WO2014035048A1

WO2014035048A1 - Item recommendation system and method

Info

Publication number: WO2014035048A1
Application number: PCT/KR2013/006184
Authority: WO
Inventors: 김민성; 윤도영; 정규환
Original assignee: 에스케이플래닛 주식회사
Priority date: 2012-08-28
Filing date: 2013-07-11
Publication date: 2014-03-06

Abstract

The present invention relates to an item recommendation system and method which can intelligently recommend items that can induce the interest of a user in various items that are provided to a user terminal by an item service device. According to the present invention, various types of recommended items can be provided by extracting and providing items which are different from those in the purchase history of the user from the purchase history of another user who has a different purchasing pattern from that of said first user, or by generating association rules between purchased items, which have been extracted by the analysis of the purchase histories of the users and, based thereon, extracting items related to each other from the purchase history of the user.

Description

Item recommendation system and method

The present invention relates to an item recommendation system and method for intelligently recommending items that can induce various interests of a user among items provided to a user by a service provider through a communication network.

With the advent of various e-commerce systems or online shopping malls, a system for purchasing necessary products or services without visiting a store is becoming common. However, in the virtual space of the Internet, it is impossible to directly view and compare products, which may cause reliability problems for products, and it is also difficult to obtain appropriate information on specific products. As a result, such a problem has caused a problem of restricting consumers to search only without purchasing a product in an internet shopping mall.

In addition, many consumers want to see and purchase various products through internet shopping malls, but due to the emergence of internet shopping malls that provide too much information, it is difficult to search for the appropriate type of products having specific reliability. . In order to solve this problem, various methods for recommending items related to the user have been developed. As a result, in recent years, items with high user relevance have been provided more intelligently. However, the item recommendation function also provides only items related to the user. As a result, the conventional item recommendation function has caused another problem in that the kind and characteristics of the product recommended to the user are limited.

The present invention has been proposed to solve the above-mentioned conventional problems, and recommends an item that can intelligently extract and provide an item capable of inducing various interests of a user among items provided to the user through a communication network by a service provider. It is intended to provide a system and method.

In particular, the present invention analyzes the purchase history of the users to calculate the heterogeneity of the purchase form between users, and based on the calculated heterogeneity degree to the user based on the purchase history of other users having the lowest similarity with the purchase form of the selected user An object recommendation system and method for recommending heterogeneous items is provided.

The present invention also provides an item recommendation system and method for generating association rules between items purchased by a user by analyzing a purchase history of users, and extracting and providing a recommendation item using the association rule.

In addition, through the introduction of a distributed data processing system, an object of the present invention is to provide an item recommendation system and method that can efficiently and quickly perform item recommendation by storing and retrieving a large amount of purchase history data.

In accordance with an aspect of the present invention, there is provided a solution to at least one of the present invention, which includes at least one user terminal having an item purchase history and receiving item recommendation information based on the item purchase history; After grouping all the user terminals, the heterogeneity degree of the purchase type between the user terminals is calculated based on the purchase history of the user terminals belonging to a specific group, and based on this, the heterogeneous purchase belongs to the same group as the specific user terminal belonging to the specific group. An item service apparatus for extracting other user terminals having a behavior and calculating item recommendation information for the specific user terminal including a list of foreign recommendation items based on the purchase history of the other user terminals and providing the same to the specific user terminal; It provides an item recommendation system comprising a.

In addition, as another solution, the present invention comprises at least one user terminal having an item purchase history, receiving the item recommendation information based on the item purchase history; Generate and store association rules that define a combination between purchased items based on history of purchasing items of the user terminals, and generate item recommendation information related to items included in a purchase history of a specific user terminal from the association rule; It provides an item recommendation system comprising an item service device provided to a specific user terminal.

The item recommendation system described above may further include a distributed processing service apparatus configured to distribute and store data necessary for generating item recommendation information.

In addition, in the first embodiment of the present invention, the item service apparatus, the device communication unit for supporting the signal processing required for the connection and purchase of at least one user terminal; Group all the user terminals into a certain number of groups, and calculate heterogeneity degree of purchase type between user terminals based on the purchase history of user terminals belonging to a specific group, and based on this, A device controller extracting other user terminals belonging to the same group having heterogeneous purchase behaviors, and generating item recommendation information including a list of heterogeneous recommended items based on the purchase history of the other user terminals and providing the same to the specific user terminal; It may include.

Here, the device controller calculates a heterogeneity degree of the purchase type between the user terminals based on a ratio of the intersection number of the products purchased by the two users to the number of the union of the products purchased by the two users.

In addition, the device control unit represents a purchase history for each user as a purchase history matrix in which rows and columns are defined as user information and product information, respectively, and the heterogeneity degree matrix between users representing heterogeneity between the calculated users in the purchase history matrix. By multiplying, it is possible to calculate a recommendation item score for each product for each user, and generate the heterogeneous recommendation item list based on the recommendation item score for each product.

In addition, the device controller divides the entire user terminals into a predetermined number of groups and allocates a group ID to each group so as to control the group ID for each user terminal and control the user for each group ID. Data of terminals may be grouped.

In addition, the device controller may be configured to compare the purchase history of each user terminal belonging to the same group with the purchase history of other user terminals belonging to the same group based on collaborative filtering to calculate a heterogeneity score in the order of low similarity. The heterogeneity may be controlled to generate a heterogeneous recommendation item list including a predetermined number of items for each score.

In addition, the item service device may further include a device storage unit for distributing and storing data necessary for generating the item recommendation information.

In addition, according to a second embodiment of the present disclosure, the item service apparatus may include: a device communication unit supporting signal processing required for accessing at least one user terminal and purchasing an item; An apparatus for generating predefined association rules based on history of purchasing items of user terminals, and generating item recommendation information corresponding to at least a part of purchase history of a specific user terminal from the association rule and providing the same to the specific user terminal. The apparatus may further include a controller, and may further include a device storage configured to store the association rules and the purchase history of the user terminals.

According to a second embodiment of the present disclosure, the device controller may check a purchase history of the specific user terminal, search for related rules corresponding to items included in the purchase history, and determine a predetermined number based on a predetermined criterion. After selecting association rules to be used as a preliminary recommendation item of, items included in a predetermined schedule ranking may be generated as item recommendation information. Here, the predetermined criterion includes at least one of support, reliability, and improvement. In this case, the support is a ratio of association rules that include two specific products of the overall association rules. The reliability is a ratio of association rules in which the specific product and the other specific product are included among all the association rules including the specific product. The degree of improvement is the ratio of association rules generated based on a history of purchasing another specific product and purchasing the specific product among the association rules including the specific product.

In addition, the present invention as a solution to the above problem, grouping all the user terminals in a certain number of groups; Calculating a heterogeneity degree for a purchase type between user terminals based on a purchase history of user terminals belonging to a specific group; Based on the heterogeneity, extracting other user terminals belonging to the same group as a specific user terminal and having heterogeneous buying behaviors, and generating a list of heterogeneous recommendation items for the specific user terminal based on the purchase history of the other user terminals. It provides an item recommendation method comprising the step of providing.

In addition, the present invention comprises the steps of generating and distributed storage by the item service device associated rules having a predetermined predetermined rule length based on the item purchase history of the user terminals; Calculating, by the item service apparatus, a predetermined number of preliminary recommended items based on a predetermined number of association rules corresponding to items included in a purchase history of a specific user terminal; Configuring, by the item service apparatus, items within a predetermined predetermined rank among the preliminary recommended items calculated as item recommendation information; The item service apparatus provides the item recommendation method comprising providing the item recommendation information to the user terminal.

According to an embodiment of the present invention, by performing heterogeneous item recommendation that the user has not easily touched or selected, the user's interest or interest is diversified, thereby supporting various forms of purchase patterns and increasing various consumption. You can do that.

In particular, the present invention further extends the user's selection range, interests and interests by not simply recommending items of a similar style, but by recommending items of a style different from those used by other users who belong to the same or similar groups as themselves. You can.

In addition, according to another embodiment of the present invention, by analyzing the purchase history of the user to create an association rule related to the user purchase pattern, and extracts the recommended item based on the generated association rule, the user is likely to purchase Item recommendation can be provided quickly, helping users to purchase items more reasonably and easily.

In addition, the present invention applies heterogeneous item recommendation by effectively storing and retrieving a large amount of data, especially a large purchase log, by applying a distributed environment, for the processing of item recommendation based on the heterogeneous item recommendation or association rule described above. You can do it faster and more effectively.

1 is a block diagram showing the overall configuration of an item recommendation system according to the present invention.

2 is a block diagram illustrating in more detail the configuration of an item service apparatus in the item recommendation system configuration according to the present invention.

3 is a block diagram illustrating a detailed configuration of a device controller for performing an item recommendation function according to a first embodiment of the present invention.

4 to 7 are diagrams for explaining an example of a process of providing a list of heterogeneous recommendation items according to the first embodiment of the present invention.

8 to 12 are flowcharts illustrating an item recommendation method according to a first embodiment of the present invention.

13 is a flowchart illustrating another example of an item recommendation method according to the first embodiment of the present invention.

14 is a diagram illustrating an example of a user purchase history for explaining an item recommendation function according to a second embodiment of the present invention.

FIG. 15 is a diagram illustrating an example of an association rule calculated from a purchase history of FIG. 14 for recommending an item according to a second embodiment of the present disclosure.

16 is a block diagram illustrating a detailed configuration of a device controller for performing an item recommendation function according to a second embodiment of the present invention.

FIG. 17 is a diagram illustrating an association rule distributed storage structure in an item recommendation system according to a second embodiment of the present invention.

FIG. 18 is a diagram for explaining distributed storing of a user purchase history in an item recommendation system according to a second embodiment of the present invention.

19 is a diagram illustrating a preliminary recommendation item searching process in the item recommendation system according to the second embodiment of the present invention.

20 is a diagram illustrating one form of association rules for an item recommendation function according to a second embodiment of the present invention.

21 and 22 illustrate example recommended items extracted based on the association rule of FIG. 20.

23 is a flowchart illustrating an item recommendation method according to a second embodiment of the present invention.

24 is a view for explaining an example of MapReduce applied to the item recommendation function of the present invention.

Hereinafter, exemplary embodiments of the present invention will be described in detail with reference to the accompanying drawings. In this case, it should be noted that like elements are denoted by like reference numerals as much as possible. In addition, detailed descriptions of well-known functions and configurations that may blur the gist of the present invention will be omitted.

On the other hand, the embodiments of the present invention disclosed in the specification and drawings are merely presented specific examples to easily explain the technical contents of the present invention and help the understanding of the present invention, and are not intended to limit the scope of the present invention. It will be apparent to those skilled in the art that other modifications based on the technical idea of the present invention can be carried out in addition to the embodiments disclosed herein.

1 is a view schematically showing the configuration of an item recommendation system according to the present invention.

Referring to FIG. 1, the item recommendation system 10 according to an exemplary embodiment of the present disclosure may be connected to the at least one user equipment 100 and the user terminals 100 by being connected through the communication network 300. It may include an item service device 200 that supports the search and purchase function. In addition, the item recommendation system 10 may further include a distributed processing service device 400.

The item recommendation system 10 includes an item including information on at least one item in order for the item service apparatus 200 to guide and sell at least one item, for example, at least one product, to the user terminal 100. A search screen can be provided. The item service device 200 may use collaborative filtering based on a purchase history of various user terminals 100 sold through the item service device 200.

In particular, in the first embodiment of the present invention, the item service apparatus 200 selects at least one item included in the purchase history of another user having the lowest similarity with the purchase history of a specific user as a heterogeneous recommendation item list. . As a result, it is possible to support a specific user terminal to enable heterogeneous recommendation items contrary to the purchase history. As a result, users can receive recommendations for items that are least relevant to their purchase history or are completely different from their purchase patterns, thereby expanding their interest in more diverse items and making it easier and more interesting to purchase new items of interest. You can proceed.

In addition, in the second embodiment of the present invention, the item service apparatus 200 generates and stores an association rule between items based on the item purchase history of the user terminals 100 and uses the specific user terminal 100. Item recommendation information may be extracted and provided from the purchase history.

In this case, the communication network 300 may support to establish a communication channel between the user terminal 100 and the item service apparatus 200. That is, the communication network 300 may support connection to the item service apparatus 200 of the user terminal 100, transmission and reception of item search screens, transmission and reception of heterogeneous recommended item information, and transmission and reception of signals for item purchase. Here, the communication network 300 may be configured as an internet network device when the user terminal 100 is a terminal such as a desktop, and a mobile communication network device when the user terminal 100 is a terminal supporting a mobile communication function. For example, it may be configured with a base station and a controller.

In addition, the communication network 300 may support the formation of a communication channel for transmitting and receiving data between the item service apparatus 200 and the distributed processing service apparatus 400. To this end, the communication network 300 may be configured as at least one of a wired network device connecting the item service device 200 and the distributed processing service device 400 by wire or a wireless network device connecting wirelessly. In this case, the communication network 300 transmits information for storing the purchase history of all the user terminals 100, transmits and receives a signal for grouping the user terminals 100, transmits and receives a signal for extracting recommended items based on collaborative filtering, and calculated items. Signal transmission and reception such as transmission of recommendation information may be supported.

The communication network 300 is, for example, an IP network, which provides a data service without interruption and a large amount of data transmission through the Internet protocol, an IP network structure that integrates different networks based on the IP all IP (All Various known technologies such as IP), Wibro (Wireless Broadband) network, Wireless LAN including Wi-Fi network, wired communication network, mobile communication network, HSDPA (High Speed Downlink Packet Access) network and satellite communication network In addition to the communication network, it can be a next generation communication network to be developed in the future.

The user terminal 100 may be connected to the communication network 300 using a communication unit, and may be connected to the item service apparatus 200 connected to the communication network 300. The user terminal 100 may receive and output information corresponding to the item search screen provided by the item service apparatus 200. That is, the user terminal 100 may output an item search screen for viewing and purchasing at least one item on the display unit. In this process, the user terminal 100 may receive and output at least one recommendation item provided by the item service apparatus 200 together with the item search screen information. The recommendation item may include heterogeneous recommendation items extracted based on a purchase history of a user having a different purchase pattern from a user, and the association rule described above with respect to a specific item such as a product, a service, or an app purchased by the user. And at least one recommendation item selected by the usability determining component. Such a recommendation item may be calculated based on an arbitrary collaborative filtering method in connection with a specific item such as a product, a service, or an app purchased by the corresponding user terminal 100. In this case, the user terminal 100 may be assigned to a specific group among all user terminals. The recommended item may be an item calculated based on a purchase history of other user terminals belonging to the group. To this end, the user terminal 100 may access the item service apparatus 200 and have a purchase history for various items provided by the item service apparatus 200. The purchase history of each user terminal 100 may be distributed and stored in the item service apparatus 200.

When an input signal for selecting a specific item among items provided on the item search screen or recommended items is generated, the user terminal 100 requests the item service apparatus 200 for detailed information of the corresponding item, and the item service apparatus 200. When the item detail information is received from the display unit, the detailed information may be output to the display unit. Thereafter, when an input signal for purchasing a specific item is generated according to a user request, the user terminal 100 may transmit a message for purchasing the item to the item service apparatus 200 and perform the item purchase according to a predefined procedure. have.

Meanwhile, when the user terminal 100 purchases specific items among the items or recommended items provided on the item search screen, the information on purchasing the item may be provided to the item service apparatus 200. In the first embodiment of the present invention, as described above, the purchase history of the items of the specific user terminal 100 provided from the user terminal 100 may be excluded from the following foreign recommendation items. Accordingly, as the user terminal 100 purchases specific items, new and various items not purchased by the user terminal and mainly purchased by other users may be recommended as foreign recommendation items. In addition, in the second embodiment of the present invention, the purchase history of the specific user terminal 100 provided from the user terminal 100 may be distributed and stored according to the association rule to contribute to item recommendation information generated later.

The item service apparatus 200 is an apparatus that provides an item search screen for the user terminal 100 to view and purchase a specific item, and supports to perform a specific item purchase at the request of the user terminal 100.

In particular, in the first embodiment of the present invention, the item service apparatus 200 has no history of purchase by the user terminal 100 and the user terminal 100 through a collaborative filtering method for the purchase history of the user terminals 100. It is possible to calculate a heterogeneous recommendation item having a very low relevance, and provide the heterogeneous recommendation item to a specific user terminal 100. Here, the item service apparatus 200 may check the purchase history of all the user terminals 100 in order to calculate a foreign item having a low relation with the purchase history of the specific user terminal 100. In this case, the item service apparatus 200 may classify all the user terminals 100 into a certain group and control the calculation of the foreign recommendation item of the specific user terminal 100 to be performed in the corresponding group. In this case, the item service apparatus 200 may operate a distributed processing system to store the purchase history of the user terminals 100, divide the group, and calculate foreign recommendation items in the divided group. That is, the item service apparatus 200 may interwork with the distributed processing service apparatus 400 that stores data and supports retrieval for calculating foreign recommendation items. The distributed processing service device 400 supports effectively storing and retrieving a large amount of data on a purchase history, thereby enabling the user to quickly and effectively recommend items based on the large amount of purchase history. In this case, the item service apparatus 200 supports the connection of the user terminal 100 and provides at least one foreign recommendation item to the user terminal 100. When the user terminal 100 is connected, the item service apparatus 200 may provide the user terminal 100 with information about an item search screen. When the user terminal 100 provides a message for purchasing a specific item, the item service apparatus 200 may support the item purchase process according to the corresponding message.

Meanwhile, the item service apparatus 200 may support heterogeneous recommendation items suitable for the corresponding user terminal 100 when the specific user terminal 100 is connected. To this end, the item service apparatus 200 may check the identification information of the specific user terminal 100 and may request a foreign recommendation item from the distributed processing service apparatus 400 based on the identification information of the specific user terminal 100.

The distributed processing service apparatus 400 may perform distributed storage of purchase histories of the user terminals 100, grouping of all user terminals 100, distributed storage of purchase histories of grouped user terminals 100, and grouping of stored user groups. It may be configured to include a plurality of data nodes to calculate the heterogeneous recommendation item based on the purchase history of them, to provide the calculated heterogeneous recommendation item to the item service device 201. Each data node supports the MapReduce function, in particular, the distributed processing service device 400 uses the first MapReduce function for grouping user terminals, and the heterogeneity for the specific user terminal 100 in the purchase history of the grouped user terminals. The second MapReduce function may be used to calculate the recommended item. Through this, the distributed processing service device 400 may calculate and provide a heterogeneous recommendation item more quickly based on a large purchase history.

As described above, in the first embodiment of the present invention, the item recommendation system 10 calculates a heterogeneous recommendation item having low relevance to the purchase history of the user terminals 100 by applying a grouping method and an arbitrary collaborative filtering method. In addition, by providing the calculated foreign recommendation items to the specific user terminal 100, a specific user may be provided with an opportunity to experience various types of items.

In the second embodiment of the present disclosure, the item service apparatus 200 may generate an association rule based on a purchase history of the user terminals 100, and may generate a specific user terminal based on the association rule and a predefined usability determination component. The item recommendation information of 100 may be generated. The item service apparatus 200 may provide the generated item recommendation information to the corresponding user terminal 100. In this case, the distributed processing service device 400 may support a distributed processing scheme for storing and using an association rule.

In other words, when a specific user terminal 100 connects, the distributed processing service apparatus 400 may provide the recommendation item to the user terminal 100 based on a purchase history and association rules of the user terminal 100. It may be configured to include a plurality of data nodes to distribute and store the purchase history of the 100, and to support the storage and retrieval of association rules.

Here, each data node supports the MapReduce function, and based on this, create and store association rules, and evaluate the association rules based on a predefined usability determination component.

The evaluation result evaluated by each data node may be used to select a preliminary recommendation item, and the item service apparatus 200 lists related rules based on the purchase history and identification information of the user terminal 100 and determines a predefined usefulness. By applying the component, association rules having a certain evaluation value among the association rules may be selected.

In addition, when the specific user terminal 100 purchases the item, the item service apparatus 200 controls to generate the item purchase information as a new association rule by applying a predefined association rule length and store it in at least one data node. can do.

As described above, according to the second embodiment of the present invention, the item recommendation system 10 classifies and stores the history of purchasing a specific item by a plurality of user terminals 100 according to a predefined association rule, and determines usefulness. After evaluating the stored association rules classified based on the element, and extracting items having a predetermined evaluation value or more, the extracted items may be provided in a predetermined number according to a predefined criterion. Accordingly, the specific user terminal 100 may obtain more reliable item recommendation information based on the purchase history of other user terminals who purchased the item using the item service apparatus 200 while using the item service apparatus 200. Can be. The present invention can support the user who purchased a particular item to properly reflect the needs of users who purchased the same item by reflecting the need for other items associated with the item.

Meanwhile, in the second embodiment of the present invention, the item service apparatus 200 stores association rules using distributed data nodes together with access of the user terminal 9100 and support for item purchase, and is evaluated based on the availability determination component. Although calculating the result as an example, the present invention is not limited thereto, and the user terminal 100 may be configured by using distributed data nodes and an item service apparatus that supports connection and purchase of an item according to a system design scheme. It may be divided into a distributed processing service device 200 that stores the purchase history of these, checks the association rules and determines the usefulness.

In addition, in the above description, the item service apparatus 200 is separately described as the item service apparatus 200 and the distributed processing service apparatus 400, but the present invention is not limited thereto. In other words, the item service apparatus 200 may provide a heterogeneous recommendation item calculation according to the connection and item purchase support of the user terminal 100, the grouping of the user terminals 100, and the application of a collaborative filtering method through one service apparatus. That is, the item service device 200 may be configured to include the functions of the distributed processing service device 400.

2 is a view showing in more detail the configuration of the item service apparatus 200 according to the present invention.

Referring to FIG. 2, the item service device 200 of the present invention may include a device communication unit 210, a device storage unit 250, and a device control unit 260.

The item service apparatus 200 of the present invention having such a configuration extracts foreign users with respect to a specific user according to the first embodiment and recommends products to a specific user by using a purchase history of a group including such foreign users. Can help create a county. In this case, in the case of a large amount of data, in order to extract other users having a different purchase history from a specific user, a similarity between the purchase history of the specific user and the purchase history of all the other user terminals may be calculated to extract a user having a similarity or less than a predetermined value. . In this case, when the total number of users is N, N X (N-1) / 2 large computational loads are required. Therefore, in the first embodiment of the present invention, the item service apparatus 200 supports grouping heterogeneous users in a Hadoop MapReduce-based distributed environment and calculating heterogeneous recommendation products based on purchase history of a specific group of heterogeneous users. .

In addition, in the second embodiment of the present invention, the association rules evaluated based on the usability determination component based on the purchase history of the user terminal 100 may be calculated and provided as item recommendation information. To this end, the item service apparatus 200 may use a distributed processing system for storing and operating an association rule.

To this end, the device communication unit 210 supports communication channel establishment with the user terminal 100 through the communication network 300. The device communication unit 210 may provide an item search screen to the user terminal 100 under the control of the device control unit 260. Thereafter, when the device communication unit 210 requests a purchase of a specific item through the item search screen, the device communication unit 210 may receive a corresponding purchase request message from the user terminal 100 and transmit the received purchase request message to the device controller 260. In particular, the device communication unit 210 may transmit to the corresponding user terminal 100 a purchase history of a user connected through the user terminal 100 and a recommendation item to which any cooperative filtering scheme is applied under the control of the device controller 260. have. The device communication unit 210 may receive item purchase request information presented through the recommended item from the user terminal 100 and provide the received item purchase request information to the device controller 260. In addition, the device communication unit 210 may support data transmission and reception required during the item purchase process.

The device storage unit 250 may store an operation program for operating the item service device 200 and a purchase history of users for calculating and providing a recommendation item according to embodiments of the present disclosure. In particular, the device storage unit 250 may store the purchase history 251 of the user terminals 100. The purchase history 251 may be managed including information such as "user ID, purchased goods ID, date ...", and the like. In addition, the device storage unit 250 may further store identification information and group ID information of the user terminal 100.

According to another embodiment of the present invention, when the device storage unit 250 stores the association rule in multiple data nodes of the distributed processing service device 400, the user log inquiry list may be used to support retrieval of information on the association rules. 253 may include. Here, the user log inquiry list 253 may include address information on which data nodes the purchase history information of a specific user terminal is stored. The device controller 260 may generate a search message for searching for related rules based on the information included in the user log inquiry list 253 and control the search message to be delivered to the data nodes. In this case, various information (eg, purchase history) related to the user terminals 100 may be distributed and stored in the data nodes of the distributed processing service apparatus 400. That is, the distributed processing service device 400 is distributed and stored in a plurality of data nodes supporting data storage and retrieval by the MapReduce function.

The device controller 260 is configured to support the overall operation of the item service device 200 according to the embodiment of the present invention in cooperation with the device communication unit 210 and the device storage unit 250. That is, the device controller 260 controls the device connection of the user terminal 100, providing an item search screen, grouping of data for all user terminals, distributed storage, searching, processing for calculating recommendation items, and providing the calculated recommendation items. can do. In this case, the detailed configuration of the device controller 260 may vary according to the recommended item extraction method.

That is, as in the first embodiment of the present invention, heterogeneous recommendation items may be extracted from the purchase history of users having different purchase patterns from the user, and as in the second embodiment, generated based on the analysis of the purchase history. Based on the association rule, the related item may be extracted as the recommended item from the purchase history of the user.

Hereinafter, detailed configurations of the device controller 260 and specific item recommendation methods thereof will be described for each of the first and second embodiments.

3 is a diagram illustrating in detail the configuration of the device controller 260 according to the first embodiment of the present invention.

Referring to FIG. 3, in the first embodiment of the present invention, the device controller 260 of the item service device 200 includes a data processor 261, a heterogeneity calculator 262, and a product recommender 263. Can be configured.

The data processor 261 may control appropriate data classification and storage before calculating the heterogeneity based on the purchase history of each user terminal 100. In particular, the data processing unit 261 collects the purchased product ID for each user ID for the entire user purchase history within a predetermined period, and has a data type such as "user ID, purchased product ID, purchased product ID, purchased product ID ...". Process as much as possible. The data processor 261 performs a first MapReduce process and a second MapReduce process.

First, in order to perform the first MapReduce process, the data processing unit 261 generates a natural number between 1 and P for each user terminal 100, and connects the generated natural numbers Q (using a concatenation operator) to a specific user terminal. It is specified by the group ID of (100). Here, P and Q are values for determining the number of all user terminals 100 and the size of the group to be determined, and are natural numbers of one or more. Here, when the total number of user terminals 100 is N, the size of each group is set to N / (P ^Q ) on average. For example, when there are 10 million user terminals 100, when P is 17 and Q is 3, an average of 2,000 user terminals may be grouped per group. Accordingly, P and Q values may be defined experimentally or statistically so that user terminals corresponding to a certain number of groups can be included.

The data processing unit 261 collects data in the form of {Key, value} = {group ID, (user ID, product ID, product ID, ...)} in the mapper stage on the MapReduce Framework. Accordingly, while performing the first MapReduce process, the data processing unit 261 may have a reducer in which user terminals having a specific group ID are collected.

While performing the second MapReduce process, the data processing unit 261 performs {Key, value} = {(user ID, product ID, product ID,…), (others of the group) for each Group generated by the first MapReduce process. User ID, product ID, product ID,…) (other user IDs of the corresponding group, product ID, product ID,…)} are outputted to Mapper. Accordingly, while performing the second MapReduce process, the data processor 261 may have a reducer in which purchase histories of other user terminals belonging to the same group as the corresponding user terminal are collected for a specific user ID.

As described above, the data processing unit 261 generates a specific group including a specific user terminal by performing a grouping process on all user terminals while performing the first MapReduce process and the second MapReduce process, and generates a specific group including the specific user terminal. It can be processed to collect the purchase history of other user terminals belonging to.

The heterogeneity calculator 262 calculates a Jaccard similarity between a specific user ID and another user (H name). That is, the heterogeneity calculator 262 may calculate H similarity between the purchase history of the specific user terminal 100 and H other user terminals. Here, the H name corresponding to the number of other user terminals may also be determined experimentally.

In this case, when H is the total number of users of the group, the heterogeneity calculator 262 calculates the total Jaccard similarity between the specific user and other users in order to calculate the heterogeneity of the specific user ID. Even in this case, if there are M users in the same group, M X (M-1) / 2 operations are performed, which significantly reduces the number of users. As a result, the computational load on the similarity calculation can be reduced. The dissimilarity value for each user terminal may be expressed as in Equation 1 below.

Equation 1

Where 1-J (UserA, UserB) is the heterogeneity between user A and user B, | A∩B | is the number of intersections of the goods purchased by user A and user B, and | A∪B | is user A The number of unions of goods purchased by and user B. That is, for collaborative filtering based on purchase history of heterogeneous user terminals, a product purchased by a user having a Jaccard similarity of 0 should receive the highest score, so the heterogeneity calculator 262 defines (1.0-jaccard Similarity) as a heterogeneity. can do. Meanwhile, when the heterogeneity calculation is completed, the heterogeneity calculator 262 may provide the values to the product recommender 263.

The heterogeneity calculator 262 may express the purchase history for each user as a matrix (hereinafter, referred to as a "purchase history matrix") in which rows and columns are represented by user information and product information, respectively, for the above-described processing.

The heterogeneity matrix between users is calculated based on Equation 1 through the above-described purchase history matrix, and then multiplied by the purchase history matrix and the heterogeneity matrix between users to recommend based on heterogeneity for each user. The score of the recommended item for each product can be calculated. Here, the score of the recommendation item means a heterogeneity-based recommendation score for each product.

The product recommendation unit 263 obtains a list of heterogeneous recommendation products for each score calculated from the heterogeneity calculator 262, and then removes the list of products already purchased by the user terminal 100 from the recommendation product list. The heterogeneity calculator 262 extracts only a list of the top K products predefined in accordance with the service requirements.

4 to 7 illustrate examples of the heterogeneity calculation calculated by the heterogeneity calculator 262 of the present invention.

4 illustrates a purchase history matrix, and illustrates three users by way of example. That is, Figure 4 shows the purchase history for the heterogeneity between the specific user of the current Reduce and the remaining H = 2 people. Meanwhile, similarity described in the present invention may be applied to similarity of other measurement methods such as cosine similarity, in addition to Jaccard similarity.

Meanwhile, referring to FIG. 4, user # 1 has a history of purchasing goods # 1, product # 2, and product # 4, and user # 2 has a history of purchasing goods # 3 and # 4. And user # 3 has a history of purchasing goods # 1, goods # 3 and goods # 4. The heterogeneity matrix for other users of user # 1 under these conditions is shown in FIG. 5. Here, the heterogeneity degree of the user # 1 may be 0, and the heterogeneity degree of the user # 1 may be 0 or more different from that of the other users # 2 and # 3. The larger the value, the greater the difference in propensity of purchase history among users.

Accordingly, as shown in FIG. 6, the dissimilar concatenation factor (dissimilar CF) may include product # 3 and product # 4 of 1.25, and product # 1 of 0.5. In this case, matrix multiplication of the heterogeneity matrix D and the purchase history matrix L may be performed to calculate a score of the recommendation item based on the heterogeneity of each user as shown in FIG. 6.

The product recommendation unit 263 extracts the recommendation list in the order of the products except the products already purchased by the user in the order shown in FIG. 6. That is, in the case of user # 1, the product recommendation unit 263 may configure a list of heterogeneous recommendation products in the order of product # 3, product # 4, product # 1, and product # 2. Here, since the user # 1 already has a history of purchasing goods # 1, # 2, and # 4, the product recommendation unit 263 excludes products having a purchase history, and the foreign recommendation product including the product # 3. You can create a list.

The product thus provided may be displayed on the screen of the user terminal 100 as a foreign recommendation item of a user having a different purchase style as shown in FIG. 7. When the foreign recommendation item is calculated, the product recommendation unit 263 may perform a screen configuration as shown in FIG. 7 and provide a corresponding screen to the user terminal 100. Accordingly, the user may be recommended for foreign recommendation items of users having a completely different purchase method from the screen confirmation as shown in FIG. 7.

As described above, the present invention supporting the heterogeneous recommendation item function of the present invention generates a list of recommended products only based on the purchasing behavior of users having similar purchase history with a specific user, but the purchase behavior is completely different from the specific user. By generating a list of heterogeneous recommendation products based on the purchase history of users and providing them through a separate menu, they can induce user interest and contribute to service activation.

8 is a view for explaining the operation of the item service device 200 according to an embodiment of the present invention, Figures 9 to 12 is a view showing each step of Figure 8 in more detail.

8 to 12, the item service apparatus 200 of the present invention collects the purchase history of all users through interworking with the user terminals 100 in step S110. In more detail, by collecting information related to a purchase transmitted from the user terminals 100, purchase history of all the users may be collected. In more detail, the item service apparatus 200 collects a purchase history including a user ID, a product ID, and a purchase date in operation S111 as shown in FIG. 9. The item service apparatus 200 extracts a user log for each period in step S112, and groups a purchase log for each user terminal in step S113. Thereafter, the item service apparatus 200 may collect data such as a user ID and a list of purchased products (item ID, product ID, product ID, ...) in step S114.

Referring back to FIG. 8, the item service apparatus 200 may perform grouping of the user terminals 100 in operation S120. Referring to the grouping process in more detail, as shown in FIG. 10, the item service apparatus 200 divides all users into a predetermined number of groups according to P and Q values predefined in operation S121, and divides each group. By assigning a group ID to the group ID for each user can be generated. Thereafter, the item service apparatus 200 may output data having a form of {Key: Value} = {Group ID: User ID, purchased product list} according to the execution of the Mapper function as in step S122. In operation S123, the item service apparatus 200 may perform grouping for each group ID. Accordingly, the item service apparatus 200 may perform purchase history grouping of each of the user terminals 100 classified by Group ID as in step S124.

8 again, the item service apparatus 200 performs data classification for each user terminal in step S130. The data classification process may be performed as shown in FIG. That is, the item service apparatus 200 performs grouping of data for each user ID in step S132, except for all users in the same group per user in the same group in step S131, and thus, the same group for each user ID as in step S133. The purchase history of other users belonging to can be classified and stored.

Next, the item service apparatus 200 performs a heterogeneity calculation for each user in operation S140, and generates a heterogeneous recommendation item list for each user terminal based on the calculated heterogeneity. In more detail, as shown in FIG. 12, the item service apparatus 200 may calculate a heterogeneity-related score for each user ID in operation S141 and generate a product list having the highest heterogeneity for each association score. have. The item service apparatus 200 may generate a heterogeneous recommendation item list by excluding the products included in the purchase history for each user ID from the product list generated in step S142.

Subsequently, when a specific user is selected in operation S150, the item service apparatus 200 may provide a list of foreign recommendation items for the selected user. That is, the item service device 200 confirms identification information of a user currently connected to the device communication unit 210 through the user terminal 100, and provides the user terminal 100 with a foreign recommendation item generated by the corresponding user ID. Can be supported.

As described above, the method for providing a heterogeneous recommendation item list according to the operation of the item service apparatus 200 of the present invention may generate and prepare a heterogeneous recommendation item list for all user terminals in advance. According to the present invention, when a specific user terminal accesses the item service apparatus 200 or determines to provide a foreign recommendation item to a specific user terminal, the user ID may be checked and a list of pre-generated foreign recommendation items may be provided. . Accordingly, the item service apparatus 200 of the present invention supports generating heterogeneous recommendation items in a time zone in which a service device is less loaded, and is previously generated without performing a heterogeneous recommendation item generation process at a time when various user terminals are connected. It can support providing a list of heterogeneous recommended items. In this case, the item service apparatus 200 may update the heterogeneous recommendation item list at regular intervals in consideration of the occurrence of the update of the purchase history according to the purchase of the items by the users.

13 is a view for explaining another operation example of the item service apparatus 200 according to the first embodiment of the present invention.

Referring to FIG. 13, the item service apparatus 200 according to the first embodiment of the present invention may collect the purchase history of the entire user terminal 100 in step S210 and perform grouping in step S220. Here, steps S210 and S220 may be the same process as the purchase history collection process and the grouping process performed by the item service apparatus 200 with respect to the user terminals in steps S110 and S120 of FIG. 8 described above.

Next, the item service apparatus 200 performs a specific user selection belonging to a specific group in step S230. That is, the item service apparatus 200 may identify identification information of a specific user currently connected through the user terminal 100, and identify a specific group to which the user belongs based on the identification information of the corresponding user. To this end, the item service apparatus 200 may store and manage a list of group IDs to which each user belongs in the process of grouping all users.

Next, the item service apparatus 200 may calculate heterogeneity with the specific user selected in operation S240 and other users belonging to the same group. That is, the item service apparatus 200 may check the purchase history of all users of the group to which the selected user belongs, and calculate heterogeneity by applying the collaboration filtering method described above to the purchase history of the selected user and other users of the group. have.

Thereafter, the item service apparatus 200 may support to generate a list of heterogeneous recommended items based on the heterogeneity calculated in operation S250 and the purchase history of the selected user terminal, and provide the same to the corresponding user terminal.

As described above, the item recommendation method based on the operation of the item service apparatus 200 according to the embodiment of the present invention is a part for generating a list of heterogeneous recommendation items when a specific user terminal is selected to access or provide a foreign recommendation item. Perform the process. Accordingly, the present invention is characterized in that the load on the service device can be reduced by performing the calculation at the time of providing the list to the user terminal. In addition, when the step S230 is performed before the step S210, the present invention can support the update time of the foreign recommendation item provided to the specific user terminal to be the latest time.

Next, an item recommendation service according to a second embodiment of the present invention will be described. In the second embodiment of the present invention, as the association rule is used for item recommendation, the association rule used in the present invention will first be described with reference to FIGS. 14 and 15.

For convenience of description, it is assumed that a plurality of users (Users 1 to 7) access the item service apparatus 200 through the user terminal 100 to purchase an item (product) as shown in FIG. 14. .

That is, the first user (User 1) has a history of purchasing Beer, Chicken, Milk, the second user (User 2) has a history of purchasing Beer, Cheese. In addition, the third user (User 3) is Cheese, Boots, the fourth user (User 4) is Beer, Chicken, Cheese, the fifth user (User 5) is Beer, Chicken, Clothes, Cheese, Milk, A user (User 6) has a history of purchasing Chicken, Clothes, Milk, and the seventh user (User 7) has purchased Chicken, Milk, Cloths, respectively.

In this case, the item service apparatus 200 according to the second embodiment of the present invention may generate an association rule as shown in FIG. 15 based on the purchase history of the above-described users. The association rule may be expressed as a set of items purchased by the same user. Accordingly, the item service apparatus 20 may generate various types of association rules according to the length of the association rule. For example, the item service apparatus 20 may generate a rule length of 2 (Beer, Chicken), (Beer, Milk), (Chicken, Milk), and the like, as shown. In addition, the item service apparatus 200 may generate (Beer, Chicken, Cheese), (Chicken, Milk, Clothes) and the like as an association rule when the length of the rule is three. In summary, the item service apparatus 200 may generate association rules corresponding to various combinations of items purchased together by the same user based on a purchase history purchased by each user, according to a predefined rule length. . Each generated association rule may be distributed and stored in a plurality of distributed processing apparatuses 400 included in the item service apparatus 200 or separately managed. In the following description, the item service apparatus 200 will be described based on an example of generating and operating an association rule based on a rule length of two.

Meanwhile, in order to determine the usefulness of the association rule, the item service apparatus 200 according to the second embodiment of the present invention may use three components. That is, the item service apparatus 200 supports to determine the usefulness of the association rule based on at least one of support, confidence, and lift. To this end, the item service apparatus 200 may derive each usefulness determination component from the statistical result of the association rules.

More specifically, if the purchase history for each user is defined as one association rule or a transaction in FIG. 14, the support may correspond to the number of transactions including a specific product A and a product B among all transactions. Here, the goods A and the goods B are for illustrating different kinds of goods and may be any one of the goods shown in FIG. 14.

The aforementioned support degree may be defined as in Equation 2 below.

Equation 2

Where s (A => B) is support, T is the number of transactions including commodity A and commodity B, and D may be the total number of transactions. In addition, A may refer to product A and B to product B, and product A and product B may be specific products included in a purchase history of users through the item service apparatus 100 as described above. Accordingly, the degree of support may be the ratio of the association rule including both goods A and B among the overall association rules.

Meanwhile, the reliability may refer to a ratio of a transaction including products A and B simultaneously with respect to a transaction including product A. That is, the reliability may be a ratio of the association rules including the product A and the product B at the same time to the association rules including the product A. This reliability can be defined by equation (3).

Equation 3

Finally, the degree of improvement represents the reliability of the product A-> B of the reliability of the product B. In other words, the degree of improvement may be the ratio of the transaction in which the purchase of the article B in the situation where the purchase of the article A of the purchase of the article B. Alternatively, the degree of improvement may be the ratio of the association rules for purchasing the commodity B in the context of purchasing the commodity A among the total association rules for purchasing the commodity B. If the improvement is less than 1, it means a negative relationship, and if it is greater than 1, it can be interpreted as a positive relationship. The degree of improvement may be defined as in Equation 4.

Equation 4

In the above, the association rule and the usefulness determination component for the item recommendation according to the exemplary embodiment of the present invention have been described. Hereinafter, the configuration of the device control unit 260 and the item recommendation method of the item service device 200 according to the second embodiment of the present invention that performs the item recommendation based on the above-described association rule and usability determination component will be described in more detail. Let's explain.

The device controller 260 of the item service apparatus 200 according to the second embodiment of the present invention collects association rules based on item purchase history information, and generates item recommendation information based on the same. It may include a configuration as shown.

16 is a view showing in more detail the configuration of the device control unit 260 of the item service device 200 configuration of the present invention. In addition, FIG. 17 is a view for explaining the distribution of association rules distributed, FIG. 18 is a view for explaining the storage process of the purchase history for each user, Figure 19 is a view for explaining the inquiry process of the association rules for each user.

16 to 19, the device controller 260 of the item service apparatus 200 according to the second embodiment of the present disclosure may include an association rule distributed storage support unit 261, an association rule inquiry unit 263 for each user, The recommendation information provider 265 may be included.

The association rule distributed storage support unit 261 generates an association rule based on the purchase history of users who use the item service apparatus 200, and generates the association rule as shown in FIG. 17 of the distributed processing service apparatus 400. This configuration supports storing data nodes. FIG. 17 shows association rules corresponding to a plurality of data nodes RDB # 1 to RDB #P when eight head product IDs and eight result product IDs each have specific support, reliability, and improvement and are provided as association rules. Indicates that they can be stored separately. In this case, association rules composed of the same head product IDs may be stored in each data node, and association rules of the same head product ID may be divided and stored in a plurality of data nodes according to the data amount of the association rule.

When the association rule distributed storage support unit 261 purchases items provided by the item service apparatus 200 through various user terminals 100, the association rule distributed storage support unit 261 may collect information about purchased items. The association rule distributed storage support unit 261 may generate an association rule based on a relationship between items based on a predefined rule length, for example, a rule length of two. In addition, the association rule distributed storage support unit 261 may distribute and store the generated association rules in data nodes in a predefined format or in a specific data node. The association rules stored at this time may be stored in a form including a head product ID, a result product ID, support, reliability, and improvement. Here, the head product ID may correspond to a search item for searching for the result product ID.

For example, when a specific user has a history of purchasing a specific product, the purchased product may correspond to a head product ID. The resulting product ID may be an item that is included in the item recommendation information to be provided to the user terminal 100 as a stored product paired with the head product ID. The degree of support, reliability, and improvement may be a determining component that determines whether a specific result product ID is included in the item recommendation information. In particular, according to an item providing service design scheme, at least one of support, reliability, and improvement may determine whether to use the resultant product ID as item recommendation information. That is, the item recommendation information may be determined based on any one of three usability determination components, two of three usability determination components, and one of three usability determination components according to an item providing service design scheme. For example, according to an item providing service design scheme, it may be determined to include the searched result product ID in the item recommendation information using only reliability.

If the user terminal 100 accesses the item service apparatus 200, the association rule inquiry unit 263 for each user may check the purchase history of the user connected through the user terminal 100 from a specific data node. . To this end, the user-specific association rule inquiry unit 263 may control to store the item purchase history of the user terminals 100 using the item service apparatus 200 in specific data nodes, as shown in FIG. 18.

In addition, the association rule inquiry unit 263 for each user extracts preliminary recommendation items having a usability determination component having a predetermined value or more based on a purchase history of a specific user. In this case, the association rule inquiry unit 263 for each user may search for an RDB, that is, a data node, to bring the actual association rule based on the hash value of the purchase item ID for each user as illustrated in FIG. 19. Referring to FIG. 19, items included in purchase histories of user ID 1 stored in data node # 1 are stored in association rules stored in data nodes RDB # 1 to RDB #P, respectively. Items which become a key value to be used and whose at least one of the usefulness determining components among the association rules stored in the respective data nodes are equal to or greater than a predetermined value may be provided as preliminary recommendation items.

In this case, the association rule inquiry unit 263 for each user may provide preliminary recommendation items corresponding to the specific item according to the purchase history to the recommendation information provider 265. In this process, the user-specific association rule inquiry unit 263 defines only a predetermined number of purchase histories as a valid purchase history when the purchase history of the user terminal 100 is greater than or equal to a certain number, and the association according to the corresponding purchase history. You can query the rules. Alternatively, the association rule inquiry unit 263 for each user may define a certain period when the purchase history is more than a certain number, and define only the items purchased within the latest or predefined corresponding period as a valid purchase history. The association rule inquiry unit 263 for each user may define all items as a valid purchase history when the number of items included in the purchase history is less than or equal to a certain number, and support to search for preliminary recommended items for each item.

When the preliminary recommendation items are delivered from the association rule inquiry unit 263 for each user, the recommendation information providing unit 265 applies the usability determination component to the delivered preliminary recommendation items and selects the recommendation items. For example, the recommendation information providing unit 265 may sort the delivered preliminary recommendation items in the order of reliability, and determine the preliminary recommendation items that fall into a predetermined rank in the order of high reliability as the recommendation item. In this process, when there are preliminary recommendation items having the same rank, the recommendation information provider 265 may set higher priority of the preliminary recommendation items of the recent purchase history item.

Meanwhile, in the item recommendation method, the first method is a method of generating a recommendation product proportional to the number of purchased products for each user, and P-association of B_u purchased product IDs of each user and each purchased product ID for each user is Head. You have a rule. Here, the recommendation item extraction criterion among the P association rules provided as the preliminary recommendation items may be used at least one of the usability determination component. For example, the item recommendation criteria may use all of the support, the reliability, and the improvement, or some of the reliability and some of the improvement.

In the following description, a description will be made based on an example of extracting a recommendation item based on reliability. First, the first method is to generate B_u recommended products for a user who has purchased B_u. The following process is repeated from the B_u first products to the B_uth products (i = 1, 2, ..., B_u). In this process, the order of all purchased products may be simply sorted by the order of the purchased product ID, or when the B_u of the purchased product history is too large. Alternatively, the association rule may be sorted in the order of purchase product ID based on user preference.

Meanwhile, the item service apparatus 200 selects a rule having the highest confidence value from the association rule of which the product ID of the head is B_i (i = 1, 2, ..., B_u) and stores the Consequent product ID. The item service apparatus 200 searches the association rule of which the product ID of the head is B_i from the next purchased product, selects the rule with the highest confidence value, and brings the Consequent product ID of the corresponding rule. 200) checks whether the imported product ID already exists in the recommended product generated from the previous purchased product. If the corresponding recommended product already exists, the item service apparatus 200 obtains the Consequent product ID of the next higher-trust rule from the association rule of which the product ID of the head is B_i and repeats the existence of the same recommended product. . On the other hand, if the association rules P extracted for each purchase history are significantly smaller than B_u or there are too many B_u, the consequent product ID does not overlap with the previous recommendation result even if all P association rules with the product ID of Head are B_i. May not be obtained. Accordingly, P to be applied to the association rules for each purchase history may be experimentally adjusted in consideration of a decrease in computation speed according to system load and the number of recommended items to be provided.

On the other hand, if there are too many B_u, for example, a user who purchases 10,000 items may be difficult to use even if they generate as many recommended products. Therefore, the item service apparatus 200 may designate and use the number K of products to be created at the maximum for each user. In this case, the item service apparatus 200 may generate as many as K pieces when K is smaller than B_u and B_u pieces in order when K is larger than B_u, and may not generate a recommendation product for subsequent products. Through this process, the item service apparatus 200 may generate as many K items as the number of B_u or B_u is too large for each user.

The second method in the item recommendation function method is a method of generating a specific number of recommended products regardless of the number of purchased products for each user. The item service apparatus 200 may collect P association rules including B_u purchased product IDs of each user and each purchased product ID for a single user. Similarly to the first method, in the second method, the item service apparatus 200 may extract association rules to be used as a preliminary recommendation item based on at least one of support, reliability, and improvement included in the usefulness determining component. The second method is to generate K recommended products for a user who purchased B_u.

Here, K is a value determined in terms of applied services and products. To select K recommended items, do the following: First, the item service apparatus 200 prepares a storage for storing up to K recommended product IDs. The item service apparatus 200 sorts the total (B_u X p) association rules in at least one of the usability determination components, for example, in order of reliability. The item service apparatus 200 starts with the association rule with the highest confidence level and repeats the following process until K pieces are added.

The item service apparatus 200 checks whether a Consequent product ID of the current rule exists in a store storing existing K pieces and adds the recommended product to the recommended product if it does not exist. When there are a plurality of association rules having the same reliability and different Head product and Consequent product IDs, and the total number of stored products does not exceed K, the item service apparatus 200 may use all Consequent product IDs of each association rule. If the number of association rules exceeds K in the process of adding all the association rules having the same reliability to a specific confidence value to the recommended product store, the item service apparatus 200 selects the number to be arbitrarily K.

For example, if K = 100 and 98 are stored in the order of reliability, and there are 4 associative rules having the next highest confidence value of 0.56, and the consequent product ID of each associative rule is different, the item service device 200 may randomly select two of the four product IDs and add them to the recommended product. Here, the item service apparatus 200 records the appearance frequency of the head product ID of the association rule used to generate the recommendation product without selecting a recommendation product arbitrarily in the process of adding the recommendation product, and has the lowest head product ID association. Can be used in order of rule. Through such an option, the item service apparatus 200 may prevent a case in which a plurality of Consequent products of the association rule in which the corresponding Head product exists are recommended as a recommendation product, because the reliability value of the association rule in which the specific product is the Head product is high.

On the other hand, if the total sum of P, the number of association rules for each product in each purchase history, is significantly smaller than K, it is difficult to generate K recommended products, so P may be designated as a multiple of K.

This ratio may be appropriately adjusted according to the K value and the type of service. For example, an item such as a book may have an average number of purchased products per user, while an item such as music content may be differentially adjusted according to a service type such as tens / day.

In addition, the number of association rules generated may vary according to the support, reliability, and improvement values, which are the usefulness judgment components that are specified when creating the association rule. When the number of association rules generated is too small, the K value is also changed. If you specify a large, too large value for P as a multiple of K, you may not get P association rules for each product. Accordingly, the designer who designs the item service apparatus 200 may experimentally adjust the usability determination component and the P value appropriately. Through the above-described process, the item service apparatus 200 may generate K recommendation products for each user.

In the above-described first and second methods, the item service apparatus 200 applies the first method of providing three recommended products to a user who has purchased three items, or has an association rule even if the three items are purchased. The second method of recommending K products may be selectively operated, or the two methods may be mixed and used according to the intention of the system designer.

20 is an example of an association rule queried according to the second embodiment of the present invention, and illustrates association rules sorted based on reliability among usability determination components. 20 described above shows association rules generated based on a purchase history in the item service apparatus 200. Based on this association rule, when the user terminal 100 purchases a certain number of goods illustrated in the head goods of FIG. 20, the item service apparatus 200 may use FIG. 21 among consequent goods by the first method. As shown in FIG. 22, a recommendation result as shown in FIG. 22 may be provided by applying the second method. K = 5.

As described above, the present invention can recommend products that are most statistically related based on their purchase history for each user. In particular, the present invention may generate and support a recommendation product having a high relation in proportion to the number of purchases per user, or generate and support the number required by the service device regardless of the number of purchases per user. The item recommendation method according to the second exemplary embodiment of the present invention provides a required number of recommendation results for users who do not have a recommendation result or a desired number of recommendation results by other collaborative filtering methods in combination with other recommendation methods. can do.

FIG. 23 is a diagram illustrating a method of operating an item service apparatus 200 according to a second embodiment of the present invention.

Referring to FIG. 23, the item service apparatus 200 may have a standby state in step S310. This means that the item service device 200 activates the device communication unit 210 so that the user terminals 100 can access and prepares for providing a service page according to the connection of the user terminals 100. In addition, the item service apparatus 200 may support performing a procedure for registering a member of each user terminal 100. Thereafter, the item service apparatus 200 checks the connection of the specific user terminal 100 in step S320. In this step, if there is no connection of the user terminal 100, the process may branch to step S310 to maintain the standby state.

In addition, when a specific user terminal 100 accesses in step S320, the item service apparatus 200 may perform authentication of the corresponding user terminal 100. The item service apparatus 200 may check the purchase history of the corresponding user terminal 100 based on the unique identification information of the user terminal 100. To this end, the item service apparatus 200 may control to store the purchase history in a specific data node in the process of purchasing an item for each user terminal 100 in advance. The item service apparatus 200 may check the purchase history of the user terminal 100 stored in the specific data node based on the identification information of the user terminal 100. When the purchase history is confirmed, the item service apparatus 200 applies the association rule previously generated from the corresponding purchase history according to at least one of the above-described first and second methods, and generates item recommendation information in operation S330. And, it can be provided to the user terminal 100.

Next, the item service apparatus 200 checks whether the corresponding user terminal 100 purchases the product at step S340, and when the product is purchased, branches to step S350 to further apply the purchased product to generate an association rule for product purchase. And store it in a data node. That is, when a specific user terminal 100 purchases an item, the item service apparatus 200 may support to generate an association rule by applying the corresponding item purchase history and apply it to the previously generated association rule.

Next, the item service apparatus 200 confirms whether or not the user terminal 100 is disconnected at step S360, and returns to step S310 when the user terminal 100 is released from connection, and supports to perform the following process again. Branching to the step S330 may support to perform the following process again.

The item service apparatus 200 of the present invention described above may use at least one distributed data node in the process of storing the association rule for the product purchase history, and in particular, when the user's data is large, It can be distributed and stored on large machines. The item service apparatus 200 of the present invention can effectively operate the item recommendation procedure for each user in a distributed environment when there are distributed association rules and a large user purchase history.

The present invention described above may be implemented on a MapReduce distributed framework. However, the present invention may obtain the same result by applying the same technology when there is the same data even in a distributed environment. MapReduce is a software framework designed by Google to support distributed computing. This framework was developed to support parallel processing in large clusters of petabytes of large data in untrusted computers. This framework consists primarily of a functional base called Map and Reduce, which is commonly used in functional programming. Hadoop is open source, supporting distributed applications that run on large clusters of computers that can handle large amounts of data. The distributed processing system of the present invention implements Hadoop Distributed File System (HDFS) and MapReduce (HDFS), which can replace the Google file system.In the form of HDFS file storage, Hadoop's large files are stored on multiple computers. A datanode corresponding to each computer may be a unit for actually storing and processing a file. And for fault tolerant, Hadoop can store the same file for up to X replications. That is, even if a file (information) stored in one data node disappears, it can be recovered by using a file stored in another data node. From the user's point of view, the file appears to be in one place, but internally it exists for as many replications. The default is usually set to 3.

Input data present in HDFS is split into smaller units through input splits. A central coordination process called Job Tracker assigns each input split to a Map task for a particular data node, taking into account the physical location of the file. The task tracker for each data node receives and processes each map task. The (key, value) values generated by a map or combiner at a particular data node are passed through the network to the reducer at a particular data node if necessary. The Reduce task sorts the (key, values) sent as the result of the map task by key, executes the reduce function, and stores the result in HDFS. The function that performs the Map function is called mapper (), and the function that performs the Reduce function is called reducer (). A single data node can run any number of mapper or reducers simultaneously.

FIG. 24 exemplarily shows Word Count for the process of applying the usefulness determination component of the association rule of the present invention.

In FIG. 24, it is assumed that three lines as inputs exist on the HDFS, and each row exists in one data node or is divided into multiple data nodes. The data node storage method may vary substantially depending on the file size. Split allocates a certain amount of data from each row or all input data to a specific mapper of a specific data node. In the figure, a total of three mapper are executed, and each mapper receives one row of data as input. The mapper reads one row, separates the data with commas (,), and outputs each word as a Key and the number of occurrences of each word (= 1) as a Value. The output (Key, Value) values are arranged so that the value for the same key can be read from the same data node. The Reducer of each data node performs the desired function by reading the incoming Value value (repeat of 1 in this example) for a particular key (word in this example). In the case illustrated in the drawings, a simple summation is illustrated. For Beer, two values of '1,1' come in, and they are simply summed up and then the data (key, value) = (Beer, 2) is output to HDFS. When all reducers have finished running, when you look at HDFS, you can see the full result.

In the above description, the related rules and the usefulness determination component are used to provide the item recommendation information to the user terminal 100 connected to the item service apparatus 200 as an example. However, the present invention is not limited thereto. . That is, the item service apparatus 200 may generate item recommendation information based on a purchase history of user terminals not connected according to a designer or administrator policy, and use the generated item recommendation information to a specific communication method to user terminals that are not connected. For example, it can also be provided through e-mail or text. In this case, the item service apparatus 200 may obtain prior consent of the user terminals 100 to provide a corresponding email or text.

The item recommendation method according to the embodiment of the present invention described above may be implemented in the form of program instructions that can be executed by various computer means, and recorded in a computer-readable recording medium. In this case, the computer-readable recording medium may include program instructions, data files, data structures, and the like, alone or in combination. Meanwhile, the program instructions recorded on the recording medium may be those specially designed and configured for the present invention, or may be known and available to those skilled in computer software.

Computer-readable recording media include magnetic media such as hard disks, floppy disks and magnetic tape, optical media such as CD-ROMs and DVDs, and magnetic disks such as floppy disks. Magnetic-Optical Media, and hardware devices specifically configured to store and execute program instructions, such as ROM, RAM, flash memory, and the like. In addition, program instructions include not only machine code generated by a compiler, but also high-level language code that can be executed by a computer using an interpreter or the like. The hardware device described above may be configured to operate as one or more software modules to perform the operations of the present invention.

In the above description of the preferred embodiments through the specification and drawings for the item recommendation service according to an embodiment of the present invention, although specific terms have been used, it is only easy to explain the technical details of the present invention and help understanding of the invention. It is only used in a general sense for this purpose, the present invention is not limited to the above-described embodiment. That is, various embodiments based on the technical idea of the present invention are possible to those skilled in the art.

The present invention relates to an item recommendation system and method, and in providing an item service such as content, an application, a product, and the like through a communication network, the present invention provides a heterogeneous recommendation item that can induce interest to a user and have a more diverse experience. To help. As a result, it is possible to support consumption promotion and diversity of users through purchase of various types of items.

In addition, the present invention can assist users to reduce the time required to search for an item by intelligently providing items that are likely to be purchased. In addition, through intelligent item recommendation, an effective item purchase can be effectively supported to optimize the item purchase.

Claims

At least one user terminal having an item purchase history and receiving item recommendation information based on the item purchase history;

After grouping all the user terminals, the heterogeneity degree of the purchase type between the user terminals is calculated based on the purchase history of the user terminals belonging to a specific group, and based on this, the heterogeneous purchase belongs to the same group as the specific user terminal belonging to the specific group. An item service apparatus for extracting other user terminals having a behavior, calculating item recommendation information for the specific user terminal including a foreign recommendation item list based on a purchase history of the other user terminals, and providing the same to the specific user terminal;

Item recommendation system comprising a.
At least one user terminal having an item purchase history and receiving item recommendation information based on the item purchase history;

Generate and store association rules that define a combination between purchased items based on history of purchasing items of the user terminals, and generate item recommendation information related to items included in a purchase history of a specific user terminal from the association rule; An item service device provided to a specific user terminal;

Item recommendation system comprising a.
The method according to claim 1 or 2,

A distributed processing service apparatus configured to distribute and store data necessary for generating item recommendation information;

Item recommendation system characterized in that it further comprises.
A device communication unit supporting signal processing required for accessing at least one user terminal and purchasing an item;

Group all the user terminals into a certain number of groups, and calculate heterogeneity degree of purchase type between user terminals based on the purchase history of user terminals belonging to a specific group, and based on this, A device controller extracting other user terminals belonging to the same group having heterogeneous purchase behaviors, and generating item recommendation information including a list of heterogeneous recommended items based on the purchase history of the other user terminals and providing the same to the specific user terminal; Item service apparatus comprising a.
The apparatus of claim 4, wherein the device controller

And a heterogeneity degree for the purchase type between the user terminals is calculated based on a ratio of the number of intersections of the products purchased by the two users to the number of unions of the products purchased by the two users.
The apparatus of claim 5, wherein the device controller

The purchase history for each user is represented by a purchase history matrix in which rows and columns are defined by user information and product information, respectively, and the purchase history matrix is multiplied by the heterogeneity matrix between users representing the degree of heterogeneity between the calculated users. And calculating a recommendation item score for each product and generating the heterogeneous recommendation item list based on the recommendation item score for each product.
The method of claim 4, wherein

And a device storage unit for distributing and storing data necessary for generating the item recommendation information.
The method of claim 4, wherein

The device control unit

After dividing the entire user terminals into a predetermined number of groups, assigning a group ID to each group, controlling each group of user terminals to have a group ID, and grouping data of the user terminals for each group ID An item service apparatus.
The method of claim 8,

The device control unit

Compares the purchase history of each user terminal belonging to the same group with the purchase history of other user terminals belonging to the same group based on collaborative filtering to calculate heterogeneity scores in the order of low similarity;

Item service device, characterized in that for controlling to generate a heterogeneous recommendation item list including a predetermined number of items for each heterogeneity score.
A device communication unit supporting signal processing required for accessing at least one user terminal and purchasing an item;

An apparatus for generating predefined association rules based on history of purchasing items of user terminals, and generating item recommendation information corresponding to at least a part of purchase history of a specific user terminal from the association rule and providing the same to the specific user terminal. Control unit;

Item service apparatus comprising a.
The method of claim 10,

A device storage unit for storing the association rules and purchase history of the user terminals;

The item service apparatus further comprises.
The method of claim 11,

The device control unit

After checking the purchase history of the specific user terminal, searching for association rules corresponding to items included in the purchase history, selecting association rules to be used as a predetermined number of preliminary recommendation items based on a predetermined criterion, and then predefined Item service device, characterized in that for generating the item recommendation information items included in the schedule ranking.
The method of claim 12,

The schedule criteria

At least one of support, reliability,

The support rate is a ratio of association rules including two specific products among the total association rules, the reliability is a ratio of association rules including the specific product and other specific products among all association rules including a specific product, The improvement degree is an item service apparatus, characterized in that the ratio of the association rules generated based on the history of the purchase of the specific product and the other specific product among the association rules including the specific product.
The method of claim 13,

The device control unit

And calculating association rules of at least one of the schedule criteria being equal to or greater than a predetermined ratio, and calculating a predetermined number of association rules of the association rules that are equal to or greater than the predetermined ratio as the preliminary recommendation item.
The item service apparatus grouping all user terminals into a certain number of groups;

Calculating, by the item service apparatus, a heterogeneity degree for a purchase type between user terminals based on a purchase history of user terminals belonging to a specific group;

The item service apparatus extracts other user terminals belonging to the same group as a specific user terminal and has heterogeneous purchase behaviors based on the heterogeneity, and recommends a heterogeneity for the specific user terminal based on the purchase history of the other user terminals. Generating and providing a list of items;

Item recommendation method comprising a.
The method of claim 15,

Providing the heterogeneous recommendation item list

After generating heterogeneous recommendation item lists for all user terminals based on heterogeneity calculation for all user terminals, searching for and providing heterogeneous recommendation item lists for the specific terminal when the specific terminal is accessed. Item recommendation method characterized in that step.
The method of claim 15,

Providing the heterogeneous recommendation item list

Excluding an item included in a purchase history of a specific user terminal provided with the heterogeneous recommendation item among items collected based on the calculated degree of heterogeneity;

Item recommendation method comprising a.
Generating and distributing, by the item service apparatus, association rules having a predetermined predetermined rule length based on the item purchase history of the user terminals;

Calculating, by the item service apparatus, a predetermined number of preliminary recommended items based on a predetermined number of association rules corresponding to items included in a purchase history of a specific user terminal;

Configuring, by the item service apparatus, items within a predetermined predetermined rank among the preliminary recommended items calculated as item recommendation information;

Providing, by the item service apparatus, the item recommendation information to the user terminal;

Item recommendation method comprising a.
The method of claim 18,

The schedule criteria

Includes at least one of support, reliability, and improvement,

The support is the ratio of association rules that include two specific products of the overall association rules,

The reliability is a ratio of association rules including the specific product and other specific products among all the association rules including the specific product,

The degree of improvement is an item recommendation method, characterized in that the ratio of the association rules generated based on the history of the purchase of the particular product and the other specific product among the association rules including the specific product.
The method of claim 18,

Generating the item recommendation information

Configuring the item service device to include the same number of items as the number of purchase history of the user terminal in the process of configuring the item recommendation information;

Configuring the item service device to include a predetermined number of predetermined items in the process of configuring the item recommendation information;

And at least one step of the item recommendation method.