KR20190100788A - Method for recommending information on wetsites customized to clients' q and a by using personalization algorithm and server using the same - Google Patents

Abstract

The present invention relates to a method of recommending a customer question and answer-customized site by using a personalized algorithm, and a personalized algorithm server. In other words, the present invention relates to an optimization site providing method using a personalized algorithm. According to the present invention, the optimization site providing method comprises: a step of allowing the personalized algorithm server to extract one or more tags from a purchase history of a first customer group and generate tag values to form a customer purchase tag table; a step of allowing the personalized algorithm server to apply a CDF function to the tag values to be matched within a predetermined range (projecting to a predetermined range); a step of allowing the personalized algorithm server to form a site tag table, which is information by purchase site, from the customer purchase tag table which is information by customer; a step of allowing the personalized algorithm server to form a customer question tag table to which question tag values are given based on answers for predetermined questions obtained from a second customer group; a step of allowing the personalized algorithm server to calculate similarity between the customer and the site from the tag values of the site tag table and the customer question tag table; and a step of allowing the personalized algorithm server to suggest a top ranking site having high similarity by each customer as a personalized and customized site based on the calculated similarity.

Description

How to use personalization algorithms to recommend customer questions and personalize algorithms. {METHOD FOR RECOMMENDING INFORMATION ON WETSITES CUSTOMIZED TO CLIENTS 'Q AND A BY USING PERSONALIZATION ALGORITHM AND SERVER USING THE SAME}

The present invention relates to a method for recommending a customer question answer customized site using a personalization algorithm, and more specifically, (a) the personalization algorithm server, the purchase history of the card users (hereinafter referred to as the first customer group) And a fashion tag class based on various purchase attributes of each purchase history of the first customer group. Flaunt consumption tag class; And a brand-specific attribute tag class; the fashion tag class includes a tag class stored in association with a corresponding purchase history when the purchase made by the first group of customers is fashion related content; The sporadic consumption tag class may include a tag class stored in association with a corresponding purchase history when the purchase purchased by the first customer group is a consumption history; And the brand-specific attribute tag class is a tag stored in association with the purchase history when the purchase made by the first customer group is related to various brand-specific attributes of the purchase history including the brand-specific attribute according to price range, style, or sector. Classify one or more purchase personalization tags included in the purchase data-based purchase personalization tag classes that include at least some of the classes and store the purchase personalization tags in association with the purchase history, and for each customer and Generating, by the purchase personalization tag, a customer purchase tag table listing tag values generated based on purchase history information of the first customer group corresponding to each purchase personalization tag; (b) the personalization algorithm server uses all purchase personalization tags associated with each purchase site tag, using a purchase site tag representing a purchase history attribute for a particular purchase site among the purchase personalization tags in the customer purchase tag table. For each purchase site for the purchase personalization tag and the corresponding tag value is listed for each of the purchase personalization tag for each customer in the first group of customers corresponding to the specific purchase personalization tag, sum the tag values, Generating a site tag table listing the values of the tag values divided by the number of customers in the first customer group greater than zero as the tag values for each purchase personalization tag; (c) for each of the predetermined questions answered by the second group of customers based on a response to the predetermined question obtained by the personalization algorithm server from the customers using a given application (hereinafter referred to as a second group of customers). Assigning a question tag value corresponding to each question, associating the purchase personalization tag matching the question, and for a brand corresponding to a question answered by the second customer group, all purchase personalization tags associated with each corresponding brand Assigning a predetermined attribute tag value to form a customer question tag table in which the brand-related purchase personalization tag and each question tag value or attribute tag value are related and listed with a tag value corresponding to the purchase personalization tag. step; (d) the personalization algorithm server calculating the similarity between the tag values in the site tag table obtained in step (b) and the tag values in the customer question tag table obtained in step (c); And (e) the personalization algorithm server optimizes a predetermined number of sites from each site having a higher ranking by sorting the sites with similarity for each customer in descending order based on the similarity calculated in the step (d). The present invention relates to a method and a personalization algorithm server.

Due to the development of the Internet, e-commerce using the Internet is exploding. As the amount of information provided through the Internet increases exponentially, users have to spend a lot of time and effort to search, search, and compare information. In other words, the era of information glut has come and consumers have to spend more time making choices and decisions because of the excess of information, services and production.

In particular, when using an online shopping mall, a user searches directly by inputting an item type or a site name that he / she wants to purchase through an internet search, and then browses the list presented by the search site or shopping mall one by one or follows some links presented. After looking at, proceed with the purchase.

On the other hand, as user requirements become more demanding and personalized requirements increase, services that provide user-specific information based on data analyzing individual patterns have been proposed in various ways.

As one of the existing methods, the method described in "Customized shopping information system and method through analysis of user tendency" disclosed in Korean Patent Publication No. 10-1419504 may provide effective shopping information through analysis of purchase tendency of each user. In order to provide customized shopping information through user disposition analysis. In the above conventional technology, the ontology module generates the at least one product tag and at least one propensity tag for each product tag by ontology processing the collected information, and the product tag generated through the ontology module. Various purchasing tendencies of customers are analyzed using tag matching storage module which maps with tags and stores them in information database.

However, the existing analysis method provides customized shopping information based on the analysis of each user's propensity and specific terminology, so that only the individual customer's inclination is provided and the term-based matching method is optimized for the customer. There is a limit to providing.

The present invention aims to solve all the above-mentioned problems.

According to the present invention, when providing an optimized site to a specific customer, even for those customers who do not have a purchase history, it is possible to provide an optimized site with high reliability to the specific customer based on answers obtained in response to questions related to the site obtained from the corresponding customer. The purpose.

The present invention also aims to provide information optimized for each customer by integrating information of multiple or all customers obtained from multiple customers.

Another object of the present invention is to provide a personalization algorithm method in which various functions are applied to attribute values and weighted differently according to each brand or attribute when providing an optimization site provided to a specific customer.

The characteristic structure of this invention for achieving the objective of this invention mentioned above, and realizing the characteristic effect of this invention mentioned later is as follows.

According to an embodiment of the present invention, a method of recommending a customer question and answer customized site using a personalization algorithm includes: (a) a personalization algorithm server, the purchase history of a first customer group and each purchase of the first customer group; Based on various purchasing attributes of the breakdown, fashion tag brackets; Flaunt consumption tag class; And a brand-specific attribute tag class; the fashion tag class includes a tag class stored in association with a corresponding purchase history when the purchase made by the first group of customers is fashion related content; The sporadic consumption tag class may include a tag class stored in association with a corresponding purchase history when the purchase purchased by the first customer group is a consumption history; And the brand-specific attribute tag class is a tag stored in association with the purchase history when the purchase made by the first customer group is related to various brand-specific attributes of the purchase history including the brand-specific attribute according to price range, style, or sector. Classify one or more purchase personalization tags included in the purchase data-based purchase personalization tag classes that include at least some of the classes and store the purchase personalization tags in association with the purchase history, and for each customer and Generating, by the purchase personalization tag, a customer purchase tag table listing tag values generated based on purchase history information of the first customer group corresponding to each purchase personalization tag; (b) the personalization algorithm server uses all purchase personalization tags associated with each purchase site tag, using a purchase site tag representing a purchase history attribute for a particular purchase site among the purchase personalization tags in the customer purchase tag table. For each purchase site for the purchase personalization tag and the corresponding tag value is listed for each of the purchase personalization tag for each customer in the first group of customers corresponding to the specific purchase personalization tag, sum the tag values, Generating a site tag table listing the values of the tag values divided by the number of customers in the first customer group greater than zero as the tag values for each purchase personalization tag; (c) the personalization algorithm server assigns a question tag value corresponding to each question to each of the predetermined questions answered by the second customer group based on the response to the predetermined question obtained from the second customer group, Associate the purchase personalization tag matching the question, and assign a predetermined attribute tag value to all purchase personalization tags associated with each corresponding brand for the brand corresponding to the question answered by the second customer group, Forming a customer question tag table listing the associated purchase personalization tag and each question tag value or attribute tag value associating with a tag value corresponding to the purchase personalization tag; (d) the personalization algorithm server calculating the similarity between the tag values in the site tag table obtained in step (b) and the tag values in the customer question tag table obtained in step (c); And (e) the personalization algorithm server optimizes a predetermined number of sites from each site having a higher ranking by sorting the sites with similarity for each customer in descending order based on the similarity calculated in the step (d). It includes; presenting the customized customer question and answer site.

In an embodiment, the step (a) may include (a1) the purchase personalization tag generated based on the purchase history of the first customer group according to each customer in the respective first customer group, and the corresponding purchase. List the number of purchases in the history as a tag value corresponding to the purchase personalization tag, and relative to which position among all tag values associated with the specific purchase personalization tags corresponding to each tag value. Generating a customer purchase tag table by modifying the tag value by applying a cumulative distribution function (CDF) function to indicate.

In an embodiment, the step (b) may include: (b1) The personalization algorithm server may use a lower weight and a site or brand or attribute frequently used by a first group of customers for each tag value of the site tag table. The site or brand or attribute that is not frequently used by the customer group is applied to a predetermined function to give a higher weight.

In one embodiment, the predetermined function is an Inverse Document Frequency (IDF) function.

In one embodiment, the step (b), (b2) the weighted tag by weighting different weights for each of the fashion tag class, the sporadic consumption tag class, the brand-specific attribute tag class to the site tag table The method further includes forming a modified site tag table reflecting the values.

In one embodiment, the step (b) is performed by (b3) the personalization algorithm server, while maintaining the tag value of the specific purchase personalization tag associated with the purchase site tag, other purchases associated with the brand-specific attribute tag class. The method further includes merging the modified site tag table to have predetermined tag values for the personalized tags.

In an embodiment, the step (c) further includes (c1) forming a modified customer question tag table reflecting the weighted tag value by assigning different weights to each purchase personalization tag.

In one embodiment, in the step (d), the similarity calculation is

, Where a, a ₁ , a ₂ ,. , a _m is a tag value of the customer question tag table and b, b ₁ , b ₂ ,. , b _m is a tag value of the site tag table, and φ function is a function representing a value multiplied by a weight according to whether or not a tag value of the customer question tag table is present.

According to another feature of the present invention, a personalization algorithm server recommending a customer question and answer customized site delivers a purchase history of each first customer group and / or a response to a predetermined question presented to each customer in the second customer group. Communication unit for receiving; And (i) a fashion tag class based on various purchase attributes of the purchase history of the first customer group and each purchase history of the first customer group; Flaunt consumption tag class; And a brand-specific attribute tag class; the fashion tag class includes a tag class stored in association with a corresponding purchase history when the purchase made by the first group of customers is fashion related content; The sporadic consumption tag class may include a tag class stored in association with a corresponding purchase history when the purchase purchased by the first customer group is a consumption history; And the brand-specific attribute tag class is a tag stored in association with the purchase history when the purchase made by the first customer group is related to various brand-specific attributes of the purchase history including the brand-specific attribute according to price range, style, or sector. Classify one or more purchase personalization tags included in the purchase data-based purchase personalization tag classes that include at least some of the classes and store the purchase personalization tags in association with the purchase history, and for each customer and Generating, by the purchase personalization tag, a customer purchase tag table listing tag values generated based on purchase history information of the first customer group corresponding to each purchase personalization tag; (ii) the purchase sites for all purchase personalization tags associated with each purchase site tag, using a purchase site tag indicating a purchase history attribute for a particular purchase site of the purchase personalization tags in the customer purchase tag table. List the purchase personalization tag and the corresponding tag value for each purchase personalization tag, and add the tag values of all the customers in the first customer group corresponding to the specific purchase personalization tag for each purchase personalization tag, and the tag values are greater than zero. Creating a site tag table listing the values divided by the number of customers in the first large customer group as tag values for each purchase personalization tag; (iii) a question tag value corresponding to each question is assigned to each of the predetermined questions answered by the second customer group based on the response to the predetermined question obtained from the second customer group, and wherein Associate a purchase personalization tag and assign a predetermined attribute tag value to all purchase personalization tags associated with each corresponding brand for a brand corresponding to a question answered by the second customer group, Forming a customer question tag table, listing each question tag value or attribute tag value associating with a tag value corresponding to the purchase personalization tag; (iv) calculating a similarity between tag values in the site tag table obtained in the process (ii) and tag values in the customer question tag table obtained in the process (iii); (v) Based on the similarity calculated in the above (iv) process, the sites with the highest similarity for each customer are arranged in descending order, and a predetermined number of sites from the site having the highest ranking are customized to the customer question and answer customized site optimized for each customer. It includes; a processor for performing;

The present invention provides a personalization algorithm that provides an optimized site with high reliability based on information on answers to shopping tendency questions even for customers who do not have a purchase history when presenting a site optimized for a customer.

In addition, the present invention provides a personalization algorithm that collects from all customers without using only the tags (attributes) of the customer, and provides an optimal site for the customer in the purchasing characteristics of the customers with similar characteristics.

In addition, the present invention provides the most optimized similarity between the customer and the site through a function that weights the inferred tag value inversely proportional to the popularity of the site, a weighting function for a specific brand, and a weighting function for each attribute. Can help you find it.

1 is a diagram illustrating the basic concept of a personalization algorithm according to the present invention.
2 is a flowchart illustrating a method for providing an optimized site using a personalization algorithm using a question and answer tag according to the present invention.
3 is a flowchart illustrating a method of modifying a tag value according to the present invention.
4 is a flowchart illustrating a process of generating a tag value for a customer question answer according to the present invention.
5 illustrates an example question list for generating a customer question answer tag in accordance with the present invention.
6 is a schematic structural diagram of a personalization algorithm server according to the present invention.

DETAILED DESCRIPTION The following detailed description of the invention refers to the accompanying drawings that show, by way of illustration, specific embodiments in which the invention may be practiced. These embodiments are described in sufficient detail to enable those skilled in the art to practice the invention. It should be understood that the various embodiments of the present invention are different but need not be mutually exclusive. For example, certain shapes, structures, and characteristics described herein may be embodied in other embodiments without departing from the spirit and scope of the invention with respect to one embodiment. In addition, it is to be understood that the location or arrangement of individual components within each disclosed embodiment may be changed without departing from the spirit and scope of the invention. The following detailed description, therefore, is not to be taken in a limiting sense, and the scope of the present invention, if properly described, is defined only by the appended claims, along with the full range of equivalents to which such claims are entitled. Like reference numerals in the drawings refer to the same or similar functions throughout the several aspects.

Hereinafter, preferred embodiments of the present invention will be described in detail with reference to the accompanying drawings so that those skilled in the art can easily implement the present invention.

1 is a diagram illustrating the basic concept of a personalization algorithm according to the present invention.

Referring to FIG. 1, the personalization algorithm according to the present invention forms a customer question tag table to which question tag values are assigned based on activity-based data of customers. Also, before that, one or more tags are extracted from the purchase history of the customer to form a purchase tag table, and based on this, the site tag table is formed. The similarity between the customer and the site is calculated from the tag values of the formed site tag table and the customer question tag table, and the optimized site is recommended for each customer.

In this case, the purchase data may be acquired through card usage history.

Meanwhile, the customers (first customer group) of the purchase history acquired to form the purchase tag table and / or the site tag table and the customers (second customer group) of the activity-based data acquired to form the question customer question tag table are mutually different. Can be different. For example, the purchase history acquired to form the purchase tag table and / or the site tag table may be card usage history (purchase history) obtained from a specific card company, and the activity-based data for forming the customer question tag table may be specified. It may be obtained from customers using the application (APP). In other words, a customer (second customer group) using a specific application (APP) and providing customer activity-based data to answer customer questions needs to be the same as a customer (first customer group) using a particular card company. none. Of course, in some cases, the first customer group and the second customer group may mean the same group.

On the other hand, in the process of forming the site tag table, based on the purchase data of the customers (first customer group) using the card company. In this process, various attributes in the purchase data (hereinafter referred to as purchase personalization tags or tags), for example, various purchase personalization tags that correspond to the fashion tag class, various purchase personalization tags that correspond to the sporadic consumption tag class, and purchase Extract various purchase personalization tags and brand-specific attribute tag classes corresponding to the brand tag class, and associate various purchase personalization tags according to the purchase history of each customer (per customer of the first customer group). Save it. Similarly, online shopping malls or product sites selling each product extract tag values associated with the same kind of purchase personalization tags. In this case, each site also includes various purchase personalization tags corresponding to the fashion tag class, various purchase personalization tags corresponding to the sporadic consumption tag class, various purchase personalization tags corresponding to the purchase brand tag class, and a variety of brand-specific attribute tag classes. Extract tag values associated with purchase personalization tags.

Meanwhile, the fashion tag class is a purchase tag class that is directly related to fashion, and is composed of purchase personalized tags including at least one attribute of a brand, a channel, a consumption region, and a consumption level, and a purchase corresponding to the fashion tag class. Examples of personalized tags include purchase brands, international online shopping malls (eg Amazon, Alibaba, etc.), luxury boutique street purchases, SPA brand purchases, online select shop purchases (eg shopbop, Fatfetch, etc.). ), A brand official mall purchase, and the like.

The sporadic consumption tag class is a tag class that indirectly reveals fashion consumption and fashion behavior, and is composed of purchase personalization tags including at least one attribute of overseas travel payment, Gourmet payment, and express hotel payment. Examples of purchasing personalization tags that fall under this category include tags such as fine dining consumption (Michelin, Jagat, Blue Ribbon, etc.), hair salon_Gangnam hair salon consumption, overseas travel_shopping_high frequency_2_20 or more. Can be.

The purchase brand tag class is a tag classified by purchase brand or site, and a specific tag name is set by each brand or site. Tags such as Buy_BRFSY1603313 and Brand Buy_BRFSY1604037 can be created and pasted.

The brand-specific attribute tag class is a tag divided by unique attributes of each purchase brand, not by purchase brand or site, and includes tags, style attributes, such as ultra high, high, medium, mid, low, low, and ultra low according to the price range attribute. According to the tag, such as Casual, Formal, Basic, Sporty, according to the professional genre, you can create and paste tags such as SPA, denim specialty, jewelry specialty.

The names or types of these tags are not fixed or limited to the above examples, and may arbitrarily select various types or conditions of tags and tag various kinds of contents and tags from purchase attributes.

2 is a flowchart illustrating a method for providing an optimized site using a personalization algorithm using a question and answer tag according to the present invention.

2, the personalization algorithm using the question and answer tag according to the present invention, by extracting one or more tags from the purchase history of the first group of customers (customers with purchase history using a card company) and generates a tag value Forming a customer purchase tag table (S10), applying a CDF function to the tag values (S20) to match a predetermined range (projection within a predetermined range), the information per customer purchase site from the customer purchase tag table that is customer-specific information Forming a site tag table (S30), a customer question tag assigned a question tag value based on a response to a predetermined question obtained from a second group of customers (customers who answered a question / answer using a specific application or site) Forming a table (S40), the similarity between the customer and the site from the tag values of the site tag table and the customer question tag table And a step (S50), and based on the calculated similarity step of presenting the sites for each customer-specific high-ranking high similarity with personalized custom site (S60) to calculate.

3 is a flowchart illustrating a method of modifying a tag value according to the present invention.

Referring to FIG. 3, in step S30 of FIG. 2, an IDF function is applied to tag values in each tag table (S31), and different weights are applied to tag values in each tag table according to a tag class (S32). And merging into the existing site tag table to have a predetermined predetermined tag value for the purchase personalization tags associated with the purchase brand unique attribute tag class (S33).

4 is a flowchart illustrating a process of generating a tag value for a customer question answer according to the present invention.

Referring to FIG. 4, the process of generating a tag value for a customer question answer may include assigning a question tag value corresponding to each question to each question answered by each customer in the second customer group, and matching the question. Associating a purchase personalization tag (S41), and assigning a predetermined attribute tag value to all purchase personalization tags related to each corresponding brand for a brand corresponding to a question answered by each customer in the second customer group (S42). ), Storing the personalized tag corresponding to the question and the personalized tag related to the brand by storing each question tag value or attribute tag value as a corresponding tag value (S43) and giving different weights to each purchased personalized tag. To reflect the weighted tag value (S44).

5 illustrates an example question list for generating a customer question answer tag in accordance with the present invention.

6 is a schematic diagram of a personalization algorithm server according to the present invention.

Referring to FIG. 6, the personalization algorithm server 100 according to the present invention includes a communication unit 10 and a processor 20.

Hereinafter, a method for recommending a site using a personalization algorithm according to the present invention will be described in detail with reference to FIGS. 1 to 6.

First, the communication unit 10 performs a function for receiving the purchase history of each first customer group and the question response history of each customer in the second customer group, and the processor 20 according to each step of FIGS. 2 to 4. It performs the process.

First, through the communication unit 10 of the personalization algorithm server 100, the purchase history of the first customer group and the reply information history of a predetermined question for personalization of the second customer group are received, and the processor 20 purchases the card users. Based on the history, extracting one or more tags and generating a tag value to form a customer purchase tag table (S10), this step, specifically, the purchase history of card users (first customer group) And a fashion tag class based on various purchase attributes of each purchase history of the first customer group. Flaunt consumption tag class; And a brand-specific attribute tag class; the fashion tag class includes a tag class stored in association with a corresponding purchase history when the purchase made by the first group of customers is fashion related content; The sporadic consumption tag class may include a tag class stored in association with a corresponding purchase history when the purchase purchased by the first customer group is a consumption history; And the brand-specific attribute tag class is a tag stored in association with the purchase history when the purchase made by the first customer group is related to various brand-specific attributes of the purchase history including the brand-specific attribute according to price range, style, or sector. Classify one or more purchase personalization tags included in the purchase data-based purchase personalization tag classes that include at least some of the classes and store the purchase personalization tags in association with the purchase history, and for each customer and For each purchase personalization tag, a customer purchase tag table listing tag values generated based on purchase history information of the first customer group corresponding to each purchase personalization tag is generated.

For example, when the processor 20 receives a plurality of purchase history information for each customer for all the customers (for example, n persons in the first customer group) through the communication unit 10, each purchase history is received. A customer purchase tag table is generated by associating one or more purchase personalization tags with each customer and each purchase information and tag values by identifying which customer's purchase history and from each purchase history.

Table 1 shows an example of a purchase personalization tag and its tag value obtained through some purchase history for customer 01 and customer 02.

Customer number TAG_NAME Tag value Customer01 Brand Purchase_BRFSY1604800_Number One Customer01 Site Purchase_FS1603990_Number One Customer01 Fashion Consumerism_Luxury Boutique Street_Purchases_Number One Customer02 Brand Purchase_BRFSY1602554_Number One Customer02 Brand Purchase_BRFSY1602871_Number One Customer02 Brand Purchase_BRFSY1603313_Number One Customer02 Brand Purchase_BRFSY1604037_Number One Customer02 Brand Purchase_BRFSY1604800_Number One Customer02 Site Purchase_FS1600284_Number 3 Customer02 Fashion consumerism One

Referring to Table 1, if the specific number of purchases (eg purchase number 01) of customer 01 relates to the purchase of a brand corresponding to the tag name BRFSY1604800, the number of brand purchases_BRFSY1604800 for the brand-specific attribute tag class Purchase personalized tag for the number of the site purchase_FS1603990 in the fashion tag class, if the specific purchase number is purchased from the site called FS1603990 If the specific purchase number is related to the fashion consumer_luxury boutique_street in the fashion tag class, the number of purchase personalized tags for the number of fashion consumer_luxury boutique_street is purchased. Store in association with tag value 1. As such, all relevant purchase personalization tags are found for each purchase number, and the corresponding purchase personalization tag name and the tag value (for example, the number of purchases) are related and stored.

In this way, the purchase personalization tag and the tag value analyzed for the total number of purchases of the customer 01 (for example, m purchases) are stored.

Meanwhile, referring to Table 1, if the specific number of purchases of customer 02 (for example, the purchase number 01) relates to the purchase of a brand corresponding to a tag name of BRFSY1602554, brand purchase_BRFSY1602554 of the brand-specific attribute class is applied. Store a personalized tag for the association with tag value 1, which is the number of purchases, and if the customer's other specific purchases (e.g., purchase number 02) are for a purchase of a brand that corresponds to a tag name of BRFSY1602871, Brand purchases for attribute class_BRFSY1602871 Store the personalized purchase tag for the number of purchases associated with tag value 1, which is the number of purchases, and another specific purchase of customer02 (for example, purchase number 03) is assigned to the tag name BRFSY1603313. If it's about buying the corresponding brand, you can add a personalization tag for the Brand Buy_ BRFSY1603313 If the store is associated with tag value 1, which is the number of purchases, and another specific purchase of customer 02 (e.g., purchase number 04) is for a purchase of a brand corresponding to the tag name BRFSY1604037, then the brand unique attribute class Brand purchase_ BRFSY1604037 Personalized tags for purchases are stored in association with tag value 1, which is the number of purchases, and another specific number of purchases for customer 02 (e.g., purchase number 05) is for the brand corresponding to the tag name BRFSY1604800. If it is about a purchase, the purchase personalization tag for the brand purchase_BRFSY1604800 for the brand-specific attribute class is stored in association with the tag value 1, which is the purchase. If the number of purchases corresponding to the number 05 of the customer 02 is purchased from the site FS1600284, the purchase personalization tag for the site purchase_FS1600284 in the fashion tag class is associated with the tag value 1, which is the number of purchases. And if the number of purchases corresponding to the number 05 of the customer 02 is related to fashion consumerism_hong target ticket_shopping in the fashion tag category, purchase personalized tag for the number of fashion consumers_hong target ticket_shopping. It is stored in association with the tag value 1, which is the number. In this way, for each purchase number of customer 02, all relevant purchase personalization tags are found and stored in association with the corresponding purchase personalization tag name and the tag value (for example, the number of purchases).

In this way, the customer purchase tag table is completed by associating and storing the total purchase personalization tag and tag values for all customers (n customers).

In the embodiment of Table 1, the example of storing only the number of purchases as tag values for each purchase history was described, but the corresponding purchases are stored by separately storing each purchase personalization tag for the number of purchases and the handling amount (purchase amount). It may be possible to store personalized tag names and tag values (eg, number of purchases and purchase amount (or handling amount)).

Then, in step S20 of applying the CDF function to the tag values and matching the predetermined values, the tag value corresponds to any position among all the tag values associated with the specific purchase personalization tags corresponding to the tag value. The tag value is modified by applying a cumulative distribution function (CDF) function to indicate relatively.

As an example of the CDF function, Equation 1 below may be applied.

Where X ₊ is the absolute value of tag value X, N ₊ is the total number of X values, t is any X value, and X = {t ₁ , t ₂ ,. , t _m }.

The CDF function is obtained by dividing the number of customers whose tag value is equal to or smaller than t in Equation 1 by the number of total customers (or all customers of the first customer group) whose tag value is greater than 0. For example, brand purchase_BRFSY1604800_ If there are 100,000 customers with a tag value greater than zero, and 70258 customers with a tag value of 7 or less of the 5.3 million customers who are all customers (or all customers in the first customer group) for the tag, The tag value for the _BRFSY1604800_ case is mapped to the value 70258/10 = 0.70258.

As such, when the CDF function is applied to tag values of the customer purchase tag table of Table 1, tag values are mapped between 0 and 1, as shown in Table 2 below.

Customer number TAG_NAME Tag value Customer01 Brand Purchase_ BRFSY1604800 _Number 0.668600435 Customer01 Site Purchase_FS1603990_Number 0.406240314 Customer01 Fashion Consumerism_Luxury Boutique Street_Purchases_Number 0.68613593 Customer02 Brand Purchase_BRFSY1602554_Number 0.773162939 Customer02 Brand Purchase_BRFSY1602871_Number 0.844827586 Customer02 Brand Purchase_BRFSY1603313_Number 0.565898436 Customer02 Brand Purchase_BRFSY1604037_Number 0.882200647 Customer02 Brand Purchase_ BRFSY1604800 _Number 0.668600435 Customer02 Site Purchase_FS1603990_Number 0.732285694 Customer02 Fashion consumerism 0.370285251

The CDF function performs on all tags in the customer purchase tag table.

Then, the step S30 of forming the site tag table, which is information for each purchase site, from the customer purchase tag table, which is information for each customer, performed by the processor 20, specifically, the purchase personalization tag of the customer purchase tag table. Of the purchase personalization tags and corresponding tag values for each purchase site for all purchase personalization tags associated with each purchase site tag, using a purchase site tag indicating a purchase history attribute of a specific purchase site. However, for each purchase personalization tag, the tag values of all the customers in the first customer group corresponding to the specific purchase personalization tag are summed, divided by the number of customers in the first customer group in which the tag values are greater than zero, for each purchase. Create a table of site tags that are listed by relating them to tag values per personalization tag. .

Specifically, in the example of Table 2, if there are purchase personalization tags related to the number of site purchase_FS1603990_number of purchase personalization tags (in bold in Table 2), the purchase personalization tags are all specific sites "FS1603990 key. And personalize the purchase personalization tags associated with the FS1603990 key, as well as all of the purchase personalization tags associated with the site for the FS1603990 key (i.e. all customers' FS1603990 key). Purchasing Personalization Tags Related to Sites) Each of the same purchasing personalization tags are grouped together to combine their tag values, and each purchase personalization tag is organized by each site.

That is, each tag value of the site tag table is the tag value (CDF) of the same purchase personalization tags of all customers (all customers of the first customer group) who have purchased on the same site using the tag values (CDF values) of the customer purchase tag table. Value) and divide by the number of customers whose tag value is greater than zero. In the bolded example of Table 2 above, suppose you want to obtain the tag value of the purchase personalization tag name for "Brand Purchase_BRFSY1604800_number" for the "FS1603990" site. If you have customers (i.e. three customers with purchase personalization tags related to site purchase_FS1603990_number) and two of them are for the "brand purchase_BRFSY1604800" tag, that is, the "brand purchase" _BRFSY1604800 Counts "There are two tag values greater than 0 (that is, two people) for the purchase personalization tag, and the values are for the" Brand Purchase_BRFSY1604800 counts "of customer 01, as shown in Table 2. The value (that is, the tag value of "Brand Purchase_BRFSY1604800" related to the FS1603990 site among the tag values for "Brand Purchase_BRFSY1604800" of Customer 01 is 0.668600435, and the value of "Brand Purchase_BRFSY1604800" of Customer 02 is If the tag value (that is, the tag value of "BuyBuy_BRFSY1604800" associated with the FS1603990 site among the tag values for "BuyBuy_BRFSY1604800" of customer 02) is 0.668600435, for the "FS1603990" site, the "Brand Buy_ The purchase personalization tag value of "BRFSY1604800" is calculated as (0.668600435+ 0.668600435 / 2 = 0.668600435.

In this way, if the CDF values of the tag values of the purchase personalization tag for each site for each site is obtained, it is derived as shown in Table 3 below. The bolded part of Table 3 shows the result of obtaining the site tag table tag value of the site corresponding to the site key FS1603990 described above by way of example.

Site key TAG_NAME Tag value FS1603990 Brand Purchase_BRFSY1602554_Number 0.773162939 FS1603990 Brand Purchase_BRFSY1602871_Number 0.844827586 FS1603990 Brand Purchase_BRFSY1603313_Number 0.565898436 FS1603990 Brand Purchase_BRFSY1604037_Number 0.882200647 FS1603990 Brand Purchase_ BRFSY1604800 _Number 0.668600435 FS1603990 Site Purchase_FS1603990_Number 0.569263004 FS1603990 Fashion consumerism 0.370285251 FS1603990 Fashion Consumerism_Luxury Boutique Street_Purchases_Number 0.68613593

The processor 20 then lowers the weight of the site or brand or attribute that customers frequently use and the weight of the site or brand or attribute that customers do not frequently use for each tag value in the site tag table. In operation S31, a predetermined function is applied to give a line. In this case, the predetermined function may be an Inverse Documents Frequency (IDF) function.

An example of the IDF function may be expressed by the following equation.

Where N is the number of customers in the entire first customer group, and | S | is the number of customers in the tagged first customer group.

The IDF value calculated through the above equation may be converted back to a logistic IDF value for scaling. The scaling equation can be expressed as follows.

In the site recommendation method using the personalization algorithm according to the present invention, an example of obtaining IDF and logistic IDF values is shown in Table 4 below.

TAG_NAME Total customers Number of customers with Tag idf logistic_idf Brand Purchase_BRFSY1602554_Number 5,300,000 8295 2.805407126 0.942967317 Brand Purchase_BRFSY1602871_Number 5,300,000 10759 2.692463598 0.936580471 Brand Purchase_BRFSY1603313_Number 5,300,000 944010 0.749298815 0.679025895 Brand Purchase_BRFSY1604037_Number 5,300,000 40943 2.112085601 0.892072298 Brand Purchase_BRFSY1604800_Number 5,300,000 73087 1.860429792 0.865347036 Site Purchase_FS1603990_Number 5,300,000 133 4.597171071 0.990020287

In the example of Table 4, when the number of customers in the entire first customer group is 5,300,000, and the number of customers associated with each purchase personalization tag is the same as the column of 'Tagged Customers' in Table 4, IDF and logistic IDF values as shown in Table 4 may be calculated.

By multiplying the logistic IDF value thus obtained with each tag value in the site tag table, the site tag table reflecting the IDF shown in Table 5 can be obtained.

Site key TAG_NAME Tag value FS1603990 Brand Purchase_BRFSY1602554_Number 0.729067382 FS1603990 Brand Purchase_BRFSY1602871_Number 0.791249018 FS1603990 Brand Purchase_BRFSY1603313_Number 0.384259692 FS1603990 Brand Purchase_BRFSY1604037_Number 0.786986758 FS1603990 Brand Purchase_BRFSY1604800_Number 0.578571405 FS1603990 Fashion consumerism 0.370285251 FS1603990 Luxury boutique street_buy_number 0.68613593

Then, the processor 20 may reflect different weights according to the tag class in the tag values in the site tag table (S32). Specifically, the fashion tag class and the sporadic consumption tag may be included in the site tag table. A process of forming a modified site tag table reflecting a weighted tag value by assigning different weights to each class and the brand-specific attribute tag class may be performed.

For example, of each purchase personalization tag in the site tag table shown in Table 5, the brand tag class associated with 'Brand Purchase _...' and 'Site Purchase _...' has a weight of 0.3, Months _... ',' Fashion Consumption _... ',' Fashion Consumption Business _... ',' Fashion Consumption Business _... ',' Fashion Consumption _... ',' Fashion Channel Overseas The fashion tag class associated with Online_ .. 'has a weight of 0.2, the brand-specific attribute class tag associated with' Properties _... 'has a weight of 0.2, and the' Gourmet _... ',' Hairroom_ .. The class of conspicuous consumption tags related to '.', 'Deluxe hotel _...' and 'overseas travel _...' can be weighted 0.3.

Site key TAG_NAME Tag value weight Weighted tag value FS1603990 Brand Purchase_BRFSY1602554_Number 0.729067382 0.3 0.218720215 FS1603990 Brand Purchase_BRFSY1602871_Number 0.791249018 0.3 0.237374706 FS1603990 Brand Purchase_BRFSY1603313_Number 0.384259692 0.3 0.115277908 FS1603990 Brand Purchase_BRFSY1604037_Number 0.786986758 0.3 0.236096028 FS1603990 Brand Purchase_BRFSY1604800_Number 0.578571405 0.3 0.173571421 FS1603990 Fashion consumerism 0.370285251 0.2 0.07405705 FS1603990 Fashion Consumerism_Luxury Boutique Street_Purchases_Number 0.68613593 0.2 0.137227186

Then, the processor 20 merges the existing site tag table with predetermined tag values for the brand-specific attribute tag class-related purchase personalization tags other than the specific purchase personalization tag associated with the purchase site tag. S33), specifically, the tag value of the specific purchase personalization tag, which is the purchase site tag indicating the purchase history attribute for the specific purchase site in relation to the card-using customer, as shown in the modified site tag table of Table 6, For example, as shown in Table 7 below, the predetermined predetermined tag value (for example, the tag value) for the predetermined purchase personalization tags for each site, among the purchase personalization tags associated with the brand-specific attribute tag class of each of all sites, in particular, 1) and merge it into the modified site tag table of Table 6.

Site key TAG_NAME FS1600012 Attribute_Price_L_Y FS1600012 Attribute_SPA_Y FS1600012 Attribute_Casual_Y FS1600012 Attribute_Basic_Y FS1600255 Attribute_Price_ML_Y FS1600255 Attribute_Casual_Y FS1600255 Attribute_Basic_Y FS1600303 Attribute_Price_ML_Y FS1600303 Attribute_Modern_Y FS1600379 Attribute_Price_MH_Y FS1600379 Attribute_Classic_Y FS1600598 Attribute_Price_ML_Y FS1600598 Attribute_Classic_Y FS1600637 Attribute_Price_MH_Y FS1600637 Attribute_Casual_Y FS1600637 Attribute_Basic_Y FS1600673 Attribute_Price_L_Y FS1600673 Attribute_Casual_Y FS1600673 Attribute_Basic_Y FS1600673 Attribute_Girlish_Y FS1600761 Attribute_Price_MH_Y FS1600761 Attribute_Casual_Y FS1600761 Attribute_Basic_Y FS1600794 Attribute_Price_ML_Y FS1600794 Attribute_SPA_Y FS1600794 Attribute_Casual_Y FS1600818 Attribute_Price_L_Y FS1601149 Attribute_Price_L_Y FS1601149 Attribute_SPA_Y FS1601149 Attribute_Casual_Y FS1601149 Attribute_Basic_Y FS1601320 Attribute_Price_L_Y FS1601320 Attribute_SPA_Y FS1601320 Attribute_Casual_Y FS1601433 Attribute_Price_L_Y FS1601433 Attribute_Casual_Y FS1601497 Attribute_Price_ML_Y FS1601497 Attribute_Casual_Y FS1601497 Attribute_Feminine_Y FS1601568 Attribute_Price_ML_Y FS1601568 Attribute_Casual_Y FS1601568 Attribute_Basic_Y FS1601568 Attribute_Classic_Y FS1601617 Attribute_Price_ML_Y FS1601617 Attribute_Classic_Y FS1601656 Attribute_Price_ML_Y FS1601656 Attribute_Modern_Y FS1601794 Attribute_Price_MH_Y FS1601794 Attribute_Formal_Y FS1601794 Property_Contemporary_Y FS1602502 Attribute_Price_ML_Y FS1602502 Attribute_clock_Y FS1602502 Attribute_Modern_Y FS1602559 Attribute_Price_ML_Y FS1602976 Attribute_Price_ML_Y FS1602976 Attribute_Casual_Y FS1602976 Attribute_Basic_Y FS1603043 Attribute_Price_MH_Y FS1603043 Attribute_Formal_Y FS1603043 Property_Sports Tech_Y FS1603548 Attribute_Price_M_Y FS1603548 Attribute_Formal_Y FS1603548 Attribute_Basic_Y FS1603548 Attribute_Modern_Y FS1603548 Property_Contemporary_Y FS1603596 Attribute_Price_ML_Y FS1603596 Attribute_Casual_Y FS1603596 Attribute_Classic_Y FS1603708 Attribute_Price_L_Y FS1603708 Attribute_SPA_Y FS1603708 Attribute_Casual_Y FS1603708 Attribute_Basic_Y FS1603718 Attribute_Price_ML_Y FS1603718 Attribute_Casual_Y FS1603759 Attribute_Price_ML_Y

For example, Table 7 may be a site-specific attribute table. That is, it extracts the purchase personalization tags associated with the unique attributes of the "FS1600012" site, extracts the purchase personalization tags associated with the unique attributes of the site not listed in Table 6, and adds a predetermined tag value (e.g., 1). After having, merge the two.

The result of merging Table 6 and Table 7 is shown in Table 8 below.

Site key TAG_NAME Tag value FS1603990 Brand Purchase_BRFSY1602554_Number 0.218720215 FS1603990 Brand Purchase_BRFSY1602871_Number 0.237374706 FS1603990 Brand Purchase_BRFSY1603313_Number 0.115277908 FS1603990 Brand Purchase_BRFSY1604037_Number 0.236096028 FS1603990 Brand Purchase_BRFSY1604800_Number 0.173571421 FS1603990 Fashion consumerism 0.07405705 FS1603990 Fashion Consumerism_Luxury Boutique Street_Purchases_Number 0.137227186 FS1600012 Attribute_Price_L_Y_number One FS1600012 Attribute_SPA_Y_number One FS1600012 Attribute_Casual_Y_Number One FS1600012 Attribute_Basic_Y_Number One FS1600255 Attribute_Price_ML_Y_number One FS1600255 Attribute_Casual_Y_Number One FS1600255 Attribute_Basic_Y_Number One FS1600303 Attribute_Price_ML_Y_number One FS1600303 Attribute_Modern_Y_number One FS1600379 Attribute_Price_MH_Y_number One FS1600379 Attribute_Classic_Y_number One FS1600598 Attribute_Price_ML_Y_number One FS1600598 Attribute_Classic_Y_number One FS1600637 Attribute_Price_MH_Y_number One FS1600637 Attribute_Casual_Y_Number One FS1600637 Attribute_Basic_Y_Number One FS1600673 Attribute_Price_L_Y_number One FS1600673 Attribute_Casual_Y_Number One FS1600673 Attribute_Basic_Y_Number One FS1600673 Attribute_Girlish_Y_Number One FS1600761 Attribute_Price_MH_Y_number One FS1600761 Attribute_Casual_Y_Number One FS1600761 Attribute_Basic_Y_Number One FS1600794 Attribute_Price_ML_Y_number One FS1600794 Attribute_SPA_Y_number One FS1600794 Attribute_Casual_Y_Number One FS1600818 Attribute_Price_L_Y_number One FS1601149 Attribute_Price_L_Y_number One FS1601149 Attribute_SPA_Y_number One FS1601149 Attribute_Casual_Y_Number One FS1601149 Attribute_Basic_Y_Number One FS1601320 Attribute_Price_L_Y_number One FS1601320 Attribute_SPA_Y_number One FS1601320 Attribute_Casual_Y_Number One FS1601433 Attribute_Price_L_Y_number One FS1601433 Attribute_Casual_Y_Number One FS1601497 Attribute_Price_ML_Y_number One FS1601497 Attribute_Casual_Y_Number One FS1601497 Attribute_Feminine_Y_number One FS1601568 Attribute_Price_ML_Y_number One FS1601568 Attribute_Casual_Y_Number One FS1601568 Attribute_Basic_Y_Number One FS1601568 Attribute_Classic_Y_number One FS1601617 Attribute_Price_ML_Y_number One FS1601617 Attribute_Classic_Y_number One FS1601656 Attribute_Price_ML_Y_number One FS1601656 Attribute_Modern_Y_number One FS1601794 Attribute_Price_MH_Y_number One FS1601794 Attribute_formal_y_number One FS1601794 Attribute_Contemporary_Y_number One FS1602502 Attribute_Price_ML_Y_number One FS1602502 Attribute_clock_Y_number One FS1602502 Attribute_Modern_Y_number One FS1602559 Attribute_Price_ML_Y_number One FS1602976 Attribute_Price_ML_Y_number One FS1602976 Attribute_Casual_Y_Number One FS1602976 Attribute_Basic_Y_Number One FS1603043 Attribute_Price_MH_Y_number One FS1603043 Attribute_formal_y_number One FS1603043 Attribute_Sports Tech_Y_Number One FS1603548 Attribute_Price_M_Y_number One FS1603548 Attribute_formal_y_number One FS1603548 Attribute_Basic_Y_Number One FS1603548 Attribute_Modern_Y_number One FS1603548 Attribute_Contemporary_Y_number One FS1603596 Attribute_Price_ML_Y_number One FS1603596 Attribute_Casual_Y_Number One FS1603596 Attribute_Classic_Y_number One

Referring back to FIG. 2, the processor 20 performs a step S40 of forming a customer question tag table to which a question tag value is assigned based on a response to a predetermined question obtained from customers. Specifically, the processor 20, based on the response to a predetermined question obtained from the customers (second customer group) using a given application, for each of the predetermined questions answered by each customer in the second customer group, Assigns a question tag value corresponding to each question, associates the purchase personalization tag that matches the question, and, for a brand corresponding to a question answered by each customer in the second customer group, all purchases associated with each corresponding brand Assigning a predetermined attribute tag value to personalization tags, and listing a customer question tag table listing the brand-related purchase personalization tag and each question tag value or attribute tag value associating with a tag value corresponding to the purchase personalization tag. Form. As described above, the predetermined application may be an application related to the card company or another application irrelevant to this. You might also use question and answer data from customers who visit specific sites, as well as customers who use specific applications.

Referring to the tag value generation process for the customer question answer of FIG. 4, for each question answered by each customer in the second customer group, a question tag value corresponding to each question is assigned, and a purchase personalization tag matching the question is provided. To associate (S41). An example of a question and answer presented to the second customer group is as shown in FIG. 5. As illustrated in FIG. 5, the questions presented to the second customer group may be questions related to a specific brand or a specific site, and if there is data in which a specific customer responds to the question, the tag value is assigned to 0 if the tag value is not 1. . This process is performed for every customer (every customer in the second customer group) to form a customer question tag table.

An example of the resulting customer question tag table is shown in Table 9.

Customer number Category 1 Category 2 Tag value TAG_NAME (tags to be matched) Customer01 Question 1 Hollister One Brand Purchase_ BRFSY1604800 _Number Customer01 Question 2 Cheongdam One Fashion Consumerism_Luxury Boutique Street_Purchases_Number Customer02 Question 1 Adidas One Brand Purchase_ BRFSN1602510 _Number Customer02 Question 1 Armani One Brand Purchase_ BRFSN1602522 _Number Customer02 Question 1 UNIQLO One Brand Purchase_ BRFSY1603313 _Number Customer02 Question 1 Apesse One Brand Purchase_ BRFSN1602657 _Number Customer02 Question 1 Hollister One Brand Purchase_ BRFSY1604800 _Number Customer02 Question 2 Hongdae One Fashion consumerism

Meanwhile, the purchase personalization tag TAG_NAME shown in bold in Table 9 are purchase personalization tags related to the brand. The processor 20 extracts the brand from the purchase personalization tag associated with the brand of Table 9, and assigns an attribute tag value to the purchase personalization tags corresponding to the brand specific attribute associated with the brand (S42).

First, Table 10 shows a complete example brand specific attribute table listing purchase personalization tags related to brand specific attributes for each brand.

Brand key POS_NAME (name of site tag to match) BRFSY1604800 Attribute_Basic_Y BRFSY1604800 Attribute_Casual_Y BRFSY1604800 Attribute_Price_L_Y BRFSN1602510 Attribute_Price_L_Y BRFSN1602522 Attribute_Price_M_Y BRFSY1603313 Attribute_Basic_Y BRFSY1603313 Attribute_Casual_Y BRFSY1603313 Attribute_Price_L_Y BRFSY1603313 Attribute_SPA_Y BRFSN1602657 Attribute_Basic_Y BRFSN1602657 Attribute_Casual_Y BRFSN1602657 Attribute_Price_M_Y BRFSY1604843 Attribute_Casual_Y BRFSY1604843 Property_Contemporary_Y BRFSY1604843 Attribute_Price_ML_Y BRFSY1604843 Attribute_denim_Y BRFSY1604845 Attribute_Casual_Y BRFSY1604845 Attribute_Price_L_Y BRFSY1604845 Property_Street_Y

Meanwhile, in bold in Table 10, among the purchase personalization tags listed in the customer question tag table in Table 9, purchase personalization tags corresponding to a brand-specific attribute associated with a specific brand (bold in Table 9). For example, in Table 9, Customer01 (Customer01 in Table 9 represents a customer in the second customer group, may be the same customer as Customer01 in Customer Table 1 (Customer01 in the first customer group), or another customer. If you answered a question related to the purchase personalization tag "Brand Purchase_BRFSY1604800_number", the purchase personalization tag is a purchase personalization tag associated with a brand with the brand key "BRFSY1604800" and the brand key is "BRFSY1604800". The purchase personalization tags corresponding to the brand-specific attribute of the in-brand indicate that there are "attribute_Basic_Y", "attribute_Casual_Y", and "attribute_Price_L_Y". That is, since the customer 01 responds to a question related to a brand whose brand key is "BRFSY1604800", the customer 01 also associates the purchase personalizations of "attribute_Basic_Y", "attribute_Casual_Y", and "attribute_Price_L_Y". .

The processor then deduces the purchase personalization tags associated with the brand specific attribute shown in bold in Table 10 and assigns the purchase personalization tags representing the brand specific attribute in Table 11 to customers associated with that brand (or brand key).

Customer number TAG_NAME Customer01 Attribute_Basic_Y_Number Customer01 Attribute_Casual_Y_Number Customer01 Attribute_Price_L_Y_number Customer02 Attribute_Basic_Y_Number Customer02 Attribute_Casual_Y_Number Customer02 Attribute_Price_L_Y_number Customer02 Attribute_Price_M_Y_number Customer02 Attribute_SPA_Y_number

Referring to Table 11, customer01 has a purchase personalization tag associated with the brand key "BRFSY1604800" in Table 9, and in Table 10, the purchase personalization tag "Attribute_Basic_Y", "Attribute_Casual_Y" associated with the brand key "BRFSY1604800" in Table 10. "," Attribute_Price_L_Y ", and customer02 has the purchase personalization tag for the brand-specific attributes associated with the brand keys" BRFSN1602510 "," BRFSN1602522 "," BRFSY1603313 "," BRFSN1602657 ", and" BRFSY1604800 "in Table 9 , Assign only one duplicate purchase personalization tag) "attribute_Basic_Y_count", "attribute_Casual_Y_count", "attribute_Price_L_Y_count", "attribute_Price_M_Y_count", "attribute_SPA_Y_count" ".

Then, in Table 11, a predetermined attribute tag value (for example, tag value 1) is assigned to all purchase personalization tags related to each corresponding brand, as shown in Table 12.

Customer number Category 2 Tag value TAG_NAME (tags to be matched) Customer01 Attribute_Basic_Y One Attribute_Basic_Y_Number Customer01 Attribute_Casual_Y One Attribute_Casual_Y_Number Customer01 Attribute_Price_L_Y One Attribute_Price_L_Y_number Customer02 Attribute_Basic_Y One Attribute_Basic_Y_Number Customer02 Attribute_Casual_Y One Attribute_Casual_Y_Number Customer02 Attribute_Price_L_Y One Attribute_Price_L_Y_number Customer02 Attribute_Price_M_Y One Attribute_Price_M_Y_number Customer02 Attribute_SPA_Y One Attribute_SPA_Y_number

The processor 20 then associates each question tag value or attribute tag value with a corresponding tag value in the purchase personalization tag corresponding to the question and the purchase personalization tag associated with the brand (S43), which are shown in Table 9 and Table. Corresponds to the integration of 12.

Table 13 shows the customer question tag table generated by incorporating the brand-specific attribute tag values into the customer question tag values in Table 9.

Customer number Category 1 Category 2 Tag value TAG_NAME (tags to be matched) Customer01 Question 1 Hollister One Brand Purchase_BRFSY1604800_Number Customer01 Question 2 Cheongdam One Fashion Consumerism_Luxury Boutique Street_Purchases_Number Customer02 Question 1 Adidas One Brand Purchase_BRFSN1602510_Number Customer02 Question 1 Armani One Brand Purchase_BRFSN1602522_Number Customer02 Question 1 UNIQLO One Brand Purchase_BRFSY1603313_Number Customer02 Question 1 Apesse One Brand Purchase_BRFSN1602657_Number Customer02 Question 1 Hollister One Brand Purchase_BRFSY1604800_Number Customer02 Question 2 Hongdae One Fashion consumerism Customer01 Attribute_Basic_Y One Attribute_Basic_Y_Number Customer01 Attribute_Casual_Y One Attribute_Casual_Y_Number Customer01 Attribute_Price_L_Y One Attribute_Price_L_Y_number Customer02 Attribute_Basic_Y One Attribute_Basic_Y_Number Customer02 Attribute_Casual_Y One Attribute_Casual_Y_Number Customer02 Attribute_Price_L_Y One Attribute_Price_L_Y_number Customer02 Attribute_Price_M_Y One Attribute_Price_M_Y_number Customer02 Attribute_SPA_Y One Attribute_SPA_Y_number

Then, the processor 20 reflects the weighted tag value by assigning different weights to each purchase personalized tag (S44). Specifically, the processor 20 reflects the modified customer question tag table (Table 15) by applying a weight value differently set for each purchase personalization tag to the customer question tag table of Table 13 as shown in Table 14. Create For example, when the tag prefix of Table 14, ie, the name of the palate personalization tag, is included among the purchase personalization tags of Table 13, the corresponding weight is reflected in the tag value to generate Table 15.

Tag prefix (included name of purchase personalized tag name) weight Brand Purchase_ 0.5 Fashion Consumption Fashion Channel Overseas Online_ International Online_ Deluxe hotel _ property_ One

Customer number TAG_NAME (tags to be matched) Tag value weight Weighted Tag Value Customer01 Brand Purchase_BRFSY1604800_Number One 0.5 0.5 Customer01 Fashion Consumerism_Luxury Boutique Street_Purchases_Number One 0.5 0.5 Customer02 Brand Purchase_BRFSN1602510_Number One 0.5 0.5 Customer02 Brand Purchase_BRFSN1602522_Number One 0.5 0.5 Customer02 Brand Purchase_BRFSY1603313_Number One 0.5 0.5 Customer02 Brand Purchase_BRFSN1602657_Number One 0.5 0.5 Customer02 Brand Purchase_BRFSY1604800_Number One 0.5 0.5 Customer02 Fashion consumerism One 0.5 0.5 Customer01 Attribute_Basic_Y_Number One One One Customer01 Attribute_Casual_Y_Number One One One Customer01 Attribute_Price_L_Y_number One One One Customer02 Attribute_Basic_Y_Number One One One Customer02 Attribute_Casual_Y_Number One One One Customer02 Attribute_Price_L_Y_number One One One Customer02 Attribute_Price_M_Y_number One One One Customer02 Attribute_SPA_Y_number One One One

Then, the processor 20 performs a process (S50) of calculating the similarity between the customer and the site from the tag values of the site tag table of Table 8 and the modified customer question tag table of Table 15.

In detail, the processor 20 may use the tag values of the site tag table of Table 8 and the customer question tag table of Table 15 by using the cosine similarity function of Equation 4 below. 2 may mean a customer within a customer group, or may mean a customer integrating both a first customer group and a second customer group) and similarity values for each site.

Where cosine similarity and cos (θ) represent the cosine similarity function,

a = (a ₁ , a ₂ ,…, a _m ) is the tag value in the customer question tag table, b = (b ₁ , b ₂ ,…, b _m ) is the tag value in the site tag table, i or 1 M represents each purchase personalization tag. In Equation 4, φ (x) is a weight value that is multiplied according to the presence or absence of purchase personalization tags and can be changed according to algorithm tuning.

Using Equation 4, using the site tag table of Table 8 and the customer question tag table of Table 15 to obtain similarity values for each customer and each site, an example is calculated as shown in the similarity result table (Table 16) below. Can be.

Customer number Site key cosine Customer01 FS1600850 0.774246889 Customer01 FS1600637 0.685794559 Customer01 FS1602976 0.662568273 Customer01 FS1603043 0.631708746 Customer01 FS1603548 0.646094571 Customer01 FS1600303 0.656305018 Customer01 FS1600673 0.672322908 Customer01 FS1603759 0.642862773 Customer01 FS1601320 0.635098523 Customer01 FS1601617 0.646846548 Customer01 FS1600761 0.641124308 Customer01 FS1603708 0.616420733 Customer01 FS1601149 0.635488542 Customer01 FS1600794 0.659833859 Customer01 FS1601794 0.677493536 Customer01 FS1603596 0.637343962 Customer01 FS1601433 0.634773774 Customer01 FS1601656 0.644593318 Customer01 FS1600598 0.614189116 ... ... ... Customer01 FS1600255 0.63716527 Customer01 FS1603718 0.638677526 ... ... ... ... ... ... ... ... ... ... ... ... ... ... ... Customer01 FS1601568 0.642685452 Customer01 FS1600012 0.644867515 Customer01 FS1603770 0.629643125 Customer01 FS1601497 0.654572545 Customer01 FS1602559 0.650797215 Customer01 FS1600379 0.668335146 Customer01 FS1604007 0.656620072 Customer01 FS1603926 0.632688153 Customer01 FS1600818 0.641262742 Customer01 FS1602502 0.659114548

In Table 16 above, only the similarity value for 'customer 01' is displayed, but the similarity value for all sites may be calculated for every customer.

Then, the processor 20 organizes the sites with similarity for each customer in descending order based on the calculated similarity, and presents a predetermined number of sites from the site having a higher ranking as a customized site optimized for each customer. (S50).

The present invention provides a personalized algorithm that provides a highly reliable optimized site based on the information on the answer to the shopping tendency question even for customers who do not have a purchase history when presenting a site optimized for the customer through the personalization algorithm as described above. To provide. In addition, the present invention can provide a personalization algorithm that collects from all customers without using only the tag (attribute) of the customer, and provides an optimal site for the customer in the purchasing characteristics of customers with similar characteristics. Functions that place tag values inversely proportional to the popularity of the site, weight a specific brand, and put different weights on each attribute can help you find the most optimized similarity between your customers and your site. .

Embodiments according to the present invention described above may be implemented in the form of program instructions that may be executed by various computer components, and may be recorded in a computer-readable recording medium. The computer-readable recording medium may include program instructions, data files, data structures, etc. alone or in combination. Program instructions recorded on the computer-readable recording medium may be those specially designed and configured for the present invention, or may be known and available to those skilled in the computer software arts. Examples of computer-readable recording media include magnetic media such as hard disks, floppy disks and magnetic tape, optical recording media such as CD-ROMs, DVDs, and magneto-optical media such as floptical disks. media), and hardware devices specifically configured to store and execute program instructions, such as ROM, RAM, flash memory, and the like. Examples of 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 may be configured to operate as one or more software modules to perform the process according to the invention, and vice versa.

Although the present invention has been described by specific embodiments such as specific components and the like, but the embodiments and the drawings are provided to assist in a more general understanding of the present invention, the present invention is not limited to the above embodiments. For those skilled in the art, various modifications and variations can be made from these descriptions.

Accordingly, the spirit of the present invention should not be limited to the above-described embodiments, and all of the equivalents or equivalents of the claims, as well as the appended claims, fall within the scope of the spirit of the present invention. I will say.

Claims (16)

In the method of recommending a customized question and answer site using a personalization algorithm,
(a) a personalization algorithm server, based on the purchase history of the first customer group and the various purchase attributes of each purchase history of the first customer group, a fashion tag class; Flaunt consumption tag class; And a brand-specific attribute tag class; the fashion tag class includes a tag class stored in association with a corresponding purchase history when the purchase made by the first group of customers is fashion related content; The sporadic consumption tag class may include a tag class stored in association with a corresponding purchase history when the purchase purchased by the first customer group is a consumption history; And the brand-specific attribute tag class is a tag stored in association with the purchase history when the purchase made by the first customer group is related to various brand-specific attributes of the purchase history including the brand-specific attribute according to price range, style, or sector. Classify one or more purchase personalization tags included in the purchase data-based purchase personalization tag classes that include at least some of the classes and store the purchase personalization tags in association with the purchase history, and for each customer and Generating, by the purchase personalization tag, a customer purchase tag table listing tag values generated based on purchase history information of the first customer group corresponding to each purchase personalization tag;
(b) the personalization algorithm server uses all purchase personalization tags associated with each purchase site tag, using a purchase site tag representing a purchase history attribute for a particular purchase site among the purchase personalization tags in the customer purchase tag table. For each purchase site for the purchase personalization tag and the corresponding tag value is listed for each of the purchase personalization tag for each customer in the first group of customers corresponding to the specific purchase personalization tag, sum the tag values, Generating a site tag table listing the values of the tag values divided by the number of customers in the first customer group greater than zero as the tag values for each purchase personalization tag;
(c) the personalization algorithm server assigns a question tag value corresponding to each question to each of the predetermined questions answered by the second customer group based on the response to the predetermined question obtained from the second customer group, Associate the purchase personalization tag matching the question, and assign a predetermined attribute tag value to all purchase personalization tags associated with each corresponding brand for the brand corresponding to the question answered by the second customer group, Forming a customer question tag table listing the associated purchase personalization tag and each question tag value or attribute tag value associating with a tag value corresponding to the purchase personalization tag;
(d) the personalization algorithm server calculating the similarity between the tag values in the site tag table obtained in step (b) and the tag values in the customer question tag table obtained in step (c); And
(e) The personalization algorithm server optimizes a predetermined number of sites from each site having a higher ranking by sorting the sites with similarity for each customer in descending order based on the similarity calculated in step (d). Presenting a customized question and answer site;
Site recommendation method comprising a. The method of claim 1,
In the step (a), (a1) the purchase personalization tag generated based on the purchase history of the first customer group according to each customer in the respective first customer group and the number of purchases of the corresponding purchase history Listed as tag values corresponding to the purchase personalization tag, the CDF (to indicate the relative position of the total tag values associated with the particular purchase personalization tag corresponding to each tag value relative) modifying the tag value by applying a cumulative distribution function) to generate a customer purchase tag table;
Site recommendation method characterized in that it further comprises. The method of claim 1,
In the step (b), (b1) the personalization algorithm server lowers the weight of the site or brand or attribute frequently used by the first group of customers for each tag value of the site tag table, and the first group of customers Applying a predetermined function to give a weight to a site or a brand or an attribute that is not frequently used;
Site recommendation method characterized in that it further comprises. The method of claim 3, wherein
And the predetermined function is an inverse document frequency (IDF) function. The method of claim 3, wherein
In step (b), the site tag table is modified to reflect weighted tag values by assigning different weights to the fashion tag class, the sporadic consumption tag class, and the brand-specific attribute tag class. And forming a site tag table. The method of claim 5,
In step (b), (b3) the personalization algorithm server maintains the tag value of the specific purchase personalization tag associated with the purchase site tag, and for other purchase personalization tags related to the brand-specific attribute tag class. And merging the modified site tag table to have a predetermined predetermined tag value. The method of claim 1,
The step (c) of claim (c1) further comprises the step of forming a modified customer question tag table reflecting the weighted tag value by assigning different weights for each purchase personalized tag. The method of claim 1,
In step (d),
The similarity calculation is as follows.

, Where a, a ₁ , a ₂ ,. , a _m is a tag value of the customer question tag table and b, b ₁ , b ₂ ,. , b _m is a tag value of the site tag table and φ function is a function indicating a value multiplied by the weight value in the presence or absence of a tag value of the customer question tag table. Customer question and answer personalization algorithm server that recommends
A communication unit for receiving a purchase history of each first customer group and / or a response result to a predetermined question presented to each customer in the second customer group; And
(i) a fashion tag class, based on various purchase attributes of the purchase history of the first customer group and each purchase history of the first customer group; Flaunt consumption tag class; And a brand-specific attribute tag class; the fashion tag class includes a tag class stored in association with a corresponding purchase history when the purchase made by the first group of customers is fashion related content; The sporadic consumption tag class may include a tag class stored in association with a corresponding purchase history when the purchase purchased by the first customer group is a consumption history; And the brand-specific attribute tag class is a tag stored in association with the purchase history when the purchase made by the first customer group is related to various brand-specific attributes of the purchase history including the brand-specific attribute according to price range, style, or sector. Classify one or more purchase personalization tags included in the purchase data-based purchase personalization tag classes that include at least some of the classes and store the purchase personalization tags in association with the purchase history, and for each customer and Generating, by the purchase personalization tag, a customer purchase tag table listing tag values generated based on purchase history information of the first customer group corresponding to each purchase personalization tag; (ii) the purchase sites for all purchase personalization tags associated with each purchase site tag, using a purchase site tag indicating a purchase history attribute for a particular purchase site of the purchase personalization tags in the customer purchase tag table. List the purchase personalization tag and the corresponding tag value for each purchase personalization tag, and add the tag values of all the customers in the first customer group corresponding to the specific purchase personalization tag for each purchase personalization tag, and the tag values are greater than zero. Creating a site tag table listing the values divided by the number of customers in the first large customer group as tag values for each purchase personalization tag; (iii) a question tag value corresponding to each question is assigned to each of the predetermined questions answered by the second customer group based on the response to the predetermined question obtained from the second customer group, and is matched with the question. Associating the purchase personalization tag and assigning a predetermined attribute tag value to all purchase personalization tags related to each corresponding brand for a brand corresponding to a question answered by the second customer group, the brand-related purchase personalization tag Forming a customer question tag table, listing each question tag value or attribute tag value associating with a tag value corresponding to the purchase personalization tag; (iv) calculating a similarity between tag values in the site tag table obtained in the process (ii) and tag values in the customer question tag table obtained in the process (iii); (v) Based on the similarity calculated in the above (iv) process, the sites with the highest similarity for each customer are arranged in descending order, and a predetermined number of sites from the site having the highest ranking are customized to the customer question and answer customized site optimized for each customer. A process for presenting;
Server comprising a. The method of claim 9,
The process (i) includes (i-1) the purchase personalization tag generated based on the purchase history of the first customer group according to each customer in the respective first customer group, and the purchase of the corresponding purchase history. List the number as tag values corresponding to the purchase personalization tag, wherein the tag value is relatively indicative of which position among the total tag values associated with the particular purchase personalization tags corresponding to each tag value And performing a process of generating a customer purchase tag table by modifying the tag value by applying a cumulative distribution function (CDF) function. The method of claim 9,
The process (ii) may include (ii-1) a site or brand or attribute frequently used by the first group of customers for each tag value in the site tag table with a lower weight and less frequently used by the first group of customers. The brand or attribute is a server, characterized in that for performing a process of applying a predetermined function to give a high weight. The method of claim 11,
The predetermined function is a server, characterized in that the IDF (Inverse Document Frequency) function. The method of claim 11,
The (ii) process may include (ii-2) different weights for the fashion tag class, the sporadic consumption tag class, and the brand-specific attribute tag class in the site tag table to reflect the weighted tag value. And further performing a process of forming a modified site tag table. The method of claim 13,
The process of (ii) maintains the tag value of a particular purchase personalization tag associated with the purchase site tag and maintains predetermined tag values for purchase personalization tags associated with other brand-specific attribute tag classes. And further merging the modified site tag table with the tag value. The method of claim 9,
The process of (iii) further comprises (iii-1) a process of forming a modified customer question tag table reflecting weighted tag values by assigning different weights to each purchase personalization tag. The method of claim 9,
In the above (iv) process,
The similarity calculation is as follows.

, Where a, a ₁ , a ₂ ,. , a _m is a tag value of the customer question tag table and b, b ₁ , b ₂ ,. , b _m is a tag value of the site tag table, and the function φ is a function representing a value multiplied by a weight according to the presence or absence of a tag value of the customer question tag table.

Images