JP2019045899A - Determination device, determination method and determination program - Google Patents

Abstract

To determine an appropriate bid unit price according to a budget.SOLUTION: A determination device has an acquisition unit and a determination unit. The acquisition unit acquires a budget spent on advertisement content. The determination unit determines a bid unit price used as reference in which the advertisement content is selected as distribution objects, on the basis of the budget acquired by the acquisition unit and a predicted value related to whether or not a user who has selected the advertisement content in the past takes an action contributing to a profit for an advertiser of the advertisement content. The determination unit determines the bid unit price used as the reference in which the advertisement content is selected as the distribution objects as a higher price, as the predicted value related to whether or not the user who has selected the advertisement content in the past takes the action contributing to the profit for the advertiser of the advertisement content is higher. Moreover, the determination unit determines the bid unit price used as the reference in which the advertisement content is selected as the distribution objects as a lower price, as the predicted value related to whether or not the user who has selected the advertisement content in the past takes the action contributing to the profit for the advertiser of the advertisement content is lower.SELECTED DRAWING: Figure 1

Description

  The present invention relates to a determination apparatus, a determination method, and a determination program.

  In recent years, advertisement distribution via the Internet has been actively performed. For example, advertisement content such as a company or a product is displayed on an advertisement space set in an advertisement medium (web page), and when such advertisement content is clicked, advertisement distribution is performed to make transition to the advertiser's web page. ing.

  In addition, in such ad delivery, a bid (CPC: Cost Per Click), which is a charge amount to be charged to an advertiser, may be used as compensation for posting advertisement content.

  Here, techniques have been proposed to provide information to help determine or optimize a bid.

Japanese Patent Application Publication No. 2013-528866

  However, in the above-described prior art, it is not always possible to determine an appropriate bid price according to the budget. Specifically, the above-described prior art only provides information that helps to determine or optimize the bid price, and it is not always possible to decide an appropriate bid price according to the budget.

  This application is made in view of the above, and an object of the present invention is to provide a determination device, a determination method, and a determination program capable of determining an appropriate bid price according to a budget.

  According to one aspect of the embodiment, it is possible to determine an appropriate bid price according to the budget.

FIG. 1 is a diagram illustrating an example of the determination process according to the embodiment. FIG. 2 is a diagram illustrating an example of the configuration of the determination system according to the embodiment. FIG. 3 is a diagram showing an example of the configuration of the determination apparatus according to the embodiment. FIG. 4 is a diagram illustrating an example of the advertisement information storage unit according to the embodiment. FIG. 5 is a diagram illustrating an example of the advertisement distribution history storage unit according to the embodiment. FIG. 6 is a diagram illustrating an example of a prediction model storage unit according to the embodiment. FIG. 7 is a diagram illustrating an example of the correspondence rule storage unit according to the embodiment. FIG. 8 is a diagram illustrating an example of the determination process according to the embodiment. FIG. 9 is a flowchart illustrating a correspondence rule generation processing procedure according to the embodiment. FIG. 10 is a flowchart of the determination process according to the embodiment. FIG. 11 is a hardware configuration diagram showing an example of a computer that implements the function of the determination device.

  Hereinafter, a determination apparatus, a determination method, and a form for implementing a determination program (hereinafter, referred to as “embodiment”) according to the present application will be described in detail with reference to the drawings. In addition, the determination apparatus, the determination method, and the determination program which concern on this application by this embodiment are not limited. Moreover, the same code | symbol is attached | subjected to the same site | part in the following each embodiment, and the overlapping description is abbreviate | omitted.

[1. Example of decision processing]
First, an example of the determination process according to the embodiment will be described with reference to FIG. FIG. 1 is a diagram illustrating an example of the determination process according to the embodiment. In FIG. 1, the advertisement content AD1 is determined based on the budget to be spent on the advertisement content AD1 in the next month by the determination apparatus 100 and the predicted value as to whether or not the user who clicks the advertisement content AD1 in the past month generates conversion. An example is shown in which a bid price that is a standard for selecting a delivery target is determined. Note that conversion is an action such as a product purchase or a document request, and refers to an action that occurs in conjunction with an action of clicking advertisement content.

  In addition, if it is an action in which the user selects advertisement content, the action is not limited to clicking on the advertisement content. For example, when the user uses a smartphone, the action of the user selecting the advertisement content corresponds to an arbitrary operation such as a tap being performed.

  The determination apparatus 100 determines a bid serving as a standard for selecting advertisement content as a distribution target, based on a budget spent on the advertisement content and a predicted value as to whether a user who clicks the advertisement content in the past causes conversion. It is an information processing apparatus.

  Hereinafter, the predicted value as to whether or not the user causes conversion is referred to as a conversion predicted value (p). The conversion prediction value (p) is a numerical value indicating the degree to which the user makes a conversion. Further, the conversion predicted value (p) is a numerical value of 0 to 1, and indicates that the larger the number, the higher the degree of conversion, and the smaller the numerical value, the lower the degree of conversion.

  Also, the result of whether the user has made a conversion is referred to as the conversion result (y). Then, when the user makes a conversion, the conversion result (y) value is set to 1. Also, if the user has not made conversion, the conversion result (y) value is set to 0.

  Hereinafter, an example of the determination process will be described with reference to FIG. In the example illustrated in FIG. 1, the determination apparatus 100 determines a higher bid price for a user who is likely to cause conversion (step S1). Specifically, the bid value corresponding to the conversion prediction value (p) of the user is determined to be higher as the value of the conversion prediction value (p) of the user who clicks on the advertisement content AD1 in the past one month is larger. In addition, when a user who clicks on the advertisement content AD1 in the past month has caused conversion, a bid corresponding to the conversion prediction value (p) of the user is determined high.

  For example, since the value of the conversion prediction value (p) is 0.8 and the value of the conversion prediction value (p) is large, the user U1 can be said to be a user who easily causes conversion. Further, the user U1 is a user who has caused conversion because the value of the conversion result (y) is one. Thus, based on the conversion forecast value of the user U1 and the conversion performance, the user whose conversion forecast value (p) has the same value (0.8) as the conversion forecast value of the user U1 has a probability of causing conversion. It is considered high. Therefore, when the conversion forecast value (p) of the user targeted for advertisement delivery is 0.8, the determination apparatus 100 determines the bid price to be high. The user U2 is also similar to the user U1.

  In addition, the determination apparatus 100 determines a bid lower for a user who is unlikely to cause conversion (step S2). Specifically, the determination device 100 determines the bid corresponding to the conversion forecast value (p) of the user as the conversion forecast value (p) of the user who clicks on the advertisement content AD1 in the past month is smaller. Decide low. In addition, when the user who clicks on the advertisement content AD1 in the past month has not caused conversion, the bid corresponding to the conversion prediction value (p) of the user is determined to be low.

  For example, since the value of the conversion prediction value (p) is 0.1 and the value of the conversion prediction value (p) is small, the user U3 can be said to be a user who is hard to cause conversion. Further, the user U3 is a user who has not caused conversion because the value of the conversion result (y) is zero. Thus, based on the conversion forecast value of the user U3 and the conversion performance, the user whose conversion forecast value (p) has the same value (0.1) as the conversion forecast value of the user U3 has a probability of causing conversion. It is considered low. Therefore, when the conversion forecast value (p) of the user targeted for advertisement delivery is 0.1, the determination apparatus 100 determines the bid price to be low. The user U4 to the user U8 are the same as the user U3.

  Then, the determination apparatus 100 determines a bid price so that the sum of bid prices matches the budget spent on the advertisement content AD1 in the next one month (step S3). Specifically, the determination apparatus 100 assumes that the total number of users who click on the advertising content AD1 in the next one month is the same as the total number 8 of users who have clicked on the advertising content AD1 in the past one month. Then, the determination apparatus 100 determines a bid price so that the sum of bid prices of eight users who click on the advertisement content AD1 in the next one month matches the budget spent on the advertisement content AD1 in the next one month.

  As described above, the determination apparatus 100 acquires a budget to spend on the advertising content AD1 in the next month. In addition, the determination apparatus 100 determines whether the advertisement content AD1 is based on a budget to be spent on the advertisement content AD1 in the next month and a conversion predicted value as to whether a user who clicks the advertisement content AD1 in the last month generates a conversion. Determine a standard bid to be selected for distribution.

  Thus, the determining apparatus 100 determines the bid by using the budget instead of using the target CPA, for example, the target CPA is too low, so that clicks leading to conversion are missed and the total number of conversions decreases. You can use your budget effectively to get conversions while avoiding problems such as Therefore, the determination device 100 can determine an appropriate bid according to the budget.

  In FIG. 1, the advertisement content AD1 is to be distributed based on the budget that the determination apparatus 100 spends on the advertisement content AD1 in the next one month and the conversion prediction value of the user who clicked on the advertisement content AD1 in the last one month. Although the example which determines the standard bid used as the standard selected is shown, the length of the period which sets a budget may not be one month. Also, the past one month may not be the latest past one month. For example, the number of users calculated by dividing the number of users who clicked on the advertisement content AD1 in the past one year by 12 months is the predicted number of users who click on the advertisement content AD1 in the next one month. It may be the number of users who have clicked on the advertising content AD1.

[2. Configuration of decision system]
Next, the configuration of the determination system 1 including the determination device 100 according to the embodiment will be described with reference to FIG. FIG. 2 is a diagram showing an example of the configuration of the determination system 1 according to the embodiment. As illustrated in FIG. 2, the determination system 1 according to the embodiment includes the user terminal 10, the access log server 20, the advertiser terminal 30, and the advertisement distribution device (determination device) 100. These various devices are communicably connected by wire or wireless via a network N (for example, the Internet). The determination system 1 illustrated in FIG. 2 may include a plurality of user terminals 10, a plurality of access log servers 20, and a plurality of advertiser terminals 30. Further, in the present embodiment, the access log server 20 and the advertisement distribution device (determination device) 100 are managed by the business operator T1.

  The user terminal 10 is a terminal device used by a user. For example, the user terminal 10 is a tablet type terminal, a PC (Personal Computer), a mobile phone, a PDA (Personal Digital Assistant), or the like. For example, the user terminal 10 acquires a web page from a predetermined web server by accessing a predetermined web server, and displays the acquired web page on a display screen.

  In addition, when the advertising space is included in the web page, the user terminal 10 acquires the advertisement content from the advertisement distribution device (determination device) 100 by accessing the advertisement distribution device (determination device) 100 and acquires the advertisement content. Display advertising content on such ad inventory.

  The access log server 20 holds access logs such as search queries of users and visit pages of users. Further, the access log server 20 acquires information on the purchase history of the user using the shopping service from a predetermined server device that provides the shopping service. Specifically, the access log server 20 performs an action (for example, purchase of an advertisement-targeted item for advertisement content, etc.) in which a user using the shopping service leads to a profit for an advertiser from a predetermined server device that provides the shopping service. Get information on whether or not you

  Advertiser terminal 30 is a terminal device used by an advertiser. For example, the advertiser terminal 30 is a tablet type terminal, a PC, a mobile phone, a PDA or the like. Further, the advertiser terminal 30 submits advertisement content to the advertisement distribution device (determination device) 100 according to the operation of the advertiser.

  For example, the advertiser terminal 30 submits advertisement content corresponding to a still image, a moving image, text data, etc. to the advertisement distribution device (determination device) 100. Also, for example, when the advertising content is selected (for example, clicked or tapped), the advertising terminal 30 distributes the advertising content corresponding to the URL (Uniform Resource Locator) of the transition destination content to be transited Device) 100 may be submitted.

  Furthermore, the advertiser terminal 30 transmits the budget spent on the advertisement content to the advertisement distribution device (determination device) 100 according to the operation of the advertiser.

  The advertisement distribution device (determination device) 100 is a server device that performs the determination process described in FIG. Further, the advertisement distribution device (determination device) 100 is a server device that distributes the advertisement content submitted from the advertiser terminal 30.

  The advertisement distribution device (determination device) 100 holds information on the user who has distributed the advertisement content. Specifically, the advertisement distribution device (determination device) 100 determines whether the user who has distributed the advertisement content clicks the advertisement content or whether the user who has distributed the advertisement content has generated a conversion. keeping.

  Although FIG. 2 shows an example in which the determination apparatus 100 acquires a user's access log from the access log server 20 that collects and holds access logs as a configuration example of the determination system 1, a search server device or a shopping server The determination apparatus 100 may acquire the access log of the user from each server device such as a device.

  Although FIG. 2 shows an example in which the determination apparatus 100 distributes advertisement content as an example of the configuration of the determination system 1, the ad server that distributes advertisement content and the determination apparatus 100 may be separate apparatuses.

  An example will be described in which the add server and the determination device 100 are separate devices. Specifically, the ad server accepts submission of advertising content from the advertiser terminal 30, and acquires a budget to spend on advertising content. Then, when the ad server receives submission of the advertisement content from the advertiser terminal 30, the ad server requests the determination device 100 to generate, for each of the received advertisement content, a correspondence rule that causes a bid price to correspond to the conversion predicted value. For example, when the ad server receives the submission of the advertisement content AD1 from the advertiser terminal 30, the ad server requests the determination apparatus 100 to generate the correspondence rule C1 of the advertisement content AD1.

  When the determining apparatus 100 receives a request for generating the correspondence rule C1 of the advertising content AD1 from the ad server, the determining apparatus 100 calculates a conversion prediction value of the user who has clicked on the advertising content AD1 in the past one month. In addition, from the ad server, a user who clicks on the advertisement content AD1 in the past one month acquires a record of whether or not conversion has occurred.

  Then, the determination apparatus 100 clicks the advertisement content AD1 in the past one month and the result (y) of whether the user who clicks on the advertisement content AD1 in the past one month has caused conversion under the predetermined constraint conditions. Based on the predicted value (p) as to whether or not the converted user causes conversion, an optimization problem is solved for finding a solution that minimizes the objective function (f).

  Then, based on the solution obtained by solving the optimization problem, the determination apparatus 100 generates a correspondence rule C1 in which the conversion bid value corresponds to a bid price serving as a reference for selecting the advertisement content AD1 as a distribution target.

  When the ad server receives a distribution request for advertisement content from the user terminal 10, the ad server selects advertisement content as a distribution candidate. Then, the ad server requests the determination apparatus 100 to determine a bid price for each piece of advertisement content as a distribution candidate.

  When the determining apparatus 100 receives a bid determination request from the ad server, the determining apparatus 100 inputs an access log such as a search query or a visiting page of a user targeted for advertisement delivery into a prediction model generated for each advertisement content of the delivery candidate. Then, the determination apparatus 100 calculates the conversion prediction value of the user to be distributed for each of the advertisement contents of the distribution candidate. For example, when the advertisement content AD1 is included in the advertisement content of the distribution candidate, the determination apparatus 100 inputs an access log such as a search query or a visit page of the user targeted for the advertisement distribution into the prediction model M1. The conversion prediction value of the user to be distributed regarding the content AD1 is calculated.

  Subsequently, the determination apparatus 100 selects the correspondence rule of the advertisement content of the distribution candidate. For example, when the advertisement content AD1 is included in the advertisement content of the delivery candidate, the determination apparatus 100 selects the correspondence rule C1 of the advertisement content AD1.

  Subsequently, the determination apparatus 100 determines a bid price for each piece of advertisement content of the distribution candidate based on the selected correspondence rule.

  Subsequently, the determining apparatus 100 transmits the bid price determined for each of the advertisement contents of the distribution candidate to the ad server.

  The ad server receives, from the determination device 100, a bid price determined for each piece of advertisement content of a delivery candidate. Then, the ad server determines, based on the bid price received from the determination apparatus 100, the advertisement content to be distributed from among the advertisement content of the distribution candidate. Then, the ad server distributes the determined advertisement content to the user terminal 10.

[3. Configuration of determination device]
Next, the advertisement distribution device (determination device) 100 according to the embodiment will be described with reference to FIG. FIG. 3 is a diagram showing a configuration example of the advertisement distribution device (determination device) 100 according to the embodiment. As shown in FIG. 3, the advertisement distribution device (determination device) 100 includes a communication unit 110, a storage unit 120, and a control unit 130.

(Communication unit 110)
The communication unit 110 is realized by, for example, a network interface card (NIC). The communication unit 110 is connected to the network N in a wired or wireless manner, and transmits and receives information between the user terminal 10 and the access log server 20, for example.

(Storage unit 120)
The storage unit 120 is realized by, for example, a semiconductor memory device such as a random access memory (RAM) or a flash memory, or a storage device such as a hard disk or an optical disk. As illustrated in FIG. 3, the storage unit 120 includes an advertisement information storage unit 121, an advertisement distribution history storage unit 122, a prediction model storage unit 123, and a correspondence rule storage unit 124.

(Advertising information storage unit 121)
The advertisement information storage unit 121 stores the advertisement content received from the advertiser terminal 30 and various information regarding the budget. FIG. 4 shows an example of the advertisement information storage unit 121 according to the embodiment. Specifically, FIG. 4 illustrates an example of the advertisement information storage unit 121. In the example illustrated in FIG. 4, the advertisement information storage unit 121 has items such as “advertiser ID”, “advertisement content ID”, and “budget”.

  The “advertiser ID” indicates identification information for identifying an advertiser of the advertiser terminal 30.

  The “advertising content ID” indicates identification information for identifying the advertising content received from the advertiser terminal 30.

  “Budget (yen)” indicates a budget to spend on advertisement content in the next month received from the advertiser terminal 30.

  In the example illustrated in FIG. 4, the first record indicates that the advertisement content identified by the advertisement content ID “AD1” has been received from the advertiser identified by the advertiser ID “C1”. Also, it shows that the budget spent on the advertising content identified by the advertising content ID “AD1” in the next one month is “150,000” yen.

(Ad distribution history storage unit 122)
The advertisement delivery history storage unit 122 stores various pieces of information regarding the delivery history of advertisement content. FIG. 5 illustrates an example of the advertisement distribution history storage unit 122 according to the embodiment. Specifically, FIG. 5 illustrates an example of the advertisement distribution history storage unit 122. In the example illustrated in FIG. 5, the advertisement distribution history storage unit 122 has items such as “advertisement content ID”, “user ID”, “presence or absence of conversion”, and “conversion predicted value”.

  "Advertising content ID" indicates identification information for identifying advertising content delivered in the past month.

  The “user ID” indicates identification information for identifying a user who has clicked on the advertisement content in the past month.

  "Conversion" indicates information on whether a user who clicks on ad content in the past month has made a conversion. In the example shown in FIG. 5, 1 indicates that conversion has occurred, and 0 indicates that conversion has not occurred.

  The “conversion forecast value” indicates a forecast value as to whether or not the user causes a conversion. Note that the conversion prediction value is a numerical value of 0 to 1, and the larger the numerical value, the higher the degree of conversion, and the smaller the numerical value, the lower the degree of conversion.

  In the example shown in FIG. 5, in the first record, the advertisement content (advertisement content AD1) identified by the advertisement content ID “AD1” in the past one month is distributed, and the user (user identified by the user ID “U1” is U1) indicates that the advertisement content AD1 has been clicked. In addition, the presence or absence of conversion being "1" indicates that the user U1 clicks on the advertising content AD1 to cause conversion. Then, that the conversion prediction value is "0.8" indicates that the prediction value regarding whether or not the user U1 causes conversion is "0.8".

(Prediction model storage unit 123)
The prediction model storage unit 123 stores various pieces of information regarding a prediction model for calculating a conversion prediction value of the user. FIG. 6 shows an example of the prediction model storage unit 123 according to the embodiment. Specifically, FIG. 6 illustrates an example of the prediction model storage unit 123. In the example illustrated in FIG. 6, the prediction model storage unit 123 has items such as “advertisement content ID” and “prediction model ID”.

  "Advertising content ID" indicates identification information for identifying advertising content.

  The “prediction model ID” indicates identification information for identifying a prediction model for which a user's conversion prediction value is calculated. The prediction model is a prediction model that calculates a conversion prediction value of a user from an access log such as a search query or a visit page of a predetermined user. The prediction model is generated for each piece of advertising content.

  In the example illustrated in FIG. 6, the first record indicates that a prediction model identified by the prediction model ID “M1” is stored for the advertisement content identified by the advertisement content ID “AD1”.

(Correspondence rule storage unit 124)
The correspondence rule storage unit 124 stores various pieces of information on a correspondence rule that associates a bid price serving as a standard for selecting advertisement content as a distribution target with the conversion prediction value. FIG. 7 shows an example of the correspondence rule storage unit 124 according to the embodiment. Specifically, FIG. 7 illustrates an example of the correspondence rule storage unit 124. In the example illustrated in FIG. 6, the correspondence rule storage unit 124 has items such as "advertisement content ID" and "corresponding rule ID".

  "Advertising content ID" indicates identification information for identifying advertising content.

  “Correspondence rule ID” indicates identification information for identifying a correspondence rule that associates a bid price serving as a standard for selecting advertisement content as a distribution target with the conversion prediction value.

  In the example illustrated in FIG. 7, the first record indicates that the correspondence rule identified by the correspondence rule ID “C1” is stored for the advertisement content identified by the advertisement content ID “AD1”.

(Control unit 130)
Returning to the explanation of FIG. 3, the control unit 130 is a controller, and for example, a storage device inside the advertisement distribution device (determination device) 100 by a central processing unit (CPU) or a micro processing unit (MPU). The various programs (corresponding to an example of a generation program) stored in the computer are realized by executing the RAM as a work area. The control unit 130 is a controller, and is realized by, for example, an integrated circuit such as an application specific integrated circuit (ASIC) or a field programmable gate array (FPGA).

  Further, the control unit 130 performs processing other than the output layer with respect to an access log such as a search query or a visit page of a predetermined user input to the input layer by information processing according to the prediction model stored in the prediction model storage unit 123. The conversion prediction value as to whether or not a predetermined user causes conversion is output from the output layer by performing an operation based on the first element and the weight of the first element with each element belonging to each layer of the first layer as the first element. Let the computer work.

  As shown in FIG. 3, the control unit 130 includes an acquisition unit 131, a determination unit 132, and a distribution unit 133, and implements or executes the action of the information processing described below. Note that the internal configuration of the control unit 130 is not limited to the configuration illustrated in FIG. 3, and may be another configuration as long as it performs the information processing described later.

(Acquisition unit 131)
The acquisition unit 131 acquires a budget to spend on advertisement content. Specifically, the acquisition unit 131 receives submission of advertising content from the advertiser terminal 30, and acquires a budget to spend on advertising content. For example, the acquisition unit 131 receives submission of the advertising content AD1 from the advertiser terminal 30, and acquires a budget to spend on the advertising content AD1 in the next one month.

(Determination unit 132)
The determination unit 132 performs the following three processes. First, the determination unit 132 generates a prediction model for calculating the conversion prediction value (p) of the user. Next, the determination unit 132 generates a correspondence rule (C) in which the bid price corresponds to the conversion prediction value (p). Finally, upon receiving a distribution request for advertisement content from the user, the determination unit 132 determines the bid price of the user targeted for advertisement distribution.

  First, the determination unit 132 generates a prediction model for calculating the conversion prediction value (p) of the user.

  First, when the acquisition unit 131 receives submission of advertisement content, the determination unit 132 generates a prediction model for calculating a conversion prediction value of the user. When generating the prediction model, the determination unit 132 stores the generated prediction model in the prediction model storage unit 123.

  Specifically, when the acquisition unit 131 receives submission of advertisement content from the advertiser terminal 30, the determination unit 132 accesses the access log server 20 for an access log such as a search query or a visit page of a user who has not clicked on the advertisement content. Get from

  Subsequently, the determination unit 132 acquires, from the access log server 20, information on whether or not the user who has not clicked on the advertisement content has taken an action (such as purchase of a product) that leads to a profit for the advertiser of the advertisement content.

  Subsequently, using the information on the user who has not clicked on the advertisement content, the determination unit 132 uses the access log such as a search query or a visit page of the predetermined user to conduct the action for the predetermined user to benefit the advertiser of the advertisement content. Generate a prediction model that calculates conversion prediction values regarding whether to cause or not.

  The determination unit 132 may generate the prediction model using any learning algorithm. For example, the determination unit 132 generates a prediction model using a learning algorithm such as a neural network, a support vector machine, clustering, and reinforcement learning. As an example, when the determination unit 132 generates a prediction model using a neural network, the prediction model includes an input layer including one or more neurons, an intermediate layer including one or more neurons, and an output including one or more neurons. And a layer.

  The prediction model is an input layer to which an access log such as a predetermined user's search query or visit page is input, an output layer, and any layer from the input layer to the output layer and belongs to a layer other than the output layer. According to an access log such as a predetermined user's search query or a visit page, which includes one element and a second element whose value is calculated based on the first element and the weight of the first element. Then, the computer is functioned to output, from the output layer, a conversion prediction value as to whether or not a predetermined user takes action that leads to a benefit to the advertiser of the advertising content.

  The prediction model is an input layer to which an access log such as a predetermined user's search query or visit page is input, an output layer, and any layer from the input layer to the output layer and belongs to a layer other than the output layer. For an access log such as a search query or a visit page of a predetermined user input to the input layer, including one element and a second element whose value is calculated based on the first element and the weight of the first element By performing an operation based on the first element and the weight of the first element with each element belonging to each layer other than the output layer as a first element, a predetermined user takes action leading to a benefit to the advertiser of the advertising content The computer functions to output the conversion prediction value regarding whether or not from the output layer.

Here, it is assumed that the prediction model is realized by a regression model represented by “y = a ₁ * x ₁ + a ₂ * x ₂ +... + A _i * x _i ”. In this case, the first element included in the prediction model corresponds to input data (x _i ) such as x ₁ and x ₂ . Also, the weight of the first element corresponds to the coefficient a _i corresponding to x _i . Here, the regression model can be regarded as a simple perceptron having an input layer and an output layer. When each model is regarded as a simple perceptron, the first element can correspond to any node of the input layer, and the second element can be regarded as a node of the output layer.

  In addition, it is assumed that the prediction model is realized by a neural network having one or more intermediate layers, such as DNN (Deep Neural Network). In this case, the first element included in the prediction model corresponds to any node possessed by the input layer or the intermediate layer. Also, the second element corresponds to the next-stage node that is a node to which a value is transmitted from the node corresponding to the first element. Also, the weight of the first element corresponds to the connection coefficient, which is a weight considered for the value transmitted from the node corresponding to the first element to the node corresponding to the second element.

  The advertisement delivery device (determination device) 100 calculates a conversion prediction value using a prediction model having an arbitrary structure, such as the above-described regression model or neural network. Specifically, the prediction model is a conversion prediction as to whether or not a given user takes an action leading to an advertisement content advertiser if a given user's search query or an access log such as a visit page is input. The coefficients are set to output a value. The advertisement delivery device (determination device) 100 uses such a prediction model to calculate a conversion prediction value.

  In the above example, when the prediction model is an access log such as a search query or a visit page of a predetermined user, conversion as to whether or not the predetermined user takes action leading to a benefit to the advertiser of the advertisement content The example which is a prediction model (it is set as model M1X) which outputs a predicted value was shown. However, the prediction model according to the embodiment may be a model generated based on a result obtained by repeating input and output of data to the model M1X. For example, the prediction model takes an access log such as a search query or a visit page of a predetermined user as an input, and a conversion prediction value as to whether or not the predetermined user output by the model M1X takes an action leading to a profit for the advertisement content advertiser It may be a model (model M1Y) that has been trained so as to be an output. Alternatively, the prediction model may be a model learned with an access log such as a predetermined user search query or a visit page as an input and an output value of the model M1Y as an output.

  In addition, when the advertisement distribution device (determination device) 100 performs calculation processing using GAN (Generative Adversalial Networks), the prediction model may be a model that constitutes a part of GAN.

  Next, the determination unit 132 generates a correspondence rule (C) in which the bid price corresponds to the conversion prediction value (p).

  First, the determination unit 132 is based on the budget acquired by the acquisition unit 131 and a predicted value as to whether or not the user who selected the advertisement content AD1 in the past takes action leading to a benefit for the advertiser of the advertisement content AD1. A bid price serving as a standard for selecting the advertisement content AD1 as a distribution target is determined. Specifically, the determination unit 132 uses the budget to be spent on the advertisement content AD1 acquired by the acquisition unit 131 in the next month and the conversion prediction value (p) of the user who selected the advertisement content AD1 in the last month. Based on the above, a bid price serving as a standard for selecting the advertisement content AD1 as a distribution target is determined.

  Further, the determination unit 132 selects the advertisement content AD1 as the distribution target as the value of the prediction value regarding whether the user who selected the advertisement content AD1 in the past takes an action leading to a profit for the advertiser of the advertisement content AD1 is larger. Higher bids. Further, the determination unit 132 selects the advertisement content AD1 as the distribution target as the value of the prediction value regarding whether the user who selected the advertisement content AD1 in the past takes action leading to a profit for the advertiser of the advertisement content AD1 is smaller. Lower the underlying bid price. Specifically, the determination unit 132 determines the bid price to be higher as the value of the conversion prediction value (p) of the user who selected the advertisement content AD1 in the past one month is larger. In addition, the determination unit 132 determines the bid price to be lower as the value of the conversion prediction value (p) of the user who selected the advertising content AD1 in the past one month is smaller.

  In addition, the determination unit 132 determines whether the user who has selected the advertisement content AD1 in the past has a result of conversion or not and the user who has selected the advertisement content AD1 in the past takes action to be profitable for the advertiser of the advertisement content AD1. Based on the prediction value regarding whether the advertisement content AD1 is to be distributed or not, a bid price serving as a standard for selecting the advertisement content AD1 as a distribution target is determined. For example, the determination unit 132 determines whether the user who has selected the advertisement content AD1 in the past month has made a conversion (y) and whether the user who has selected the advertisement content AD1 in the last month has a conversion or not Determine a bid based on the prediction value (p) regarding

  In addition, when the user who has selected the advertisement content AD1 in the past is causing conversion, the determination unit 132 determines a bid price corresponding to the predicted value of the user to be high. Specifically, when the user who has selected the advertisement content AD1 in the past month has caused conversion (y = 1), the determination unit 132 bids for the conversion prediction value (p) of the user. Determine the unit price high.

  In addition, when the user who has selected the advertisement content AD1 in the past has not caused conversion, the determination unit 132 determines the bid price corresponding to the predicted value of the user to be low. Specifically, if the user who selected the advertisement content AD1 in the past one month has not caused conversion (y = 0), the determination unit 132 bids for the conversion prediction value (p) of the user. Decide the unit price lower.

  In addition, the determination unit 132 determines a solution that minimizes an objective function having as an independent variable a predicted value as to whether or not the user who selected the advertisement content AD1 in the past takes action leading to a benefit for the advertiser of the advertisement content AD1. Solve the optimization problem. Then, based on the solution obtained by solving the optimization problem, the determination unit 132 generates a correspondence rule that makes the bid price correspond to the predicted value.

  Specifically, the determination unit 132 determines the optimization problem for finding a solution that minimizes the objective function (f) whose independent variable is the conversion prediction value (p) of the user who selected the advertisement content AD1 in the past one month. solve. Then, based on the solution obtained by solving the optimization problem, the determination unit 132 generates a correspondence rule (C) in which the bid price is made to correspond to the conversion prediction value (p).

  In addition, the determination unit 132 selects the advertisement content AD1 as a distribution target based on the correspondence rule in which the bid price corresponds to the predicted value regarding whether the user takes action leading to a profit for the advertiser of the advertisement content AD1. Determine a baseline bid. Specifically, the determination unit 132 determines a bid price based on a corresponding rule (C) in which a bid price is associated with the conversion predicted value (p).

  In addition, the determination unit 132 solves the optimization problem with the constraint that the sum of the values of the objective function matches the budget acquired by the acquisition unit 131. Specifically, the determination unit 132 solves the optimization problem under the constraint that the sum of the values of the objective function (f) matches the budget spent on the advertisement content AD1 in the next one month.

  In addition, the determination unit 132 solves the optimization problem under the constraint that the value of the objective function does not exceed a predetermined threshold. Further, the determination unit 132 solves the optimization problem under the constraint that the value of the objective function does not fall below a predetermined threshold. Specifically, the determination unit 132 solves the optimization problem under the constraint that the value of the objective function (f) does not exceed the predetermined threshold b. In addition, the determination unit 132 solves the optimization problem under the constraint that the value of the objective function (f) does not fall below the predetermined threshold value a.

  Further, the determination unit 132 solves the optimization problem on the condition that the value of the objective function (f) monotonously increases in accordance with the conversion prediction value (p).

  Hereinafter, an example of the determination process will be specifically described with reference to FIG. In the example illustrated in FIG. 8, under the constraint that the sum of the values of the objective function (f) matches the acquired budget, the determination unit 132 converts the user who clicks on the advertising content AD1 in the past month. The solution that minimizes the objective function (f) is based on the performance (y) of whether or not it has occurred and the predicted value (p) as to whether the user who clicks on the advertising content AD1 in the past month has caused conversion The desired optimization problem is solved (step S11).

  Then, since the value of the conversion prediction value (p) is 0.8 and the value of the conversion prediction value (p) is large, the user U1 can be said to be a user who easily causes conversion. Further, the user U1 is a user who has caused conversion because the value of the conversion result (y) is one. Thus, based on the conversion forecast value of the user U1 and the conversion performance, the user whose conversion forecast value (p) has the same value (0.8) as the conversion forecast value of the user U1 has a probability of causing conversion. It is considered high. Therefore, when the conversion prediction value (p) of the user targeted for advertisement delivery is 0.8, the determination unit 132 determines the bid price to be high. The user U2 is also similar to the user U1.

  Therefore, the determination unit 132 solves an optimization problem in which the value of the objective function (f) is increased as the value of the conversion prediction value (p) is increased. In addition, the determination unit 132 increases the value of the objective function (f) corresponding to the conversion prediction value (p) for the conversion prediction value (p) of the user whose conversion performance (y) value is 1. Solve the optimal optimization problem. Then, part of the optimization problem is expressed by the following equation (1).

  Further, since the value of the conversion prediction value (p) is 0.1 and the value of the conversion prediction value (p) is small, the user U3 can be said to be a user who is hard to cause conversion. Further, the user U3 is a user who has not caused conversion because the value of the conversion result (y) is zero. Thus, based on the conversion forecast value of the user U3 and the conversion performance, the user whose conversion forecast value (p) has the same value (0.1) as the conversion forecast value of the user U3 has a probability of causing conversion. It is considered low. Therefore, when the conversion prediction value (p) of the user targeted for advertisement delivery is 0.1, the determination unit 132 determines the bid price to be low. The user U4 to the user U8 are the same as the user U3.

  Therefore, the determination unit 132 solves an optimization problem that reduces the value of the objective function (f) as the value of the conversion prediction value (p) decreases. In addition, for the conversion prediction value (p) of the user whose conversion result (y) has a value of 0, the determination unit 132 reduces the value of the objective function (f) corresponding to the conversion prediction value (p). Solve the optimal optimization problem. Then, part of the optimization problem is expressed by the following equation (2).

  If both are added together in order to satisfy both Formula (1) and Formula (2), it will become like following formula (3). Therefore, the determination unit 132 solves an optimization problem that minimizes the objective function (f) under predetermined constraints. That is, the determination unit 132 solves the following Expression (3) under a predetermined constraint condition.

  Here, the determination unit 132 sets upper and lower limits to the value of the objective function (f) as a constraint of the optimization problem. Assuming that the lower limit value of the value of the objective function (f) is a and the upper limit value is b, the first constraint is expressed by the following equation (4). Here, a corresponds to the lower limit value of the bid price, and b corresponds to the upper limit value of the bid price.

  Further, the determination unit 132 sets a constraint that the sum of the values of the objective function (f) matches the budget spent on the advertisement content AD1 in the next one month acquired by the acquisition unit 131 as a constraint condition of the optimization problem. . Here, the determination unit 132 assumes that the total number of users who click on the advertising content AD1 in the next one month is the same as the total number 8 of users who have clicked on the advertising content AD1 in the past one month. Then, the sum of Expression (5) is the sum of the values of the objective function (f) for all users who click on the advertisement content AD1 in the next one month, and all the users who have clicked on the advertisement content AD1 in the past one month The sum of the values of the objective function (f) for Assuming that the budget is N, the second constraint is expressed by the following equation (5).

  In addition, the determination unit 132 imposes a constraint that the value of the objective function (f) monotonously increases in accordance with the conversion prediction value (p). Then, the third constraint is expressed by the following equation (6).

  As described above, the determination unit 132 solves an optimization problem for finding a solution that minimizes the objective function (f) under three constraints. In other words, the determination unit 132 solves Equation (3) under Equations (4) to (6). Then, the determination unit 132 obtains a solution of the optimization problem by solving the optimization problem as described above.

  Subsequently, based on the solution obtained by solving the optimization problem, the determination unit 132 generates a correspondence rule C1 in which the conversion bid value corresponds to a bid price serving as a reference for selecting the advertisement content AD1 as a distribution target. (Step S12).

  In addition, the determination part 132 will store the produced | generated corresponding | compatible rule in the corresponding | compatible rule memory | storage part 124, if it produces a corresponding | compatible rule which makes a bid price correspond to the predicted value regarding whether to take action leading to a profit for the advertiser of advertising content. .

  Finally, upon receiving a distribution request for advertisement content from the user, the determination unit 132 determines the bid price of the user targeted for advertisement distribution.

  When the determination unit 132 receives a distribution request for advertisement content from the user, the determination unit 132 inputs an access log such as a search query or a visit page of the user targeted for advertisement distribution to the prediction model M1, thereby converting the conversion forecast value of the user targeted for advertisement distribution. Calculate (step S13).

  Subsequently, the determination unit 132 determines, based on the conversion prediction value of the user targeted for advertisement distribution, a bid price serving as a standard for selecting the advertisement content AD1 as a distribution target based on the correspondence rule C1 generated based on the solution of the optimization problem. Are determined (step S14).

  Although not described intentionally in FIG. 8, when the determination unit 132 receives a distribution request for advertisement content from the user, the determination unit 132 determines advertisement content to be distributed for advertisement. Then, the determination unit 132 determines the bid price of the user targeted for advertisement delivery for each piece of advertisement content targeted for advertisement delivery.

  Specifically, upon receiving a distribution request for advertisement content from the user terminal 10, the determination unit 132 selects advertisement content as a distribution candidate. For example, when the nearest search query of the distribution request source user is “notebook computer”, the determination unit 132 sets the advertisement content including “notebook computer” in the keyword set in the advertisement content as the advertisement content of the distribution candidate. select. For example, the determination unit 132 selects the advertisement content AD1 and the advertisement content AD2 as advertisement content of the distribution candidate.

  Subsequently, the determination unit 132 calculates, for each piece of advertisement content of distribution candidate, a conversion prediction value of a user targeted for advertisement distribution. Specifically, the determination unit 132 refers to the advertisement content ID of the prediction model storage unit 123, and selects a prediction model corresponding to the advertisement content ID of the advertisement content of the distribution candidate. For example, the determination unit 132 refers to the advertisement content ID of the prediction model storage unit 123, and selects the prediction model M1 corresponding to the advertisement content ID “AD1” of the advertisement content AD1 of the distribution candidate. For example, the determination unit 132 refers to the advertisement content ID of the prediction model storage unit 123, and selects the prediction model M2 corresponding to the advertisement content ID “AD2” of the advertisement content AD2 of the distribution candidate.

  Then, the determination unit 132 calculates a conversion prediction value using the selected prediction model. Specifically, the determination unit 132 calculates the conversion prediction value of the advertisement delivery target user by inputting an access log such as a search query or a visit page of the advertisement delivery target user into the selected prediction model. For example, for the advertisement content AD1, the determination unit 132 inputs an access log such as a search query or a visit page of the user targeted for advertisement delivery into the selected prediction model M1 to thereby obtain the conversion prediction value F1 of the user targeted for advertisement delivery. calculate. For example, for the advertisement content AD2, the determination unit 132 inputs an access log such as a search query or a visit page of the user targeted for advertisement delivery into the selected prediction model M2 to obtain the conversion prediction value F2 of the user targeted for advertisement delivery. calculate.

  Subsequently, when the determination unit 132 calculates the conversion prediction value of the user who is the target of the advertisement distribution for each of the advertisement contents for the distribution candidate, the advertisement content of the advertisement contents for the distribution candidate is referred to Select the corresponding rule corresponding to the ID. For example, the determination unit 132 refers to the advertisement content ID of the correspondence rule storage unit 124, and selects the correspondence rule C1 corresponding to the advertisement content ID “AD1” of the advertisement content AD1 of the distribution candidate. For example, the determination unit 132 refers to the advertisement content ID of the correspondence rule storage unit 124, and selects the correspondence rule C2 corresponding to the advertisement content ID “AD2” of the advertisement content AD2 of the distribution candidate.

  Subsequently, based on the selected correspondence rule, the determination unit 132 determines a bid price corresponding to the conversion prediction value of the user targeted for advertisement delivery for each piece of advertisement content of the delivery candidate. For example, the determination unit 132 determines a bid B1 corresponding to the conversion prediction value F1 of the user targeted for advertisement delivery based on the selected correspondence rule C1. For example, based on the selected correspondence rule C2, the determination unit 132 determines a bid B2 corresponding to the conversion prediction value F2 of the user targeted for advertisement delivery.

  Subsequently, the determination unit 132 extracts the advertising content to be distributed from the distribution contents of the advertising content. Specifically, the determination unit 132 extracts the advertisement content to be distributed, based on eCPM (effective Cost Per Mill) obtained by multiplying the determined bid price and the click rate (CTR: Click Through Rate).

  For example, it is assumed that a bid B1 of advertisement content AD1 is "1,875" and a bid B2 of advertisement content AD2 is "4,500". Further, for example, it is assumed that the CTR of the advertising content AD1 is "0.05" and the CTR of the advertising content AD2 is "0.03". In this case, the determination unit 132 calculates eCPM “94” of the advertising content AD1 by multiplying the bid B1 “1,875” of the advertising content AD1 and the CTR “0.05” of the advertising content AD1. Further, the determination unit 132 calculates eCPM "135" of the advertisement content AD2 by multiplying the bid B2 "4,500" of the advertisement content AD2 and the CTR "0.03" of the advertisement content AD2. Then, the determination unit 132 determines, of the two eCPMs calculated as described above, the advertisement content AD2 for which the eCPM having a large value is calculated as the advertisement content to be distributed.

(Distribution unit 134)
The distribution unit 134 distributes the advertising content determined by the determination unit 132 to the user terminal 10 of the distribution request source. For example, the distribution unit 134 distributes the advertisement content AD2 determined by the determination unit 132 based on the eCPM to the user terminal 10 of the distribution request source.

[4. Flow of correspondence rule generation process]
Next, the procedure of the correspondence rule generation process according to the embodiment will be described with reference to FIG. FIG. 9 is a flowchart illustrating a correspondence rule generation processing procedure according to the embodiment.

  As shown in FIG. 9, the advertisement distribution device (determination device) 100 acquires a budget and advertisement content from the advertiser (step S101).

  Subsequently, the advertisement distribution device (determination device) 100 calculates a conversion prediction value of the user who has clicked on the advertisement content in the past one month (step S102).

  Subsequently, the advertisement distribution device (determination device) 100 solves an optimization problem for finding a solution that minimizes the objective function whose conversion variable is the independent variable within the budget (step S103).

  Subsequently, the advertisement delivery device (determination device) 100 generates a correspondence rule that makes the bid price correspond to the conversion prediction value, based on the solution obtained by solving the optimization problem (step S104).

[5. Flow of decision processing]
Next, the procedure of the determination process according to the embodiment will be described with reference to FIG. FIG. 10 is a flowchart of the determination process according to the embodiment.

  As shown in FIG. 10, the advertisement distribution device (determination device) 100 determines whether a distribution request for advertisement content has been received from the user (step S201). When the advertisement distribution device (determination device) 100 has not received a distribution request for advertisement content from the user (Step S201; No), the advertisement distribution device (determination device) 100 waits until the distribution request for advertisement content is received from the user.

  When the advertisement distribution device (determination device) 100 receives a distribution request for advertisement content from the user (step S201; Yes), the advertisement distribution device (determination device) 100 selects advertisement content which is a distribution candidate (step S202).

  Subsequently, the advertisement distribution device (determination device) 100 calculates the conversion prediction value of the user targeted for advertisement distribution for each of the advertisement contents of the distribution candidate (step S203).

  Subsequently, the advertisement distribution device (determination device) 100 selects the correspondence rule of the advertisement content of the distribution candidate (step S204).

  Subsequently, the advertisement distribution device (determination device) 100 determines a bid price for each piece of advertisement content that is a distribution candidate (step S205).

Subsequently, the advertisement distribution device (determination device) 100 determines the advertisement content to be distributed from among the advertisement content of the distribution candidate (step S206).

  Subsequently, the advertisement distribution device (determination device) 100 distributes the advertisement content (step S207).

[6. Modified example]
The determination system 1 according to the above-described embodiment may be implemented in various different forms other than the above-described embodiment. So, below, other embodiments of the determination system 1 are described.

  Although FIG. 1 shows an example in which the determination apparatus 100 determines a bid price serving as a reference for selecting advertisement content as a distribution target based on the conversion prediction value of the user who has clicked the advertisement content in the past, Based on the conversion rate of the user who clicks on the content, a bid may be determined which serves as a criterion for selecting the advertisement content for distribution. In this case, the conversion prediction value in step S11 to step S14 in FIG. 1 is replaced with the conversion rate.

  Specifically, when using the conversion rate instead of the conversion prediction value, the determination unit 132 performs the following four processes. First, the determination unit 132 generates a prediction model that calculates the conversion prediction value of the user. Second, the determination unit 132 generates a conversion model that converts a conversion prediction value into a conversion rate by correcting the prediction model. Third, the determination unit 132 generates a correspondence rule (C) that makes a bid correspond to the conversion rate. Fourth, when the determination unit 132 receives a distribution request for advertisement content from the user, the determination unit 132 determines the advertisement content to be distributed.

  A process in which the determination unit 132 generates a conversion model that converts a conversion prediction value into a conversion rate by correcting the prediction model will be described. Specifically, the determination unit 132 calculates the conversion prediction value of the user who has clicked on the advertisement content by inputting an access log such as a search query or a visit page of the user who clicked on the advertisement content in the generated prediction model. Then, the determination unit 132 generates, from the conversion prediction value of the user who clicks on the advertisement content, a conversion model that predicts the conversion rate regarding whether or not the user who clicks on the advertisement content causes a conversion.

  For example, when the determination unit 132 calculates the conversion prediction value of the click user who clicks the advertisement content, it is assumed that there are 10 users whose conversion prediction value is 0.1. Then, it is assumed that among 10 users whose conversion predicted value is 0.1, one user has made conversion. In other words, it is assumed that among 10 users with a conversion prediction value of 0.1, 9 users did not cause conversion. In this case, the determination unit 132 generates a conversion model that predicts that the conversion rate regarding whether the conversion prediction value is 0.1 is 10%. Note that the example in which the determination unit 132 generates a conversion model is not limited to this example.

  The determination unit 132 may generate a transformation model using any learning algorithm. For example, the determination unit 132 generates a conversion model using a learning algorithm such as a neural network, a support vector machine, clustering, and reinforcement learning. As an example, when the determination unit 132 generates a transformation model using a neural network, the transformation model includes an input layer including one or more neurons, an intermediate layer including one or more neurons, and an output including one or more neurons. And a layer.

  The conversion model is a conversion forecast value calculated using a prediction model generated based on the information of the user who has taken action (product purchase etc.) that leads to a benefit for the advertiser of the advertisement content without clicking on the advertisement content. In the input layer, the output layer, and any layer from the input layer to the output layer, in which a conversion forecast value as to whether or not a predetermined user takes action leading to a benefit to the advertiser of the advertisement content is input And a second element whose value is calculated based on the first element and the weight of the first element, the first element belonging to a layer other than the output layer, and the conversion predicted value input to the input layer Accordingly, whether or not a given user takes action (conversion) that leads to a benefit for the advertiser in conjunction with the action of clicking on the ad content That to output the conversion rate from the output layer, causing a computer to function.

  The conversion model is a conversion forecast value calculated using a prediction model generated based on the information of the user who has taken action (product purchase etc.) that leads to a benefit for the advertiser of the advertisement content without clicking on the advertisement content. There is an input layer, an output layer, and from the input layer to the output layer, into which a conversion forecast value regarding whether or not a predetermined user causes an action (product purchase etc.) leading to a profit for the advertiser of the advertisement content is input. Is input to the input layer, including a first element belonging to any layer but a layer other than the output layer, and a second element whose value is calculated based on the first element and the weight of the first element By performing an operation based on the first element and the weight of the first element, each element belonging to each layer other than the output layer is used as the first element on the conversion conversion predicted value to obtain a predetermined value. As the seat is output from the output layer of the conversion rate on whether or not to take action that leads to profits for the advertiser in conjunction with the action to click on the ad content (conversion), causing a computer to function.

Here, the transformation model is represented by “y = 1 / (1 + exp (−t)), t = a ₀ + a ₁ * x ₁ + a ₂ * x ₂ +... + A _i * x _i ” It is assumed that it is realized by a logistic regression model. In this case, the first element included in the prediction model corresponds to input data (x _i ) such as x ₁ and x ₂ . Here, the input data (x _i ) is a conversion prediction value of a predetermined user. Moreover, y is a conversion rate which is a prediction probability (value of 0 to 1). For example, y is a conversion rate (conversion rate) related to whether or not an action (conversion) that leads to a benefit to an advertiser of advertisement content occurs in conjunction with an action of a predetermined user clicking on advertisement content. Also, the weight of the first element corresponds to the coefficient a _i corresponding to x _i . Here, the logistic regression model can be expressed as “y = σ (t) = σ (a ₁ * x ₁ + a ₂ * x ₂ +... + A _i * x _i )” using the standard sigmoid function σ It can be expressed. Therefore, the logistic regression model can be regarded as a simple perceptron having an input layer and an output layer. When each model is regarded as a simple perceptron, the first element can correspond to any node of the input layer, and the second element can be regarded as a node of the output layer. In logistic regression, the parameter (coefficient a _i ) is estimated by the maximum likelihood method. Specifically, in the maximum likelihood method, a likelihood function for estimating the parameters (coefficients a _i ) of the model is derived. The parameter (coefficient a _i ) is estimated by repeating the iterative method so that the negative log likelihood function is minimized.

  The advertisement distribution device (determination device) 100 is a model used to predict the occurrence probability, such as the above-described logistic regression model, and calculates the conversion rate using a conversion model having an arbitrary structure. Specifically, the conversion model is calculated using a prediction model generated based on information of a user who has taken action (such as purchase of a product) that leads to a benefit to the advertiser of the advertisement content without clicking on the advertisement content. If the predetermined conversion forecast value is input as to whether or not the predetermined user causes an action (product purchase etc.) for the advertiser of the advertising content to be profitable, the predetermined user may A coefficient is set to output a conversion rate as to whether or not an action (conversion) leading to profit for the advertiser occurs in conjunction with the click action. The advertisement distribution device (determination device) 100 uses such a conversion model to calculate the conversion rate.

[7. effect〕
As described above, the determination apparatus 100 according to the embodiment includes the acquisition unit 131 and the determination unit 132. The acquisition unit 131 acquires a budget to spend on advertisement content. The determination unit 132 distributes the advertisement content based on the budget acquired by the acquisition unit 131 and the predicted value as to whether the user who has selected the advertisement content in the past takes action to lead to a benefit for the advertiser of the advertisement content. Determine the standard bid to be selected as.

  Thereby, the determining apparatus 100 according to the embodiment determines the bid price using the budget instead of using the target CPA, for example, the target CPA is too low and misses the selection leading to conversion, and the total conversion You can use your budget effectively to get conversions while avoiding the problem of diminishing numbers. Therefore, the determination device 100 can determine an appropriate bid according to the budget.

  In addition, the determination unit 132 determines that the advertisement content is selected as a distribution target as the value of the prediction value regarding whether the user who selected the advertisement content in the past takes action leading to a profit for the advertiser of the advertisement content is larger. And the higher the bid. In addition, the determination unit 132 determines that the advertisement content is selected as the distribution target as the value of the prediction value regarding whether the user who selected the advertisement content in the past takes action leading to a profit for the advertiser of the advertisement content is smaller. Lower the bid.

  As a result, the determination apparatus 100 can use the budget effectively by making the user sharp and clear according to the ease of causing the conversion. In addition, it is possible to gain more conversions within the budget by making a clear bid for each user depending on the propensity to cause conversion and deciding a bid.

  In addition, the determination unit 132 relates to a result of whether the user who has selected the advertisement content in the past has made a conversion and whether the user who has selected the advertisement content in the past takes an action leading to a benefit for the advertiser of the advertisement content. Based on the predicted value, a bid that is a standard for selecting advertisement content as a distribution target is determined. Specifically, when a user who has selected advertisement content in the past has caused conversion, the determination unit 132 determines a bid price corresponding to the predicted value of the user to be high. In addition, when the user who has selected the advertisement content in the past has not caused conversion, the determination unit 132 determines the bid price corresponding to the predicted value of the user to be low.

  As a result, the determination apparatus 100 can use the budget effectively by providing a sharp sense for each user according to the latest conversion results, in addition to the predicted value of the likelihood of causing conversion. In addition, by setting a clear bid for each user and determining a bid, more conversions can be obtained within the budget.

  In addition, the determination unit 132 is optimized to obtain a solution that minimizes an objective function in which a predicted value as to whether a user who has selected advertisement content in the past takes action leading to a benefit to the advertiser of the advertisement content is minimized. Solve the problem. Then, based on the solution obtained by solving the optimization problem, the determination unit 132 generates a correspondence rule that makes the bid price correspond to the predicted value. In addition, the determination unit 132 sets a criterion for selecting the advertisement content as a distribution target based on a correspondence rule that makes a bid price correspond to a predicted value regarding whether the user takes an action leading to a profit for the advertiser of the advertisement content. Determine the bid. In addition, the determination unit 132 solves the optimization problem with the constraint that the sum of the values of the objective function matches the budget acquired by the acquisition unit 131.

  Thus, the determining apparatus 100 determines the bid by using the budget instead of using the target CPA, for example, the target CPA is too low, and thus the selection leading to conversion is missed and the total number of conversions decreases. You can use your budget effectively to get conversions while avoiding problems such as Therefore, the determination device 100 can determine an appropriate bid according to the budget.

  In addition, the determination unit 132 solves the optimization problem under the constraint that the value of the objective function does not exceed a predetermined threshold.

  Thereby, the determination apparatus 100 can determine an appropriate bid according to the budget so that the bid does not exceed the upper limit.

  Further, the determination unit 132 solves the optimization problem under the constraint that the value of the objective function does not fall below a predetermined threshold.

  Thereby, the determination apparatus 100 can determine an appropriate bid according to the budget so that the bid will not fall below the lower limit.

[8. Hardware configuration]
Further, the determination apparatus 100 according to the embodiment described above is realized by, for example, a computer 1000 configured as shown in FIG. FIG. 11 is a hardware configuration diagram showing an example of a computer that implements the function of the determination apparatus 100. The computer 1000 includes a CPU 1100, a RAM 1200, a ROM 1300, an HDD 1400, a communication interface (I / F) 1500, an input / output interface (I / F) 1600, and a media interface (I / F) 1700.

  The CPU 1100 operates based on a program stored in the ROM 1300 or the HDD 1400 to control each part. The ROM 1300 stores a boot program executed by the CPU 1100 when the computer 1000 starts up, a program depending on the hardware of the computer 1000, and the like.

  The HDD 1400 stores a program executed by the CPU 1100, data used by the program, and the like. The communication interface 1500 receives data from another device via a predetermined communication network, sends the data to the CPU 1100, and transmits data generated by the CPU 1100 to the other device via the predetermined communication network.

  The CPU 1100 controls an output device such as a display or a printer and an input device such as a keyboard or a mouse via the input / output interface 1600. The CPU 1100 acquires data from an input device via the input / output interface 1600. The CPU 1100 also outputs the generated data to the output device via the input / output interface 1600.

  The media interface 1700 reads a program or data stored in the recording medium 1800 and provides the CPU 1100 with the program via the RAM 1200. The CPU 1100 loads such a program from the recording medium 1800 onto the RAM 1200 via the media interface 1700 and executes the loaded program. The recording medium 1800 is, for example, an optical recording medium such as a digital versatile disc (DVD) or a phase change rewritable disc (PD), a magneto-optical recording medium such as a magneto-optical disk (MO), a tape medium, a magnetic recording medium, or a semiconductor memory. Etc.

  For example, when the computer 1000 functions as the determination device 100 according to the embodiment, the CPU 1100 of the computer 1000 executes the program or data (for example, a prediction model) loaded on the RAM 1200 to perform the function of the control unit 130. To realize. The CPU 1100 of the computer 1000 reads and executes these programs or data (for example, a prediction model) from the recording medium 1800, but as another example, these programs or data from another device via a predetermined communication network For example, a prediction model may be acquired.

  Although some of the embodiments of the present application have been described above in detail based on the drawings, these are only examples, and various modifications can be made based on the knowledge of those skilled in the art, including the aspects described in the section of the description of the invention. It is possible to implement the invention in other improved forms.

[9. Other]
Further, among the processes described in the embodiment and the modification, all or part of the process described as being automatically performed can be manually performed, or it is described as manually performed. All or part of the treatment may be performed automatically by known methods. In addition, information including processing procedures, specific names, various data and parameters shown in the above-mentioned documents and drawings can be arbitrarily changed unless otherwise specified. For example, the various information shown in each figure is not limited to the illustrated information.

  Further, each component of each device illustrated is functionally conceptual, and does not necessarily have to be physically configured as illustrated. That is, the specific form of the distribution and integration of each device is not limited to the illustrated one, and all or a part thereof may be functionally or physically dispersed in any unit depending on various loads, usage conditions, etc. It can be integrated and configured.

  Moreover, it is possible to combine suitably the embodiment and modification which were mentioned above in the range which does not make process content contradictory.

  In addition, the "section (module, unit)" described above can be read as "means" or "circuit". For example, the determination unit can be read as a determination means or a determination circuit.

DESCRIPTION OF SYMBOLS 1 determination system 10 user terminal 20 access log server 30 advertiser terminal 100 determination device 121 advertisement information storage unit 122 advertisement distribution history storage unit 123 prediction model storage unit 124 correspondence rule storage unit 131 acquisition unit 132 determination unit 133 distribution unit

Claims (12)

An acquisition unit for acquiring a budget to spend on advertising content;
The advertisement content is to be distributed based on the budget acquired by the acquisition unit and a predicted value as to whether the user who has selected the advertisement content in the past takes action leading to a benefit for the advertiser of the advertisement content. A determination unit that determines a standard bid price to be selected;
A determination apparatus comprising: The determination unit is
The determination apparatus according to claim 1, wherein the bid price is determined to be higher as the value of the prediction value is larger, and the bid price is determined to be lower as the value of the prediction value is smaller. The determination unit is
The apparatus according to claim 1 or 2, wherein the bid price is determined based on the result of whether or not the user who selected the advertisement content has caused conversion in the past and the predicted value. The determination unit is
The determination apparatus according to claim 3, wherein, when a user who has selected the advertisement content in the past has caused conversion, a bid price corresponding to the predicted value of the user is determined high. The determination unit is
The determination apparatus according to claim 3, wherein when a user who has selected the advertisement content in the past has not caused conversion, a bid price corresponding to the predicted value of the user is determined to be low. The determination unit is
Solve the optimization problem for finding a solution that minimizes the objective function having the prediction value as an independent variable, and use a correspondence rule to make the bid price correspond to the prediction value based on the solution obtained by solving the optimization problem. The determination apparatus according to any one of claims 1 to 5, wherein the determination apparatus generates the information. The determination unit is
The determination apparatus according to claim 6, wherein a bid price serving as a standard for selecting the advertisement content as a distribution target is determined based on the correspondence rule. The determination unit is
The determination apparatus according to claim 6, wherein the optimization problem is solved under a constraint that the sum of the values of the objective function matches the budget acquired by the acquisition unit. The determination unit is
The determination apparatus according to claim 6, wherein the optimization problem is solved under a constraint that the value of the objective function does not exceed a predetermined threshold. The determination unit is
The determination apparatus according to claim 6, wherein the optimization problem is solved under a constraint that the value of the objective function does not fall below a predetermined threshold. A decision method that the computer executes,
An acquisition process for acquiring a budget to spend on advertising content;
The advertisement content is selected as a distribution target based on the budget acquired by the acquisition step and a predicted value as to whether the user who selected the advertisement content in the past takes action leading to a benefit for the advertiser of the advertisement content. A decision process for determining a standard bid price to be
A determination method comprising: Acquisition procedure to acquire the budget to spend on advertising content,
The advertisement content is selected as a distribution target based on the budget acquired by the acquisition procedure and a predicted value as to whether the user who selected the advertisement content in the past takes action leading to a benefit for the advertiser of the advertisement content. A determination procedure for determining a standard bid to be
A decision program for causing a computer to execute.

Images