JP2013182595A - Advertisement distribution device, advertisement distribution method, and program - Google Patents

Abstract

PROBLEM TO BE SOLVED: To enable evaluation of a position for displaying an advertisement.SOLUTION: A bias estimation part 217 in a Web server 20 divides advertisements with a similar click through rate into groups, calculates a position-by-position click through rate for each group, and calculates a proportion of the position-by-position click through rate to the click through rate of a position 1. The bias estimation part 217 obtains a bias by averaging the proportion by position.

Description

  The present invention relates to an advertisement distribution device, an advertisement distribution method, and a program.

  Advertisement distribution (listing) to Web pages and the like is widely performed. In conventional advertisement distribution, an advertisement having a high advertisement evaluation, such as CTR (Click Through Rate) or CPM (Cost Per Mille), is selected from many advertisements. It is designed to be displayed on the side. Further, Patent Document 1 describes that an advertisement is promoted from a less conspicuous position to a more conspicuous position according to economic value or the like.

Special table 2007-524915

  In recent years, the content of contents and the layout of Web pages have been diversified, and the upper part of the Web page is not necessarily conspicuous. However, Patent Document 1 is based on the premise that a prominent position on a Web page is set in advance, and it is not possible to evaluate which position on the Web page is prominent.

  The present invention has been made in view of such a background, and an object thereof is to provide a bias estimation apparatus, a bias estimation method, and a program capable of evaluating a position where an advertisement is displayed.

  A main invention of the present invention for solving the above-described problem is an apparatus for estimating a bias that is an evaluation value of a position where an advertisement is displayed, and the advertisement is viewed for each position where the advertisement can be displayed and for each advertisement. An action number information acquisition unit for acquiring action number information for determining the number of actions that is the number of times a person has taken action according to the advertisement, an advertisement evaluation value that is an evaluation value of the advertisement, and the advertisement An advertisement group generation unit that divides the advertisement into a plurality of groups according to an evaluation value, and at least one of the groups, for each position, the advertisement belonging to the group and the action frequency information corresponding to the position And a bias estimation unit that estimates the bias according to the total number of actions determined based on the behavior.

  The bias estimation apparatus of the present invention further includes a display count information acquisition unit that acquires display count information for calculating the display count of the advertisement at the position for each of the advertisement and the position, and the bias estimation unit For each position, the total of the number of actions determined based on the advertisement belonging to the group and the action count information corresponding to the position corresponds to the advertisement belonging to the group and the position. An action rate divided by the total number of display times determined based on the display number information may be calculated, and the bias may be estimated according to the action rate.

  In the bias estimation apparatus of the present invention, the bias estimation unit may calculate a ratio of the behavior rate for each position to the behavior rate of the reference advertising space as the bias.

  According to another aspect of the present invention, there is provided an apparatus for estimating a bias, which is an evaluation value of a position where an advertisement is displayed. In order to calculate the number of times of display of the advertisement at the position for each of the advertisement and the position, the number-of-actions information acquisition unit for acquiring the number of times of action for determining the number of actions that is the number of actions that have occurred A display frequency information acquisition unit for acquiring the display frequency information, and a total of the action times determined based on the action frequency information corresponding to the advertisement is determined based on the display frequency information corresponding to the advertisement. An advertisement evaluation unit that calculates an advertisement evaluation value, which is an evaluation value of the advertisement, divided by the total number of display times, and for each of the position and the advertisement, Calculating the action rate obtained by dividing the total number of actions determined based on the action number information to be divided by the total number of display times determined based on the display number information corresponding to the advertisement and the position; Bias for estimating the bias so that the sum of the difference between the advertisement evaluation value of the advertisement multiplied by the bias of the position and the action rate for the advertisement and the position is minimized. An estimation unit.

として、

をＬが最小となるように、ａ，ｂおよびｒ_ｋを推計するようにしてもよい。 Further, in the bias estimation apparatus of the present invention, the bias estimation unit sets the advertisement evaluation value for the advertisement i to x _i , the action rate for the advertisement i at the position k to y _ki , and the position k to the position k. the bias and _{r k,}

As

The so L is minimized, it is also possible to estimate the a, b and r _k.

のＬが最小となるように、ａ，ｂを推計し、（ｂ）式

により、ｋ≠１のすべてのｋについてｒ_ｋを算出し、（ａ）および（ｂ）を複数回繰り返すことにより、前記バイアスｒ_ｋを推定するようにしてもよい。 Also, bias estimation apparatus of the present invention, the bias estimation unit for all of the position k, and initialize the bias r _k to 1, (a) Formula

A and b are estimated so as to minimize L of Eq. (B)

Accordingly, for all k for k ≠ 1 calculates r _k, by repeating a plurality of times (a) and (b), may be estimated the bias r _k.

  According to another aspect of the present invention, there is provided a method for estimating a bias, which is an evaluation value of a position where an advertisement is displayed, wherein the computer displays a viewer of the advertisement for each position where the advertisement can be displayed and the advertisement. Acquiring action number information for determining the number of actions that is the number of times the action has been taken according to the advertisement, acquiring an advertisement evaluation value that is an evaluation value of the advertisement, and according to the advertisement evaluation value , Dividing the advertisement into a plurality of groups, and, for at least one of the groups, for each position, the number of actions determined based on the advertisement belonging to the group and the action count information corresponding to the position And estimating the bias according to the sum of.

  According to another aspect of the present invention, there is provided a method for estimating a bias, which is an evaluation value of a position where an advertisement is displayed. Obtaining the number of actions for determining the number of actions that is the number of times the action has been taken according to the advertisement, and calculating the number of times the advertisement is displayed at the position for each of the advertisement and the position. A step of obtaining display number information; and a sum of the number of actions determined based on the action number information corresponding to the advertisement is calculated based on the number of display times determined based on the display number information corresponding to the advertisement. Calculating an advertisement evaluation value that is an evaluation value of the advertisement by dividing by the total, and corresponding to the advertisement and the position for each of the position and the advertisement Calculating a behavior rate obtained by dividing the total number of behaviors determined based on the behavior number information by the total number of display times determined based on the display number information corresponding to the advertisement and the position; Estimating the bias so that a sum of a difference between a value obtained by multiplying the advertisement evaluation value of the advertisement by the bias of the position and the action rate for the advertisement and the position is minimized; , Will be executed.

  According to another aspect of the present invention, there is provided a program for estimating a bias that is an evaluation value of a position where an advertisement is displayed. A step of obtaining action number information for determining the number of actions, which is the number of times that a viewer has acted according to the advertisement, an advertisement evaluation value that is an evaluation value of the advertisement, and obtaining the advertisement evaluation value Accordingly, the advertisement is divided into a plurality of groups, and at least one of the groups is determined for each position based on the advertisement belonging to the group and the action frequency information corresponding to the position. The step of estimating the bias according to the total number of actions is executed.

  According to another aspect of the present invention, there is provided a program for estimating a bias that is an evaluation value of a position where an advertisement is displayed. A step of obtaining action number information for determining the number of actions that is the number of times a viewer has acted according to the advertisement, and calculating the number of times the advertisement is displayed at the position for each of the advertisement and the position. Obtaining the display frequency information for the display, and determining the total of the action times determined based on the action frequency information corresponding to the advertisement based on the display frequency information corresponding to the advertisement Dividing an advertisement evaluation value that is an evaluation value of the advertisement by dividing by the total number of times, and for each of the position and the advertisement, the advertisement and the position A behavior rate is calculated by dividing the total number of behaviors determined based on the behavior number information corresponding to, by the total number of display times determined based on the display number information corresponding to the advertisement and the position. The bias is estimated so that the sum of the difference between the advertisement evaluation value of the advertisement multiplied by the bias of the position and the action rate for the advertisement and the position is minimized. And the step of performing.

  Other problems and solutions to be disclosed by the present application will be made clear by the embodiments of the invention and the drawings.

  According to the present invention, a position where an advertisement is displayed can be evaluated.

It is a figure which shows an example of the web page 1 displayed in the advertising system of this embodiment. It is a figure which shows the structural example of the advertisement delivery system of this embodiment. 2 is a diagram illustrating a hardware configuration example of a Web server 20. FIG. 3 is a diagram illustrating an example of a software configuration of a Web server 20. FIG. 5 is a diagram illustrating a configuration example of layout information stored in a layout database 252. FIG. It is a figure which shows the structural example of the advertisement information memorize | stored in the advertisement database. 6 is a diagram illustrating a configuration example of bias information stored in a bias database 254. FIG. It is a figure which shows the structural example of the impression log memorize | stored in the impression log memory | storage part 255. FIG. 6 is a diagram illustrating a configuration example of a click log stored in a click log storage unit 256. FIG. It is a figure which shows the flow of the process which estimates a bias. FIG. 4 is a diagram showing a flow of processing performed by a Web server 20. It is a figure which shows the structural example of the advertisement list. It is a figure which shows the flow of the process which performs selection of the display advertisement in step S509, and allocation of a position. It is a figure for demonstrating the allocation of a position. It is a figure which shows the flow of the process which estimates the ratio of the click rate about a reference | standard advertisement as a bias. It is a figure which shows the flow of the process which estimates a bias from the average click rate according to position for every group by a click rate. It is a figure which shows an example of the graph 54 which shows the click rate and the bias for every position. It is a figure explaining the flow of the estimation process of a CTR model. It is a figure which shows the flow of a calculation process of click rate.

== (1) Summary of Invention ==
Hereinafter, an advertisement distribution system according to an embodiment of the present invention will be described. In the present embodiment, an advertisement is assumed to be an advertisement (referred to as a listing advertisement) that is displayed on a Web page according to the content or the like. FIG. 1 is a diagram showing an example of a web page 1 displayed in the advertising system of the present embodiment. The Web page 1 is provided with a plurality of advertisement frames 2 to 4 together with display frames for various information (contents). The advertisement distribution system according to the present embodiment distributes advertisements to be displayed in these advertisement frames 2 to 4, and puts advertisements in information (hereinafter referred to as screen information) for displaying the Web page 1. Is assumed to be transmitted.

  The advertisement distribution system of the present embodiment is intended to increase the advertising effect by advertisement. Examples of metrics that indicate advertising effectiveness include click-through rate (CTR), advertising revenue (CTR x bid amount), CPM (Cost Per Mille), eCPM (effective cost per mille; revenue per 1000 page views) In this embodiment, in order to simplify the description, it is assumed that the advertisement effect is indicated by a click rate (CTR) in the entire page. Then, the advertisement distribution system according to the present embodiment obtains a value (hereinafter referred to as a bias) indicating ease of clicking for each advertisement space, and displays an advertisement with a high predicted click rate in an advertisement space with a large bias. The click rate of the entire Web page 1 is increased.

In the following description, information indicating the advertising space is referred to as “position”. In the present embodiment, it is assumed that the position is a number indicating the number of advertisement frames from the top in the screen information. The screen information is described in HTML, and the advertising space is text (including a tag and its contents; hereinafter referred to as an advertising tag) for displaying an advertisement in the screen information.
== (2) Overall system configuration ==
FIG. 2 is a diagram illustrating a configuration example of the advertisement distribution system according to the present embodiment. The advertisement distribution system according to the present embodiment includes a Web server 20 that is communicably connected to the user terminal 10 via a communication network 30.

  The communication network 30 is, for example, the Internet or a LAN (Local Area Network) constructed by a public telephone line network, a dedicated communication line network, a mobile phone line network, a wireless communication network, Ethernet (registered trademark), or the like.

  The Web server 20 is a computer such as a workstation or a personal computer that provides various contents such as images, moving images, and texts according to HTTP (HyperText Transfer Protocol). The user terminal 10 is a computer such as a personal computer, a tablet computer, a mobile phone terminal, a smartphone, or a PDA (Personal Digital Assistant) that operates a browser that browses various Web pages provided by the Web server 20.

  The user terminal 10 transmits an HTTP request to the Web server 20, receives a Web page described in HTML (HyperText Markup Language) transmitted from the Web server 20 in response to the request, and displays the Web page on the display.

== (3) Hardware Configuration of Web Server 20 ==
FIG. 3 is a diagram illustrating a hardware configuration example of the Web server 20. The Web server 20 includes a CPU 201, a memory 202, a storage device 203, a communication interface 204, an input device 205, and an output device 206. The storage device 203 is, for example, a hard disk drive, a solid state drive, or a flash memory that stores various data and programs. The communication interface 204 is an interface for connecting to the communication network 30, for example, an adapter for connecting to Ethernet (registered trademark), a modem for connecting to a public telephone line network, and a wireless communication device for performing wireless communication A USB (Universal Serial Bus) connector for serial communication, an RS232C connector, or the like. The input device 205 is, for example, a keyboard, a mouse, a touch panel, a button, or a microphone that inputs data. The output device 206 is, for example, a display, a printer, or a speaker that outputs data.

== (4) Software Configuration of Web Server 20 ==
FIG. 4 is a diagram illustrating a software configuration example of the Web server 20. The web server 20 includes an HTTP processing unit 211, an advertisement allocation unit 212, a listing processing unit 213, a position determination unit 214, a web page creation unit 215, a log collection unit 216, a bias estimation unit 217, a web information database 251, a layout database 252, An advertisement database 253, a bias database 254, an impression log storage unit 255, and a click log storage unit 256 are provided.

  The CPU 201 of the Web server 20 stores the functional units of the HTTP processing unit 211, the advertisement allocation unit 212, the listing processing unit 213, the position determination unit 214, the Web page creation unit 215, and the bias estimation unit 217 in the storage device 202. The layout database 252, the advertisement database 253, the bias database 254, the impression log storage unit 255, and the click log storage unit 256 are stored in the memory 202 and the storage included in the Web server 20. This is realized as a part of a storage area provided by the device 203.

== (5) Description of Storage Unit ==
The web information database 251 stores information for creating a web page (hereinafter referred to as web information). The Web information is, for example, a file described in HTML, a program file storing a program executed by a mechanism such as CGI (Common Gateway Interface), an image file, or the like. The Web information database 251 can be realized by, for example, a file system directory or folder provided by an operating system operating on the Web server 20.

  The layout database 252 stores information related to the placement of advertisements on the Web page (hereinafter referred to as layout information). FIG. 5 is a diagram illustrating a configuration example of layout information stored in the layout database 252. The layout information includes information (layout ID) indicating the layout of the Web page in association with a URL (Uniform Resource Locator) pattern indicating the Web page. An advertisement frame is provided at the same position on Web pages having the same layout ID.

  The advertisement database 253 stores information for displaying advertisements (hereinafter referred to as advertisement information). FIG. 6 is a diagram illustrating a configuration example of advertisement information stored in the advertisement database 252. The advertisement information is associated with the advertisement ID that is the identification information of the advertisement, the URL of the advertisement category, the URL of information related to the advertisement (called a landing page), the title displayed as the advertisement, the bid amount, the click rate (CTR), etc. Is included.

In the present embodiment, a link to a predetermined URL (the Web server 20 is specified as the host name, hereinafter referred to as a redirect URL) in which the position indicating the advertisement ID and the advertisement space is specified is set. It is assumed that the title character string is displayed as an advertisement. When the advertisement is clicked, a request including the advertisement ID is transmitted to the web server 20, and a log indicating that the click is performed by the log collection unit 216 described later is collected. Redirected to a URL (for example, a response specifying the URL as a redirect destination is transmitted). Thereby, the user's action (click) according to the advertisement can be detected.
The bid amount is a unit price paid by the advertiser when the advertisement is clicked, and is determined by the advertiser. The click rate is a value (actual click rate) obtained by dividing the number of clicks on the advertisement in the past by the number of times the advertisement was displayed in the past (referred to as the number of impressions). The number of impressions and the number of clicks are managed in an impression log storage unit 255 and a click log storage unit 256 described later. In this embodiment, the actual click rate included in the advertisement information is used as the predicted click rate as it is, but the predicted click rate is calculated based on the actual click rate in consideration of factors other than the actual click rate. You may do it.
In addition, you may make it display an image, a moving image, etc. as an advertisement, or you may make it output an audio | voice, and in that case, URL which shows the data of the image displayed as an advertisement, or a moving image is added, for example Can be.

  The bias database 254 stores information related to bias (hereinafter referred to as bias information). FIG. 7 is a diagram illustrating a configuration example of the bias information stored in the bias database 254. The bias information includes a bias for the advertising space indicated by the position on the Web page of the layout indicated by the layout ID in association with the layout ID and the position. In this embodiment, the click rate at the position “1” is “1.0”, and the ratio of the click rate at the other position to the click rate at the position “1” is registered as a bias. The bias information managed in the bias database 254 is updated by a bias estimation unit 211 described later.

The impression log storage unit 255 stores information (hereinafter referred to as an impression log) indicating that an advertisement is displayed (referred to as impression, which corresponds to display number information of the present invention). FIG. 8 is a diagram illustrating a configuration example of an impression log stored in the impression log storage unit 255. The impression log includes the date and time (delivery date and time) when the advertisement is displayed on the Web page 1, the URL indicating the Web page 1, the position where the advertisement is displayed, and the advertisement ID indicating the displayed advertisement.
In the example of the Web page 1 in FIG. 1, since three advertisements are displayed, when the Web page 1 is browsed once, three impression logs are registered.

  The click log storage unit 256 stores information indicating that the advertisement displayed on the Web page 1 has been clicked (hereinafter referred to as a click log, which corresponds to the action frequency information of the present invention). FIG. 9 is a diagram illustrating a configuration example of the click log stored in the click log storage unit 256. The click log includes an advertisement ID indicating the clicked advertisement, the URL of the Web page 1 on which the advertisement is displayed, the position where the advertisement is displayed, and the date and time when the advertisement was clicked (click date and time).

== (6) Description of Functional Units ==
The HTTP processing unit 211 performs HTTP processing. The HTTP processing unit 212 receives an HTTP request transmitted from the user terminal 10 and performs response processing according to the request.
When a redirect URL is specified in the request, the HTTP processing unit 212 calls a log collection unit 216 described later to register a click log, and then advertises advertisement information corresponding to the advertisement ID included in the request. A response that is read from the database 253 and specifies the URL included in the read advertisement information as a redirect destination is returned to the user terminal 10.
When a URL indicating a Web page other than the redirect URL is specified in the request, the HTTP processing unit 211 calls information for displaying the Web page by calling the Web page creation unit 212 (hereinafter referred to as screen information). The information is, for example, an HTML file.), The image information created by the web page creation unit 212 is responded to the user terminal 10, and the log collection unit 216 is called to register the impression log.

  The advertisement allocation unit 212 selects advertisement candidates to be displayed. The advertisement selection unit 212 reads out a predetermined number (for example, any number such as 10, 50, 100, etc.) of advertisement information from the advertisement database 253 as advertisement candidates. In the present embodiment, in order to simplify the description, the advertisement allocation unit 212 determines a predetermined number of pieces of advertisement information as advertisement candidates in descending order of the predicted click rate with the click rate included in the advertisement information as the predicted click rate. However, for example, advertisement candidates may be determined according to user attributes, or evaluation is performed according to CTR, CPM, eCPM, etc. for each advertisement, and a predetermined number of advertisement information is assigned to advertisement candidates in descending order of evaluation. May be determined as Note that a general advertisement selection process can be used for the process in which the advertisement allocation unit 212 selects a predetermined number of advertisements from among a number of advertisements.

  The listing processing unit 213 determines an advertisement to be displayed (hereinafter referred to as a display advertisement) from the advertisement candidates determined by the advertisement allocation unit 212. In the present embodiment, in order to simplify the description, among the advertisement candidates determined by the advertisement allocation unit 212, n (Web page) in descending order of the value obtained by multiplying the predicted click rate by the bid amount (expected profit value) The predetermined number is determined according to the layout of the advertisement information). The listing processing unit 213, for example, includes the content of the advertisement displayed on the Web page 1 together with the advertisement and the attribute of the user who has accessed the character (a character string described in the advertisement category or advertisement title, The ad may be selected according to the degree of matching (such as the character string described on the landing page), or when the search engine responds with the search result, the content of the ad matches the search keyword You may make it select an advertisement according to a degree. In addition, for the advertisement determination process by the listing processing unit 213, a process used for another general search-linked advertisement (listing advertisement) or content matching advertisement may be adopted.

  The position determination unit 214 determines in which advertisement frame the display advertisement determined by the listing processing unit 213 is displayed. The position determination unit 214 assigns a position to a display advertisement so that an advertisement with a higher predicted click rate is displayed in an advertisement space with a larger bias. For example, the position determination unit 214 may assign a position to the display advertisement so that the total value of the values obtained by multiplying the predicted click rate of the display advertisement by the bias and the bid amount becomes the maximum.

  The web page creation unit 215 creates screen information in response to the request received by the HTTP processing unit 211. The web page creation unit 215 reads web information corresponding to the URL specified in the request from the web information database 251, creates image information based on the read web information, and displays advertisements determined by the listing processing unit 213. However, the advertisement information is added to the screen information so as to be displayed in the advertisement frame indicated by the position assigned by the position determination unit 214. Further, the web page creation unit 215 inserts, for example, a character string in which a link tag is set in the screen information so as to access the redirect URL when the display advertisement is clicked. In addition, about the process in which the web page preparation part 216 produces various contents other than an advertisement, the process by a general web server is used, and description is abbreviate | omitted here.

The log collection unit 216 registers an impression log and a click log.
When the redirect URL is specified in the request, the log collection unit 216 displays the advertisement ID and position specified in the redirect URL, and the referrer included in the request (URL indicating the Web page on which the clicked advertisement was displayed). In addition, a click log in which the current date and time is set is created and registered in the click log storage unit 256. In addition, the log collection unit 216 updates the click rate of the advertisement information corresponding to the advertisement ID. That is, the log collection unit 216 counts the number of click logs and the number of impression logs corresponding to the advertisement ID from the click log storage unit 256 and the impression log storage unit 255, respectively, and calculates the number of click logs as the number of impression logs. The divided click rate is calculated, and the advertisement database 253 is updated so as to set the click rate of the advertisement information corresponding to the advertisement ID. Note that the log collection unit 216 calculates the click rate when registering the click log and updates the advertisement information. However, when the advertisement allocation unit 212 selects an advertisement candidate, the impression log storage unit 255 and The predicted click rate may be calculated with reference to the click log storage unit 256.
When a URL other than the redirect URL is specified in the request, the log collection unit 216 displays the current date and time, the URL specified in the request, and the position determination unit 214 for each display advertisement determined by the listing processing unit 213. An impression log including the position assigned to the advertisement and the advertisement ID indicating the advertisement is created and registered in the impression log storage unit 255.

  The bias estimation unit 217 (corresponding to the advertisement evaluation unit, the advertisement group generation unit, and the bias estimation unit of the present invention) estimates the bias based on the impression log and the click log. As described above, the bias is the ease of clicking for each advertising space, and in the present embodiment, when the advertisement is displayed in the advertising space indicated by the position of “1” (hereinafter referred to as a reference advertising space). This is the ratio of the click rate when an advertisement is displayed on another ad space to the click rate. In the present embodiment, the click rate of the reference advertising space is a value “1”.

== (7) Bias estimation processing (average CTR ratio by position) ==
FIG. 10 is a diagram showing a flow of processing for estimating the position bias. The bias estimation unit 217 performs the following processing for each layout.
The bias estimation unit 217 reads a pattern corresponding to the layout ID of the layout from the layout database 252 (S401), and initializes the position k to 1 (S402).

  The bias estimation unit 217 counts the number of impression logs whose URL matches the above pattern and the position is k from the impression log storage unit 255 (S403), and from the click log storage unit 256, The number of click logs whose URL matches the above pattern and whose position is k is counted as the number of clicks (S404). The bias estimation unit 217 sets a quotient obtained by dividing the number of clicks by the number of impressions as a click rate for the position k (S405).

  The bias estimation unit 217 divides the click rate for the position k by the click rate for the position 1 to obtain a bias (S406). The bias estimation unit 217 associates the position k with the calculated bias and registers them in the bias database 254 (S407). The bias estimation unit 217 increments k (S408), and until k exceeds the number of advertising spaces n in the layout (where n is given for each layout) (S409: NO), step The processing from S403 is repeated.

  By performing the above processing for each layout, the bias for each position is registered in the bias database 254 for each layout.

== (8) Request processing ==
FIG. 11 is a diagram illustrating a flow of request processing performed by the Web server 20.
When receiving an HTTP request from the user terminal 10 (S501), the HTTP processing unit 211 determines whether or not an advertisement has been clicked depending on whether or not a redirect URL is specified in the request (S502).

  When it is determined that the advertisement is clicked (S502: YES), the log collection unit 216 reads advertisement information corresponding to the advertisement ID set in the redirect URL from the advertisement database 252, and the advertisement ID set in the request, A click log including the URL indicating the Web page on which the advertisement indicated by the advertisement ID is displayed, the position set in the request, and the current date and time is created and registered in the click log storage unit 256 (S503). Note that the URL indicating the Web page on which the advertisement was displayed can be obtained by reading the referrer set in the request, for example. Further, for example, the Web server 20 establishes a session for each user terminal 10, records the URL of the Web page 1 previously displayed as the session information, and the URL where the advertisement is displayed as the URL recorded as the session information. Or the URL of the Web page on which the advertisement was displayed may be encoded and set in the redirect URL.

The log collection unit 216 counts the number of click logs corresponding to the advertisement ID set in the redirect URL from the click log storage unit 256 (S504), and from the impression log storage unit 255, the impression corresponding to the advertisement ID. The number of logs is counted (S505), and the click rate is calculated by dividing the number of clicks by the number of impressions (S506). The log collection unit 216 updates the click rate of the advertisement information corresponding to the advertisement ID in the advertisement database 253 to the calculated click rate (S507).
The HTTP processing unit 211 transmits an HTTP response to the user terminal 10 so as to redirect to the URL of the advertisement information (S508).

  On the other hand, if it is determined that the advertisement has not been clicked (S502: NO), a display advertisement is determined and a position is assigned to each display advertisement by the process of FIG. 13 described later (S509). The details of the advertisement information determination and position allocation processing will be described later, and FIG. 12 shows a configuration example of the advertisement list 51 created by this processing. Each record of the advertisement list 51 includes an advertisement ID indicating a display advertisement, a title of the display advertisement, and a position assigned to the display advertisement.

  The web page creation unit 215 creates screen information such as an HTML file for displaying various contents according to the URL set in the request (hereinafter referred to as page URL) (S510). Note that the screen information creation processing in step S510 can employ processing by a general Web server.

  The web page creation unit 215 creates, for each record in the advertisement list 51, a redirect URL in which the advertisement ID and position included in the record are set as arguments (S511), and an advertisement is displayed in the advertisement space corresponding to the position included in the record. As shown, a link tag to a redirect URL containing the title contained in the record as the content (for example, “<a href="http://www.foobar.com/ad?id=xxxx"> title </ a> ") is inserted into the screen information (S512).

  The HTTP processing unit 211 transmits the created screen information to the user terminal 10 as a response to the request (S513). For each record included in the advertisement list 51, the log collection unit 216 creates an impression log including the current date and time, the page URL, and the advertisement ID and position included in the record, and registers them in the impression log storage unit 255 (S514). .

== (9) Determination of display advertisement and allocation of position ==
FIG. 13 is a diagram showing a flow of processing for selecting a display advertisement and assigning a position in step S509.
The advertisement allocation unit 212 extracts a predetermined number m pieces of advertisement information from the advertisement database 253 and sets it as a candidate list (S521). Note that what kind of advertisement information is extracted from the advertisement database 253 can be arbitrarily set. For example, m pieces of advertisement information may be extracted at random, or m pieces of advertisements may be extracted in descending order of predicted click rate. Further, an advertisement that matches the content displayed on the Web page may be extracted, or an advertisement that matches the user's attribute may be extracted.

  The listing processing unit 213 reads a layout ID whose pattern matches the page URL from the layout database 252 (S522), acquires a position and a bias corresponding to the read layout ID from the bias estimation unit 217 (S523), and executes the layout. The number n of advertisement spaces in is acquired (S524). Note that n is given for each layout. The listing processing unit 213 uses the click rate of the advertisement information as the predicted click rate, and extracts n pieces of advertisement information from the candidate list in descending order of the predicted click rate as a candidate list (S525).

  The position determination unit 214 assigns positions so that the higher the predicted click rate, the higher the bias (S526). For example, as shown in FIG. 14, when three advertisements are extracted as the candidate list 52 and the bias is as shown in the list 53, “A0001” having the highest predicted click rate has the highest bias. Position “3” is assigned, “A0003” with the second highest predicted click-through rate is assigned position “1” with the second highest bias, and the last “A0002” advertisement is biased third. The position “2” having a higher value is assigned.

For each piece of advertising information included in the candidate list, the listing processing unit 213 associates the advertising ID and title included in the advertising information with the position assigned by the position determination unit 214 and adds the information to the advertising list. (S527).
As described above, display advertisement selection and position assignment are performed.

== (10) Example of effects ==
As described above, according to the advertisement distribution system of the present embodiment, a bias can be estimated for each advertisement space (position). Depending on the layout of the web page, it is not always easy to click the upper part of the page, but by grasping the bias for each advertisement frame, the position where the web page is easily clicked can be easily grasped. Therefore, for example, it is possible to correct the evaluation of the advertisement in consideration of the bias, and it is possible to more accurately evaluate the advertisement displayed in the advertisement space.

  In the advertisement distribution system according to the present embodiment, the bias is the ratio of the click rate of each ad space based on the click rate of the reference ad space, so that the correction of the predicted click rate corresponding to the ad space is easily calculated. can do. For example, it is possible to easily calculate the correction of the predicted click rate of the advertisement corresponding to the advertisement space by multiplying the predicted click rate of the advertisement by the bias of the advertisement space in which the advertisement is displayed.

  Also, according to the advertisement distribution system of the present embodiment, since the advertisement with the higher predicted click rate is assigned to the advertising space with higher bias, the click rate of the entire Web page can be improved. Therefore, it is possible for the advertiser to improve the advertising effect, and for the publisher of the advertisement (operator of the Web server 20), if the advertising fee is generated by the click, the profit from the advertisement can be improved. It becomes possible.

  In the advertisement distribution system according to the present embodiment, an advertisement frame bias can be estimated for each layout, and an advertisement with a high predicted click rate can be displayed in an advertisement frame with a high bias. Therefore, it is possible to improve the click rate in the entire Web site that distributes a plurality of Web pages with various layouts. Therefore, it is possible to improve the advertising effect on the entire website and improve the profits from the advertisement.

  In the advertisement distribution system according to the present embodiment, an average value of the click rate is calculated for each advertising space (position), and the bias is estimated based on the average click rate. Therefore, a log that can calculate the click rate for each position (in this embodiment, a click log and an impression log are used, but the present invention is not limited to this, and for example, a log that associates the number of clicks with the number of impressions for each position may be used). If there is, the bias can be easily calculated.

== (11) Modification 1: Average CTR ratio by position for the standard advertisement ==
In the bias estimation process shown in FIG. 10, the ratio of the click rate for each position for all advertisements is calculated as a bias. However, the ratio of the click rate for each position is calculated only for a specific advertisement. You may do it.
FIG. 15 is a diagram illustrating a flow of processing for estimating a click rate ratio for a reference advertisement (hereinafter referred to as a reference advertisement) as a bias.

  The bias estimation unit 217 first determines one reference advertisement, sets the advertisement ID of the reference advertisement as the reference advertisement ID (S421), and then, for each layout, a pattern corresponding to the layout ID of the layout, as in FIG. Is read from the layout database 252 (S401), and the position k is initialized to 1 (S402).

  The bias estimation unit 217 counts the number of impressions from the impression log storage unit 255 instead of step S403 by counting the number of impression logs whose URL matches the pattern, the position is k, and the advertisement ID is the reference advertisement ID. (S422). Also, the bias estimation unit 217 counts the number of click logs in which the URL matches the pattern, the position is k, and the advertisement ID is the reference advertisement ID from the click log storage unit 256 instead of step S404. The number of clicks is set (S423). Steps S405 to S409 are the same as those in FIG.

  As described above, how much the click rate for the reference advertisement changes depending on the advertisement space can be calculated by using the ratio of the click rate of the reference advertisement space as 1, which can be used as a bias. In this case, the ratio of the click rate for each advertisement space can be set as a bias based on the logs for the same advertisement. It is known that the click rate varies greatly depending on the ad. By using the same ad, the influence of the ad itself on the click rate is eliminated, and the influence of the ad space on the click rate is evaluated as a correct bias. Can do.

  In addition, you may make it prepare one new advertisement for bias measurement as a standard advertisement, and when there are advertisements distributed in various advertising spaces, the advertisement is set as a standard advertisement. May be.

  When newly preparing a reference advertisement for bias measurement, the advertisement allocation unit 212 and the listing processing unit 213 always determine the reference advertisement as an advertisement candidate and a display advertisement, and the position determination unit 214 randomly determines the reference advertisement. With respect to other advertisements, advertisement spaces other than the advertisement space allocated to the reference advertisement can be allocated in descending order of bias. Thus, the advertisement can be distributed for a predetermined period so that the reference advertisement is displayed in a random advertisement space. By using the reference advertisement, the click rate for each advertisement space can be measured using the same advertisement, so that the bias can be accurately grasped.

  When the advertisements distributed in various advertising spaces are used as reference advertisements, the possibility that the advertisement effect of the reference advertisements themselves will be reduced compared to the case where newly prepared reference advertisements are randomly distributed. can do.

  A plurality of reference advertisements may be set. In this case, in steps S422 and S423, the number of impression logs and click logs in which the URL matches the pattern, the position is k, and the advertisement ID matches one of the selected plurality of reference advertisement IDs is counted. What should I do?

== (12) Modification 2: Average of average CTR ratio by position for each group ==
In the first modification, the ratio of the average click rate for each position with respect to the reference advertisement is set as a bias, but the advertisements are grouped according to the click rate, and the bias for each position is calculated for each group as in the first modification. Alternatively, the bias may be obtained by averaging this for each position. FIG. 16 is a diagram showing a flow of processing for estimating the bias from the average click rate for each position for each group based on the click rate.

  Similarly to FIG. 10, the bias estimation unit 217 reads a pattern corresponding to the layout ID of the layout from the layout database 252 (S401) and initializes the position k to 1 (S402).

  The bias estimation unit 217 sets the minimum click rate (may be 0) in the advertisement database 253 as the click rate x (S441). The click rate is equal to or greater than x from the advertisement database 253, and the click rate is x. An advertisement ID of advertisement information that is less than a value obtained by adding a predetermined step value to is read as a reference advertisement ID (S442). For example, an arbitrary value such as 0.01% or 0.1% may be set as the step value, or a value obtained by subtracting the minimum value from the maximum value of the click rate is divided by a predetermined number of groups. It may be a value.

  The bias estimation unit 217 counts the number of impression logs in which the URL matches the pattern, the position is k, and the advertisement ID is included in the reference advertisement ID from the impression log storage unit 255 instead of step S403 in FIG. And set as the number of impressions (S443). Also, the bias estimation unit 217 counts the number of click logs in which the URL matches the pattern, the position is k, and the advertisement ID is included in the reference advertisement ID from the click log storage unit 256 instead of step S404. To the number of clicks (S444).

  Instead of step 405, the bias estimation unit 217 sets a quotient obtained by dividing the number of clicks by the number of impressions as a click rate (k, x) for each position k and click rate x (S445), and this click rate (k, x ) Is divided by the click rate (1, x) of the position “1” corresponding to the click rate x as a bias B (x) corresponding to the click rate x (S446).

  The bias estimation unit 217 adds the step value to x (S447), and until x exceeds the maximum value of the click rate in the advertisement database 253 (which may be a given maximum value) (S448: NO), step The processing from 442 is repeated.

  When x exceeds the maximum value (S448: YES), the bias estimation unit 217 averages the bias B (x) for each click rate to obtain a bias corresponding to the position k (S449), and steps S407 to S409 are shown in FIG. 10 is the same.

  FIG. 17 is a diagram illustrating an example of a graph 54 showing the click rate x and the bias B (k, x) for each position k. In the graph 54, the bias for each position k is plotted for each click rate x for every 0.01%. An average of the biases for each click rate x for each position k is registered in the bias database 254.

  As described above, the advertisements are grouped according to the click rate, the bias for each position is calculated for each group as in the first modification, and the bias can be estimated by averaging the bias for each position. Since advertisements with similar click rates are divided into groups, the degree of influence of the advertising space on the click rate can be evaluated as a correct bias without worrying about the impact of the click rate due to the advertisement itself.

  In addition, since the number of logs used to determine the bias can be increased compared to the case where the reference advertisement is specified, it becomes possible to improve the reliability of the bias, and a new advertisement is measured for the bias measurement. The bias can be estimated based on the click log and the impression log without delivery.

  Also, when ads related to click logs and impression logs are delivered, if ads with a high predicted click rate are distributed in order from the top (in order of smaller position), an ad with a higher predicted click rate is displayed as the position is smaller. If you simply average the click rate for each position, there will be a bias that the smaller the position, the higher the click rate will be evaluated. By averaging the click rate for each position only for advertisements with similar rates, the bias can be estimated so as not to cause the above-described bias. Therefore, the accuracy of bias can be improved.

== (13) Modification 3: Model Estimation ==
In Modification 2, the ratio of the click rate and the click rate for each position is averaged for each position and calculated as a bias. However, the bias may be calculated by regression analysis. In the third modification, a bias is determined by estimating a statistical model (hereinafter referred to as a CTR model) for a click rate for each position.

ここで、関数ｆは

である。 First, it is assumed that the click rate for each position is the same if the advertisement has the same click rate. When the click rate of the advertisement i is x _i and the click rate of the advertisement i at the position k is y _ki , the click rate y _1i of the position “1” indicating the reference advertisement space is expressed by the following equation (1).

Here, the function f is

It is.

このＣＴＲモデルを実際のクリック率から推計する。なお、ポジション「１」のバイアスｒ_１は１であるものとして推計を行う。 The other position k bias is a ratio to the position 1 bias. Therefore, CTR model bias position k as _{r k,} representing the CTR _{y ki} position k is expressed by the following equation (3).

The CTR model is estimated from the actual click rate. Note that the estimation is performed assuming that the bias r ₁ of the position “1” is 1.

式（４）においてＬが最小となるａ，ｂおよびｒ_ｋを求める。本実施形態では、式（４）についてｆおよびｒを交互に最適化していく。 In this embodiment, the least square method is used for estimation of the CTR model. The objective function to be minimized is given by the following equation (4).

In Equation (4), a, b, and rk that minimize L are _obtained . In the present embodiment, f and r are alternately optimized for Expression (4).

すなわち、広告ごとのクリック率ｘ_ｉについて、重みをｒ_ｋ ^２とし、ポジションごとのクリック率ｙ_ｋｉに１／ｒ_ｋを乗じた上で、ｆを最適化することになる。 First, r is fixed and f is optimized. Equation (4) can be modified as follows.

That is, the CTR _{x i} of each ad, the weight and _r ^{k 2,} after having multiplied by 1 / _{r k} CTR _{y ki} for each position, so that the optimizing f.

Next, f is fixed and r is optimized. Equation (5) is differentiated with respect to _{r k} 0 far away and the following equation is (6).

By repeatedly performing calculation of the formula optimization of f for (5) and Equation (6) alternately, to calculate the final bias r _k. In this embodiment, for all positions k other than 1 to the amount of change in the bias r _k converges becomes less than a predetermined value, performing the calculation of equation (5) Optimization and formula for (6) alternately For example, the calculation process may be terminated when a predetermined number of calculation processes have been performed.

FIG. 18 is a diagram for explaining the flow of CTR model estimation processing. The bias estimation unit 217 performs the following processing for each layout.
The bias estimation unit 217 reads a pattern corresponding to the layout ID of the layout from the layout database 252 (S601), and performs a click rate calculation process shown in FIG. 19 (S602).

Ambiguity estimation unit 217, for all ad i, reads CTR included from the advertisement database 253 the advertisement information, the click rate _{x i} for ad i (S621). For all positions k, the bias estimation unit 217 counts the number of impression logs from the impression log storage unit 255 where the URL matches the above pattern, the position is k, and the advertisement ID indicates the advertisement i. k and the number of impressions P _ki corresponding to the advertisement i are set (S622). The bias estimation unit 217 counts the number of click logs in which the URL matches the above pattern, the position is k, and the advertisement ID indicates the advertisement i from the click log storage unit 256, and corresponds to the position k and the advertisement i. The number of clicks C _{ki to be made} is set as S _ki (S623). The bias estimation unit 217 sets a quotient obtained by dividing the click number C _ki by the impression number P _ki as the click rate y _ki for the position k and the advertisement i (S624).

Returning to Figure 18, the bias estimation section 217 initializes to 1 a bias _{r k} for all positions k (S603).
The bias estimation unit 217 optimizes f with respect to Expression (5) (S604). That is, the bias estimation unit 217, the CTR _{x i} of each advertisement i, the weights and _r ^{k 2,} after having multiplied by 1 / _{r k} CTR _{y ki} for each position k, such that L is minimized Thus, the coefficients a and b of f are estimated. The optimization process is based on a general least square method. In addition, the bias estimation unit 217 optimizes Expression (5) after rounding _xi and _yki for each predetermined step value such as 0.001% or 0.005%.

Ambiguity estimation unit 217, for each position k of 2 or more, calculating a bias _{r k} by equation (6) (S605). That is, the bias estimation unit 217, optimized coefficients a, b to the results gave CTR x _i of each advertisement i to f (x _i) that is set in equation (2), for each ad i and position k The quotient obtained by dividing the sum of the values obtained by multiplying the click rate y _ki by the sum of the squares of f (x _i ) given the click rate x _i of each advertisement i is calculated as the bias r _k .

Ambiguity estimation unit 217, to bias _{r k} converges (S606: NO), repeats the process of step S604～S605. For example, the bias estimation unit 217, the determination processing in the first step S606 determines that it has not converged, is stored in the memory 202 to bias r _k as the previous value _k, in the second and subsequent determination processing, each position for k, it calculates a difference between the previous value _k stored in the r _k and memory 202 which is currently calculated at step S605, the convergence when it becomes within the predetermined threshold for the magnitude all positions k of the difference and the determined and, if not within one threshold of the magnitude of the difference, stores a bias r _k in the memory 202 as the previous value, so as to determined not to converge. It should be noted that it may be determined that the process has converged when the processes of steps S604 and S605 are calculated a predetermined number of times (for example, an arbitrary number such as 5 or 10).

Ambiguity estimation unit 217, the bias _{r k} converges (S606: NO), for each position k, and registers the bias database 254 in association with the position k and the bias _{r k} (S607).

As described above, it is possible to estimate the bias r _k by estimating the CTR model. In Modification 3, since the approximate expression is calculated using the CTR model, the bias can be accurately estimated even if the click rate variation of the displayed advertisement is biased depending on the position.

  For example, in the Web server 20 displaying a high advertisement such as CTR, CPM, eCPM in order from the top, an advertisement with a high click rate is always displayed in the advertisement frame located at the top of the screen, and the advertisement frame located at the bottom of the screen May always display an advertisement with a low click rate. When analyzing past impression logs and click logs accumulated in the web server 20, an advertisement with a high click rate is displayed at the bottom of the screen. It can also be assumed that there are few displayed logs. However, even when the log is biased in this way, a more accurate bias can be estimated by estimating the CTR model and obtaining the bias.

Further, in the modified example 3, f a (x _i) in terms of the linear equations as in Equation (1), bias by repeatedly calculating the equation (5) and (6), to converge the bias r _k since the estimated r _k, it can be compared with the case of performing calculations to optimize directly the objective function of equation (4), to reduce the processing load of the optimization process.

However, the bias estimation unit 217 may directly optimize Equation (4). In this case, it is possible to estimate a more accurate bias r _k.

== (14) Other Modifications ==
In the present embodiment, an advertisement distribution system that uses a bias for advertisement distribution has been described. However, the bias can be easily applied to other systems that evaluate advertisements. For example, when analyzing an advertisement evaluation value (predicted click rate, advertisement revenue, CPM, eCPM, etc.), the evaluation value can be corrected by multiplying the advertisement evaluation value by the bias of the displayed advertising space. In this case, the advertisement can be evaluated with higher accuracy. In addition, when evaluating the performance of a matching engine that performs matching between content and advertisement, the matching engine can be evaluated more accurately by using the corrected evaluation value of the advertisement.

  In the present embodiment, the advertisement distribution system is configured by one Web server 20, but is not limited thereto, and may be configured by a plurality of computers. For example, the HTTP processing unit 211 is realized by the Web server 20, the functional units of the advertisement allocation unit 212 to the bias estimation unit 217 are realized by the application server, and the storage units of the Web information database 251 to the click log storage unit 256 are stored in the database. It can also be configured to be realized by a server.

  Further, for example, the bias estimation processing may be made independent as a bias estimation device different from the Web server 20. The bias estimation device includes a bias estimation unit 217, an advertisement information acquisition unit that acquires advertisement information from the advertisement database 253 of the Web server 20, an impression log acquisition unit that acquires an impression log from the impression log storage unit 255, and a click log storage unit A click log acquisition unit that acquires a click log from 256, and the bias can be estimated by the processing shown in FIG. 10, FIG. 15, FIG. 16, or FIG. In this case, since the user can easily grasp the bias for each advertising space, the bias can be used for various types of analysis. For example, as described above, when the advertisement is evaluated with higher accuracy or when the matching engine is evaluated, the evaluation accuracy can be improved by using a bias.

In addition, for example, an advertisement distribution server that provides a display advertisement to the Web server 20 is provided separately, and the advertisement distribution server includes an advertisement allocation unit 212, a listing processing unit 213, a position determination unit 214, a log bias estimation unit 217, and the like. You may do it.
In this case, in the Web server 20, the Web server 20 that creates the screen information does not select an advertisement, and the advertisements received from the advertisement distribution server are arranged in the upper advertisement frame in descending order of the predicted click rate. If they are arranged, make corrections by multiplying the advertising revenue (predicted click rate x bid amount) and predicted click rate by the bias of the assigned position, and add the corrected ad revenue and predicted click rate to the ad. Can be. As a result, the screen information is created so that the advertisement is correctly displayed on the Web server 20 at a position considering the bias.

  In addition, the advertisement is displayed on a part of the web page. However, the present invention is not limited to this, and a screen on which only the advertisement is displayed may be displayed. May be displayed.

  In this embodiment, the predicted click rate (CTR) of each advertisement is a value obtained by dividing the number of past clicks by the number of impressions. However, the present invention is not limited to this. For example, the advertisement allocation unit 212 and the listing processing unit 213 may calculate the predicted click rate by dividing the number of clicks corresponding to the context by the number of impressions according to the context (the content of the content delivered together with the attribute of the browsing user and the advertisement). In addition, the click rate may be predicted using various statistical methods, and the predicted value may be used as the predicted click rate. In addition, the number of clicks may be predicted based on a click log or the like using various statistical methods, and the number of impressions may be predicted based on an impression log or the like. Further, the predicted click rate for each position may be calculated for each position, and this may be averaged for each position to calculate the predicted click rate for each position.

  Further, in this embodiment, the higher biased advertisement is assigned so that an advertisement with a higher predicted click-through rate is displayed. However, the advertising space with higher bias has an evaluation value of other advertisements such as advertising revenue, CPM, and eCPM. You may make it allocate an advertisement with high. In this case, it is possible to maximize the advertising revenue as the entire Web page.

  In the present embodiment, the advertisement is evaluated and the advertisement space is evaluated using the predicted click rate. However, the predicted click rate may be used instead of the predicted click rate.

  In the present embodiment, the clicked logs are accumulated in the click log. However, the present invention is not limited to this. For example, when an advertisement viewer applies for an e-mail or a viewer applies for a service or When a product is purchased, etc., it is received from the service provider or product seller, and the history of the fact that the viewer has taken some action in response to viewing the advertisement is stored instead of the click log. You may make it do. In this case, for example, the advertisement is displayed as a coupon with a number including the advertisement ID or position, and information on the used coupon is stored as a log. It shall be possible to identify whether or not. Thereby, it becomes possible to evaluate an advertisement and an advertising space from various actions of the viewer other than the click.

  In the present embodiment, the impression log is accumulated when the advertisement is distributed. However, the impression log may be accumulated when actually displayed on the screen of the user terminal 10. In this case, for example, together with the screen information, a program for notifying that the advertisement is displayed on the user terminal 10 (hereinafter referred to as a display notification program) is transmitted to the user terminal 10, and the web page 1 is displayed on the user terminal 10. And the display notification program is executed. When an advertisement is displayed on the screen of the user terminal 10 by executing the display notification program, the advertisement ID of the advertisement displayed on the screen and the position indicating the advertisement space displayed on the screen are transmitted to the Web server 20. Thus, the log collection unit 216 creates and registers an impression log when the advertisement ID and the position are received. Thereby, even if the advertisement is output on a large Web page that does not display all advertisements on the screen at once, the click rate can be calculated based on the number that is surely viewed by the user. Therefore, the reliability of the click rate can be improved.

  Although the present embodiment has been described above, the above embodiment is intended to facilitate understanding of the present invention and is not intended to limit the present invention. The present invention can be changed and improved without departing from the gist thereof, and the present invention includes equivalents thereof.

1 screen 2 advertisement 10 user terminal 20 web server 30 communication network 211 HTTP processing unit 212 advertisement allocation unit 213 listing processing unit 214 position determination unit 215 web page creation unit 216 log collection unit 217 bias estimation unit 251 web information database 252 layout database 253 Advertising database 254 Bias database 255 Impression log storage unit 256 Click log storage unit

Claims (10)

A device for estimating a bias, which is an evaluation value of a position where an advertisement is displayed,
For each position where an advertisement can be displayed and the advertisement, an action number information acquisition unit that acquires action number information for determining the number of actions that is the number of times that a viewer of the advertisement has taken action according to the advertisement;
An advertisement group generation unit that acquires an advertisement evaluation value that is an evaluation value of the advertisement and divides the advertisement into a plurality of groups according to the advertisement evaluation value;
For at least one group, for each position, a bias estimation that estimates the bias according to the total number of actions determined based on the advertisement belonging to the group and the action count information corresponding to the position And
A bias estimation apparatus comprising: The bias estimation apparatus according to claim 1, wherein
For each of the advertisements and the position, a display number information acquisition unit that acquires display number information for calculating the number of times the advertisement is displayed at the position is further provided.
The bias estimation unit, for each of the positions, calculates the total number of actions determined based on the advertisement belonging to the group and the action number information corresponding to the position, and the advertisement belonging to the group and the Estimating the bias according to an action rate that is a value divided by the total number of display times determined based on the display number information corresponding to a position;
A bias estimation apparatus. The bias estimation apparatus according to claim 2, wherein
The bias estimation unit calculates, as the bias, a ratio of the behavior rate for each position to the behavior rate of the advertising space serving as a reference;
A bias estimation apparatus. A device for estimating a bias, which is an evaluation value of a position where an advertisement is displayed,
For each position where an advertisement can be displayed and the advertisement, an action number information acquisition unit that acquires action number information for determining the number of actions that is the number of times that a viewer of the advertisement has taken action according to the advertisement;
For each of the advertisement and the position, a display count information acquisition unit that acquires display count information for calculating the display count of the advertisement at the position;
The total value of the number of actions determined based on the number of times of action corresponding to the advertisement is divided by the total number of times of display determined based on the number of times of display corresponding to the advertisement. An ad evaluation unit that calculates an ad evaluation value of
For each of the position and the advertisement, a total of the number of actions determined based on the advertisement and the action number information corresponding to the position is determined based on the display number information corresponding to the advertisement and the position. And calculating the action rate divided by the total number of display times, and multiplying the advertisement evaluation value of the advertisement by the bias of the position and the difference between the action rate for the advertisement and the position A bias estimator for estimating the bias so that the total sum is minimized;
A bias estimation apparatus comprising: The bias estimation apparatus according to claim 4, wherein
The bias estimation unit, the advertisement evaluation value x _i for the advertisement _i, the action rate y _ki for the advertisement i in the position _k, the bias for the position k and r _k,

As

The so L is minimized, to estimate the a, b and r _k,
A bias estimation apparatus. The bias estimation apparatus according to claim 5, wherein
The bias estimation unit includes:
For all the positions k, initialize the bias r _k to 1;
(A) Formula

A and b are estimated so that L of the
(B) Formula

By, to calculate the _{r k} for all k of k ≠ 1,
(A) and by a plurality of times (b), estimating the bias _{r k,}
A bias estimation apparatus. A method of estimating a bias, which is a rating of where an ad is displayed,
Computer
For each position where an advertisement can be displayed and for each advertisement, a step of obtaining action number information for determining the number of actions that is the number of times a viewer of the advertisement has taken action according to the advertisement;
Obtaining an advertisement evaluation value, which is an evaluation value of the advertisement, and dividing the advertisement into a plurality of groups according to the advertisement evaluation value;
Estimating at least one group for each position according to a total of the number of actions determined based on the advertisement belonging to the group and the action number information corresponding to the position; ,
The bias estimation method characterized by performing. A method of estimating a bias, which is a rating of where an ad is displayed,
Computer
For each position where an advertisement can be displayed and for each advertisement, a step of obtaining action number information for determining the number of actions that is the number of times a viewer of the advertisement has taken action according to the advertisement;
Obtaining display frequency information for calculating the display frequency of the advertisement at the position for each of the advertisement and the position;
The total value of the number of actions determined based on the number of times of action corresponding to the advertisement is divided by the total number of times of display determined based on the number of times of display corresponding to the advertisement. Calculating an ad rating value of
For each of the position and the advertisement, a total of the number of actions determined based on the advertisement and the action number information corresponding to the position is determined based on the display number information corresponding to the advertisement and the position. And calculating the action rate divided by the total number of display times, and multiplying the advertisement evaluation value of the advertisement by the bias of the position and the difference between the action rate for the advertisement and the position Estimating the bias such that the total sum is minimized;
The bias estimation method characterized by performing. A program for estimating a bias, which is an evaluation value of a position where an advertisement is displayed,
On the computer,
For each position where an advertisement can be displayed and for each advertisement, a step of obtaining action number information for determining the number of actions that is the number of times a viewer of the advertisement has taken action according to the advertisement;
Obtaining an advertisement evaluation value, which is an evaluation value of the advertisement, and dividing the advertisement into a plurality of groups according to the advertisement evaluation value;
Estimating at least one group for each position according to a total of the number of actions determined based on the advertisement belonging to the group and the action number information corresponding to the position; ,
A bias estimation method characterized in that is executed. A program for estimating a bias, which is an evaluation value of a position where an advertisement is displayed,
On the computer,
For each position where an advertisement can be displayed and for each advertisement, a step of obtaining action number information for determining the number of actions that is the number of times a viewer of the advertisement has taken action according to the advertisement;
Obtaining display frequency information for calculating the display frequency of the advertisement at the position for each of the advertisement and the position;
The total value of the number of actions determined based on the number of times of action corresponding to the advertisement is divided by the total number of times of display determined based on the number of times of display corresponding to the advertisement. Calculating an ad rating value of
For each of the position and the advertisement, a total of the number of actions determined based on the advertisement and the action number information corresponding to the position is determined based on the display number information corresponding to the advertisement and the position. And calculating the action rate divided by the total number of display times, and multiplying the advertisement evaluation value of the advertisement by the bias of the position and the difference between the action rate for the advertisement and the position Estimating the bias such that the total sum is minimized;
A bias estimation method characterized in that is executed.

Images