this method is used for security and network It is not efficient in terms of load, and since each shopping mall operator uses a different information collection method, the collection information is limited and the amount of data is not large, so the accuracy is low.

Therefore, the development of a new marketing technology that can easily acquire big data from a plurality of e-commerce-based shopping malls and recommend reliable product information that customers want based on big data in a way suitable for customers. This is an urgent need.

No. 2013-0157475

The present invention is to solve this problem, and to build big data from a plurality of shopping malls, and based on the built big data, product recommendation method and apparatus using big data that can perform product recommendation optimized for customers and to provide a system for its purpose.

In order to achieve this object, the present invention provides a product recommendation method in one aspect (Aspect). The product recommendation method is performed by a server that interworks with a plurality of e-commerce based shopping malls, and provides a plurality of individual shopping mall customer identifiers and a plurality of personal information element groups corresponding to the plurality of individual shopping mall identifiers from the plurality of shopping malls. receiving; generating an integrated customer identifier for each customer based on the plurality of personal information element groups, wherein the same customer in the plurality of shopping malls is identified as one integrated customer identifier; collecting order history data of a plurality of customers identified as the plurality of integrated customer identifiers from the plurality of shopping malls; converting the collected order history data into data in the form of a graph consisting of at least three node hierarchies; storing the data in the form of a graph in a database; and based on the stored data in the form of a graph, determining and returning a degree of similarity based on the number of connection paths between at least two nodes among nodes included in the at least three node hierarchies in response to a query about similar products. includes

Personal information elements included in the plurality of personal information element groups may be de-identified personal information. The de-identified personal information includes customer email, customer name, and customer phone number. It may be associated with at least two of customer nickname, customer address, customer date of birth, customer gender, and customer age.

The generating of the integrated customer identifier for each customer may include specifying a first customer based on the plurality of personal information element groups; and generating an integrated customer identifier corresponding to the first customer. The step of specifying the first customer may include: calculating a personal information similarity between the plurality of personal information element groups; and specifying the first customer based on the calculated personal information similarity.

Calculating the personal information similarity may include calculating the number of personal information elements included in the personal information element group and the number of personal information elements matching between the personal information element groups. Specifying the first customer based on the calculated personal information similarity may include specifying the first customer as the same customer if the calculated similarity is greater than or equal to a predetermined value.

The at least three node hierarchies may include a first node hierarchy associated with products, a second node hierarchy associated with orders, and a third node hierarchy associated with customers. The step of determining and returning the similarity may include at least one of at least one node of the first node hierarchy and at least one node of the third node hierarchy in response to a query about the similar product, on the stored graph data. The method may include determining the degree of similarity based on the number of connection paths between two nodes.

In the data in the form of a graph, at least one second node of a second node hierarchy is connected to a first node of a first node hierarchy, and a third node of a third node hierarchy is connected to the at least one second node. As it is connected, it can be in the form of a radial graph.

At least two of the nodes of the at least one first node hierarchy and the nodes of the at least one third node hierarchy are extracted by extracting connection paths belonging to a first distance range among at least one connection paths to which at least two nodes can be connected. The higher the number of connection paths, the higher the degree of similarity.

Products of the first node hierarchy may be categorized and defined together with product category identifiers. At least one of a second node of a second node hierarchy and a third node of a third node hierarchy directly or indirectly connected to the first product in response to an inquiry about a similar product to the first product of the first node hierarchy; Among related products, based on the first product category identifier of the first product, only products of a second product category having a similar correspondence thereto may be considered as a connection path.

Products of the first node may be categorized and defined together with product category identifiers. Corresponding to the query of the first node for a similar product to the first product, the first product is associated with at least one of a second node of a second node hierarchy and a third node of a third node hierarchy directly or indirectly connected to the first product. Based on the first product category identifier of the first product among the products, a connection path with a product of a second product category having a similar correspondence thereto may be excluded.

The product of the first node is categorized and defined together with a product category identifier, and in response to a query about a similar product to the first product of the first node, a second node hierarchy directly or indirectly connected to the first product Based on the first product category identifier of the first product among the products associated with at least one of the second node of and the third node of the third node hierarchy, a product of a third product category having a corresponding relationship by combining with the first product category identifier, and connection path can be considered.

In the step of determining and returning the similarity based on the number of connection paths, in calculating the number of connection paths, the distance between two directly connected nodes is defined as a distance of 1 hop, and the at least one first The number of connection paths may be adjusted in consideration of a distance of a connection path between at least two nodes among nodes of the first node layer and at least one node of the third node layer.

A node of the first node hierarchy is defined as p, and a node of the third node hierarchy is defined as m, so that a query for a similar product and a connection path between corresponding nodes are defined as either m or p. and an index number for a specific node layer identifier to create a query.

The product recommendation method may further include training an artificial intelligence model to improve accuracy of similarity determination. First order history data including a first node serving as a criterion for querying similar products and second order history data including a second node related to a product for which an actual order was made among similar products returned by similarity determination The artificial intelligence model can be trained by creating a learning data set.

When calculating the number of connection paths, only connection paths within the first distance range are calculated, but based on the learning data set, the artificial intelligence model can be trained while changing the first distance range, which is the critical range of connection path calculation. . Based on the learning data set, an artificial intelligence model may be trained while changing whether to use a class of product category division and a similar category corresponding to the class or a category by combination.

On the other hand, in order to achieve the above object of the present invention, the present invention provides a product recommendation device in another aspect. The product recommendation device may include a communication unit communicating with at least one of wired and wireless networks; Receiving a plurality of individual shopping mall customer identifiers and a plurality of personal information element groups corresponding to the plurality of individual shopping mall identifiers from the plurality of shopping malls through the communication unit, and based on the plurality of personal information element groups, integrated customers for each customer Create an identifier -the same customer in the plurality of shopping malls is identified as one integrated customer identifier-, collect order history data of a plurality of customers identified as the plurality of integrated customer identifiers from the plurality of shopping malls, and collect the converts the order history data into graph-type data composed of at least three node hierarchies, stores the graph-type data in a database, and responds to queries about similar products based on the stored graph-type data a processor for determining and returning a degree of similarity based on the number of connection paths between at least two nodes among nodes included in the at least three node hierarchies; and a memory for storing data necessary for the operation of the processor.

Personal information elements included in the plurality of personal information element groups may be de-identified personal information. The de-identified personal information includes customer email, customer name, and customer phone number. It may be associated with at least two of customer nickname, customer address, customer date of birth, customer gender, and customer age.

The at least three node hierarchies may include a first node hierarchy associated with products, a second node hierarchy associated with orders, and a third node hierarchy associated with customers. The processor determines a connection path between at least two nodes among at least one node of the first node hierarchy and at least one node of the third node hierarchy in response to the query about the similar product, on the stored data in the form of a graph. The degree of similarity may be determined based on the number of .

Meanwhile, in order to achieve the above object of the present invention, the present invention provides a product recommendation system in another aspect. The product recommendation system may include a plurality of shopping malls; and receiving a plurality of individual shopping mall customer identifiers and a plurality of personal information element groups corresponding to the plurality of individual shopping mall identifiers from the plurality of shopping malls, and generating an integrated customer identifier for each customer based on the plurality of personal information element groups. And -the same customer in the plurality of shopping malls is identified as one integrated customer identifier-, collects order history data of a plurality of customers identified as the plurality of integrated customer identifiers from the plurality of shopping malls, and the collected order history Convert data into graph-type data composed of at least three node hierarchies, store the graph-type data in a database, and respond to queries about similar products based on the stored graph-type data It may include a server that determines and returns a degree of similarity based on the number of connection paths between at least two nodes among nodes included in the node hierarchy.

Meanwhile, in order to achieve the above object of the present invention, in another aspect, the present invention receives a plurality of individual shopping mall customer identifiers and a plurality of personal information element groups corresponding to the plurality of individual shopping mall identifiers from a plurality of shopping malls in a computer. doing; generating an integrated customer identifier for each customer based on the plurality of personal information element groups, wherein the same customer in the plurality of shopping malls is identified as one integrated customer identifier; collecting order history data of a plurality of customers identified as the plurality of integrated customer identifiers from the plurality of shopping malls; converting the collected order history data into data in the form of a graph consisting of at least three node hierarchies; storing the data in the form of a graph in a database; and based on the stored data in the form of a graph, determining and returning a degree of similarity based on the number of connection paths between at least two nodes among nodes included in the at least three node hierarchies in response to a query about similar products. A computer program stored in a medium may be provided to execute the.

As described above, according to the present invention, by integrating and managing customer activity information scattered across a plurality of e-commerce shopping malls, big data is built, and the big data is used to recommend products optimized for individuals based on a graph database. Alternatively, product recommendation between group customers and group products can be efficiently performed.

1 is a block diagram showing the configuration of a system for constructing customer big data based on generation of an integrated customer identifier and product recommendation based thereon according to a preferred embodiment of the present invention.
2 is a flowchart for explaining a customer big data construction process according to a preferred embodiment of the present invention.
FIG. 3 is a flowchart illustrating an embodiment of a process of specifying a first customer shown in FIG. 2 .
4 is an exemplary diagram illustrating a table stored and managed in the database 140 by way of example.
5 is a conceptual diagram illustrating a product recommendation method based on similarity.
6 is a graph based on product, order, and customer nodes used in a product recommendation method using a graph database according to an embodiment of the present invention.
7 is a conceptual diagram for explaining direct correlation and indirect correlation between products.
8 is a flowchart illustrating a product recommendation method using a graph database according to an embodiment of the present invention.
9 is a conceptual diagram showing a state in which a plurality of product-centered radial graphs are intertwined.
10 is a diagram for explaining proper use of same/similar categories and combined categories in a product recommendation method using a graph database according to an embodiment of the present invention.
11 is a conceptual diagram illustrating an artificial intelligence learning method related to a recommendation algorithm used in a product recommendation method using a graph database according to an embodiment of the present invention.
12 is a conceptual diagram illustrating a connection path indexing method of a product recommendation method using a graph database according to an embodiment of the present invention.
13A to 13D are diagrams illustrating various embodiments of finding similar products and customers with similar tendencies as outputs by specifying product and customer nodes as inputs.
14 is a diagram showing a result of implementing a product recommendation method using a graph database according to an embodiment of the present invention on a web page.
15 is a flowchart illustrating a product recommendation method using a graph database according to another embodiment of the present invention.
16A and 16B are diagrams illustrating an example of product route prediction in a product recommendation method using a graph database according to another embodiment of the present invention.
17A and 17B are diagrams illustrating an example of product route prediction in a situation in which a next product is retrieved after product inquiry in the embodiment of FIG. 16A.
18 is a conceptual diagram illustrating a situation in which promotion is required for products whose sales are low according to product sales activity analysis.
19A and 19B are conceptual views illustrating a matching method for providing a special promotion by matching a specific number of customer groups and a specific number of product groups when there are 5 customers and 6 products.
20A and 20B are flowcharts illustrating a product basket recommendation method according to another embodiment of the present invention.
21 is a flowchart illustrating a product-based basket recommendation method according to another embodiment of the present invention. This relates to method (1) of FIG. 19A.
22A to 22C are tables for explaining in detail a method of generating a matrix for each step of the basket recommendation method of FIG. 21 .
23 is a flowchart illustrating a customer-based basket recommendation method according to another embodiment of the present invention.
24A to 24C are tables for explaining in detail a method of generating a matrix for each step of the basket recommendation method of FIG. 23 .
25 is an apparatus for constructing customer big data based on generation of an integrated customer identifier and performing a graph database-based product recommendation method and/or product basket recommendation method based on the built big data according to an embodiment of the present invention. It is a schematic block diagram.

Since the present invention can make various changes and have various embodiments, specific embodiments are illustrated in the drawings and described in detail. However, this is not intended to limit the present invention to specific embodiments, and should be understood to include all modifications, equivalents, and substitutes included in the spirit and scope of the present invention.

Terms such as first and second may be used to describe various components, but the components should not be limited by the terms. These terms are only used for the purpose of distinguishing one component from another. For example, a first element may be termed a second element, and similarly, a second element may be termed a first element, without departing from the scope of the present invention. The terms and/or include any combination of a plurality of related recited items or any of a plurality of related recited items.

It is understood that when a component is referred to as being 'connected' or 'connected' to another component, it may be directly connected or connected to the other component, but other components may exist in the middle. It should be. On the other hand, when a component is referred to as being 'directly connected' or 'directly connected' to another component, it should be understood that no other component exists in the middle.

Terms used in this application are only used to describe specific embodiments, and are not intended to limit the present invention. Singular expressions include plural expressions unless the context clearly dictates otherwise. In this application, terms such as 'include' or 'have' are intended to designate that there is a feature, number, step, operation, component, part, or combination thereof described in the specification, but one or more other features It should be understood that the presence or addition of numbers, steps, operations, components, parts, or combinations thereof is not precluded.

Hereinafter, with reference to the accompanying drawings, preferred embodiments of the present invention will be described in more detail. In order to facilitate overall understanding in the description of the present invention, the same reference numerals are used for the same components in the drawings, and redundant descriptions of the same components are omitted.

First, the construction of customer big data based on the generation of integrated customer identifiers will be described.

1 is a block diagram showing the configuration of a system for constructing customer big data based on generation of an integrated customer identifier and product recommendation based thereon according to a preferred embodiment of the present invention.

As shown in FIG. 1 , the system may include a plurality of shopping malls 110-1 to 110-N, a network 120, a server 130, a database 140, and a user terminal 150. The user terminal 150 may be a customer terminal using a shopping mall.

The server 130 may interwork with a plurality of shopping malls 110-1 to 110-N based on e-commerce through the network 120. Each of the shopping malls 110-1 to 110-N may be, for example, a web server providing an e-commerce site. That is, the shopping malls 110-1 to 110-N refer to online shopping malls that sell products or services through the Internet. The shopping malls 110-1 to 110-N may sell products to visitors to the corresponding address on the Internet with a specific URL. The shopping mall manager operates the shopping mall by acquiring at least a part of the price of the sold product through his/her own account associated with the shopping mall. For example, the shopping malls 110-1 to 110-N may be not only clothing shopping malls and open markets, but also game sites that sell goods or services for a fee, social media sites, community sites, broadcasting sites, and multimedia providing sites. there is.

In the embodiment shown in FIG. 1, the shopping malls 110-1 to 110-N are, as described above, a web page associated with a specific address (eg, URL) on the Internet, a server or shopping mall that handles information related to the shopping mall. It can be implemented as a terminal of an operator of The terminal may include a wired/wireless networking computing device such as a smart phone or a PC.

The plurality of shopping malls 110-1 to 110-N may be operated by different types of companies. For example, the first shopping mall 110-1 may be a web server for providing an Internet shopping mall operated by an e-commerce company A, and the second shopping mall 110-2 is completely unrelated to the e-commerce company A. B may be a service web server for providing an Internet shopping mall operated by an e-commerce company.

The plurality of shopping malls 110-1 to 110-N may interwork with the server 130 through the network 120. Here, the server 130 is a device that obtains information from a plurality of shopping malls 110-1 to 110-N and recommends a product in response to a product recommendation query.

The server 130 may be a server (not shown) distributing and managing a platform that supports the opening and operation of a shopping mall or a server interoperating with the server. The server 130 according to a preferred embodiment of the present invention does not handle only information related to one shopping mall 110-1 based on generating and managing an integrated customer identifier, but a plurality of shopping malls 110-1 to 110-1. 110-N) and can analyze and learn various information based on big data.

The server 130 may be implemented with one or a plurality of computing devices. The computing device referred to herein may be implemented as a server-class computer terminal. The computing device may include an input device of a typical computer terminal, a display device, a networking device, a hard disk, a memory for storing programs, and a processor for executing programs stored in the memory. However, it is not necessarily implemented as a server-class computer terminal.

The server 130 generates an integrated customer identifier capable of identifying the same customer in the plurality of shopping malls 110-1 to 110 to N, and uses the integrated customer identifier from the plurality of shopping malls 110-1 to 110 to N. Log records of identified customers may be collected, and based on the collected log records, marketing may be performed through product recommendation tailored to the customer.

In the conventional case, the collection of user's Internet use details is performed by individual e-commerce companies. For example, an individual e-commerce company records a cookie or session of a user who accesses a shopping mall operated by the e-commerce company, and then when the user accesses the shopping mall of the corresponding e-commerce company again, information related to the user's previous usage history is recorded. Product advertisements are displayed on the web page. However, since the target advertisement performed for each e-commerce company is based on the customer's use details of a single shopping mall, the amount of collected data is small and reliability is low.

If you are trying to collect Internet usage history from multiple shopping malls in the past, since each web server has a different operator and data collection policy, the format and type of user usage history information collected are different, making it very difficult to integrate them. In particular, since user IDs, i.e., individual shopping mall customer identifiers corresponding to customers, are often different for each shopping mall even for the same customer, it is practically impossible to integrate the order details of the same customer in a plurality of shopping malls.

However, in a preferred embodiment of the present invention, the server 130 provides a plurality of individual shopping mall customer identifiers and Receiving a plurality of personal information element groups corresponding to a plurality of individual shopping mall customer identifiers, specifying a customer based on the plurality of personal information element groups, generating an integrated customer identifier corresponding to the specified customer, and responding to the customer A plurality of shopping malls (110-1 ~ 110 ~ N) based on the integrated customer identifier are stored in the database 140 and managed by associating the integrated customer identifier with a plurality of individual shopping mall customer identifiers and a plurality of personal information element groups. It is possible to build big data by collecting log records of specified customers from The integrated customer identifier may be a customer identifier for identifying the same customer in a plurality of shopping malls 110-1 to 110-N as one identifier.

The network 120 includes a plurality of shopping malls 110-1 to 110-N, a server 130, a server 130 and a user terminal 150, a plurality of shopping malls 110-1 to 110-N, and a user terminal ( 150) may mean a communication network that connects. For example, the network 120 may include a wired network or a wireless network (Internet).

2 is a flowchart for explaining a customer big data construction process according to a preferred embodiment of the present invention.

As shown in FIG. 2 , the server 130 may access a plurality of shopping malls 110-1 to 110-N. The server 130 provides a plurality of individual shopping mall customer identifiers corresponding to the plurality of shopping malls 110-1 to 110-N from the plurality of shopping malls 110-1 to 110-N and a plurality of individual shopping mall customer identifiers corresponding to the plurality of shopping mall customer identifiers. It is possible to receive two personal information element groups (S1).

The individual shopping mall customer identifier may be a customer identifier used to identify a customer in an individual shopping mall. The personal information element group may refer to personal information elements corresponding to the individual shopping mall customer identifier. That is, it may refer to a set of personal information elements registered and managed in an individual shopping mall corresponding to an individual customer identifier. For example, the first shopping mall 110-1 identifies the first customer based on an individual shopping mall customer identifier corresponding to the first customer, and in response to the individual shopping mall customer identifier, the first shopping mall 110-1 identifies the first customer. You can manage the elements of your personal information.

Personal information elements included in the personal information element group may be non-identified personal information elements. The non-identified personal information element may refer to a personal information element other than information used to identify a customer in a shopping mall, such as an identifier (ID) and a resident registration number. For example, the de-identified personal information elements include customer email, customer name, and customer phone number. This could be customer nickname, customer address, customer date of birth, customer gender, customer age, etc.

Next, the server 130 may specify a first customer based on the plurality of personal information element groups (S2). This can be said to identify customers by comparing non-identified personal information element groups corresponding to the respective shopping malls 110-1 to 110-N and considering customers having personal information element groups with high similarity as the same customer. .

FIG. 3 is a flowchart illustrating an embodiment of a process of specifying a first customer shown in FIG. 2 .

As shown in FIG. 3 , the server 130 may calculate a personal information similarity between the plurality of personal information element groups (S11). For example, the server 130 may calculate the number of personal information elements included in the personal information element group and the number of personal information elements matched between the personal information element groups.

Subsequently, the server 1230 may specify the first customer based on the calculated personal information similarity. For example, the server 140 determines whether the calculated personal information similarity is equal to or greater than a predetermined value (S12), and identifies the same customer if the personal information similarity is greater than or equal to a predetermined value (S13).

For example, if the number of personal information elements included in one personal information element group is 10 and the personal information similarity value determined to determine the same customer, that is, the standard value is 0.7, then there are 7 personal information elements that match each other. Customers corresponding to the above personal information element groups may be determined as the same customer and identified as one customer.

Meanwhile, according to another embodiment of the present invention, the server 130 may input the received data of the plurality of personal information element groups to an artificial intelligence engine learned through deep learning and classify the data according to the same customer. When the data of a plurality of non-identified personal information element groups is input, the artificial intelligence engine classifies personal information element groups having a high personal information similarity to each other and corresponds to the same customer, thereby learning to classify by the same customer there is.

Next, the server 130 may generate an integrated customer identifier corresponding to the first customer specified above (S3). As mentioned above, the integrated customer identifier may be a customer identifier used to identify the same customer in the plurality of shopping malls 110-1 to 110-N as one identifier. The server 130 may associate the generated customer identifier corresponding to the first customer with the plurality of individual shopping mall customer identifiers and the plurality of personal information element groups and store them in the database 140 (S4).

4 is an exemplary diagram illustrating a table stored and managed in the database 140 by way of example.

As shown in FIG. 4, the first customers are customers of shopping malls A, B, C, and D, and each shopping mall uses a different individual shopping mall customer identifier. That is, "AAAA" in shopping mall A, "BBBB" in shopping mall B, "CCCC" in shopping mall C, and "DDDD" in shopping mall D are used as individual shopping mall customer identifiers, "Mall-ID".

Each shopping mall corresponds to a personal information element group including non-identified personal information elements, such as e-mail, phone number, name, nickname, age, and gender of the first customer.

The server 130 determines the personal information similarity between the personal information element groups, and determines whether "AAAA" of shopping mall A, "BBBB" of shopping mall B, "CCCC" of shopping mall C, and "DDDD" of shopping mall D are the same for the first customer. and creates "CUAA01" as an integrated customer identifier "EC-ID" capable of identifying the first customer in the four shopping malls, and associates it with a plurality of individual shopping mall customer identifiers and personal information element groups, as shown in FIG. The same table as in the example may be stored and managed in the database 140 .

On the other hand, when a new second customer accesses one of a plurality of shopping malls, for example, the first shopping mall 110-1 and becomes a customer, the server 13 receives the first shopping mall (from the first shopping mall 110-1). In 110-1), an individual shopping mall customer identifier capable of identifying the second customer and a personal information element group corresponding to the individual shopping mall customer identifier of the second customer may be received.

The server 130 may determine whether an integrated customer identifier corresponding to the second customer exists in the database 140 based on the individual shopping mall customer identifier of the second customer. Here, since the second customer is a newly introduced customer, the integrated customer identifier of the second customer does not exist in the database 140 .

Therefore, when the integrated customer identifier corresponding to the second customer does not exist in the database 140, the server 130 determines the information corresponding to the second customer based on the personal information element group corresponding to the second customer. An integrated customer identifier may be generated, and the integrated customer identifier corresponding to the second customer may be associated with an individual shopping mall customer identifier of the second customer and a personal information element group corresponding to the individual shopping mall customer identifier, and stored in the database 140.

Meanwhile, if it is assumed that the second customer is not a new customer and an integrated customer identifier already exists, since an integrated customer identifier corresponding to the second customer exists in the database 140, the server 130 Based on the comparison of the personal information element group of the second customer stored in the database 140 and the received personal information element group, it is determined whether there is a change, and if there is a change as a result of the comparison, based on the changed content The content of the personal information element group of the second customer may be updated with .

Referring back to FIG. 2 , the server 130 records a log record of a first customer from a plurality of shopping malls 110-1 to 110-N based on the integrated customer identifier stored and managed in the database 140 as described above. It can be collected (S6). That is, big data related to customer order history, such as purchase details, access details, and return details performed in the plurality of shopping malls 110-1 to 110-N for each customer, may be collected.

The server 130 collects the collected order history data of customers to build big data, and then provides customized products corresponding to customers based on at least one of product recommendation according to the first method and product recommendation according to the second method. Recommendations can be performed.

Product recommendation by the first method may be product recommendation based on similarity using, for example, a graph database. Product recommendation by the second method may be product recommendation between group customers and group products, for example, basket recommendation.

Hereinafter, the recommendation by the first method will be described. The first method is product recommendation based on similarity using a graph database. Here, the degree of similarity is different from the degree of similarity of personal information between a plurality of groups of personal information elements calculated when generating the aforementioned integrated customer identifier. Similarity using a graph database will be described in detail below.

Referring back to FIG. 1 , the server 130 receives information from the shopping mall management server (not shown) that manages the overall information of the shopping malls 110-1 to 110-N and/or the shopping malls 110-1 to 110-N. Obtain order history data and/or product inquiry data. In the above data, the customer is identified by the combined customer identifier. That is, the server 130 collects order history data, product inquiry data, and the like of a plurality of customers identified as the plurality of integrated customer identifiers from the plurality of shopping malls.

The server 130 converts the acquired data into a graph form having a plurality of node hierarchies. Data converted into a graph form is stored in the database 140 outside the server 130 . However, the database 140 does not necessarily have to be outside the server 130 . may exist inside.

The server 130 responds to a product recommendation request (which may be called a "recommended product search request", "query", etc.) from the user terminal 150 by using a graph composed of three node hierarchies, from specific products to similar products. Process queries for searches. At this time, on the converted graph data, connection paths between products and products, customers and customers, or products and customers are extracted, and the degree of similarity between them can be determined by considering the distance and / or number of extracted connection paths. . In addition, when product inquiry data is used, queries for similar products are processed by analyzing the inquiry history for a certain step before or after product search on the graph. A part related to a technique for processing queries about similar products will be described in more detail below.

In another embodiment of the present invention, the server 130 uses the product based on the presence of an index (which may include a similarity using the above graph database) indicating the purchasing possibility between an individual customer and an individual product. You can optimize your basket. That is, if the product basket recommendation method according to the embodiment of the present invention is used, a plurality of product groups optimized for at least one customer (in a plurality of cases, a customer group) can be appropriately promoted.

The user terminal 150 may be a terminal of a purchaser of the shopping mall 110-1. The buyer may transmit a product recommendation request to the server 130 or a separate shopping mall management server (not shown) using the user terminal 150 . If a separate shopping mall management server exists, the shopping mall management server transfers the received product recommendation request to the server 130 .

The product recommendation request may include a query about similar products. Furthermore, it may include queries about similar customers. Then, when processing of the query is completed using the graph-based data analysis method and/or basket recommendation method, the processed data is provided to the user terminal 150 so that the buyer can check it and select a product based on it. make it possible The user terminal 150 may provide a product recommendation request to the server 130 using the network 120 (a specific web page within the network).

According to another embodiment of the present invention, the terminal or user terminal 150 associated with the shopping malls 110-1 to 110-N may download an application provided by the server 130 and perform automatic generation through the application. . That is, in the terminal or user terminal 150 associated with the shopping malls 110-1 to 110-N, when a user inputs a query about similar products or similar customers, a graph-based product recommendation process or product basket recommendation process is performed within the terminal. can be done in Without a separate server device, a processor (not shown) of the terminal may execute the downloaded application to recommend an optimal product through data transformation and analysis of the transformed data.

5 is a conceptual diagram illustrating a product recommendation method based on similarity. The term "apparatus" or "product recommendation device" mentioned below may be interpreted as referring to the server 130 shown in FIG. 1 .

Referring to FIG. 5 , the device generally recommends products based on the similarity of images and attributes of products. That is, first, the similarity between product images may be calculated using an image analysis module to be used as a purchase possibility index. In the embodiment of FIG. 5 , the device may extract a shoe product 2 having the most similar color and shape from among numerous product images stored in the database by image analysis of color and shape of shoe product 1 . Alternatively, based on the product attributes of shoe product 1, shoe product 2 having similar properties may be extracted. For example, if the color of shoe product 1 is red, the size is 250, the brand is "Brand A", and the shape is "shoes", shoe product 2 with the same or similar properties can be searched and extracted from the database. . However, as described above, the method of using the similarity based on image and/or product attributes is applicable to the case of purchasing similar products at the same time, so the possibility of purchase is not very high, and there is a case where the buyer's demand cannot be satisfied. Occurs.

According to an embodiment of the present invention, the apparatus extracts the attribute of a product purchased a lot based on a purchase history of a specific customer, and sets a high purchasing possibility index for a product having a corresponding attribute. Using this point, a purchase possibility index may be calculated by categorizing customers with attributes of products purchased by the customer. That is, it is possible to find customers having a similar disposition to customers who have purchased the corresponding product, and to recommend a product purchased by the similar disposition customer as a similar product.

As described above, the purchase possibility index based on the analysis of product images, product attributes, and/or customer attributes in various methods may be used as an element entering an empty column of a matrix in a product basket recommendation method in the future.

6 is a graph based on product, order, and customer nodes used in a product recommendation method using a graph database according to an embodiment of the present invention.

Referring to FIG. 6 , the device may collect order history data and generate a graph composed of at least three node hierarchies. The order history data basically includes information that "a specific customer A ordered product C using order B at a specific shopping mall". In other words, it is data that includes relationships between customers, orders, and products.

The device creates a radial graph by forming products as a first node layer, orders as a second node layer, and customers as a third node layer based on each order history data. In the example of FIG. 6 , product A 310 is included in six orders 320 (orders A through F), each order obtained from a different customer 330 .

In this case, the device connects order A 320 to order F around product A 310 using a connection line, and customer A 330 requesting order A 320 is connected to order A 320. Each is connected so that a radial graph is created from the product. In other words, the product-order connection line represents the product included in the order, and the order-customer connection line represents the customer who placed the order. At this time, among each node hierarchy of the graph, the first node hierarchy (product hierarchy) can be represented by red color, the second node hierarchy (order hierarchy) by green color, and the third node hierarchy (customer hierarchy) by blue color. The graph thus generated may be visualized and displayed through a display device.

As shown in the embodiment of FIG. 6 , the device may display multiple orders including one product 310 (product A) and a customer who placed the order. At this time, it can be seen that between the customers 330 included in the graph of FIG. 6, there is one path connected by a length of distance 4 (customer → order → product → order → customer). Therefore, it can be seen that these customers show similarities in purchasing propensity compared to customers who do not. The meaning of the distance 4 will be described in more detail with reference to FIG. 7 below.

7 is a conceptual diagram for explaining direct correlation and indirect correlation between products.

Referring to (a) of FIG. 7 , there are not many people who simultaneously order two very similar products (which may be products belonging to a top). That is, in the case of a connection path having a high direct correlation and a distance of 2, it is desirable to determine that the degree of similarity is low.

However, referring to (b) of FIG. 7 , when looking at another product (bottoms (product 3)) as a medium, many customers purchase a combination of (product 1-product 3) and (product 2-product 3). may exist. That is, the initial direct correlation (distance 2) (see the embodiment of (a) in FIG. 7) shows a low similarity, but when the distance becomes distant by a certain interval (eg, distance 4), an indirect correlation can be observed, It is desirable that this indirect correlation be implemented as a high index in terms of purchase possibility. In other words, the number of paths through which two products (product 1 and product 2) can be connected at a predetermined distance (distance 4 in the embodiment of FIG. 7(b)) can be an index indicating similarity.

Returning to FIG. 6 again, in view of the above correlation, customers connected to the distance 4 may have a relationship of “other customers who purchased products purchased by a specific customer”. For a specific customer, if you look for the relationship of "customer who purchased the same product" as above, you can extract a path connecting a specific customer and other customers through different products. According to an embodiment of the present invention, the device may determine customer B as a customer with the most connected routes and having the most similar purchasing propensity to a specific customer (customer A).

Accordingly, the device may recommend to customer B products that a specific customer A purchased but similar customer B did not purchase, based on the graph database. Similarly, a product purchased by customer B but not purchased by customer A may be recommended to A. That is, the device may connect customer A and customer B through the integrated product A 310, and analyze the degree of similarity in consideration of the length and number of connection paths.

8 is a flowchart illustrating a product recommendation method using a graph database according to an embodiment of the present invention.

First, a server (which may be referred to as a "device") receives a plurality of individual shopping mall customer identifiers and a plurality of personal information element groups corresponding to the plurality of individual shopping mall identifiers from a plurality of shopping malls, and receives the plurality of personal information element groups. Based on this, it is possible to generate an integrated customer identifier for each customer. The same customer in the plurality of shopping malls may be identified as one integrated customer identifier.

Referring to FIG. 8 , the server collects order history data from an online shopping mall or a server managing the shopping mall (S510). For example, the server collects order history data of a plurality of customers identified as the plurality of integrated customer identifiers from a plurality of shopping malls. At this time, the obtained data includes information about a relationship between a product, an order including the corresponding product, and a customer who performed the order.

The device converts the collected order history data into graph-type data (S512). The graph is composed of three node hierarchies, the first node hierarchy, products, is connected to the second node hierarchy, orders, through a connection line, and the order is connected to the third node hierarchy, customers, through a connection line, resulting in the first node hierarchy. It may have a radial graph form with a center. Then, the device stores the data converted into a graph form in a database (S514).

The client terminal, that is, the user terminal inputs a recommended product search request (S516). The input request information is transmitted to the device (server) side. At this time, the recommended product search request may be input after being divided into (i) a product-centered search request and (ii) a customer-centered search request. This will be described in more detail with reference to FIGS. 16A to 16D.

When a recommendation product search request is input, the device analyzes the correlation centering on the target product or target customer using the converted data in the form of a graph (S518). Then, with the target product or target customer at the center, a connection path to other products or other customers is created based on the analyzed correlation (S520). At this time, based on the distance of the created connection path, it is determined whether the connection path exists within a certain distance range (S522).

As described above, it is preferable to have a larger distance than the distance 2, which indicates a direct correlation (eg, a = 2), and as the distance increases, the number of connected paths increases proportionally, so having an excessively large distance value It is desirable to exclude the connection path (eg, b = 10). The a and b values defining the distance range can be set through user settings and can be arbitrarily changed. In addition, the values of a and b may be changed in the form of finding an optimal solution through machine learning. A connection path having a distance less than or exceeding the specified range is excluded (S524).

According to an embodiment of the present invention, the device may adjust the number of connection paths in consideration of the distance of the connection path between two nodes. As described above, the distance between two directly connected nodes is defined as a distance of 1 hop. In this case, when the distance of the first connection path is n1 (n1 is a natural number) and the distance of the second connection path is n2 (n2 is a natural number), the number of first connection paths can be calculated by adjusting to 1/n1. . Also, the number of second connection paths may be calculated by adjusting to 1/n2. That is, except for the direct correlation of distance 2, since nodes connected by a short distance may have a high similarity, the number of connection paths can be adjusted by dividing by the corresponding distance considering this more important. For example, the number of connection paths between nodes at a distance of 4 is considered to be 1/4, which may be considered larger (important) than 1/8, which is the number of connection paths at a distance of 8.

The device calculates the number of connection paths existing within the set distance range based on the above method (S526). Then, products or customers having the largest number of connection paths are extracted (S528). At this time, the top several products or customers may be extracted as products or customers with high similarity by sorting based on the number of connection paths. A high degree of similarity means that the purchasing propensity is similar. In particular, when the customer-to-customer similarity is high, from the point of view of the customers judged to be similar, the possibility that the target customer will purchase the remaining products minus the integrated product within the first order including the integrated product through the integrated product is high Since it can be judged to be high, these products can be derived as recommended products.

In this way, product and customer information obtained as a result of similarity, and information on other products in the first order, excluding integrated products based on the corresponding customer, are provided to the client terminal. The terminal may display the provided information to the buyer by outputting it (S530).

9 is a conceptual diagram showing a state in which a plurality of product-centered radial graphs are intertwined.

Referring to FIG. 9 , a plurality of radial graphs 610, 620, 630, and 640 may be entangled. In the device, a first radial graph 610 centered on product A and a second radial graph 620 centered on product B may be connected to each other via order A or customer A. These two radial graphs 610 and 620 can be visualized and displayed on a display screen in two dimensions. Furthermore, the first radial graph 610 and the third radial graph 630 centered on product C may be connected to each other via order B or customer B, and in this case, the second radial graph 620 and the third radial graph 630 Graph 630 may also be linked to each other via order C or customer C. As such, when three or more radial graphs 620 are connected, it is difficult to visualize and express a correlation between each radial graph in two dimensions.

Thus, in one embodiment of the present invention, the device may display radial graphs in three dimensions. That is, if the first and second radial graphs 610 and 620 are displayed on the xy plane, the first and third radial graphs 610 and 630 may be displayed with a first angle in the z-axis direction, and The second and fourth radial graphs 620 and 640 may be displayed at an angle. Accordingly, the graphs can be pulled through the zoom-in/zoom-out function of the user interface, and the connection relationship between a plurality of radial graphs can be seen by adjusting the three-dimensional angle using the panning function. can be visualized. Also, when a specific node is selected, the selected node is brought to the center of the screen so that a plurality of radial graphs connected to the center node are displayed in all directions.

According to another embodiment of the present invention, the relationship type does not necessarily have to be "customer-order-product-order-customer", and the relationship of "product-order-customer-order-product" centered on the customer. can be specified. In other words, it is okay to take various standards of relationship.

10 is a diagram for explaining proper use of same/similar categories and combined categories in a product recommendation method using a graph database according to an embodiment of the present invention.

Referring to FIG. 10 , the device may utilize product category information when determining the degree of similarity and when deriving recommended products after extracting similar products/customers. For example, in finding similar products, similar products can be found only within the same category (eg tops), or other categories (categories by competition or combination described above) (eg tops and bottoms, bottoms and shoes, etc.) ) to find the corresponding product.

In particular, by classifying product categories, the first class is divided into the largest categories such as tops, bottoms, shoes, and accessories, and the same/similar categories and combined categories are classified, and the second class is classified within the first class. can be defined in detail. Each product category is defined through an identifier, and a similar/combined relationship between identifiers is coded so that it is possible to identify what kind of relationship each product category has. In the case of the second class, for example, within the top, it can be divided into a shirt, a men's shirt, a T-shirt, a hood, a vest, and the like. The third class is defined as a subdivided category of the second class.

According to an embodiment of the present invention, assuming that there is class classification up to the third class, similar products can be derived using the same/similar product category in all classes from the first to the third class. In this case, for "T-shirt", the "Southern" product category can be selected as a product category similar to "T-shirt" while being "Top", and a product belonging to the "Southern" product category within a specific order is derived as a similar product. It can be.

Alternatively, similar products may be derived by using the combination category in the first class. In this case, the "bottoms" category is selected for "T-shirt", and a product belonging to the bottom product category can be derived as a similar product of the T-shirt. In the case of a T-shirt, since it is often purchased in combination with a bottom, this setting may be appropriate.

In another embodiment, the same/similar category may be used in the first class, but a combined category may be used in the second class. In this case, for “T-shirt”, the “top” category may be selected as the same/similar category in the first class, and the “jumper” category combined with “T-shirt” may be selected in the second class. That is, a product belonging to the product category of “jumper” within a specific order may be derived as a similar product.

As such, in deriving similar products using a graph, determining whether to use the same/similar product category or a combined product category is a factor that is considered important because the resulting value varies. In addition, the result value differs depending on which class uses the similar category and the combined category. The device may define these settings through user settings and allow the user to arbitrarily change them. Or, we can use big data-based artificial intelligence to optimally adjust these areas. A method of utilizing the artificial intelligence model will be described in detail with reference to FIG. 11 .

11 is a conceptual diagram illustrating an artificial intelligence learning method related to a recommendation algorithm used in a product recommendation method using a graph database according to an embodiment of the present invention.

Referring to FIG. 11 , determining whether to use the “same/similar category” or the “combined category” in a specific divided class can be said to be very important because the result value is completely different. Therefore, after learning an AI model using an AI learning module (not shown), the device may set whether to use a similar or combined category in a specific class based on the AI model that has been trained.

According to an embodiment of the present invention, the artificial intelligence model used in the device includes a deep learning convolutional neural network (CNN) model. However, it is not limited thereto, and may include any one of a deep neural network (DNN), a recurrent neural network (RNN), and a long short term memory network (LSTM).

The device generates order data 1 and order data 2 as training data sets. At this time, the two order data are the previous order data (order data subject to the similarity search request) of the selection process according to the graph-based derivation of similar products according to an embodiment of the present invention and the actual order data after product recommendation. It has a relationship with order data. That is, the customer who ordered order data 1 is recommended a plurality of products using the graph-based similar product derivation method according to an embodiment of the present invention, and combines order data 2 made by selecting one of the recommended data. Create a training data set. At this time, in deriving graph-based similar products for product recommendation, similar products are derived while changing whether similar or combined categories are used in the first class and whether similar or combined categories are used in the second class, Based on the order data after derivation, artificial intelligence is trained in the direction of placing an order in the product derived at the top through the graph-based similar product derivation process.

In addition, while changing the threshold values (a and b values in FIG. 5) for the optimal distance range, the optimal threshold value is found while accumulating the resultant value based on the learning data sets of order data 1 and order data 2. let it be

At this time, the node layer serving as a medium between order data 1 and order data 2 forming the learning data does not necessarily have to be "product", which is the first node layer, but may be "customer", which is the third node layer. Also, the second node layer may be a reference.

12 is a conceptual diagram illustrating a connection path indexing method of a product recommendation method using a graph database according to an embodiment of the present invention.

Referring to FIG. 12 , the device performs indexing to identify connection paths between products, customers and customers, or products and customers. At this time, the connection path may be composed of a node layer identifier and an index number. The node layer identifier is composed of a combination of the identifier of the target node, which is the center of the search, and the identifier of the final node. In this case, the first node layer (product node) may be identified as p, and the third target node layer (customer node) may be identified as m. After identifying the node hierarchy of the departure node and arrival node in this way, an indexing number is attached thereafter. An indexing number is related to a query and may have a special meaning. For example, if the index "0100" is attached to the pp node layer identifier, this means a query to find other products purchased by customers who purchased the product, and the path associated with this is bundled by customers who ordered the first product (start node) Means other products ordered. A description of a query according to the meaning of an index attached directly following a node layer identifier will be described in more detail with reference to FIG. 13 .

In this way, in the state where the index is attached, each connection path is identified in the form of "-1". That is, the first path associated with the pp0100 query can be dataized as "pp0100-1". After that, the number at the end of "-" is increased by 1 to identify the path. For example, if there are a plurality of other products ordered by a customer who ordered the first product in a bundle, it may be defined as pp0100-1, pp0100-2, pp0100-3, and the like.

In the same way, the connection path from the first customer to the second customer is expressed as mm0100-1, mm0100-2, or mm0101-1, mm0101-2, and the like. A connection path from the first product to the first customer is expressed as pm0100-1, pm0100-2, and the like. A connection path from the first customer to the first product is expressed as mp0100-1, mp0100-2, and the like.

In this way, the connection path can be converted into data and retained by the device, and the number of connection paths within a specific distance range is calculated using this data.

Additionally, the device may control the conversion from the implementation in the graph form of FIG. 9 to the form of a connection path list. That is, while visualizing a plurality of radial graphs, if there is a request for converting information related to a connection path between a specific departure node and a specific arrival node into a list form, the connection path index between the departure node and the arrival node is converted into a list. can be displayed In addition, the distance of the connection path (eg, pp0100-1) of the first index may also be displayed in the corresponding list. Accordingly, when a connection path having a specific index is clicked, the corresponding path is displayed in a highlighted state on the graph, and it is visualized to intuitively see which path is connected from the departure node to the destination node.

In addition, in the case of pp0100-1 and pp0101, they can be indexed overlapping. To prevent this, when extracting all connection paths for pp0100, pp0100-1 is extracted and all connection paths for pp0101 are extracted. In this case, pp0101-1 is extracted to prevent duplication, and when all routes are extracted, pp0101-1, which expresses a more specific connection route, can be identified as pp0101-1, except for overlapping pp0100-1.

13A to 13D are diagrams illustrating various embodiments of finding similar products and customers with similar tendencies as outputs by specifying product and customer nodes as inputs.

The device generates a query to find similar products and customers by specifying product or customer nodes as inputs.

Referring to FIG. 13A , a query for finding a product similar to a specific product is created. A customer who ordered a product might create a pp0100 query that lists other products that have been bundled. In addition, a connection path can be derived using the pp0101 query indicating products in the same category among other products that the customer who ordered the product bundled ordered.

In addition, a pp0200 query indicating a product purchased by a customer who ordered a product may be used, and a pp0201 query indicating a product within the same category in the other order may be used.

Referring to FIG. 13B , a query for finding a customer similar to a specific customer is as follows. A query mm0100 is available which indicates a second customer who ordered the same product as the first customer. Furthermore, the mm0101 query may be used to find a path to a second customer who ordered a second product of the same category, but not the exact same product as the first product ordered by the first customer.

Referring to FIG. 13C , in relation to a query for finding a customer similar to a specific product, a pm0000 query indicating a path to a customer who purchased a bundled product with the corresponding product, and among the bundled products, the corresponding product You can use the query pm0100 that represents the path to a customer who purchased a product in the same category as .

Referring to FIG. 13D , in relation to a query for finding a product similar to a specific customer, an mp0000 query indicating a path to bundled products such as the product ordered by the customer, and among the bundled products, the corresponding An mp0001 query representing a path to a product in the same category as the product can be used.

According to an embodiment of the present invention, in determining the similarity between products, the above query exists, and the connection path for the query is also expressed as ppXXXX-X. In particular, although the connection path ppXXXX-X between products and products can be used, a combination of the pmXXXX-X path, which represents the path related to the similarity between the product and the customer, and the mpXXXX-X path, which represents the connection path related to the similarity between the customer and the product is available. That is, similarity between products may be determined using pm0000-1 + mp0000-1. In addition, similarity between products may be determined using three or more similarity-related connection paths, such as using pm0000-1 + mm0100-1 + mp0000-1.

According to an embodiment of the present invention, when using the path of (i) ppXXXX-X and (ii) pmXXXX-X + mmXXXX-X + mpXXXX-X, whether optimal product recommendation is possible is also shown in FIG. Accuracy can be improved by learning while changing the parameters (including weights) associated with the path using the artificial intelligence learning model of .

14 is a diagram showing a result of implementing a product recommendation method using a graph database according to an embodiment of the present invention on a web page.

Referring to FIG. 14, the device enables execution of the above algorithm by accessing using REST API. The device designates which logic (relationship) to apply using a URL (Uniform Resource Locator). In the embodiment of FIG. 14 , pp0100 is designated to derive other products purchased by members who have purchased the corresponding product.

In the URL path, a query parameter such as pp0100 may be followed by a query parameter, which may include a central product number, the number of data to be returned, and a member ID. That is, only customer or product information can be used according to the recommendation logic. Additional parameters such as product code and customer number are transmitted through URL encoding.

In response to this query, as query output, the number of data (n), product or member (node), importance (hits: number of paths searched), and encryption query (Cypher query for debugging purposes) will be provided. can

The device may return the top n products (n=5, ie, 5 in this embodiment) and the number of paths for each product in the order of high similarity, that is, the number of connection paths.

In the case of using the product purchase history, it may be too late to identify customer preferences, and in the case of purchase cases, it is difficult to see statistical correlation because there is not enough data. that can be considered

15 is a flowchart illustrating a product recommendation method using a graph database according to another embodiment of the present invention.

Referring to FIG. 15 , the device (server) collects inquiry history data related to product inquiry from shopping malls and/or a server managing the shopping malls (S1210). This can be achieved by extracting product information search histories of users from web log files. The search history can be converted into sequential product codes and converted into data. For example, product 1 is viewed, product 2 is inquired, product 4 is inquired, and product 1 is viewed again and the logout history is converted into a product code in consideration of a time-series part, thereby generating inquiry history data.

Then, the device converts the inquiry history data into a graph form (S1212) and stores it in the database (S1214). In this case, the structure of the graph may have a structure of “session-page-product” to correspond to the preceding “product-order-customer”.

The client (user terminal) may input a recommended product search request (S1216). As described in FIG. 14, this can be done using a URL.

The device extracts the searched product within the range of x number of web pages before or after searching for a specific product (S1218). The x value may be determined by user settings and may be arbitrarily changed.

The device may recommend products to the purchaser in the order of products with a high number of views among the products viewed (S1220). The client may receive and output recommended product information (S1222).

16A and 16B are diagrams illustrating an example of product route prediction in a product recommendation method using a graph database according to another embodiment of the present invention.

Referring to FIGS. 16A and 13B , when the number n of recommended products is set to 10 and there is a request for recommended products, the device creates a graph formed by at least three node hierarchies. This may include a web page (1320: second node hierarchy) that the user has searched for based on the session 1330, which is the first node hierarchy, and a product 1310 (third node hierarchy) searched for in the corresponding web page. The node hierarchy may be changed so that the product becomes the first node hierarchy, the page becomes the second node hierarchy, and the session becomes the third node hierarchy.

The device connects the web page 1320 associated with the specific session 1330, and configures the product 1310 searched within the web page 1320 to be connected to the web page 1320. In addition, it is preferable to compose a graph by distinguishing preceding and succeeding operations by using arrows (→) instead of connecting lines (-) formed in a straight line so as to grasp the time-sequential order.

After constructing the graph, the device retrieves and recommends 10 products retrieved within a range of 3 pages in the order of the highest number of hits after product 24522 (1310) is retrieved. Contents related to the following three web pages will be described with reference to FIGS. 17A and 17B.

17A and 17B are diagrams illustrating an example of product route prediction in a situation in which a next product is retrieved after product inquiry in the embodiment of FIG. 16A.

17A and 17B, after the customer inquires product 24522 (not shown) and then inquires product 25799 (1410) through web page 1420, using web page 1422, The fact that product 25622 (1412) was inquired can be confirmed based on the converted graph data. In this way, the client of FIGS. 16A to 16B searches for and returns 10 products searched within the range of a specific number of pages (in this embodiment, it may be 3 pages) after a specific product in the order of the highest number of hits. requests from can be processed. However, at this time, the three pages may be set to be pages on which product search is performed. That is, it is preferable to exclude web pages without product searches from calculating the number of page strokes, calculate three pages based on pages on which products are searched, and recommend top products with many inquiries within the range of the page.

According to another embodiment of the present invention, it is not necessary to recommend products by limiting only pages, but by limiting sessions, it is possible to recommend products based on the number of product views made within three sessions. Furthermore, only products in a competing category or products in a similar category may be recommended using a category identifier of a product recommended due to a high number of views. The use of similar or competing categories and the use of classes can also be optimized using the artificial intelligence model of FIG. 11 . At this time, the learning data set may be a set of inquiry data 1, which is the object of the search request, and inquiry data 2 related to a product actually purchased through a graph-based product recommendation process according to an embodiment of the present invention.

In an embodiment of the present invention, the device may mix and use a first graph database based on product purchase history data and a second graph database based on product inquiry history data. For example, after searching for the top 10 products with the largest number of hits using the second graph database, the degree of similarity between the 10 recommended products and the requested product may be calculated using the first graph database. Accordingly, the recommendation order of 10 recommended candidates may be changed based on products having a high degree of similarity using the first graph database. The reverse is also possible.

Additionally, by assigning a weight of 1 to the recommended product candidate calculated based on the first graph database and by assigning a weight of 2 to the recommended product candidate calculated based on the second graph database, a product having the highest similarity based on the weight is finally determined. How to select a candidate can be considered.

Hereinafter, product recommendation between group customers and group products, for example, basket recommendation, as product recommendation according to the second method will be described.

18 is a conceptual diagram illustrating a situation in which promotion is required for products whose sales are low according to product sales activity analysis.

Referring to FIG. 18, the apparatus performs promotion on some of 10 products (products with an activity lower than 0.2) (products at the lagger stage) whose sales are low by analyzing the sales volume and sales activity of the product. this may be requested. That is, when a promotion is to be performed by selecting 4 out of 10 low-selling products, the product basket recommendation method according to an embodiment of the present invention may be used.

19A and 19B are conceptual views illustrating a matching method for providing a special promotion by matching a specific number of customer groups and a specific number of product groups when there are 5 customers and 6 products.

Referring to FIG. 19A , when there are 5 customers and 6 products, and 3 products are to be promoted among 3 customers, a decision to optimize salability is required. To this end, the device may perform optimized promotion through two methods.

(1) The first method is a method of matching multiple product groups with multiple customer groups by first selecting product groups that are optimal for five out of six products and then secondarily selecting optimal customers for the selected products (see FIG. 21). .

(2) The second method is a method of matching a plurality of product groups with a plurality of customer groups by first selecting a customer group that is optimal for six products for five customers, and secondarily selecting an optimal product for the selected customers (FIG. 23). reference).

The device may perform a determination to optimize salability using at least one of the above two methods, and may proceed with a promotion in the state shown in FIG. 19B based on the determination.

Referring to FIG. 19B , the device promotes a package consisting of Product 2, Product 3 and Product 6 to Customer 1, Customer 3 and Customer 4. The promotion may be carried out in the form of providing a special discount on products included in the corresponding package only to the corresponding customer. Alternatively, the product of the corresponding package may be displayed on the first page of the shopping mall in a pop-up method, or the package product may be arranged in the most conspicuous area to induce sales.

20A and 20B are flowcharts illustrating a product basket recommendation method according to another embodiment of the present invention.

Referring to FIGS. 20A and 20B , when the apparatus selects and promotes customers to recommend to each of a plurality of products (eg, 10 products), the following procedure is followed.

The device first obtains product and customer data that are candidates for promotion (S1710). In the embodiment of FIG. 20B , it is assumed that there are 10 product candidates and about 100 customer candidates, and data for these are obtained. At this time, in the customer candidates, buyers who have pre-purchased a specific product are excluded or the data is set to 0 and collected. The collected data preferably takes the form of a matrix. At this time, products are formed as rows and customers are formed as columns. Of course, the opposite is also possible. And, in the empty cells of the matrix, indicators related to each product and each customer's purchasing potential are entered. The indicator related to the purchase possibility may use an indicator calculated based on product image analysis, product attribute analysis, and propensity according to customer purchase history. Alternatively, a similarity based on a graph database according to an embodiment of the present invention may be used as the indicator. In the embodiment of FIG. 20B , a similarity calculated as the number of connection paths using a graph database is used as the purchase possibility index.

After obtaining the data and generating a matrix, the device calculates the total number of connection paths for each product (S1720) and selects high-ranking products with a large number of connection paths (S1730). The number of selected products follows the number of promotional products. This is a situation that a user (eg, shopping mall operator) can select through a user interface. Alternatively, all products having a number of connection paths greater than or equal to a threshold value may be selected without setting a specific number of products as a reference value. For example, products having a connection path greater than 700 connection paths may be selected without limiting the number. A small number of connection paths may be determined to have little relevance. In this case, processing may be performed in the direction of increasing the discount by selecting a connection path having a small number. In the embodiment of FIG. 20B , the device selects the top 4 products with a high number of connection paths.

The device reconstructs the matrix into a new matrix by leaving only the top four selected products (which can be changed through user settings) and deleting the remaining products (S1740). Then, 6 customers with a high sum of connection paths for the remaining 4 products are selected as promotion target customers. The device promotes the selected 4 product packages to 6 customers of the group of customers to be promoted (S1750).

21 is a flowchart illustrating a product-based basket recommendation method according to another embodiment of the present invention. This relates to method (1) of FIG. 19A.

Referring to FIG. 21 , after acquiring data on a purchase possibility index associated with a product to be a promotion candidate and a customer, the apparatus creates a matrix with product as a column and customer as a row (S1810). An index of purchasing possibility between a product and a customer is input in the blank of the matrix.

Then, the device calculates the sum of availability for each row (S1820). Then, the products are ranked in the order of the highest purchase availability, and products having the highest purchase availability total are selected as many as the number of products to be recommended by promotion, and the remaining products are deleted to reconstruct the matrix (S1830).

After the product is selected, the device calculates subtotals for each customer for each column (S1840). Then, after ranking each customer in the ascending order of subtotals, customers with a higher order of purchasing possibility subtotals are selected as many as the number of customers to be promoted, and the rest of the customers are excluded (S1850).

When the selection of the customer group and the product group is completed, the device forms the corresponding product group into a package for the corresponding customer group and conducts a promotion (S1860).

22A to 22C are tables for explaining in detail a method of generating a matrix for each step of the basket recommendation method of FIG. 21 .

Referring to FIG. 22A , the apparatus generates a matrix by configuring customers as rows and products as columns based on the purchasing availability index for customers 1 to 5 and products 1 to 6.

Referring to FIG. 22B , the device ranks products by calculating partial sums for each product through row summation. According to the embodiment of FIG. 22B , product 2 ranks first with a value of 69, product 3 ranks second with a value of 55, and product 6 ranks third with a value of 53.

Referring to FIG. 22C , the top three products (product 2, product 3, and product 6) are selected, and subtotals for each customer are calculated by summing columns. Then, rank them. According to the embodiment of FIG. 22C , customer 1 ranks first with a value of 48, customer 3 ranks second with a value of 43, and customer 4 ranks third with a value of 33.

Accordingly, the device selects customer 1, customer 3, and customer 4 as promotion target customers, and selects product 2, product 3, and product 6 as promotion target products, and proceeds with the promotion.

23 is a flowchart illustrating a customer-based basket recommendation method according to another embodiment of the present invention. This relates to method (2) of FIG. 19A.

Referring to FIG. 23 , after obtaining data on purchaseability indicators associated with products to be promotional candidates and customers, the apparatus creates a matrix with products as columns and customers as rows (S2010). An index of purchasing possibility between a product and a customer is input in the blank of the matrix.

Then, the device calculates the sum of availability for each column (S2020). Then, the customers are ranked in the order of the highest purchase possibility sum, and the customer having the highest purchase possibility sum rank is selected by the number of customers to be recommended, and the remaining customers are deleted to reconstruct the matrix (S2030).

After the customer is selected, the device calculates subtotals for each product for each row (S2040). Then, after ranking in ascending order of subtotals for each product, a product having a higher rank of subtotal availability as many as the number of products to be promoted is selected, and the remaining products are excluded (S2050).

When the selection of the customer group and the product group is completed, the device forms the product group into a package for the corresponding customer group and conducts a promotion (S2060).

24A to 24C are tables for explaining in detail a method of generating a matrix for each step of the basket recommendation method of FIG. 23 .

Referring to FIG. 24A , the apparatus generates a matrix by configuring customers as rows and products as columns based on the purchasing availability index for customers 1 to 5 and products 1 to 6.

Referring to FIG. 24B , the device calculates subtotals for each customer through column summation and ranks them. According to the embodiment of FIG. 24B , customer 1 ranks first with a value of 77, customer 3 ranks second with a value of 73, and customer 4 ranks third with a value of 72.

Referring to FIG. 24C , the top three customers (customer 1, customer 3, and customer 4) are selected, and a partial sum for each product is calculated by summing rows. Then, rank them. According to the embodiment of FIG. 24C , product 2 ranks first with a value of 47, product 3 ranks second with a value of 41, and product 5 ranks third with a value of 37.

Accordingly, the device selects customer 1, customer 3, and customer 4 as promotion target customers, and selects product 2, product 3, and product 5 as promotion target products, and proceeds with the promotion.

According to the product-based recommendation method according to the embodiment of FIG. 20 and the customer-based recommendation method according to the embodiment of FIG. 22, the product is the same but the product is different. As such, depending on which method is used, the result may be completely different.

25 is an apparatus for constructing customer big data based on generation of an integrated customer identifier and performing a graph database-based product recommendation method and/or product basket recommendation method based on the built big data according to an embodiment of the present invention. It is a schematic block diagram.

As shown in FIG. 25 , the product recommendation device according to an embodiment of the present invention may include a communication unit 2210, a display unit 2220, a processor 2230, a memory 2240, and an input unit 2250. .

Referring to FIG. 25 , the communication unit 2210 is a component for communicating with another device by being connected to at least one of wired and wireless networks. The communication unit 2210 may be implemented with an antenna, a communication-related chip, or the like.

The display unit 2220 is a component that displays information input/output through the communication unit 2210 and information processed through the processor 2230 . It can be implemented as one of monitors, TVs and other panels.

Based on the communication unit, the processor 2230 may receive a plurality of individual shopping mall customer identifiers corresponding to the plurality of shopping malls and a plurality of personal information element groups corresponding to the plurality of individual shopping mall customer identifiers from the plurality of shopping malls. The processor 2230 identifies a first customer based on the plurality of personal information element groups, generates an integrated customer identifier corresponding to the first customer, and generates the integrated customer identifier corresponding to the first customer. Big data is stored in a database in association with a plurality of individual shopping mall customer identifiers and the plurality of personal information element groups, and by collecting log records of the first customer from the plurality of shopping malls in the database based on the integrated customer identifier can be built The processor 2230 may be implemented as a microprocessor.

The processor 2230 may calculate a similarity between a plurality of personal information element groups and specify a first customer based on the calculated similarity. The processor 2230 may calculate the number of personal information elements included in the personal information element group and the number of personal information elements matched between the personal information element groups. The processor 2230 may specify the same customer if the calculated similarity is greater than or equal to a predetermined value. The processor 2230 may input data of a plurality of personal information element groups into an artificial intelligence engine learned through deep learning and classify the data according to the same customer.

The processor 2230 receives, from a first shopping mall, a first individual shopping mall customer identifier capable of identifying a second customer in the first shopping mall and a first personal information element group corresponding to the first individual shopping mall customer identifier; It is determined whether an integrated customer identifier corresponding to the second customer exists in the database based on the first individual shopping mall customer identifier, and if the integrated customer identifier corresponding to the second customer does not exist in the database, the An integrated customer identifier corresponding to the second customer is generated based on the first personal information element group, and the integrated customer identifier corresponding to the second customer is combined with the first individual shopping mall customer identifier and the first personal information element group. It can be associated with and stored in the database.

The processor 2230 compares the personal information element group of the second customer stored in the database with the first personal information element group when an integrated customer identifier corresponding to the second customer exists in the database. Thus, it is possible to determine whether there has been any change. At this time, if there is a changed content as a result of the comparison, the content of the personal information element group of the second customer may be updated based on the changed content.

The processor 2230 recommends a product corresponding to the first customer based on at least one of product recommendation and basket recommendation based on a graph database based on the collected log records of the first customer. can do. The memory 2230 may store data necessary for the operation of the processor 2230.

The processor 2230 receives a plurality of individual shopping mall customer identifiers and a plurality of personal information element groups corresponding to the plurality of individual shopping mall identifiers from a plurality of shopping malls, and integrates for each customer based on the plurality of personal information element groups. Create a customer identifier, collect order history data of a plurality of customers identified as the plurality of integrated customer identifiers from the plurality of shopping malls, and convert the collected order history data into graph-type data composed of at least three node hierarchies. Convert, store the data in the form of a graph in a database, and based on the data in the form of a graph, connect between at least two nodes among the nodes included in the hierarchy of the at least three nodes in response to a query about a similar product Based on the number of routes, similarity can be determined and returned.

Meanwhile, from another point of view, the processor 2230 is a component that executes a graph-based similarity determination and product basket recommendation algorithm according to an embodiment of the present invention. The processor 2230 converts the order history data and inquiry history data input through the communication unit 2210 into graph-type data, and uses the converted graph-type data in response to a query about similar products from a client. to return information on similar products and similar customers. In this case, based on the number of connection paths from the first node to the second node (which may be limited to having a certain range of distance), products and/or customers may be sorted and returned in order of highest similarity. .

In addition, the processor 2230 may sort and return in order of the most searched products in pages within a certain range before or after the first product inquiry. The returned result data may be displayed through the display unit 2220 or provided to the requested terminal through the communication unit 2210.

The processor 2230 obtains information related to an individual product and an individual customer's purchasing possibility index based on graph-based similarity information or other similarity indicators, and generates a matrix having customers and products as columns and rows, respectively. Then, subtotals of rows and columns of the generated matrix are calculated, and the calculated values are ranked to create an optimal product package for a specific customer group, so that an appropriate promotion can be performed.

According to an embodiment of the present invention, the processor 2230 may include an artificial intelligence model learning module (not shown) and an artificial intelligence model execution module (not shown), and converts data of a plurality of personal information element groups based on similarity. It can be trained to classify by the same customer, and can perform classification by customer to respond to the integrated customer identifier. In addition, it is possible to train an artificial intelligence model to increase the accuracy of graph-based similarity determination and to improve the accuracy of actual product recommendation by executing the learned artificial intelligence model.

The memory 2240 is a storage for storing instructions of a program running on the processor 2230. The memory 2240 stores information received or transmitted through the communication unit 2210 . In addition, at least some of the information processed through the processor 2230 may be stored.

The input unit 2250 is a component that receives a user's input and may be implemented as a keyboard, mouse, or touch panel. Various parameters related to optimal product recommendation may be directly input through the input unit 2250 .

The system or apparatus described above may be implemented as hardware components, software components, and/or a combination of hardware components and software components. For example, systems, devices and components described in the embodiments may include, for example, a processor, a controller, an arithmetic logic unit (ALU), a digital signal processor, a microcomputer, a field programmable array (FPA) ), programmable logic unit (PLU), microprocessor, or any other device capable of executing and responding to instructions. A processing device may run an operating system (OS) and one or more software applications running on the operating system. A processing device may also access, store, manipulate, process, and generate data in response to execution of software. For convenience of understanding, there are cases in which one processing device is used, but those skilled in the art will understand that the processing device includes a plurality of processing elements and/or a plurality of types of processing elements. It can be seen that it can include. For example, a processing device may include a plurality of processors or a processor and a controller. Other processing configurations are also possible, such as parallel processors.

Software may include a computer program, code, instructions, or a combination of one or more of the foregoing, which configures a processing device to operate as desired or processes independently or collectively. You can command the device. Software and/or data may be any tangible machine, component, physical device, virtual equipment, computer storage medium or device, intended to be interpreted by or provide instructions or data to a processing device. , or may be permanently or temporarily embodied in a transmitted signal wave. Software may be distributed on networked computer systems and stored or executed in a distributed manner. Software and data may be stored on one or more computer readable media.

The method according to the embodiments may be implemented in the form of program instructions that can be executed through various computer means and recorded on a computer readable medium. The computer readable medium may include program instructions, data files, data structures, etc. alone or in combination. Program commands recorded on the medium may be specially designed and configured for the embodiment or may be known and usable to those skilled in computer software. Examples of computer-readable recording media include magnetic media such as hard disks, floppy disks and magnetic tapes, optical media such as CD-ROMs and DVDs, and magnetic media such as floptical disks. - includes hardware devices specially configured to store and execute program instructions, such as magneto-optical media, and ROM, RAM, flash memory, and the like. Examples of program instructions include high-level language codes that can be executed by a computer using an interpreter, as well as machine language codes such as those produced by a compiler. The hardware devices described above may be configured to operate as one or more software modules to perform the operations of the embodiments, and vice versa.

As described above, although the embodiments have been described with limited examples and drawings, those skilled in the art can make various modifications and variations from the above description. For example, the described techniques may be performed in an order different from the method described, and/or components of the described system, structure, device, circuit, etc. may be combined or combined in a different form than the method described, or other components may be used. Or even if it is replaced or substituted by equivalents, appropriate results can be achieved.

Therefore, other implementations, other embodiments, and equivalents of the claims are within the scope of the following claims.

110-1, 110-2, ... , 110-N: shopping mall
120: network
130: server
140: database
150: user terminal

Claims (22)

It is performed by a server that links with a plurality of e-commerce based shopping malls,
receiving a plurality of individual shopping mall customer identifiers and a plurality of personal information element groups corresponding to the plurality of individual shopping mall identifiers from the plurality of shopping malls;
generating an integrated customer identifier for each customer based on the plurality of personal information element groups, wherein the same customer in the plurality of shopping malls is identified as one integrated customer identifier;
collecting order history data of a plurality of customers identified as the plurality of integrated customer identifiers from the plurality of shopping malls;
converting the collected order history data into data in the form of a graph consisting of at least three node hierarchies;
storing the data in the form of a graph in a database; and
Based on the stored data in the form of a graph, determining and returning a degree of similarity based on the number of connection paths between at least two nodes among nodes included in the at least three node hierarchies in response to a query about similar products. include,
The step of generating an integrated customer identifier for each customer,
specifying a first customer based on the plurality of personal information element groups; and
generating an integrated customer identifier corresponding to the first customer;
The step of specifying the first customer,
Calculating a personal information similarity between the plurality of personal information element groups based on calculating the number of personal information elements included in the personal information element group and the number of personal information element groups that match the personal information element groups; and
If the calculated personal information similarity is greater than or equal to a predetermined value, specifying as the same customer;
The at least three node hierarchies include a first node hierarchy associated with products, a second node hierarchy associated with orders, and a third node hierarchy associated with customers;
Further comprising the step of learning an artificial intelligence model to improve the accuracy of similarity judgment,
First order history data including a first node serving as a criterion for querying similar products and second order history data including a second node related to a product for which an actual order was made among similar products returned by similarity determination Create a learning data set to train the artificial intelligence model,
When calculating the number of connection paths, only connection paths within a first distance range are calculated, and the artificial intelligence model is trained while changing the first distance range, which is a critical range of connection path calculation, based on the learning data set,
Based on the learning data set, the product recommendation method characterized in that the artificial intelligence model is trained while changing whether to use a class of product category division and a similar category corresponding to the class or a category by combination.
delete delete delete delete delete delete delete delete delete delete delete delete delete delete delete delete In the product recommendation device,
a communication unit communicating with at least one of a wired and wireless network;
Receiving a plurality of individual shopping mall customer identifiers and a plurality of personal information element groups corresponding to the plurality of individual shopping mall identifiers from a plurality of shopping malls through the communication unit, and based on the plurality of personal information element groups, an integrated customer identifier for each customer -The same customer in the plurality of shopping malls is identified as one integrated customer identifier-, collects order history data of a plurality of customers identified as the plurality of integrated customer identifiers from the plurality of shopping malls, and the collected Convert order history data into graph-type data consisting of at least three node hierarchies, store the graph-type data in a database, and respond to queries about similar products based on the stored graph-type data a processor for determining and returning a similarity based on the number of connection paths between at least two nodes among nodes included in at least three node hierarchies; and
A memory for storing data necessary for the operation of the processor;
The processor identifies a first customer based on the plurality of personal information element groups and generates an integrated customer identifier corresponding to the first customer;
Based on calculating the number of personal information elements included in the personal information element group and the number of matching personal information elements between the personal information element groups, the personal information similarity between the plurality of personal information element groups is calculated, and the calculated If the similarity of the personal information is greater than or equal to a predetermined value, the same customer is identified,
The at least three node hierarchies include a first node hierarchy associated with products, a second node hierarchy associated with orders, and a third node hierarchy associated with customers;
the processor,
An artificial intelligence model to improve the accuracy of similarity judgment is trained,
First order history data including a first node serving as a criterion for querying similar products and second order history data including a second node related to a product for which an actual order was made among similar products returned by similarity determination Create a learning data set to train the artificial intelligence model,
When calculating the number of connection paths, only connection paths within a first distance range are calculated, and the artificial intelligence model is trained while changing the first distance range, which is a critical range of connection path calculation, based on the learning data set,
Based on the learning data set, the product recommendation device is characterized in that the artificial intelligence model is trained while changing whether to use a class of product category division and a similar category corresponding to the class or a category by combination. delete delete delete delete

Images