WO2020175732A1

WO2020175732A1 - Commercial product recommendation method and device therefor

Info

Publication number: WO2020175732A1
Application number: PCT/KR2019/003839
Authority: WO
Inventors: 강상우; 서철우
Original assignee: 주식회사 틸투원
Priority date: 2019-02-26
Filing date: 2019-04-02
Publication date: 2020-09-03
Also published as: KR20200104013A

Abstract

Provided are a commercial product recommendation method and a device therefor. The method, which is a commercial product recommendation method for a commercial product recommendation device operated by at least one processor, comprises the steps of: generating a user matrix in which multiple attributes and multiple attribute values corresponding to the multiple attributes are two-dimensionally arranged; providing a user with multiple images tagged with an attribute value selected from the multiple attribute values for each of the multiple attributes; acquiring preference level responses, each selected by the user from a plurality of preference level responses set for each of the multiple images; generating a user profile where matrix values in the user matrix are set to the preference level responses; calculating respective preference ratings on a plurality of commercial products by selecting matrix values corresponding to the intersections of one or more attributes with one or more attribute values from the user profile on the basis of information about the plurality of commercial products and finding the sum of the selected matrix values; and determining a recommended commercial product from the information about the plurality of commercial products on the basis of the preference ratings.

Description

Product recommendation method and device

The present invention relates to a product recommendation method and apparatus thereof, and to a technology for recommending products predicted to be preferred by a user in consideration of a user's taste.

As the number of Internet shopping malls increases with the development of the Internet, the number of products that customers can purchase through shopping malls is increasing in various ways.

Customers select products by referring to evaluations or recommendations on products that others have experienced or possessed. For example, when a product is searched on a search service, evaluation or recommendations for various products are presented as a result of the search, and a user often selects a product suitable for him based on such information. Accordingly, the customer spends a considerable amount of time searching for the desired product.

In addition, by exploiting such a situation, advertisement information and the like are scattered, and the reliability of evaluation or recommendation for a product is deteriorating. In addition, since each person has a different personality, there is a problem in that the taste of the product that they prefer is different accordingly, and when selecting a product according to the evaluation or recommendation of the product, a product that is not actually suitable for one's taste is selected. For this reason, many recommendation systems have been developed that can recommend products suitable to customers.

The recommendation system, as one of the personalization technologies, recommends content or information that the user may be interested in to the user. Among various recommendation technologies, a recommendation method based on a collaborative filtering algorithm is widely used.

The recommendation method using the collaborative filtering algorithm is a personalization technique that recommends related products to customers by using the preferences of other customers similar to the patterns that the customers prefer.

However, the recommendation method using such a collaborative filtering algorithm improves the accuracy of data only when large-scale customer behavioral information of existing users is accumulated, so data collection and analysis are expensive. In addition, when the amount of data is large, the time required to calculate the similarity increases exponentially, which can cause problems in service scalability.

In addition, although there is little interest, there is a disadvantage of low reliability due to the occurrence of a long tail economy that does not reflect information on products used by many. In particular, the fashion industry is a multi-kind, small-scale product industry in which the consumption of trends is very fast and individual tastes are very diverse. Therefore, the effectiveness compared to the investment for obtaining enormous amount of information is low.

In addition, when the collaborative recommendation algorithm is used, there is a cold start problem that is difficult to cope with new customers, new products, and new situations that have not been accumulated in advance.

In addition, the product database of conventional Korean fashion companies was constructed by a number of MDs under their own supervision. Therefore, there is a problem that the accuracy of recommendation is lowered because the consistency of the product database is inferior regardless of the type of recommendation algorithm.

The problem to be solved by the present invention is a method for improving the purchase conversion rate by recommending the user's preferred product through a content-based filtering method by expressing the user's preference for a product in a matrix and calculating the correlation between item-item. To provide a device.

According to one feature of the present invention, a product recommendation method is a product recommendation method of a product recommendation device operated by at least one processor, wherein a plurality of attributes and a plurality of attribute values respectively corresponding to the plurality of attributes are two-dimensional. Generating a user matrix arranged in a manner, presenting a plurality of images tagged with one attribute value selected from among the plurality of attribute values for each of the plurality of attributes to the user, among a plurality of preference responses set for each of the plurality of images Acquiring preference responses selected by a user, generating a user profile in which the preference responses are set as matrix values of the user matrix, summing at least one matrix value selected from the user matrix based on a plurality of product information And calculating a preference score for each product, and recommending at least one product from among the plurality of product information to a user based on the preference score.

The plurality of product information includes at least one attribute set for each product and at least one attribute value corresponding to the at least one attribute,

The at least one attribute is selected from the plurality of attributes,

The at least one attribute value is selected from the plurality of attribute values,

The calculating may include checking the at least one attribute and at least one matrix value corresponding to the at least one attribute value from the user matrix, and adding the checked at least one matrix value to a preference score for each product. It may include the step of calculating.

The generating of the user profile may include, in the user matrix, a preference selected by the user for the first image by selecting a first attribute tagged to a first image and a matrix value corresponding to a first attribute value set in the first attribute. Setting a response value as a response value, and in the user matrix, the first attribute tagged to the second image after the first image and the matrix value corresponding to the first attribute value are selected by the user for the second image And setting a preference response value, wherein the matrix value may include a preference response value for the second image accumulated in the preference response value for the first image.

The recommending may include allocating different weights to the plurality of products based on the user's body data, applying the weights to each preference score calculated for the plurality of products, and the weight being Based on the applied preference score, determining at least one product from among the plurality of product information as a recommended product.

The plurality of attributes may be set as variables that influence a user's fashion taste.

The plurality of attribute values may include a plurality of scores set at predetermined intervals within a specific score range.

The plurality of images may be default images previously determined in consideration of a plurality of attributes and a plurality of attribute values that determine a user's fashion taste.

The preference responses may include scores set for each user's preference for the plurality of images.

According to another feature of the present invention, the product recommendation device includes a communication device that transmits and receives data through a network, a memory that stores a product recommendation program, and at least one processor that executes the product recommendation program, and the product recommendation program Is, by presenting a plurality of images tagged with a plurality of attributes that determine the user's taste to the user to obtain user preference responses for the plurality of images, and for each product based on the preference responses and the plurality of attributes. It calculates a user's preference score, and includes instructions for determining at least one recommended product from among a plurality of products based on the preference score.

The product recommendation program generates a user matrix in which the plurality of attributes and a plurality of attribute values respectively corresponding to the plurality of attributes are arranged in two dimensions, and a matrix value is set as the preference responses, and the plurality of attributes and A plurality of products in which a plurality of attribute values corresponding to the plurality of attributes are arranged in two dimensions, and a point where the plurality of attributes and a plurality of attribute values corresponding to the plurality of attributes intersect is selectively activated and deactivated for each product The matrix may include instructions for comparing the user matrix and the plurality of product matrices to calculate a preference score for each product based on matrix values of the user matrix corresponding to activated points for each product matrix.

The product recommendation program calculates a user preference score for the first product by summing at least one attribute activated in the first product matrix and at least one matrix value of the user matrix corresponding to the at least one attribute value. May contain instructions.

The product recommendation program selects at least one attribute to be assigned a weight based on user information from among the at least one activated attribute, and applies the weight to a matrix value corresponding to the selected attribute to calculate the user preference score. It may contain instructions to calculate.

The product recommendation program sets a plurality of attributes tagged to a first image and matrix values of the user matrix corresponding to a plurality of attribute values as a preference response to the first image, and then to the user after the first image. And commands for setting matrix values of the user matrix corresponding to the plurality of attributes and the plurality of attribute values tagged to the presented second image as a preference response for the second image, wherein the matrix values are the first The preference response for the second image may be added to the preference response for the image.

The preference response is a first value indicating a user's negative reaction to the image, a second value indicating a user's positive reaction to the image, and a second value indicating a moderate reaction to the image that is neither positive nor negative. It may include a value according to the user's selection among 3 values.

The plurality of images may include a plurality of clothing product images in which a plurality of attributes that influence the user's fashion taste are tagged differently.

According to an embodiment of the present invention, it is possible to minimize expensive user data and secure a relatively high conversion rate at low cost through a scoring system reflecting the quality characteristics of products.

In addition, it is possible to recommend a customer based on similarity between products, and in particular, a high probability recommendation is possible even for a new customer who does not have a purchase history, thereby solving the conventional cold start problem.

In addition, compared to the collaborative recommendation algorithm, it is possible to obtain data that is meaningful for decision-making such as trend analysis and production/distribution at a relatively low cost.

1 is a block diagram showing the configuration of a product recommendation device according to an embodiment of the present invention.

2 is an example of attribute values according to an embodiment of the present invention.

3 is an example of a default image according to an embodiment of the present invention.

4 is an example of default images tagged with attribute values for each attribute according to an embodiment of the present invention.

5 is a flowchart showing a product recommendation method according to an embodiment of the present invention.

6 is an example of creating a user profile according to an embodiment of the present invention.

7 is an example of creating a user profile according to another embodiment of the present invention.

8 is an example of creating a user profile according to another embodiment of the present invention.

9 is an exemplary diagram for calculating a preference score according to an embodiment of the present invention.

10 is an exemplary diagram for calculating a preference score according to another embodiment of the present invention.

11 is an exemplary diagram of a method for calculating a preference score according to an embodiment of the present invention.

12 is an exemplary view of a product recommendation screen according to an embodiment of the present invention.

13 is a block diagram showing a hardware configuration of a server device to which an embodiment of the present invention can be applied.

14 is a block diagram showing a hardware configuration of a terminal device to which an embodiment of the present invention can be applied.

Hereinafter, embodiments of the present invention will be described in detail with reference to the accompanying drawings so that those of ordinary skill in the art can easily implement the embodiments of the present invention. However, the present invention may be implemented in various different forms and is not limited to the embodiments described herein. In the drawings, parts irrelevant to the description are omitted in order to clearly describe the present invention, and similar reference numerals are assigned to similar parts throughout the specification.

Throughout the specification, when a part "includes" a certain component, it means that other components may be further included rather than excluding other components unless specifically stated to the contrary.

In addition, terms such as "... unit", "... group", and "... module" described in the specification mean a unit that processes at least one function or operation, which can be implemented by hardware or software or a combination of hardware and software. I can.

Now, a product recommendation method and apparatus thereof according to an embodiment of the present invention will be described with reference to the drawings.

1 is a block diagram showing the configuration of a product recommendation device according to an embodiment of the present invention, FIG. 2 is an example of attribute values according to an embodiment of the present invention, and FIG. 3 is a default image according to an embodiment of the present invention. 4 is an example of default images tagged with attribute values for each attribute according to an embodiment of the present invention.

The product recommendation device 100 is a device that recommends customized products that will lead to purchase to target customers. Referring to FIG. 1, a default image database 101, a user profile database 103, and a product database ( 105), a user profile generation unit 107, a product recommendation unit 109, and a default image setting unit 111. Here, the product recommendation device 100 may be implemented as a server device. Alternatively, the product recommendation device 100 may be downloaded to the terminal and executed.

The default image database 101 stores a plurality of product images. Since such product images are set in advance, they are referred to as default images. In the default images, one attribute value selected from a plurality of attribute values for each of a plurality of attributes is tagged. Default images are images selected to identify the user's taste, and different attributes are tagged. According to one example, the number of default images may be 12, and this number may be the minimum number of images required to determine the user's taste.

Here, attributes are qualitative characteristics of a product image, and qualitative characteristics are defined as variables that influence fashion tastes or fashion tendencies. Qualitative characteristics are item type, style, time (Time), place (place), situation (Occasion) by TPO recommendations, color, pattern, body type, season, weather, fashion boldness (boldness), keywords, mood It is determined and scored based on etc. These scored values are attribute values.

Qualitative characteristics can be changed according to new fashion trends, new fashion styles, and new products. Qualitative characteristics not only represent general fashion tendencies, but may also be related to products to be recommended.

In one default image, scores are assigned for each of the qualitative characteristics of nine predefined fashions, and an example showing this is shown in Table 1.

속성property	속성값(-6, -4, -2, 0, +2, +4, +6 중에서 택일)Property value (choose from -6, -4, -2, 0, +2, +4, +6)
AA	00
BB	-4-4
CC	+6+6
DD	+2+2
EE	00
FF	00
GG	00
HH	00
II	-2-2

The attribute values assigned for each of the nine attributes shown in Table 1 are set differently for each default image.

The attribute value may include a plurality of scores set at regular intervals within a specific score range. That is, the attribute values may be'-6, -4, -2, 0, +2, +4, +6'.

Referring to FIG. 2, when the qualitative characteristic is Elegance, it is defined to be closer to Elegance as it goes from '0' to'+6' and closer to active as it goes from '0' to'-6'.

Referring to FIG. 3, a product P201 is displayed as a default image P200, and three different types of preference responses P203 are set. The preference responses P203 may be YES (P205), SO SO (P207), and NO (P209). In addition, scores are assigned to each of the preference responses P205, P207, and P209 as shown in Table 2 below.

구분division	점수score
YESYES	+1+1
SO SOSO SO	00
NONO	-1-One

Referring to Table 2, YES indicates the user's positive reaction to the default image, NO indicates the user's negative reaction to the default image, and SO SO indicates a moderate reaction to the default image, not positive or negative. Show.

The default images may include a plurality of clothing product images tagged differently from a plurality of attributes that influence a user's fashion taste.

Referring to FIG. 4, in image 1, each attribute value (0, -4, +6, …, +2) is set for each of a plurality of attributes (A, B, C, ..., I). In image 2, each attribute value (-4, -4, +2, ..., 0) is set for each of a plurality of attributes (A, B, C, ..., I). In the image N, each attribute value (0, -4, +6, ..., +4) is set for each of a plurality of attributes (A, B, C, ..., I).

In this case, the user may select a different preference response for each image. For example,'YES' may be selected for image 1,'NO' may be selected for image 2, and'SO SO' may be selected for image N.

The user profile database 103 stores a user profile created by the user profile generating unit 107. The user profile is a matrix in which a plurality of attributes and a plurality of attribute values respectively corresponding to the plurality of attributes are arranged in two dimensions, and a score corresponding to the preference response selected by the user is set as the matrix value. That is, the horizontal rows of the user profile are A, B, C,… It is set to I, and the column is set to'-6, -4, -2, 0, +2, +4, +6'. In addition, as for the matrix value at the intersection of'A' and'-6', the score of the preference response selected by the user is set for images in which'-6' is set for attribute'A' among default images. This will be described in more detail with reference to FIGS. 6 to 8.

In addition, the user profile database 103 includes user information input by a user when registering as a member. The user information may include user body data such as user characteristics related to the user's taste, body type data, and skin tone data. These user characteristics and user body data may be used to allocate weights for each attribute when calculating a user preference score. User body data may be collected by non-face-to-face body measurement technology.

The product database 105 includes a product matrix for each product. In the product matrix, a plurality of attribute values corresponding to each attribute are arranged in two dimensions for each of a plurality of attributes, and the matrix values are selectively activated and deactivated. That is, the horizontal rows of the product matrix are A, B, C, ... It is set to I, and the column is set to'-6, -4, -2, 0, +2, +4, +6'. In addition, the matrix value at the intersection of'A' and'-6' is activated, and the matrix value at the intersection of'A' and'-4' is deactivated.

The user profile generating unit 107 obtains user preference responses for the plurality of images by presenting a plurality of images tagged with a plurality of attributes that determine the user's taste to the user. Then, a user matrix is created, and a user profile in which the matrix values of the user matrix are set as a response to each preference for the images is created.

The product recommendation unit 109 calculates a user's preference score for each product by comparing the user profile and the product matrices, respectively. The product recommendation unit 109 calculates a user preference score for a product by summing at least one attribute activated in the product matrix and at least one matrix value of a user matrix corresponding to the at least one attribute value.

The product recommendation unit 109 determines a recommended product based on the preference score calculated for each product. In this case, the product recommendation unit 109 may allocate different weights to the plurality of products based on the user's body data, and apply the weights assigned to each preference score calculated for the plurality of products, respectively. The product recommendation unit 109 may determine a recommended product based on a preference score to which each weight is applied.

The product recommendation unit 109 selects at least one attribute to be assigned a weight based on user information from among at least one attribute activated in the product matrix, and applies a weight to a matrix value corresponding to the selected attribute to calculate a user preference score. Can be calculated. For example, when the A attribute is activated in the product matrix, the preference score can be calculated by applying the weight a to the A attribute.

The product recommendation unit 109 may determine at least one product as a recommended product in the order of the highest preference score. Alternatively, at least one product having a preference score equal to or greater than a predetermined standard may be determined as a recommended product. Alternatively, products may be listed in order of preference score and provided.

The default image setting unit 111 may set a plurality of default images according to a user input. In addition, the default image setting unit 111 may set a product matrix for each product.

According to an embodiment of the present invention, more customer information may be acquired through acquiring a user preference response compared to a simple search. And based on such customer information, a user profile of a two dimensional profile can be configured instead of the existing linear profile information.

In addition, user data in the fashion industry has the characteristic that individual deviations and time series changes according to trends appear dramatically. In an embodiment of the present invention, based on the user information input in the member registration process and the preference responses selected by the user for default images, user preferences corresponding to nine qualitative characteristics for each product are quantified to correlate user profiles and products. Calculate. Therefore, it is possible to design to flexibly respond to new trends, new styles, and products by using qualitative characteristics composed of variables that consider the characteristics of fashion products.

Referring to FIG. 5, the user profile generation unit 107 presents a plurality of default images set by the default image setting unit 111 and stored in the image database 101 to the user (S101). The user profile generation unit 107 may provide step S101 upon initial access after registering as a member.

The user profile generator 107 receives user preference responses for each of a plurality of default images (S103).

The user profile creation unit 107 is a user who sets a preference response value (one of 0, -1, 1) as a qualitative characteristic value tagged to each default image for each default image, that is, a matrix value of attribute values tagged for each attribute. The matrix is created as a user profile (S105).

The product recommendation unit 109 calculates a user preference score for each product by comparing the product matrix and the user profile in which the matrix value of the attribute value matched for each attribute is active (On) or inactive (Off) (S107) . And, based on the user preference score, at least one product predicted to be preferred by the user is filtered and recommended to the user (S109).

6, in image 1, A=0, B=-4, C=+6, D=0, E=+4, F=-6, G=-2, H=0, I=+2 It is tagged as, and the user selects'YES'.

The user profile generation unit 107 sets the matrix value of the user matrix P1 corresponding to A=0 to'+1' (①).

The user profile generation unit 107 sets the matrix value of the user matrix P1 corresponding to B=-4 to'+1' (2).

The user profile generation unit 107 sets the matrix value of the user matrix P1 corresponding to C=+6 to'+1' (③).

The user profile generation unit 107 sets the matrix value of the user matrix P1 corresponding to D=0 to'+1' (4).

The user profile generation unit 107 sets the matrix value of the user matrix P1 corresponding to E=+4 to'+1' (5).

The user profile generation unit 107 sets the matrix value of the user matrix P1 corresponding to F=-6 to'+1' (6).

The user profile generation unit 107 sets the matrix value of the user matrix P1 corresponding to G=-2 to'+1' (7).

The user profile generation unit 107 sets the matrix value of the user matrix P1 corresponding to H=0 to'+1' (8).

The user profile generation unit 107 sets the matrix value of the user matrix P1 corresponding to I=+2 to'+1' (⑨).

Referring to FIG. 7, in image 2, A=-4, B=-4, C=+2, D=-2, E=+4, F=+6, G=-2, H=0, I= It is tagged as 0, and the user has selected'NO'.

The user profile generation unit 107 sets the matrix value of the user matrix P3 corresponding to A=-4 to'-1' (①).

The user profile generating unit 107 has already set the matrix value of the user matrix P3 corresponding to B=-4 to'+1', so it adds'-1' to the preset'+1' to '0' Set to'(②).

The user profile generation unit 107 sets the matrix value of the user matrix P3 corresponding to C=+2 to'-1' (③).

The user profile generation unit 107 sets the matrix value of the user matrix P3 corresponding to D=-2 to'-1' (4).

The user profile generation unit 107 has already set the matrix value of the user matrix P3 corresponding to E=+4 to'+1', and thus adds'-1' to the preset'+1' '(⑤).

The user profile generation unit 107 sets the matrix value of the user matrix P3 corresponding to F=+6 to'-1' (6).

The user profile generating unit 107 has already set the matrix value of the user matrix P3 corresponding to G=-2 to'+1', so it adds'-1' to the preset'+1' to '0' '(⑦).

The user profile generating unit 107 has already set the matrix value of the user matrix P3 corresponding to H=0 to'+1', and thus '0' by adding'-1' to the preset'+1'. Set to (⑧).

The user profile generation unit 107 sets the matrix value of the user matrix P3 corresponding to I = 0 to'-1' (⑨).

8, in image 3, A=0, B=-4, C=+6, D=-6, E=+2, F=-2, G=0, H=-4, I=+ It is tagged as 4, and the user selects'SO SO'.

The user profile generation unit 107 sets the matrix value of the user matrix P5 corresponding to A=0 to'+1' (①). At this time, if the preset matrix value is'+1' and '0' corresponding to'SO SO' is added up, it is set to'+1'.

The user profile generation unit 107 sets the matrix value of the user matrix P5 corresponding to B=-4 to “0” corresponding to “SO SO” (②).

The user profile generation unit 107 sets the matrix value of the user matrix P5 corresponding to C=+6 to'+1' (③). At this time, if the preset matrix value is'+1' and '0' corresponding to'SO SO' is added up, it is set to'+1'.

The user profile generation unit 107 sets the matrix value of the user matrix P5 corresponding to D=-6 to '0' corresponding to'SO SO' (④).

The user profile generation unit 107 sets the matrix value of the user matrix P5 corresponding to E=+2 to 0 corresponding to'SO SO' (⑤).

The user profile generation unit 107 sets the matrix value of the user matrix P5 corresponding to F=-2 to '0' corresponding to'SO SO' (⑥).

The user profile generation unit 107 sets the matrix value of the user matrix P5 corresponding to G=0 to “0” corresponding to “SO SO” (7).

The user profile generation unit 107 sets the matrix value of the user matrix P5 corresponding to H=-4 to “0” corresponding to “SO SO” (8).

The user profile generator 107 sets the matrix value of the user matrix P5 corresponding to I = +4 to '0' corresponding to'SO SO' (⑨).

Referring to FIG. 9, the product recommendation unit 109 compares the user profile P7 generated through the processes of FIGS. 6, 7 and 8 with the matrix P9 of the product 1.

The matrix P9 of product 1 is the same two-dimensional matrix as the user profile, but the matrix values are selectively activated and deactivated. At this time, A=-4, B=0, C=-6, D=-6, E=+2, F=+4, G=-2, H=0, I=+4 are activated and all others It is deactivated.

The product recommendation unit 109 generates a preference score matrix P11 by comparing the user profile P7 with the matrix P9 of the product 1. At this time, the preference score matrix P11 is the attribute values matched for each activated attribute, that is, A=-4, B=0, C=-6, D=-6, E=+2, F=+4, G Each matrix value of =-2, H=0, and I=+4 is set as the corresponding matrix value of the user profile P7. That is, user profiles corresponding to each of A=-4, B=0, C=-6, D=-6, E=+2, F=+4, G=-2, H=0, I=+4 Matrix values of (P7) are'+1','-2','+1','+2','+6', '0','+2','+3','+1' Each is set to. The product recommendation unit 109 adds up each set matrix value, that is, calculates +1-2+1+2+6+0+2+3+1=14 as the user's preference score for product 1.

Referring to FIG. 10, the product recommendation unit 109 compares the user profile P7 generated through the processes of FIGS. 6, 7 and 8 with the matrix P13 of the product N.

The matrix P13 of the product N is the same two-dimensional matrix as the user profile, but the matrix values are selectively activated and deactivated. At this time, A=-2, B=+6, C=+2, D=-6, E=-6, F=-4, G=+4, H=0, I=-2 are activated and the rest All are disabled.

The product recommendation unit 109 generates a preference score matrix P15 by comparing the user profile P7 with the matrix P13 of the product N. At this time, the preference score matrix P15 is the attribute values matched for each activated attribute, that is, A=-2, B=+6, C=+2, D=-6, E=-6, F=-4, The matrix values of G=+4, H=0, and I=-2 are set as the corresponding matrix values of the user profile P7. That is, users corresponding to each of A=-2, B=+6, C=+2, D=-6, E=-6, F=-4, G=+4, H=0, and I=-2 The matrix values of the profile P7 are '0','+1','+4','+2', '0','+1','+5','+3','+1'. Each is set to. The product recommendation unit 109 adds up the matrix values set in this way. The summed value, that is, 0+1+4+2+0+1+5+3+1=17, is calculated as the user's preference score for product N.

10 and 11, the product recommendation unit 109 may determine a product N having a relatively higher preference score as the user's recommended product. Alternatively, a list of recommended products may be provided to the user in the order of product N and product 1.

Referring to FIG. 11, the product recommendation unit 109 determines a weight for each attribute based on the user's characteristic data or the user's body data (S201). Here, the weight may be predetermined. For example, when the user is skinny, a weight of a=2 can be applied to the attribute A preferred by the skinny body. Accordingly, the product recommendation unit 109 calculates a preference score obtained by applying a weight to the attribute A (S203). That is, each of the matrix values of A=-4, B=0, C=-6, D=-6, E=+2, F=+4, G=-2, H=0, I=+4 is' When calculating the preference score as +1','-2','+1','+2','+6', '0','+2','+3','+1', ( 1×2)-2+1+2+6+0+2+3+1=15.

Referring to FIG. 12, the product recommendation unit 109 may provide a product recommendation screen in which a plurality of product images P17 are listed. The plurality of product images P17 are arranged in the order of the highest user preference scores for each product.

Meanwhile, the product recommendation device 100 described above may be implemented as a web server, a smart phone application server, or the like.

13 is a block diagram showing a hardware configuration of a server device to which an embodiment of the present invention can be applied, and shows an embodiment configuration of the product recommendation device 100 described in FIGS. 1 to 12.

Referring to FIG. 13, the server device 200 is connected to the user terminal 400 through a communication network 300 and provides product recommendations described in FIGS. 1 to 12.

The communication network 300 may be formed of a wired communication network, a mobile communication network, a wireless network, and a combination thereof. For example, the wired communication network includes a wired network based on Internet Protocol (IP) such as Voice over Internet Protocol (VoIP). Mobile communication networks include 5G networks, long term evolution (LTE) networks, and wideband code division multiple access (WCDMA) networks. Wireless networks include various types of wireless networks such as Wi-Fi networks. However, the communication network 300 is not limited to a specific technology, since it may include all networks developed in the future as well as the networks already established as described above.

According to an embodiment, the user terminal 400 may access the server device 200 by driving a PC version or a mobile version of a web browser. According to another embodiment, the user terminal 400 may access the server device 200 by executing a smartphone application, an app, a mobile application, a web application, etc. installed in the user terminal 400.

The server device 200 includes a communication device 201, a memory 203, a storage device 205, and at least one processor 207. The communication device 201 is connected to the processor 207 and transmits and/or receives data through the communication network 300. The memory 203 is connected to the processor 207 and stores a program including instructions for executing the configuration and/or method according to the embodiments described with reference to FIGS. 1 to 12. The processor 207 executes a program stored in the memory 203. The program includes instructions for implementing the operation method of the present invention described with reference to FIGS. 1 to 11, and the processor 207 executes the present invention by combining it with hardware such as a memory 203.

At this time, the user terminal 400 displays a default image received from the server device 200 on the screen while being connected to the server device 200, and transmits a preference response input by the user to the server device 200, A series of user interface operations for receiving recommended product information from the server device 200 and outputting it to a screen are implemented.

14 is a block diagram showing a hardware configuration of a terminal to which an embodiment of the present invention can be applied, and shows an embodiment configuration of the product recommendation apparatus 100 described in FIGS. 1 to 12.

Referring to FIG. 14, the terminal 500 is connected to the server 600 through a communication network 300 and provides product recommendations described in FIGS. 1 to 12.

At this time, the server 600 includes a default image database 101, a user profile database 103, a product database 105, and a default image setting unit 111 of FIG. 1. And the terminal 500 includes a user profile generating unit 107 and a product recommendation unit 109, and in a state connected to the server 600, the configuration and/or method according to the embodiments described in FIGS. 1 to 12 Run.

Referring to FIG. 14, the terminal 500 includes a communication device 501, a memory device 503, a display 505, an input device 507, and at least one processor 509. The communication device 501 is connected to the processor 509 and transmits and/or receives data through the communication network 300. The memory device 503 is connected to the processor 509 and stores a program including instructions for executing the configuration and/or method according to the embodiments described in FIGS. 1 to 12. The display 505 is connected to the processor 509 and outputs data according to the embodiments described in FIGS. 1 to 12 on a screen. The input device 507 is connected to the processor 509 and is a means for a user input operation according to the embodiments described with reference to FIGS. 1 to 12. The processor 509 executes a program stored in the memory device 503. The program includes instructions for implementing the operation method of the present invention described with reference to FIGS. 1 to 12, and the processor 509 executes the present invention in combination with hardware such as a memory 503.

The embodiments of the present invention described above are not implemented only through an apparatus and a method, but may be implemented through a program that realizes a function corresponding to the configuration of the embodiment of the present invention or a recording medium on which the program is recorded.

Although the embodiments of the present invention have been described in detail above, the scope of the present invention is not limited thereto, and various modifications and improvements by those skilled in the art using the basic concept of the present invention defined in the following claims are also provided. It belongs to the scope of rights.

Claims

A product recommendation method of a product recommendation device operated by at least one processor,

Generating a user matrix in which a plurality of attributes and a plurality of attribute values respectively corresponding to the plurality of attributes are arranged in two dimensions,

Presenting a plurality of images tagged with one attribute value selected from the plurality of attribute values for each of the plurality of attributes to a user,

Obtaining preference responses selected by a user from among a plurality of preference responses set for each of the plurality of images,

Generating a user profile in which the preference responses are set as a matrix value of the user matrix,

Calculating a preference score for each product by summing at least one matrix value selected from the user matrix based on a plurality of product information, and

Recommending at least one product from among the plurality of product information to a user based on the preference score

Including, product recommendation method.
In claim 1,

The plurality of product information,

At least one attribute set for each product and at least one attribute value corresponding to the at least one attribute,

The at least one attribute is selected from the plurality of attributes,

The at least one attribute value is selected from the plurality of attribute values,

The calculating step,

Checking the at least one attribute and at least one matrix value corresponding to the at least one attribute value from the user matrix, and

Calculating a preference score for each product by summing the checked at least one matrix value

Including, product recommendation method.
In paragraph 2,

The step of creating the user profile,

In the user matrix, setting a first attribute tagged to a first image and a matrix value corresponding to a first attribute value set in the first attribute as a preference response value selected by the user for the first image, and

In the user matrix, setting a matrix value corresponding to the first attribute and the first attribute value tagged to a second image after the first image as a preference response value selected by the user for the second image Including,

The matrix value is

A product recommendation method, wherein a preference response value for the second image is accumulated in a preference response value for the first image.
In paragraph 2,

The recommending step,

Allocating different weights to the plurality of products based on the user's body data,

Applying the weight to each preference score calculated for the plurality of products, and

Determining at least one product from among the plurality of product information as a recommended product based on the weighted preference score

Including, product recommendation method.
In claim 1,

The plurality of attributes,

A product recommendation method that is set by variables that influence a user's fashion taste.
In claim 1,

The plurality of attribute values,

Product recommendation method including a plurality of points set at regular intervals in a specific score range.
In claim 1,

The plurality of images,

A product recommendation method, which is default images previously determined in consideration of a plurality of attributes and a plurality of attribute values that determine a user's fashion taste.
In claim 1,

The preference responses are:

A product recommendation method including scores set for each user's preference for the plurality of images.
A communication device that transmits and receives data through a network,

A memory for storing product recommendation programs, and

At least one processor for executing the product recommendation program,

The product recommendation program above,

A plurality of images tagged with a plurality of attributes that determine the user's taste are presented to the user to obtain user's preference responses for the plurality of images, and based on the preference responses and the plurality of attributes, Calculate the preference score,

A product recommendation device comprising instructions for determining at least one recommended product from among a plurality of products based on the preference score.
In claim 9,

The product recommendation program above,

Generating a user matrix in which the plurality of attributes and a plurality of attribute values respectively corresponding to the plurality of attributes are arranged in two dimensions, and the matrix value is set as the preference responses,

The plurality of attributes and a plurality of attribute values corresponding to the plurality of attributes are arranged in two dimensions, and a point where the plurality of attributes and a plurality of attribute values corresponding to the plurality of attributes intersect is selectively activated for each product, and Create a matrix of multiple products that are disabled,

And instructions for calculating a preference score for each product based on matrix values of a user matrix corresponding to points activated for each product matrix by comparing the user matrix with the plurality of product matrices.
In claim 10,

The product recommendation program above,

A product comprising instructions for calculating a user preference score for the first product by summing at least one attribute activated in the first product matrix and at least one matrix value of the user matrix corresponding to the at least one attribute value Recommended device.
In clause 11,

The product recommendation program above,

Including instructions for calculating the user preference score by selecting at least one attribute to be assigned a weight based on user information from among the at least one activated attribute, and applying the weight to a matrix value corresponding to the selected attribute , Product recommendation device.
In claim 10,

The product recommendation program above,

A plurality of attributes tagged to the first image and matrix values of the user matrix corresponding to the plurality of attribute values are set as a preference response for the first image,

Including commands for setting matrix values of the user matrix corresponding to a plurality of attributes and a plurality of attribute values tagged to a second image presented to a user after the first image as a preference response for the second image,

The matrix values are,

The product recommendation device, wherein the preference response for the second image is added to the preference response for the first image.
In claim 13,

The preference response is,

A first value indicating the negative reaction of the user to the image,

A second value indicating a positive reaction of the user to the image, and

A product recommendation device comprising a value according to a user's selection from among third values representing a moderate reaction of the user to the image.
In claim 9,

The plurality of images,

A product recommendation device comprising a plurality of clothing product images in which a plurality of attributes that influence a user's fashion taste are differently tagged.