WO2007105909A1

WO2007105909A1 - Method for targeting web advertisement clickers based on click pattern by using a collaborative filtering system with neural networks and system thereof

Info

Publication number: WO2007105909A1
Application number: PCT/KR2007/001250
Authority: WO
Inventors: Jinwoo Baek
Original assignee: Nhn Corporation
Priority date: 2006-03-16
Filing date: 2007-03-14
Publication date: 2007-09-20

Abstract

An advertisement recommendation method and system which can collect advertisement click information, calculate a relevance value between a plurality of advertisements, and recommend an advertisement having a greater relevance value with respect to previously clicked advertisement is provided. The advertisement recommendation method includes the steps of: collecting advertisement click information about each of a plurality of advertisements; calculating a relevance value between the plurality of advertisements by using the advertisement click information; and recommending a predetermined advertisement having a greater relevance value with respect to previously clicked advertisement by using the relevance value between the plurality of advertisements. Specifically, since an advertisement recommendation method and system which can calculate a relevance value between a plurality of advertisements by using advertisement click information, and recommend an advertisement having a greater relevance value with respect to previously clicked advertisement, it is possible to improve advertising effects.

Description

METHOD FOR TARGETING WEB ADVERTISEMENT CLICKERS BASED

ON CLICK PATTERN BY USING A COLLABORATIVE FILTERING SYSTEM

WITH NEURAL NETWORKS AND SYSTEM THEREOF

Technical Field

The present invention relates to a method and system for recommending an advertisement, and more particularly, to an advertisement recommendation method and system which can collect a user's advertisement click information, calculate a relevance value between a plurality of advertisements, and recommend an advertisement having a greater relevance value with respect to the user's previously clicked advertisement.

Background Art

Various types of services, such as electronic commerce, electronic advertisement, Internet phones, and the like, are being provided based on an Internet web. Accordingly, users readily utilize services which used to be performed in a physical way and thus Internet services are highlighted. As described above, since the Internet is filled with a huge amount of information, a service of retrieving particular data or contents in the Internet is essential to effectively utilize the Internet.

Specifically, when a user retrieves particular data or content without knowing a Universal Resource Locator (URL) or an Internet Protocol (IP) address of a corresponding site, the user may access a site of providing an Internet search service using an information search system, and enter a keyword associated with the data or the content and thereby utilize the desired data.

Sites of providing an Internet search service generally generate revenue by displaying a banner advertisement for a user who desires to utilize the Internet search service. In this instance, when the user frequently clicks the displayed banner advertisement, a click rate may be increased, and thus the sites may generate more revenue.

However, in a conventional banner advertisement method, unless a user enters a search keyword, a search service may not identify a user's interest or concern. Accordingly, a banner advertisement which the user is not much interested in may be provided. Consequently, since the user may not click the banner advertisement, a click rate of the banner advertisement, which has a closer relation to advertisement revenues, may be reduced.

Accordingly, there is a need for an advertisement recommendation method which can increase a click rate of a user for a banner advertisement.

Disclosure of Invention

Technical Goals

An aspect of the present invention provides a method and system which can recommend an advertisement, having a greater relevance value with respect to a user's previously clicked advertisement, to increase a click rate of a banner advertisement.

Another aspect of the present invention also provides a method and system which can recommend an advertisement, having a greater similarity value with respect to a user's previously clicked advertisement, to increase a click rate of a banner advertisement.

Technical solutions

According to an aspect of the present invention, there is provided a method of recommending an advertisement, the method including the steps of: collecting advertisement click information about each of a plurality of advertisements; calculating a relevance value between the plurality of advertisements by using the collected advertisement click information; and recommending a predetermined advertisement having a greater relevance value with respect to a user's previously clicked advertisement by using the calculated relevance value between the plurality of advertisements. According to another aspect of the present invention, there is provided a method of recommending an advertisement, the method including the steps of: collecting advertisement click information about each of a plurality of advertisements; calculating a relevance value between the plurality of advertisements by using the collected advertisement click information; training a neural network with the calculated relevance value between the plurality of advertisements; and recommending an advertisement having a greater relevance value with respect to a user's previously clicked advertisement by using the calculated relevance value between the plurality of advertisements and a result of the training.

According to still another aspect of the present invention, there is provided a system for recommending an advertisement, the system including: an advertisement information collector configured to collect clicked advertisement information and cookie information of a user which clicks the advertisement; a pattern extractor configured to extract an advertisement pattern by using the advertisement information and the user's cookie information; a relevance value calculator configured to calculate a relevance value between a plurality of advertisements by using the extracted advertisement pattern; a comparison component configured to compare a relevance value between the user's previously clicked advertisement and another advertisement by using the calculated relevance value between the plurality of advertisements; and a recommendation component configured to recommend an advertisement having a greater relevance value with respect to the user's previously clicked advertisement, based on a result of the comparison. According to yet another aspect of the present invention, there is provided a system for recommending an advertisement, the system including: an advertisement information collector configured to collect clicked advertisement information, and cookie information of a user which clicks the advertisement; a pattern extractor configured to extract an advertisement pattern by using the clicked advertisement information and the user's cookie information; a relevance value calculator configured to calculate a relevance value between the plurality of advertisements by using the extracted advertisement pattern; a neural network configured to train an SOM to find a similar advertisement using the calculated relevance value between the plurality of advertisements; a comparison component configured to compare the relevancy value between the plurality of advertisements by using the user's previously clicked advertisement and a result of the training of the neural network; and a recommendation component configured to recommend an advertisement having a greater relevance value with respect to the user's previously clicked advertisement based on a result of the comparison.

Brief Description of Drawings

FIG. 1 illustrates a relation between a user terminal and an advertisement recommendation system according to an exemplary embodiment of the present invention;

FIG. 2 is a flowchart illustrating a method of recommending an advertisement according to an exemplary embodiment of the present invention; FIG. 3 is a flowchart illustrating a process of selecting an advertisement having a greater relevance value according to an exemplary embodiment of the present invention;

FIG. 4 illustrates an example of determining whether a user clicks an advertisement according to an exemplary embodiment of the present invention; FIG. 5 is a flowchart illustrating a process of recommending an advertisement based on a result of a comparison of a relevance value between a plurality of advertisements according to an exemplary embodiment of the present invention;

FIG. 6 illustrates an example of recommending an advertisement according to an exemplary embodiment of the present invention; FIG. 7 is a block diagram illustrating a configuration of an advertisement recommendation system according to an exemplary embodiment of the present invention;

FIG. 8 is a flowchart illustrating a method of recommending an advertisement according to another exemplary embodiment of the present invention; FIG. 9 is a flowchart illustrating a process of selecting an advertisement having a greater relevance value according to another exemplary embodiment of the present invention;

FIG. 10 is a flowchart illustrating a process of recommending an advertisement based on a result of a comparison of a similarity value or a relevance value between a plurality of advertisement according to another exemplary embodiment of the present invention; and

FIG. 11 is a block diagram illustrating a configuration of an advertisement recommendation system according to another exemplary embodiment of the present invention.

Best Mode for Carrying Out the Invention

Reference will now be made in detail to embodiments of the present invention, examples of which are illustrated in the accompanying drawings, wherein like reference numerals refer to the like elements throughout. The embodiments are described below in order to explain the present invention by referring to the figures.

FIG. 1 illustrates a relation between a user terminal 120 and an advertisement recommendation system 100 according to an exemplary embodiment of the present invention.

Referring to FIG. 1, the advertisement recommendation system 100 provides a banner advertisement, requested by an advertiser, to the user terminal 120 via a communication network 110. In this instance, the user terminal 120 accesses a webpage of an advertisement recommendation site. Also, the advertisement recommendation system 100 collects click information of users for the banner advertisement and calculates a relevance value between a plurality of banner advertisements by using the collected click information.

Specifically, the advertisement recommendation system 100 identifies user information by using cookie information of the user terminal 120 when the user terminal 120 accesses the webpage of the advertisement recommendation site. Also, the advertisement recommendation system 100 provides a banner advertisement, having a greater relevance value with respect to the user's previously clicked advertisement, to the user terminal 120. Also, when the user accesses the webpage of the advertisement recommendation site, the advertisement recommendation system 100 identifies the user by using the cookie information, searches a database for the user's previously clicked advertisement, selects an advertisement having a greater similarity value with respect to the retrieved advertisement, and then provides the selected advertisement to the user terminal 120 via the webpage.

Accordingly, the advertisement recommendation system 100 recommends a banner advertisement having a greater relevance value or a greater similarity value with respect to the user's previously clicked banner advertisement for the user of the accessed user terminal 120 via the communication network 110. As described above, the advertisement recommendation system 100 may calculate a user's advertisement click information and provide an advertisement having a greater relevance value or a greater similarity value with respect to the user's previously clicked advertisement. Accordingly, it is possible to increase the user's click rate for an advertisement and thereby improve advertising effects.

The communication network 110 indicates a wired/wireless network which transmits various types of data between the advertisement recommendation system 100 and the user terminal 120.

The user terminal 120 indicates a device which can access the advertisement recommendation system 100 via the communication network 110 by including a memory and a microprocessor and thereby having a computation capability, such as a desktop personal computer (PC), a notebook PC, a personal digital assistant (PDA), a mobile communication terminal, and the like.

The user terminal 120 accesses the advertisement recommendation system 100 via the communication network 110, and transmits cookie information of a user and information about whether the user clicks an advertisement provided from the advertisement recommendation system 100, to the advertisement recommendation system 100. Also, the user terminal 120 receives a recommendation advertisement from the advertisement recommendation system 100 via the communication network 110.

FIG. 2 is a flowchart illustrating a method of recommending an advertisement according to an exemplary embodiment of the present invention. Referring to FIGS. 1 and 2, in operation 210, the advertisement recommendation system 100 collects advertisement click information about each of a plurality of advertisements, which are transmitted from the user terminal 120 via the communication network 110. In this instance, the advertisement click information includes information about whether any of the plurality of advertisement is clicked by a user, and also includes user information of the user which clicks the advertisement.

Specifically, in operation 210, when the user of the user terminal 210 clicks the advertisement corresponding to a banner advertisement, provided from the advertisement recommendation system 100 via the communication network 110, the advertisement recommendation system 100 may collect a user identifier included in cookie information by using the user's cookie information, and store the user identifier and a banner advertisement identifier in an advertisement database. In this instance, the cookie information may include a cookie identifier for the user identifier. In operation 220, the advertisement recommendation system 100 calculates a relevance value between the plurality of advertisements by using the collected advertisement click information. Specifically, in operation 220, the advertisement recommendation system 100 extracts an advertisement pattern vector by using the collected advertisement click information, and calculates the relevance value between the plurality of advertisements by using the extracted advertisement pattern vector.

Also, in operation 220, when a sufficient number of advertisement pattern vectors is extracted, the advertisement recommendation system 100 calculates the relevance value between the plurality of advertisements, based on a collaborative filtering algorithm, using the extracted advertisement pattern vector.

The collaborative filtering algorithm according to the present exemplary embodiment indicates a technique capable of identifying advertisements, which a large number of users are interested in, or advertisements with a similar pattern, based on the extracted advertisement pattern vector. The collaborative filtering algorithm is used to alternatively recommend pre-clicked advertisements to users with similar interests or to recommend an advertisement associated with a user's classified interest.

In operation 230, the advertisement recommendation system 100 recommends a predetermined advertisement having a greater relevance value with respect to a user's previously clicked advertisement by using the calculated relevance value between the plurality of advertisements.

Specifically, in operation 230, the advertisement recommendation system 100 stores the calculated relevance value between the plurality of advertisements, identifies the user and the user's previously clicked advertisement, compares a relevance value between the identified advertisement and the stored advertisement, and recommends an advertisement having a greater relevance value with respect to the identified advertisement, based on a result of the comparison.

Accordingly, the advertisement recommendation system 100 recommends the advertisement having either a greater relevance value or a greater similarity value with respect to the user's previously clicked advertisement. As an example, when a user clicks a first advertisement and a second advertisement having a greater relevance value with respect to the first advertisement is provided to the user as a recommendation advertisement, the advertisement recommendation system 100 may improve a click rate of the second advertisement in comparison with providing a third advertisement having a lower relevance value with the first advertisement.

As described above, a method of recommending an advertisement according to an exemplary embodiment may collect advertisement click information of users, analyze the collected advertisement click information, and thereby calculate a relevance value between a plurality of advertisements, and recommends an advertisement having a greater relevance value with respect to the user's previously clicked advertisement. Accordingly, it is possible to improve advertising effects. FIG. 3 is a flowchart illustrating a process of selecting an advertisement having a greater relevance value according to an exemplary embodiment of the present invention.

Referring to FIGS. 1 and 3, in operation 310, the advertisement recommendation system 100 receives a cookie identifier of a user from the user terminal 120, accessing an advertisement recommendation site, via the communication network 110. In this instance, when the user terminal 120 accesses the advertisement recommendation site, the user terminal 120 transmits a user identifier and a password to the advertisement recommendation system 100 via the communication network 110 as cookie information. The user identifier and the password are entered from the user to log into the advertisement recommendation site.

Also, the advertisement recommendation system 100 displays various types of advertisements on the accessed user terminal 120 through a webpage of the advertisement recommendation site. When the user clicks the advertisement of interest to the user from among the plurality of advertisements provided from the advertisement recommendation site, the user terminal 120 transmits click information about the clicked advertisement to the advertisement recommendation system 100 via the communication network 110. In this instance, the click information indicates information about whether the user clicks the provided advertisement.

In operation 320, the advertisement recommendation system 100 receives advertisement click information from the user terminal 120 accessing the advertisement recommendation site.

In operation 330, the advertisement recommendation system 100 extracts an advertisement click pattern by using the cookie identifier and the advertisement click information. In this instance, the advertisement click pattern is configured as a matrix which includes the cookie identifier corresponding to user information, and the advertisement click information corresponding to information about the clicked advertisement.

In operation 340, the advertisement recommendation system 100 determines whether a number of advertisement click vectors according to the extracted advertisement click pattern is sufficient to calculate the relevance value between the plurality of advertisements. Specifically, when an insufficient number of advertisement click vectors is extracted, a rule to calculate the relevance value between the plurality of advertisements may not be acquired. Accordingly, in operation 340, the advertisement recommendation system 100 determines whether a number of advertisement click vectors is sufficient to calculate the relevance value between the plurality of advertisements. In operation 350, the advertisement recommendation system 100 selects an appropriate advertisement click vector from the extracted advertisement click pattern. Specifically, in operation 350, the advertisement recommendation system 100 selects the appropriate advertisement click vector to calculate the relevance value between the plurality of advertisements, from the extracted advertisement click pattern. In operation 360, the advertisement recommendation system 100 calculates the relevance value between the plurality of advertisements by using the selected advertisement click vector. Specifically, in operation 360, the advertisement recommendation system 100 calculates the relevance value between the plurality of advertisements by using a collaborative filtering algorithm with respect to the selected advertisement click vector.

As described above, a method of recommending an advertisement according to an exemplary embodiment of the present invention compares the selected advertisement click vector based on the collaborative filtering algorithm. When a pattern of the selected advertisement click vector is similar, the method may calculate the relevance value between the plurality of advertisements to be comparatively greater. Conversely, when a pattern of the selected advertisement click vector is different, the method may calculate the relevance value between the plurality of advertisements to be comparatively smaller.

Input data to perform the collaborative filtering algorithm according to the present invention is indicated as an m x n (user-advertisement) matrix which includes an m number of user identifiers and an n number of advertisement identifiers. In the matrix, as shown in FIG. 4, when the user clicks an advertisement, a corresponding value is indicated as T. Conversely, when the user does not click an advertisement, a corresponding value is indicated as ¹O'.

To identify the advertisement click pattern between the plurality of advertisements, the collaborative filtering algorithm may group the advertisement identifiers, as show in FIG. 4, into, for example, "{1,0,1,...}, {0,1,0,...}, {1,0,1,...}, {0,0,0,...}, ...", and then may calculate the relevance value between the plurality of advertisements by comparing advertisement click patterns between the plurality of advertisements.

In operation 360, for example, when users clicking an advertisement '7}' generally click an advertisement '¹^' based on user identifiers, the advertisement recommendation system 100 may calculate a relevance value between the advertisements '?}' and 'ty to be comparatively greater. Conversely, when the users clicking the advertisement '7p do not generally click an advertisement 'M"', the advertisement recommendation system 100 may calculate the relevance value between the advertisements '7p and '¹^' to be comparatively smaller.

In operation 370, the advertisement recommendation system 100 stores the calculated relevance value between the plurality of advertisements. Specifically, in operation 370, the advertisement recommendation system 100 stores the calculated relevance value between the plurality of advertisements with respect to a plurality of advertisements click information and thereby creates a database.

In operation 380, the advertisement recommendation system 100 selects a top N number of advertisements with the greater relevance value. Specifically, in operation 380, the advertisement recommendation system 100 selects the top N number of advertisements having the greater relevance value between the plurality of advertisements, to recommend an advertisement having a greater relevance value with respect to the user's previously clicked advertisement.

As described above, a method of recommending an advertisement according to an exemplary embodiment of the present invention may extract advertisement click information about each of a plurality of advertisements, calculate a relevance value between the plurality of advertisements by using the extracted advertisement click information, and thereby select a top N number of advertisements having a greater relevance value between the plurality of advertisements.

FIG. 5 is a flowchart illustrating a process of recommending an advertisement based on a result of a comparison of a relevance value between a plurality of advertisements according to an exemplary embodiment of the present invention.

Referring to FIGS. 1 and 5, in operation 510, the advertisement recommendation system 100 receives a cookie identifier of a user from the user terminal 120 via the communication network 110. In this instance, the user terminal 120 accesses a webpage of an advertisement recommendation site. Specifically, in operation 510, the advertisement recommendation system 100 receives cookie information from the user terminal 120, accessing the advertisement recommendation site, via the communication network 110, extracts a cookie identifier from the cookie information and thereby identifies the user information.

In operation 520, when a user of the user terminal 120 clicks a banner advertisement provided from the advertisement recommendation site, the advertisement recommendation site 100 receives advertisement click information about the clicked banner advertisement. Specifically, in operation 520, when the banner advertisement is provided from the webpage of the advertisement recommendation site and clicked by the user, the advertisement recommendation system 100 receives the advertisement click information including an identifier of the clicked banner advertisement.

Accordingly, the advertisement recommendation system 100 may identify information about the user, clicking the banner advertisement, and information about the clicked advertisement through operations 510 and 520.

In operation 530, the advertisement recommendation system 100 extracts an advertisement click vector by using the cookie identifier and the advertisement click information. In operation 540, the advertisement recommendation system 100 compares a relevance value with the clicked advertisement by using the extracted advertisement click vector. Specifically, in operation 540, the advertisement recommendation system 100 may compare the extracted advertisement click vector and the stored relevance value and thereby select an advertisement having a greater relevance value with the clicked advertisement.

In operation 550, the advertisement recommendation system 100 recommends the selected advertisement, and provides the recommended advertisement to the user terminal 120 accessing the webpage of the advertisement recommendation site. Specifically, in operation 550, the advertisement recommendation system 100 provides an advertisement, having a greater relevance value with respect to the user's previously clicked advertisement, as a recommendation advertisement, to the user terminal 120. Accordingly, the user of the user terminal 120 may verify the recommendation advertisement provided from the webpage of the advertisement recommendation site. When the recommendation advertisement corresponds to the user's interest field, the user may click the advertisement. Conversely, when the recommendation advertisement is out of the user's interest field, the user may not click the advertisement. In operation 560, the advertisement recommendation system 100 determines whether the user of the accessed user terminal 120 clicks the recommendation advertisement provided through the webpage of the advertisement recommendation site.

When the user does not click the recommendation advertisement provided through the webpage of the advertisement recommendation site, the advertisement recommendation system 100 again performs operation 550 and recommends another advertisement, having a greater relevance value with respect to the clicked advertisement, through the webpage of the advertisement recommendation site. In this instance, the advertisement recommendation system 100 may sequentially provide at least one advertisement having a greater relevance value with respect to the clicked advertisement to the user terminal 120 until the user of the user terminal 120 clicks the recommendation advertisement provided through the webpage of the advertisement recommendation site.

Also, when the user clicks the recommendation advertisement provided through the webpage of the advertisement recommendation site, the advertisement recommendation system 100 terminates the advertisement recommendation process.

As described above, a method of recommending an advertisement according to the present invention may recommend an advertisement having a greater relevance value with respect to a user's previously clicked advertisement and thereby increase a click rate of a recommendation advertisement and also improve advertising effects.

FIG. 6 illustrates an example of recommending an advertisement according to an exemplary embodiment of the present invention. Referring to FIG. 6, a first advertisement 610 corresponds to a banner advertisement to advertise a premium shopping mall of Renault Samsung, and a second advertisement 620 corresponds to a banner advertisement to advertise new Santa Fe of Hyundai motors. As shown in FIG. 6, the first advertisement 610 and the second advertisement 620 are associated with vehicles and thus have a comparatively greater relevance value therebetween.

Accordingly, when a user clicks the first advertisement 610, the advertisement recommendation system 100 recommends the second advertisement 620 having a greater relevance value with respect to the first advertisement 610. Specifically, the second advertisement 620 is generally clicked by users which click the first advertisement 610. Also, since the second advertisement 620 has the greater relevance value with respect to the first advertisement 610, the second advertisement 620 is provided, as a recommendation advertisement, to the user which previously clicked the first advertisement 610 and thus a probability that the user may click the second advertisement 620 may be increased. Therefore, according to the present invention, it is possible to increase advertising effects.

FIG. 7 is a block diagram illustrating a configuration of an advertisement recommendation system 700 according to an exemplary embodiment of the present invention.

Referring to FIG. 7, the advertisement recommendation system 700 includes an advertisement information collector 710, a pattern extractor 720, a relevance value calculator 730, a comparison component 740, and a recommendation component 750.

The advertisement information collector 710 collects advertisement click information of a clicked advertisement and user information of a user which clicks the advertisement. Specifically, to advertise a predetermined advertisement, the advertisement information collector 710 collects a plurality of advertisement click information about whether a recommendation advertisement is clicked, and cookie information of the user which clicks the advertisement. The pattern extractor 720 extracts an advertisement pattern by using the advertisement click information and the user's cookie information. Specifically, the pattern extractor 720 extracts an advertisement pattern vector from a matrix as shown in

FIG. 4. In this instance, the matrix uses the user's cookie information and the advertisement information for a line and a column respectively.

The relevance value calculator 730 calculates a relevance value between a plurality of advertisements by using the extracted advertisement pattern. Specifically, the relevance value calculator 730 calculates the relevance value between the plurality of advertisements by using the extracted advertisement pattern vector. Also, the relevance value calculator 730 calculates the relevance value between the plurality of advertisements, based on a collaborative filtering algorithm, using the extracted advertising pattern vector.

The comparison component 740 compares a relevance value between the user's previously clicked advertisement and another advertisement by using the calculated relevance value between the plurality of advertisements. Specifically, the comparison component 740 compares a relevance value between advertisements, enrolled as a banner advertisement in response to a request from an advertiser, and the user's previously clicked advertisement by using the calculated relevance value between the plurality of advertisements. The recommendation component 750 recommends an advertisement having a greater relevance value with respect to the user's previously clicked advertisement among the enrolled advertisements, based on a result of the comparison.

As described above, the advertisement recommendation system 700 according to the present exemplary embodiment may collect advertisement click information of users about a banner advertisement, calculate a relevance value between a plurality of banner advertisements by using the collected advertisement click information, and recommend a banner advertisement having a greater relevance value with respect to the user's previously clicked banner advertisement. Accordingly, it is possible to improve a click rate of a recommendation advertisement and improve advertising effects. FIG. 8 is a flowchart illustrating a method of recommending an advertisement according to another exemplary embodiment of the present invention.

Referring to FIGS. 1 and 8, in operation 810, the advertisement recommendation system 100 collects advertisement click information about each of a plurality of advertisements, which are transmitted from the user terminal 120 via the communication network 110. In this instance, the advertisement click information includes information about whether any of the plurality of advertisement is clicked by a user, and user information of the user which clicks the advertisement.

Specifically, in operation 810, when the user of the user terminal 210 clicks the advertisement corresponding to a banner advertisement, provided from the advertisement recommendation system 100, via the communication network 110, the advertisement recommendation system 100 may collect a user identifier included in cookie information by using the user's cookie information and store the user identifier and a banner advertisement identifier in an advertisement database. In this instance, the cookie information may include a cookie identifier for the user identifier.

In operation 820, the advertisement recommendation system 100 calculates a relevance value between the plurality of advertisements by using the collected advertisement click information. Specifically, in operation 820, the advertisement recommendation system 100 extracts an advertisement pattern vector by using the collected advertisement click information, and calculates the relevance value between the plurality of advertisements by using the extracted advertisement pattern vector.

Also, in operation 820, when a sufficient number of advertisement pattern vectors is extracted, the advertisement recommendation system 100 calculates the relevance value between the plurality of advertisements, based on a collaborative filtering algorithm, using the extracted advertisement pattern vector.

In operation 830, the advertisement recommendation system 100 trains a neural network with the calculated relevance value between the plurality of advertisements. Specifically, in operation 830, the advertisement recommendation system 100 trains a self-organizing map (SOM) to find a similar advertisement by using the calculated relevance value between the plurality of advertisements. In this instance, the SOM detects similar data, based on predetermined data, through the training using an artificial intelligence neural network.

Accordingly, in operation 830, if the advertisement recommendation system 100 trains the SOM by using the calculated relevance value between the plurality of advertisements and the neural network, the advertisement recommendation system 100 may recommend an advertisement having a greater similarity value with respect to the advertisement.

In operation 840, the advertisement recommendation system 100 recommends an advertisement having a greater relevance value with respect to a user's previously clicked advertisement by using the calculated relevance value between the plurality of advertisements and a result of the training.

Specifically, the advertisement recommendation system 100 stores the relevance value between the plurality of advertisements and the result of the training, identifies the user and the user's previously clicked advertisement, compares a relevance value between the identified advertisement and the stored advertisement, or the result of the training, and recommends an advertisement having a greater relevance value with respect to the identified advertisement, based on a result of the comparison.

Accordingly, the advertisement recommendation system 100 recommends an advertisement having either the greater relevance value or the greater similarity value with respect to the clicked advertisement and thereby provides the recommendation advertisement to the user terminal 120 via the communication network 110.

Referring again to FIG. 6, for example, the advertisement recommendation system 100 provides the second advertisement 620, having the greater relevance value with respect to the first advertisement 610, as a recommendation advertisement, to a user which clicks the first advertisement 610. As described above, the first advertisement 610 and the second advertisement 620 are associated with vehicles and thus have a comparatively greater relevance value therebetween, and also has a greater similarity value therebetween. Accordingly, when the user clicks the first advertisement 610, the advertisement recommendation system 100 recommends the second advertisement 620 having the greater relevance value or the greater similarity value with respect to the first advertisement 610. Specifically, the second advertisement 620 is generally clicked by users which click the first advertisement 610. Also, since the second advertisement 620 has the greater similarity value with respect to the first advertisement 610, the second advertisement 620 is provided, as the recommendation advertisement, to the user which previously clicked the first advertisement 610 and thus a probability that the user may click the second advertisement 620 may be increased.

As described above, the advertisement recommendation system 100 may provide the second advertisement 620, having the greater relevance value to the first advertisement 610, to the user which clicked the first advertisement 610. Accordingly, it is possible to increase a click rate for the second advertisement 620 compared with a provided third advertisement having a lower relevance value with the first advertisement 610, as the recommendation advertisement.

As described above, a method of recommending an advertisement according to an exemplary embodiment may collect advertisement click information of users, analyze the collected advertisement click information, and thereby calculate a relevance value between a plurality of advertisements, and recommends an advertisement having a greater relevance value with respect to the user's previously clicked advertisement. Accordingly, it is possible to improve advertising effects.

FIG. 9 is a flowchart illustrating a process of selecting an advertisement having a greater relevance value according to another exemplary embodiment of the present invention.

Referring to FIGS. 1 and 9, in operation 910, the advertisement recommendation system 100 receives a cookie identifier of a user from the user terminal 120, accessing an advertisement recommendation site, via the communication network 110. In this instance, when the user terminal 120 accesses the advertisement recommendation site, the user terminal 120 transmits a user identifier and a password to the advertisement recommendation system 100 via the communication network 110 as cookie information. The user identifier and the password are entered from the user to log into the advertisement recommendation site. Also, the advertisement recommendation system 100 displays various types of advertisements on the accessed user terminal 120 through a webpage of the advertisement recommendation site. When the user clicks an advertisement of interest to the user from among the plurality of advertisements provided from the advertisement recommendation site, the user terminal 120 transmits click information for the clicked advertisement to the advertisement recommendation system 100 via the communication network 110. In this instance, the click information indicates information about whether the user clicks the provided advertisement.

In operation 920, the advertisement recommendation system 100 receives advertisement click information from the user terminal 120 accessing the advertisement recommendation site.

In operation 930, the advertisement recommendation system 100 extracts an advertisement click pattern by using the cookie identifier and the advertisement click information. In this instance, the advertisement click pattern is configured as a matrix which includes the cookie identifier corresponding to user information, and the advertisement click information corresponding to information about the clicked advertisement. In operation 940, the advertisement recommendation system 100 determines whether a number of advertisement click vectors according to the extracted advertisement click pattern is sufficient to calculate the relevance value between the plurality of advertisements. Specifically, when a smaller number of advertisement click vectors is extracted, a rule to calculate the relevance value between the plurality of advertisements may not be acquired. Accordingly, in operation 940, the advertisement recommendation system 100 determines whether a number of advertisement click vectors is sufficient to calculate the relevance value between the plurality of advertisements.

In operation 950, the advertisement recommendation system 100 selects an appropriate advertisement click vector from the extracted advertisement click pattern. Specifically, in operation 950, the advertisement recommendation system 100 selects the appropriate advertisement click vector to calculate the relevance value between the plurality of advertisements, from the extracted advertisement click pattern.

In operation 960, the advertisement recommendation system 100 calculates the relevance value between the plurality of advertisements by using the selected advertisement click vector. Specifically, in operation 960, the advertisement recommendation system 100 calculates the relevance value between the plurality of advertisements by using a collaborative filtering algorithm with respect to the selected advertisement click vector.

As described above, a method of recommending an advertisement according to an exemplary embodiment of the present invention compares the selected advertisement click vector based on the collaborative filtering algorithm. When a pattern of the selected advertisement click vector is similar, the method may calculate the relevance value between the plurality of advertisements to be comparatively greater. Conversely, when a pattern of the selected advertisement click vector is dissimilar, the method may calculate the relevance value between the plurality of advertisements to be comparatively smaller.

Input data to perform the collaborative filtering algorithm according to the present invention is indicated as an m x n (user-advertisement) matrix which includes an m number of user identifiers and an n number of advertisement identifiers. In the matrix, as shown in FIG. 4, when the user clicks an advertisement, a corresponding value is indicated as ¹I'. Conversely, when the user does not click an advertisement, a corresponding value is indicated as ¹O'.

To identify the advertisement click pattern between the plurality of advertisements, the collaborative filtering algorithm may group the advertisement identifiers, as shown in FIG. 4, into, for example, "{1,0,1,...}, {0,1,0,...}, {1,0,1,...}, {0,0,0,...}, ...", and then may calculate the relevance value between the plurality of advertisements by comparing advertisement click patterns between the plurality of advertisements.

In operation 960, for example, when users clicking an advertisement '7p generally click an advertisement '^f' based on user identifiers, the advertisement recommendation system 100 may calculate a relevance value between the advertisements '7p and ¹^-' to be comparatively greater. Conversely, when the users clicking the advertisement '7p do not generally click an advertisement ¹M-', the advertisement recommendation system 100 may calculate the relevance value between the advertisements '7]-' and 'M-' to be comparatively smaller. In operation 970, the advertisement recommendation system 100 trains a neural network with the calculated relevance value between the plurality of advertisements. Also, in operation 970, the advertisement recommendation system 100 finds an advertisement with the greater similarity value through an SOM using the calculated relevance value between the plurality of advertisements. Specifically, in operation 970, the advertisement recommendation system 100 may train the SOM using the neural network and the calculated relevance value between the plurality of advertisements, and thereby link similar advertisements. Accordingly, in operation 970, the advertisement recommendation system 100 may classify advertisements having a greater relevance value or a greater similarity value with respect to, for example, a first advertisement, using the SOM based on the neural network.

In operation 980, the advertisement recommendation system 100 stores the calculated relevance value between the plurality of advertisements and a result of the training based on the neural network. Specifically, in operation 980, the advertisement recommendation system 100 stores the calculated relevance value between the plurality of advertisements a plurality of advertisements click information and the result of the training, and thereby makes it a database. In operation 990, the advertisement recommendation system 100 selects a top N number of advertisements with the greater relevance value. Specifically, in operation 990, the advertisement recommendation system 100 selects the top N number of advertisements having the greater relevance value between the plurality of advertisements, to recommend an advertisement having a greater relevance value with respect to the user's previously clicked advertisement.

FIG. 10 is a flowchart illustrating a process of recommending an advertisement based on a result of comparison of a similarity value or a relevance value between a plurality of advertisement according to another exemplary embodiment of the present invention.

Referring to FIGS. 1 and 10, in operation 1010, the advertisement recommendation system 100 receives a cookie identifier of a user from the user terminal 120 via the communication network 110. In this instance, the user terminal 120 accesses a webpage of an advertisement recommendation site. Specifically, in operation 1010, the advertisement recommendation system 100 receives cookie information from the user terminal 120, accessing the advertisement recommendation site via the communication network 110, extracts a cookie identifier from the cookie information and thereby identifies the user information.

In operation 1020, when a user of the user terminal 120 clicks a banner advertisement provided from the advertisement recommendation site, the advertisement recommendation site 100 receives advertisement click information about the clicked banner advertisement. Specifically, in operation 1020, when the banner advertisement is provided from the webpage of the advertisement recommendation site and clicked by the user, the advertisement recommendation system 100 receives the advertisement click information including an identifier of the clicked banner advertisement.

Accordingly, the advertisement recommendation system 100 may identify information about the user clicking the banner advertisement and also identify information about the clicked advertisement through operations 1010 and 1020.

In operation 1030, the advertisement recommendation system 100 extracts an advertisement click vector by using the cookie identifier and the advertisement click information. In operation 1040, the advertisement recommendation system 100 compares a relevance value or a similarity value with to the clicked advertisement by using the extracted advertisement click vector. Specifically, in operation 1040, the advertisement recommendation system 100 may compare the extracted advertisement click vector and the stored relevance value or the similarity value, and thereby select an advertisement having a greater relevance value or a greater similarity value with respect to the clicked advertisement.

In operation 1050, the advertisement recommendation system 100 recommends the selected advertisement, and provides the recommended advertisement to the user terminal 120 accessing the webpage of the advertisement recommendation site. Specifically, in operation 1050, the advertisement recommendation system 100 provides an advertisement, having a greater relevance value or a greater similarity value with respect to the user's previously clicked advertisement, as a recommendation advertisement, to the user terminal 120. Accordingly, the user of the user terminal 120 may verify the recommendation advertisement provided from the webpage of the advertisement recommendation site. When the recommendation advertisement corresponds to the user's interest field, the user may click the advertisement. Conversely, when the recommendation advertisement is out of the user's interest field, the user may not click the advertisement.

In operation 1060, the advertisement recommendation system 100 determines whether the user of the accessed user terminal 120 clicks the recommendation advertisement provided through the webpage of the advertisement recommendation site. When the user does not click the recommendation advertisement provided through the webpage of the advertisement recommendation site, the advertisement recommendation system 100 again performs operation 1050 and recommends another advertisement, having a greater relevance value or a greater similarity value with respect to the clicked advertisement through the webpage of the advertisement recommendation site. In this instance, the advertisement recommendation system 100 may sequentially provide at least one advertisement having a greater relevance value with or a greater similarity value with respect to the clicked advertisement to the user terminal 120 until the user of the user terminal 120 clicks the recommendation advertisement provided through the webpage of the advertisement recommendation site. Also, when the user clicks the recommendation advertisement provided through the webpage of the advertisement recommendation site, the advertisement recommendation system 100 terminates the advertisement recommendation process.

As described above, a method of recommending an advertisement according to the present invention may recommend an advertisement having a greater relevance value or a similarity value with respect to a user's previously clicked advertisement and thereby increase a click rate of a recommendation advertisement and also improve advertising effects.

FIG. 11 illustrates a configuration of an advertisement recommendation system 1100 according to another exemplary embodiment of the present invention. Referring to FIG. 11, the advertisement recommendation system 1100 includes an advertisement information collector 1110, a pattern extractor 1120, a relevance value calculator 1130, a neural network 1140, a comparison component 1150, and a recommendation component 1160.

The advertisement information collector 1110 collects advertisement click information of an advertisement and user information of a user which clicks the advertisement. Specifically, to advertise an advertisement, the advertisement information collector 1110 collects a plurality of advertisement click information about whether a recommendation advertisement is clicked, and cookie information of the user which clicks the advertisement.

The pattern extractor 1120 extracts an advertisement pattern by using the advertisement click information and the user's cookie information. Specifically, the pattern extractor 1120 extracts an advertisement pattern vector from a matrix as shown in FIG. 4. In this instance, the matrix uses the user's cookie information and the advertisement information for a line and a column respectively.

The relevance value calculator 1130 calculates a relevance value between a plurality of advertisements by using the extracted advertisement pattern. Specifically, the relevance value calculator 1130 calculates the relevance value between the plurality of advertisements by using the extracted advertisement pattern vector.

Also, the relevance value calculator 1130 calculates the relevance value between the plurality of advertisements, based on a collaborative filtering algorithm, using the extracted advertising pattern vector. The neural network 1140 is trained with the relevance value between the plurality of advertisements to find a similar advertisement. Specifically, the neural network trains an SOM to find an advertisement having a greater similarity by using the calculated relevance value between the plurality of advertisements.

The comparison component 1150 compares the relevancy value between the plurality of advertisements by using the user's previously clicked advertisement and a result of the training of the neural network. Specifically, the comparison component 1150 compares an advertisement having a greater relevance value or a greater similarity value with respect to the user's previously clicked advertisements among advertisements enrolled as banner advertisements in response to a request from an advertiser. The recommendation component 1160 recommends an advertisement having a greater relevance value with respect to the user's previously clicked advertisement among the enrolled advertisements, based on a result of the comparison. Also, the recommendation component 1160 recommends an advertisement having a greater relevance value or a greater similarity value with respect to the user's previously clicked advertisement among the enrolled advertisements, based on the result of the comparison. As described above, the advertisement recommendation system 1100 according to the present exemplary embodiment may collect advertisement click information of users about a banner advertisement, calculate a relevance value between a plurality of banner advertisements by using the collected advertisement click information, and recommend a banner advertisement having a greater relevance value with respect to the user's previously clicked banner advertisement. Accordingly, it is possible to improve a click rate of a recommendation advertisement and improve advertising effects.

The exemplary embodiments of the present invention include computer-readable media including program instructions to implement various operations embodied by a computer. The media may also include, alone or in combination with the program instructions, data files, data structures, tables, and the like. The media and program instructions may be those specially designed and constructed for the purposes of the present invention, or they may be of the kind well known and available to those having skill in the computer software arts. Examples of computer-readable media include magnetic media such as hard disks, floppy disks, and magnetic tape; optical media such as CD ROM disks; magneto-optical media such as floptical disks; and hardware devices that are specially configured to store and perform program instructions, such as readonly memory devices (ROM) and random access memory (RAM). The media may also be a transmission medium such as optical or metallic lines, wave guides, etc. including a carrier wave transmitting signals specifying the program instructions, data structures, etc. Examples of program instructions include both machine code, such as produced by a compiler, and files containing higher level code that may be executed by the computer using an interpreter.

According to the present invention, there is provided an advertisement recommendation method and system which can calculate a relevance value between a plurality of advertisements by using advertisement click information, and recommend an advertisement having a greater relevance value with respect to a user's previously clicked advertisement based on the calculated relevance value between the plurality of advertisements, and thereby can improve advertising effects.

Also, according to the present invention, there is provided an advertisement recommendation method which can calculate a relevance value between a plurality of advertisements by using advertisement click information, and train an SOM neural network by using the calculated relevance value between the plurality of advertisements, and thereby recommend an advertisement having a greater similarity value with respect to a user's previously clicked advertisement.

Although a few embodiments of the present invention have been shown and described, the present invention is not limited to the described embodiments. Instead, it would be appreciated by those skilled in the art that changes may be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the claims and their equivalents.

Claims

1. A method of recommending an advertisement, the method comprising the steps of: collecting advertisement click information about each of a plurality of advertisements; calculating a relevance value between the plurality of advertisements by using the collected advertisement click information; and recommending a predetermined advertisement having a greater relevance value with respect to a user's previously clicked advertisement by using the calculated relevance value between the plurality of advertisements.

2. The method of claim 1, wherein the step of calculating the relevance value between the plurality of advertisements by using the collected advertisement click information comprises the steps of: extracting an advertisement click vector by using the collected advertisement click information; and calculating the relevance value between the plurality of advertisements by using the extracted advertisement click vector.

3. The method of claim 2, wherein the step of calculating the relevance value between the plurality of advertisements by using the extracted advertisement click vector comprises the step of: calculating the relevance value between the plurality of advertisements, based on a collaborative filtering algorithm, using the extracted advertisement click vector when a sufficient number of advertisement click vectors is extracted.

4. The method of claim 1, further comprising the step of: storing the calculated relevance value between the plurality of advertisements, wherein the step of recommending the advertisement having the greater relevance value with respect to the user's previously clicked advertisement comprises the steps of: identifying the user and the user's previously clicked advertisement; comparing a relevance value between the identified advertisement and the stored advertisement; and recommending an advertisement having a greater relevance value with respect to the identified advertisement, based on a result of the comparison.

5. The method of claim 4, wherein the step of recommending the advertisement having the greater relevance value with respect to the identified advertisement comprises the step of: sequentially recommending the advertisement having the greater relevance value with respect to the identified advertisement, based on the result of the comparison, until the user clicks the recommended advertisement.

6. A method of recommending an advertisement, the method comprising the steps of: collecting advertisement click information about each of a plurality of advertisements; calculating a relevance value between the plurality of advertisements by using the collected advertisement click information; training a neural network with the calculated relevance value between the plurality of advertisements; and recommending an advertisement having a greater relevance value with respect to a user's previously clicked advertisement by using the calculated relevance value between the plurality of advertisements and a result of the training.

7. The method of claim 6, wherein the step of calculating the relevance value between the plurality of advertisements by using the collected advertisement click information comprises the steps of: extracting an advertisement pattern vector by using the collected advertisement click information; and calculating the relevance value between the plurality of advertisements by using the extracted advertisement pattern vector.

8. The method of claim 7, wherein the step of calculating the relevance value between the plurality of advertisements by using the extracted advertisement pattern vector comprises the step of: calculating the relevance value between the plurality of advertisements, based on a collaborative filtering algorithm, using the extracted advertising pattern vector when a sufficient number of advertisement pattern vectors is extracted.

9. The method of claim 6, further comprising the step of: storing the calculated relevance value between the plurality of advertisements and the result of the training, wherein the step of recommending the advertisement having a greater relevance value with respect to the user's previously clicked comprises the steps of: identifying the user and the user's previously clicked advertisement; comparing a relevance value between the identified advertisement and the stored advertisement, or the result of the training; and recommending an advertisement having a greater relevance value with respect to the identified advertisement, based on a result of the comparison.

10. The method of claim 6, wherein the step of training of the neural network with the calculated relevance value between the plurality of advertisements comprises the step of: training a self-organizing map (SOM) to find an advertisement having a greater similarity value by using the calculated relevance value between the plurality of advertisements.

11. The method of claim 10, wherein the step of recommending the advertisement having the greater relevance value with the user's previously clicked advertisement comprises the step of: recommending an advertisement having a greater similarity value with respect to the user's previously clicked advertisement.

12. A computer-readable recording medium storing a program implementing the method according to any one of claims 1 through 11.

13. A system for recommending an advertisement, the system comprising: an advertisement information collector configured to collect clicked advertisement information and cookie information of a user which clicks the advertisement; a pattern extractor configured to extract an advertisement pattern by using the advertisement information and the user's cookie information; a relevance value calculator configured to calculate a relevance value between a plurality of advertisements by using the extracted advertisement pattern; a comparison component configured to compare a relevance value between the user's previously clicked advertisement and another advertisement by using the calculated relevance value between the plurality of advertisements; and a recommendation component configured to recommend an advertisement having a greater relevance value with respect to the user's previously clicked advertisement, based on a result of the comparison.

14. The system of claim 13, wherein the pattern extractor extracts an advertisement pattern vector from a matrix, the matrix using the user's cookie information and the advertisement information for a line and a column respectively.

15. The system of claim 13, wherein the relevance value calculator calculates the relevance value between the plurality of advertisements by using the extracted advertisement pattern vector.

16. The system of claim 15, wherein the relevance value calculator calculates the relevance value between the plurality of advertisements, based on a collaborative filtering algorithm, using the extracted advertising pattern vector.

17. A system for recommending an advertisement, the system comprising: an advertisement information collector configured to collect clicked advertisement information, and cookie information of a user which clicks the advertisement; a pattern extractor configured to extract an advertisement pattern by using the clicked advertisement information and the user's cookie information; a relevance value calculator configured to calculate a relevance value between the plurality of advertisements by using the extracted advertisement pattern; a neural network configured to train an SOM to find a similar advertisement using the calculated relevance value between the plurality of advertisements; a comparison component configured to compare the relevancy value between the plurality of advertisements by using the user's previously clicked advertisement and a result of the training of the neural network; and a recommendation component configured to recommend an advertisement having a greater relevance value with respect to the user's previously clicked advertisement based on a result of the comparison.

18. The system of claim 17, wherein the pattern extractor extracts an advertisement pattern vector from a matrix, the matrix using the user's cookie information and the advertisement information for a line and a column respectively.

19. The system of claim 18, wherein the relevance value calculator calculates the relevance value between the plurality of advertisements by using the extracted advertisement pattern vector.

20. The system of claim 18, wherein the relevance value calculator calculates the relevance value between the plurality of advertisements, based on a collaborative filtering algorithm, using the extracted advertising pattern vector.

21. The system of claim 17, wherein the neural network is used to train an SOM to find a similar advertisement by using the calculated relevancy value between the plurality of advertisements.

22. The system of claim 21, wherein the recommendation component recommends an advertisement having either a greater relevance value or a greater similarity value with respect to the clicked advertisement based on the result of the comparison.