Information processing device, information processing method, and information processing program

By calculating attribution scores for multiple advertisements and generating a prediction model that considers their combined impact, the device addresses the limitations of last-touch models, enhancing advertisement conversion rate prediction and bidding optimization.

JP2024166862APending Publication Date: 2024-11-29LY CORP

Patent Information

Application Number
JP2023083249
Authority / Receiving Office
JP · JP
Patent Type
Applications
Current Assignee / Owner
Filing Date
2023-05-19
Publication Date
2024-11-29

AI Technical Summary

Technical Problem

Conventional methods struggle to accurately predict the conversion rate of advertisements due to reliance on last-touch models, which fail to consider the relative contributions of multiple advertisements provided before conversion.

Method used

An information processing device that calculates attribution scores for each advertisement based on their relative contribution to conversion, using a prediction model that considers the sequence and combination of advertisements provided to a user, and generates a second prediction model to accurately forecast conversion rates.

Benefits of technology

The solution enables more accurate prediction of advertisement conversion rates, optimizing bidding strategies by considering the cumulative impact of multiple advertisements, leading to reduced costs per conversion and improved advertising efficiency.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure 2024166862000001_ABST
    Figure 2024166862000001_ABST
Patent Text Reader

Abstract

To provide an information processing device, an information processing method, and an information processing program that allow generation of a prediction model for appropriately predicting a conversion ratio of an advertisement.SOLUTION: An information processing device includes an acquisition unit and a generation unit. The acquisition unit obtains a score of the degree of relative contribution for conversion of each of a plurality of advertisements which are provided to a user before completion of the conversion by the user. The generation unit generates a prediction model for predicting a conversion ratio of an advertisement to be predicted, based on the score of each of the plurality of advertisement which is obtained by the acquisition unit.SELECTED DRAWING: Figure 3
Need to check novelty before this filing date? Find Prior Art

Description

[Technical Field]

[0001] The present invention relates to an information processing device, an information processing method, and an information processing program. [Background technology]

[0002] In recent years, with the rapid spread of the Internet, advertisement distribution via the Internet has become popular. As an example of such advertisement distribution, a technology has been proposed that maximizes advertising profits by controlling bid amounts for advertisement spaces based on past bidding records (see, for example, Patent Document 1).

[0003] In automated bidding for this type of advertising space, the bid amount is determined by multiplying the cost per acquisition set by the advertiser, such as a threshold CPA (Cost Per Action), by a predicted CVR (Conversion Rate). [Prior art documents] [Patent documents]

[0004] [Patent Document 1] Japanese Patent Application Publication No. 2019-220052 Summary of the Invention [Problem to be solved by the invention]

[0005] However, in conventional technologies, only the last ad served before conversion, the so-called last-touch ad, is used to train a prediction model that predicts CVR. Therefore, for example, in cases where multiple ads are served to a user before conversion, it can be difficult to generate a prediction model that accurately predicts CVR.

[0006] The present application has been made in consideration of the above, and aims to provide an information processing device, an information processing method, and an information processing program that can generate a prediction model that appropriately predicts the conversion rate of an advertisement. [Means for solving the problem]

[0007] The information processing device according to the present disclosure includes an acquisition unit and a generation unit. The acquisition unit acquires a score relating to a relative contribution to conversion of each of a plurality of advertisements provided to a user until the user reaches a conversion. The generation unit generates a prediction model for predicting a conversion rate of the advertisement to be predicted based on the scores of each of the plurality of advertisements acquired by the acquisition unit. [Effects of the Invention]

[0008] According to one aspect of the embodiment, it is possible to generate a prediction model that appropriately predicts the conversion rate of an advertisement. [Brief explanation of the drawings]

[0009] [Figure 1] FIG. 1 is a diagram illustrating an example of information processing according to an embodiment. [Figure 2] FIG. 2 is a diagram illustrating an example of the configuration of an information processing system according to the embodiment. [Figure 3] FIG. 3 is a diagram illustrating an example of the configuration of the information processing device according to the embodiment. [Figure 4] FIG. 4 is a diagram illustrating an example of a user information table stored in the user information storage unit of the information processing device according to the embodiment. [Figure 5] FIG. 5 is a diagram illustrating an example of an advertisement information table stored in the advertisement information storage unit of the information processing device according to the embodiment. [Figure 6] FIG. 6 is a diagram illustrating an example of an advertisement provision history table stored in the advertisement provision history storage unit of the information processing device according to the embodiment. [Figure 7]FIG. 7 is a flowchart showing an example of information processing by the processing unit of the information processing device according to the embodiment. [Figure 8] FIG. 8 is a hardware configuration diagram illustrating an example of a computer that realizes the functions of the information processing device according to the embodiment. DETAILED DESCRIPTION OF THE INVENTION

[0010] Hereinafter, modes for implementing an information processing device, an information processing method, and an information processing program according to the present application (hereinafter referred to as "embodiments") will be described in detail with reference to the drawings. Note that the information processing device, the information processing method, and the information processing program according to the present application are not limited to these embodiments. Furthermore, the respective embodiments can be appropriately combined within the scope of not causing any contradiction in the processing content. Furthermore, the same components in the following embodiments will be assigned the same reference numerals, and redundant explanations will be omitted.

[0011] [1. An example of information processing] First, an example of information processing according to the embodiment will be described with reference to Fig. 1. Fig. 1 is a diagram showing an example of information processing according to the embodiment, which is executed by an information processing device 1.

[0012] The information processing device 1 generates a prediction model for predicting the conversion rate of an advertisement to be predicted based on an attribution score, which is a score relating to the relative contribution of each of a plurality of advertisements provided to a user until the user reaches a conversion. A conversion is, for example, a user behavior targeted by an advertisement, and may be any user behavior.

[0013] For example, a conversion may be, but is not limited to, a purchase of the transaction object that is the subject of the advertisement (hereinafter sometimes referred to as the "advertised object"), use of the transaction object that is the subject of the advertisement, registration for a membership service that is the subject of the advertisement, or viewing (viewing) of a site that is the subject of the advertisement. The transaction object may be, for example, a product or service.

[0014] For example, a conversion may be signing up for an email distribution service that is the subject of the advertisement, registering an account for an online service that is the subject of the advertisement, requesting materials that are the subject of the advertisement, visiting a physical store that is the subject of the advertisement, or sharing on a social networking service (SNS) that is the subject of the advertisement, or it may be some other action.

[0015] What kind of user behavior is considered to be a conversion can be set arbitrarily. For example, what kind of user behavior is considered to be a conversion for each advertisement is individually set by the administrator of the information processing device 1, the advertiser, or the like.

[0016] An advertisement is content for advertising an advertising target, and is also called advertising content. Examples of such advertisements include, but are not limited to, banner advertisements, listing advertisements, native advertisements, social media advertisements, and video advertisements.

[0017] An example of information processing will be described below with reference to Fig. 1. In Fig. 1, a user's purchase of product X is taken as an example of a conversion, and the information processing device 1 calculates an attribution score, which is a score indicating the relative contribution of each advertisement, using the number of purchases of product X as the number of conversions.

[0018] In the following, a user who has achieved conversion may be referred to as a "relevant user." For simplicity, FIG. 1 illustrates an example in which three advertisements, Ads A, B, and C, are provided. For example, Ads A, B, and C are advertisements targeting Product X and included in a single advertising campaign, but the example is not limited to this.

[0019] 1 illustrates the calculation of a score when at most three advertisements, Ads A, B, and C, are provided to the user, but the number of advertisements provided to the user may be four or more, or may be just two. Ad A may be, for example, a broad distribution advertisement, Ad B may be, for example, an interest targeting advertisement, and Ad C may be, for example, a retargeting advertisement, but these examples are not limited to these.

[0020] Broadband advertising is, for example, advertising that is provided without narrowing down to a specific target group. Interest-targeted advertising is, for example, advertising that is provided based on a user's interests and concerns. Retargeted advertising is, for example, advertising that is provided to users who have visited the website or web page of the transaction that is the subject of the advertisement. Note that providing an advertisement means, for example, presenting the advertisement to a user, but may also include transitioning to a link destination of the advertisement when the user selects (e.g., clicks) the advertisement.

[0021] First, the information processing device 1 acquires, from an internal storage unit or the like, provided advertisement information including information on advertisements provided to a user who has achieved a conversion for each conversion (step S1).

[0022] For example, the information processing device 1 acquires provided advertisement information including information about conversions for each combination of advertisements provided from the time when the user purchased the product X until a predetermined period TA. Hereinafter, the combination of advertisements may be referred to as an advertisement combination.

[0023] The predetermined period TA is, for example, one week, but is not limited to this example and can be set appropriately depending on, for example, the attributes of the user, the category of product X, the content of the conversion, and the like.

[0024] There are seven types of advertising combinations: advertising combination of only advertisement A, advertising combination of only advertisement B, advertising combination of only advertisement C, advertising combination of advertisements A and B, advertising combination of advertisements A and C, advertising combination of advertisements B and C, and advertising combination of advertisements A, B, and C.

[0025] Next, the information processing device 1 generates a first prediction model, which is a prediction model that predicts the number of conversions corresponding to the advertisement combination, which is a combination of advertisements provided to the relevant user, based on the provided advertisement information acquired in step S1 (step S2).

[0026] The first prediction model generated in step S2 is expressed, for example, by the following formula (1): In the following formula (1), the feature amount is information on the advertisement combination, and pCV is the predicted number of conversions. f(feature) = pCV (1)

[0027] In the process of step S2, the information processing device 1 calculates the number of conversions for each advertisement combination based on the provided advertisement information acquired in step S1. For example, by tallying the number of conversions when the advertisement combination is a combination of advertisements A, B, and C, the number of conversions when the advertisement combination is a combination of advertisements A, B, and C is calculated.

[0028] Then, the information processing device 1 generates learning information including information on the advertisement combination and information indicating the number of conversions for each advertisement combination, and generates a first prediction model using the learning information.

[0029] Next, the information processing device 1 calculates an attribution score, which is a score indicating the relative contribution of advertisements A, B, and C, using the first prediction model generated in step S2 (step S3).

[0030] For example, the information processing device 1 performs a calculation process to calculate the attribution score of each advertisement A, B, and C by inputting information about the advertising combination into a first prediction model and obtaining the predicted number of conversions output from the first prediction model.

[0031] For example, assume that the first prediction model is an RNN such as LSTM or GRU. In this case, the information processing device 1 inputs information on advertisement combinations to the first prediction model, and calculates output values ​​of a softmax function in the first prediction model at this time as attribution scores of advertisements A, B, and C.

[0032] The output value of the softmax function is obtained by applying the softmax function to the output of hierarchical attention, and is expressed as the probability distribution of advertisements A, B, and C. Hierarchical attention is an example of an attention mechanism. In the example shown in Figure 1(b), the attribution score of advertisement A is "0.2", the attribution score of advertisement B is "0.3", and the attribution score of advertisement C is "0.5".

[0033] Furthermore, it is assumed that the first prediction model is a classification model such as a linear regression model or GBDT. In this case, the information processing device 1 inputs information about the advertisement combination into the first prediction model and performs a process of obtaining pCV, which is the predicted number of conversions, from the first prediction model for each advertisement combination. Then, the information processing device 1 performs a process of calculating an attribution score for each advertisement A, B, and C for each advertisement combination using the pCV for each advertisement combination.

[0034] For example, assume that the advertisement combination is a combination of advertisements A, B, and C. In this case, the information processing device 1 performs the calculations of the following formulas (2) to (4). v(A) is a function related to the contribution degree of advertisement A, v(B) is a function related to the contribution degree of advertisement B, and v(C) is a function related to the contribution degree of advertisement C. v(A)={f(φ)+(f(A)-f(φ)) ···(2) v(B)={f(φ)+(f(A,B)-f(A)) ···(3) v(C)={f(φ)+(f(A,B,C)-f(A,B)) ···(4)

[0035] In the above formulas (2) to (4), f(φ) is the pCV when none of advertisements A, B, and C are provided to the user, and f(A) is the pCV when only advertisement A is provided to the user. In the above formulas (3) and (4), f(A, B) is the pCV when advertisements A and B are provided to the user. In the above formula (4), f(A, B, C) is the pCV when advertisements A, B, and C are provided to the user. Note that "φ" in f(φ) means an empty set.

[0036] In this way, the functions v(A), v(B), and v(C) are expressed in a form that includes an equation that subtracts the pCV of an advertising combination that does not include one advertisement from the pCV of an advertising combination in which one advertisement is added to an advertising combination that does not include one advertisement.

[0037] The information processing device 1 performs the calculations of the following equations (5) to (8) based on the calculation results of the above equations (2) to (4). Sc(A) is the attribution score of advertisement A, Sc(B) is the attribution score of advertisement B, and Sc(C) is the attribution score of advertisement C. v(A,B,C)=v(A)+v(B)+v(C) ···(5) Sc(A)=v(A) / v(A,B,C) (6) Sc(B)=v(B) / v(A,B,C) (7) Sc(C)=v(C) / v(A,B,C) (8)

[0038] The information processing device 1 similarly calculates the attribution scores Sc(A), Sc(B), and Sc(C) of advertisements A, B, and C for advertisements A, B, and C for other advertisement combinations. In this way, the information processing device 1 performs a calculation process to calculate the attribution scores Sc(A), Sc(B), and Sc(C) of each advertisement A, B, and C for each advertisement combination. Hereinafter, when multiple attribution scores including Sc(A), Sc(B), and Sc(C) are referred to without distinction, they may be referred to as the attribution score Sc.

[0039] In the above example, the information processing device 1 calculates the attribution score Sc of each advertisement for each advertisement combination, but is not limited to this example. For example, the information processing device 1 can also calculate the attribution score Sc of each advertisement for each advertisement presentation order.

[0040] The advertisement presentation order is the order in which advertisements are presented to a user. For example, if advertisements X, Y, and Z are presented to the user in this order from the time the user purchases product X until a predetermined period TA, the advertisement presentation order is advertisement X, Y, and Z. Advertisements X, Y, and Z may be advertisements A, B, and C, for example, but some or all of advertisements X, Y, and Z may be the same advertisement.

[0041] Furthermore, if one or more advertisements provided to the user in a predetermined period TA are advertisement X, advertisement Y in that order, the advertisement presentation order is advertisement X, advertisement Y. Furthermore, if one or more advertisements provided to the user in a predetermined period TA are advertisement X only (advertisement X is any of advertisements A, B, and C), the advertisement presentation order is advertisement X only.

[0042] The information processing device 1 calculates the number of conversions for each advertisement presentation order. For example, the information processing device 1 calculates the number of conversions when the advertisement presentation order is advertisement A, advertisement B, and advertisement C by tallying up the number of conversions when the advertisement presentation order is advertisement A, advertisement B, and advertisement C.

[0043] Then, the information processing device 1 generates learning information including information on the advertisement presentation order and information indicating the number of conversions for each advertisement presentation order, and generates a first prediction model using the learning information. Hereinafter, when the advertisement combination and the advertisement presentation order are not distinguished from each other, they may be referred to as advertisement presentation types.

[0044] Furthermore, the feature quantities of the first prediction model may include user attribute information in addition to information on the type of advertisement presentation. In this case, the user attribute information is information on the attributes of the user, such as the user's age and gender, but is not limited to such examples and may include, for example, occupation, place of residence, and annual income. In this case, the information processing device 1 generates learning information including, for each combination of the user attribute information and the advertisement presentation order, information on the user attribute information, information on the advertisement presentation order, and information indicating the number of conversions, and generates the first prediction model using the learning information.

[0045] Next, the information processing device 1 generates a second prediction model that is a prediction model for predicting the CVR of the advertisement to be predicted, based on the attribution score Sc of each advertisement calculated in step S3 (step S4).

[0046] For example, the information processing device 1 generates a second prediction model represented by the following formula (9): The feature represented by the following formula (9) is a feature obtained from, for example, one or more pieces of information among user attribute information, user behavior history, and advertisement display environment information, and may be referred to as a target feature hereinafter. f(feature) = pCVR (9)

[0047] The user attribute information is information on the user's demographic attributes, but may also include information on the user's psychographic attributes instead of or in addition to the user's demographic attribute information. The user's behavioral history includes, for example, information on the user's behavioral history in online services. The advertisement display environment information includes, for example, information such as the category of the website on which the advertisement is provided, the type of application on which the advertisement is provided, and the display position of the advertisement.

[0048] The information processing device 1 acquires, from an internal storage unit or the like, advertisement information including target feature amounts and information indicating whether or not a conversion has occurred (an example of second advertisement provision information, which will be described later). For example, the information processing device 1 acquires information including target feature amounts when an advertisement is presented within a predetermined period TB and information indicating whether or not a conversion has occurred by a user to whom the advertisement has been presented within the predetermined period TB. The predetermined period TB is the same period as the predetermined period TA, but may be a period different from the predetermined period TA.

[0049] For example, assume that advertisements A, B, and C are provided to a user within a predetermined period TB before the user converts. In this case, the information processing device 1 acquires information including target features when the advertisements are provided within the predetermined period TB and information indicating whether or not the user who was provided with the advertisements within the predetermined period TB converted, for each advertisement A, B, and C or for each combination of advertisements A, B, and C.

[0050] If the second provided advertising information includes information indicating whether or not a conversion has occurred for each of advertisements A, B, and C, the information indicating whether or not a conversion has occurred will include information indicating whether a conversion has occurred in the case of the last advertisement provided before the conversion occurred, and information indicating whether or not a conversion has occurred in other cases, but is not limited to such examples.

[0051] Here, when multiple ads are provided to a user before conversion, the tendency of the last touch, which is the last ad provided to the user, is learned. In this case, the learning information used to train the second prediction model is such that the last ad provided to the user among the multiple ads provided to the user before conversion is labeled with information indicating that conversion occurred, and the remaining ads are labeled with information indicating that conversion did not occur.

[0052] For example, if the last advertisement served to the user was advertisement C, as shown in (a) of Figure 1, the conversion presence / absence information indicating whether or not a conversion occurred by the user to whom advertisements A, B, and C were served will be "0" for advertisement A, "0" for advertisement B, and "1" for advertisement C. In the conversion presence / absence information, "0" is information indicating no conversion (information without conversion), and "1" is information indicating a conversion (information with conversion).

[0053] When learning last touch trends, it may be difficult to appropriately predict the conversion rate of an advertisement because the contribution of advertisements A and B to conversion is not reflected. Therefore, the information processing device 1 generates a second prediction model using learning information that uses, as labels, the attribution scores Sc of multiple advertisements provided to the user for a predetermined period TB until conversion.

[0054] For example, suppose that when ads A, B, and C are provided to a user within a specified period TA before the user converts, the attribution scores Sc(A), Sc(B), and Sc(C) of ads A, B, and C are "0.2," "0.3," and "0.5."

[0055] In this case, instead of the conversion presence / absence information of Advertisement A "0", Advertisement B "0", and Advertisement C "1" shown in (a) of Figure 1, the information processing device 1 generates information including the attribution scores Sc of "0.2" (Sc(A)), "0.3" (Sc(B)), and "0.5" (Sc(C)) as training information (training data), as shown in (b) of Figure 1.

[0056] The information processing device 1 generates learning information including target features when advertisements are provided to users during a predetermined period TB and the attribution score Sc of each advertisement, and including multiple pieces of training information in which the attribution score Sc of each advertisement is used as a label.

[0057] The information processing device 1 generates a second prediction model by machine learning using learning information including the plurality of pieces of teacher information generated as described above. This allows the information processing device 1 to generate a prediction model that reflects the contribution to conversion provided to the user up to the conversion, and can generate a prediction model that appropriately predicts the conversion rate of each advertisement.

[0058] Next, the information processing device 1 receives an advertisement request transmitted from the user terminal 2 of the user or the like (step S5). When the information processing device 1 receives the advertisement request, it calculates the pCVR of advertisements A, B, and C using the second prediction model generated in step S4 (step S6).

[0059] The information processing device 1 inputs the target features corresponding to the advertisement request received in step S5 into the second prediction model, and obtains the pCVR of each of advertisements A, B, and C output from the second prediction model, thereby calculating the pCVR of advertisements A, B, and C. The target features corresponding to the advertisement request include, for example, features obtained from attribute information and behavior history of the user of user terminal 2 that sent the advertisement request, and features obtained from advertisement display environment information.

[0060] When a CVR prediction model that has learned last-touch trends is used for the advertising request received in step S5, in the example shown in Figure 1(a), the pCVR of advertisement A is "0.01", the pCVR of advertisement B is "0.05", and the pCVR of advertisement C is "0.3".

[0061] On the other hand, when the second prediction model, which is a CVR prediction model trained using the attribution score Sc, is used for the advertising request received in step S5, in the example shown in (b) of Figure 1, the pCVR of advertisement A is "0.05", the pCVR of advertisement B is "0.1", and the pCVR of advertisement C is "0.2".

[0062] In this way, the information processing device 1 can calculate a pCVR that reflects the contribution to conversion provided to the user up to the conversion.

[0063] Next, the information processing device 1 calculates the bid amount (Bid) for each of the advertisements A, B, and C using the pCVR determined in step S6 (step S7). The information processing device 1 calculates the bid amount (Bid) for each of the advertisements A, B, and C using, for example, the following formula (10). Bid amount (Bid) = pCVR × thCPA (10)

[0064] In the above formula (10), thCPA is a threshold CPA (Cost Per Acquisition), and is set by, for example, the advertisers of advertisements A, B, and C or the operator of the information processing device 1.

[0065] In the example shown in FIG. 1, the thCPA of advertisement A is set to "200 yen," the thCPA of advertisement B is set to "200 yen," and the thCPA of advertisement C is set to "400 yen."

[0066] When a CVR prediction model that has learned last-touch trends is used for the ad request received in step S5, the pCVRs of ads A, B, and C are "0.01," "0.05," and "0.3." Therefore, the bid amounts for ads A, B, and C are "2 yen (= 0.01 x 200 yen)," "10 yen (= 0.05 x 200 yen)," and "120 yen (= 0.3 x 400 yen)."

[0067] On the other hand, when the second prediction model, which is a CVR prediction model trained using the attribution score Sc, is used for the advertising request received in step S5, the pCVRs of advertisements A, B, and C are "0.05," "0.1," and "0.2." Therefore, the bid amounts for advertisements A, B, and C are "10 yen (= 0.05 x 200 yen)," "20 yen (= 0.1 x 200 yen)," and "80 yen (= 0.2 x 400 yen)."

[0068] Therefore, for example, if only Ad C is used, and a CVR prediction model that has learned about last touch trends is used, the CPA per CV will be 120 yen or less. On the other hand, if Ads A, B, and C are used, and a second prediction model that has learned using the attribution score Sc is used, the CPA per CV will be 110 yen or less (= 10 yen + 20 yen + 80 yen).

[0069] When using a CVR prediction model that has learned last touch trends, bidding will be performed using pCVR based on a model that has been trained without considering, for example, ads A and B that were served up to ad C (for example, a combination of ads A, B, and C, or the path (order of ad presentation) of ad A → ad B → ad C), and bidding will be optimized for ad requests related to ad C only. In other words, when using a CVR prediction model that has learned last touch trends, bidding will be optimized on an ad request basis.

[0070] On the other hand, when a CVR prediction model trained using the attribution score Sc is used, bidding will be performed using pCVR based on a model trained in consideration of advertisements A, B, etc. provided up to advertisement C (for example, a combination of advertisements A, B, and C or the path of advertisement A → advertisement B → advertisement C). Therefore, advertisements A, B, etc. provided up to advertisement C (for example, a combination of advertisements A, B, and C or the path of advertisement A → advertisement B → advertisement C) are considered for each user, and a pCVR according to the relative contribution of each advertisement A, B, and C to the user can be used when providing advertisements (delivering advertisements). Therefore, the information processing device 1 can optimize CPA on a per-user basis.

[0071] The reason why the CPA is 120 yen or less or 110 yen or less is because it is an automatic bidding method called second price action, in which the second bid amount is adopted, and the cost is less than the bid amount.

[0072] In this way, the information processing device 1 generates a prediction model that predicts the conversion rate of an advertisement to be predicted based on the attribution score Sc, which is a score related to the relative contribution of each of a plurality of advertisements provided to a user until the user reaches conversion. This allows the information processing device 1 to generate a prediction model that appropriately predicts the conversion rate of an advertisement.

[0073] The configuration of an information processing system including the information processing device 1 and user terminal 2 that perform such processing will be described in detail below.

[0074] [2. Information Processing System Configuration] 2 is a diagram illustrating an example of a configuration of an information processing system according to an embodiment. As illustrated in FIG. 2, the information processing system 100 according to the embodiment includes an information processing device 1, a plurality of user terminals 2, and a plurality of advertiser terminals 3.

[0075] The multiple user terminals 2 are used by different users U. The multiple advertiser terminals 3 are used by different advertisers O. Each of the user terminals 2 and advertiser terminals 3 is, for example, a notebook PC (Personal Computer), a desktop PC, a smartphone, a tablet PC, or a wearable device. The wearable device is, for example, smart glasses or a smart watch, but is not limited to such examples. The user U is a user who receives advertisements provided from an information processing device 1 or the like.

[0076] The information processing device 1, the user terminal 2, and the advertiser terminal 3 are connected to each other via a network N so as to be able to communicate with each other by wire or wirelessly. Note that the information processing system 100 shown in FIG. 2 may include a plurality of information processing devices 1, etc.

[0077] The network N includes, for example, a WAN (Wide Area Network) such as the Internet and a mobile communication network such as LTE (Long Term Evolution), 4G (4th Generation), and 5G (5th Generation: 5th generation mobile communication system).

[0078] The user terminal 2 and the advertiser terminal 3 can connect to the network N via short-range wireless communication such as a mobile communication network, Bluetooth (registered trademark), or wireless LAN (Local Area Network) and communicate with the information processing device 1.

[0079] 3. Configuration of Information Processing Device 1 3 is a diagram showing an example of the configuration of the information processing device 1 according to the embodiment. As shown in FIG. 3, the information processing device 1 includes a communication unit 10, a storage unit 11, and a processing unit 12.

[0080] [3.1. Communication Unit 10] The communication unit 10 is realized by, for example, a communication module or a NIC (Network Interface Card). The communication unit 10 is connected to a network N by wire or wirelessly, and transmits and receives information to and from various other devices. For example, the communication unit 10 transmits and receives information to and from the user terminal 2 and the advertiser terminal 3 via the network N.

[0081] [3.2. Storage section 11] The storage unit 11 is realized by, for example, a semiconductor memory element such as a random access memory (RAM) or a flash memory, or a storage device such as a hard disk or an optical disk. The storage unit 11 has a user information storage unit 20, an advertisement information storage unit 21, an advertisement provision history storage unit 22, and an attribution score storage unit 23.

[0082] 3.2.1. User Information Storage Unit 20 The user information storage unit 20 stores various types of information related to the user U. Fig. 4 is a diagram showing an example of a user information table stored in the user information storage unit 20 of the information processing device 1 according to the embodiment.

[0083] 4, the user information table stored in the user information storage unit 20 includes information items such as "user ID (identifier)," "attribute information," and "history information." The "user ID" is an identifier that identifies a user U, and is information assigned to each user U.

[0084] "Attribute information" is attribute information indicating the attributes of user U associated with "user ID." The attributes of user U include, for example, demographic attributes, psychographic attributes, etc. Demographic attributes are demographic attributes and include multiple attribute items such as age, gender, occupation, place of residence, annual income, and family structure.

[0085] Psychographic attributes are psychological attributes and include, for example, multiple attribute items related to lifestyle, values, interests, etc. For example, each of the multiple attribute items in the psychographic attributes is an object of interest to the user U, such as cars, clothes, travel, games, camping, motorcycles, trains, home appliances, or computers.

[0086] "History information" includes information on the behavioral history of user U associated with "user ID." User U's behavioral history includes, for example, information on user U's behavioral history in online services. Information on user U's behavioral history in online services includes, for example, search history information, browsing history information, distribution history information, and transaction history information in online services of user U.

[0087] The search history information of the user U includes, for example, information on the search history of the user U in a web search service. The browsing history information of the user U includes, for example, information on the browsing history of content by the user U in an online service. The distribution history information includes information on the distribution history of content distributed to the user U's user terminal 2. The transaction history information includes information on the transaction history of products by the user U in an online service.

[0088] 3.2.2. Advertising Information Storage Unit 21 The advertisement information storage unit 21 stores various types of information related to advertisements received from the advertiser terminal 3. Fig. 5 is a diagram showing an example of an advertisement information table stored in the advertisement information storage unit 21 of the information processing device 1 according to the embodiment.

[0089] 5, the advertisement information table stored in the advertisement information storage unit 21 includes information on items such as "advertisement ID," "advertisement type," "advertisement," "distribution period," and "budget." The "advertisement ID" is an identifier that identifies an advertisement, and is information assigned to each advertisement.

[0090] The "advertising type" is information indicating the type of advertisement associated with the "advertising ID." The type of advertisement may be, for example, a broad distribution advertisement, an interest targeting advertisement, or a retargeting advertisement, but is not limited to these examples.

[0091] An "advertisement" is an advertisement associated with an "advertising ID," and examples thereof include, but are not limited to, banner advertisements, listing advertisements, native advertisements, social media advertisements, and video advertisements.

[0092] "Distribution period" is information about the distribution period of the advertisement associated with the "advertising ID." "Budget" is information about the budget of the advertisement associated with the "advertising ID." Note that the information included in the advertisement information table is not limited to the information described above, and may also include, for example, advertiser identification information, advertising target information, and campaign information.

[0093] The advertiser identification information is information that identifies the advertiser O of the advertisement associated with the "advertising ID", for example, an advertiser ID. The advertising target information is information about the advertising target that is the subject of the advertisement associated with the "advertising ID", for example, information about the transaction target that is the subject of the advertisement. The campaign information is information about the advertising campaign associated with the "advertising ID". Note that the advertising information storage unit 21 is not limited to the above and can store various types of information depending on the purpose.

[0094] 3.2.3. Advertisement Serving History Storage Unit 22 The advertisement provision history storage unit 22 stores various information related to the advertisement distribution history. Fig. 6 is a diagram showing an example of an advertisement provision history table stored in the advertisement provision history storage unit 22 of the information processing device 1 according to the embodiment.

[0095] 6, the advertisement provision history table stored in the advertisement provision history storage unit 22 includes information items such as "User ID," "Provided Advertisement," and "Conversion." The "User ID" is the same as the user ID shown in FIG. 4, and is information about the identifier of the user U to whom the advertisement is provided.

[0096] The "provided advertisement" includes, for example, the advertisement ID of the advertisement provided (distributed) to the user U associated with the "user ID", and information indicating the date and time the advertisement was provided to the user U. The "conversion" includes, for example, information indicating the content of the conversion performed by the user U associated with the "user ID", and information on the date and time of the conversion by the user U (conversion date and time).

[0097] The content of the conversion may be, for example, but not limited to, the purchase of a transaction object, the use of an advertised transaction object, the registration of an advertised membership, or the viewing (display) of an advertised site. The transaction object may be, for example, a product or service.

[0098] The advertisement provision history storage unit 22 can also store advertisements provided to the user U for a predetermined period TA in association with the user U who has achieved conversion for one transaction object.

[0099] For example, the advertisement provision history storage unit 22 associates advertisements corresponding to the advertisement target with a user U who has achieved a conversion corresponding to the advertisement target, and stores advertisements corresponding to the advertisement target among the advertisements provided to the user U during a specified period TA, in association with the combination or provision order thereof.

[0100] For example, the advertisement provision history storage unit 22 stores, for a product X, one or more advertisements that advertise the product X among those that were provided during a predetermined period (such as the most recent week) before the date and time when the user U purchased the product X, in association with the user U who purchased the product X, and stores the advertisements in association with the combinations and the order of provision thereof.

[0101] The advertisement provision history storage unit 22 also includes attribute information and history information of the user U at the time the advertisement was provided, as well as advertisement display environment information of the advertisement when it was provided to the user U. As described above, the advertisement display environment information includes information such as the category of the website on which the advertisement is provided, the type of application of the user terminal 2 on which the advertisement is provided, and the display position of the advertisement.

[0102] [3.2.4. Attribution Score Storage Unit 23] The attribution score storage unit 23 stores, for each advertisement group, attribution score information, which is information on the attribution score Sc calculated by the processing unit 12. An advertisement group includes multiple advertisements. The multiple advertisements included in an advertisement group are, for example, multiple advertisements included in the same campaign, but are not limited to this example.

[0103] The attribution score information includes, for example, information on the advertisement presentation type and the attribution scores Sc of the multiple advertisements for each advertisement presentation type. The information on the advertisement presentation type is, for example, information on the combination of advertisements or information on the advertisement presentation order.

[0104] [3.3. Processing Unit 12] The processing unit 12 is a controller, and is realized, for example, by a CPU (Central Processing Unit) or an MPU (Micro Processing Unit) executing various programs stored in a storage device inside the information processing device 1 using RAM as a working area.

[0105] The processing unit 12 may be partially or entirely realized by an integrated circuit such as an ASIC (Application Specific Integrated Circuit) or an FPGA (Field Programmable Gate Array).

[0106] 3, the processing unit 12 has an acquisition unit 30, a reception unit 31, a determination unit 32, a calculation unit 33, a generation unit 34, a determination unit 35, and a provision unit 36, and realizes or executes the functions and actions of information processing described below. Note that the internal configuration of the processing unit 12 is not limited to the configuration shown in FIG. 3, and may have any other configuration as long as it performs the information processing described below.

[0107] [3.3.1. Acquisition part 30] The acquisition unit 30 acquires various pieces of information from an external information processing device, the user terminal 2, the advertiser terminal 3, or the like via the communication unit 10, and stores the acquired information in the storage unit 11.

[0108] For example, the acquisition unit 30 acquires user information, which is information about the user U, from an external information processing device or the user terminal 2 via the communication unit 10, and adds the acquired user information to the user information table in the user information storage unit 20.

[0109] Furthermore, the acquisition unit 30 acquires various types of information from the storage unit 11. For example, the acquisition unit 30 acquires user information, which is information about the user U, from the user information storage unit 20 or the like. The user information acquired by the acquisition unit 30 includes, for example, part or all of at least one of the above-described attribute information and history information.

[0110] Furthermore, the acquisition unit 30 acquires the advertisement provision history from the advertisement provision history storage unit 22. For example, when the reception unit 31 receives a model generation request from the advertiser terminal 3, the acquisition unit 30 acquires provided advertisement information including information on a plurality of advertisements provided to the user U until the user U achieves conversion. The provided advertisement information includes, for example, first provided advertisement information acquired to generate the first prediction model and second provided advertisement information acquired to generate the second prediction model.

[0111] The first provided advertisement information includes, for example, information on one or more advertisements provided to the user U for a predetermined period TA from the time when the user U achieved conversion, for each conversion. For example, the first provided advertisement information includes information on conversion for each combination of advertisements provided to the user U for a predetermined period TA from the time when the user U achieved conversion.

[0112] For example, the first provided advertisement information may be information such as [whether advertisement A is provided, whether advertisement B is provided, whether advertisement C is provided], but is not limited to such an example. For example, the first provided advertisement information may be information such as [the number of times advertisement A is provided, the number of times advertisement B is provided, the number of times advertisement C is provided].

[0113] The first provided advertisement information may be information on a plurality of advertisements provided to the user U until the user U achieves a conversion, and may include, for each conversion, information on the order in which the plurality of advertisements were provided to the user U. In this case, the acquisition unit 30 can convert the advertisement provision history into time-series information in a predetermined format for each conversion. The time-series information in the predetermined format includes, for example, information indicating the order in which the advertisements were provided.

[0114] Here, it is assumed that the prediction model generated by the calculation unit 33 is an RNN such as LSTM or GRU. In this case, the time-series information in a predetermined format is, for example, information including unit information, in chronological order, each unit information including information indicating whether or not each advertisement is provided. For example, the unit information is information such as [whether or not advertisement A is provided, whether or not advertisement B is provided, whether or not advertisement C is provided].

[0115] Furthermore, when the prediction model generated by the calculation unit 33 is a linear regression model, GBDT, or the like, the time-series information in a predetermined format is, for example, information including a plurality of pieces of information indicating the order in which each advertisement is presented in chronological order. The information indicating the order in which each advertisement is presented may be, for example, information such as [presentation order 1 (advertisement A, advertisement B, advertisement C), . . ., presentation order n (advertisement A, advertisement B, advertisement C)]. n is, for example, the maximum number of advertisements presented to each user U.

[0116] The first provided advertisement information may include information obtained from one or more of information about the user U (for example, attribute information or history information) and advertisement display environment information.

[0117] The second provided advertisement information includes, for example, information about the user U (for example, attribute information and history information), advertisement display environment information, and advertisement presentation type information for each conversion. The advertisement presentation type information is, for example, the above-mentioned advertisement combination information or the above-mentioned advertisement presentation order information.

[0118] The acquisition unit 30 also acquires information on an attribution score Sc, which is a score related to the relative contribution of each of a plurality of advertisements provided to the user U until the user U reaches conversion. For example, the acquisition unit 30 acquires the above-described attribution score information from the attribution score storage unit 23.

[0119] [3.3.2. Reception Unit 31] The reception unit 31 receives various requests and information from the user terminal 2 and the advertiser terminal 3 via the communication unit 10.

[0120] For example, the requests received by the reception unit 31 include, but are not limited to, an advertisement request transmitted from the user terminal 2 and a model generation request transmitted from the advertiser terminal 3. The reception unit 31 also receives advertisements transmitted to the information processing device 1 from the advertiser terminal 3 and the like, and adds the received advertisements to the advertisement information table in the advertisement information storage unit 21.

[0121] [3.3.3. Judgment unit 32] When a model generation request sent from an advertiser terminal 3 is received by the receiving unit 31, the determining unit 32 determines whether an advertisement group including multiple advertisements of the advertiser O of the advertiser terminal 3 that sent the model generation request is an advertisement group in which the number of conversions is equal to or greater than a threshold value Th.

[0122] The plurality of advertisements included in the advertisement group may be, for example, advertisements A, B, and C included in the same campaign, but are not limited to such examples. The number of conversions may be, for example, the number of conversions within a predetermined period TA. The threshold value Th is the number required for the prediction accuracy of the first prediction model to be equal to or greater than the accuracy estimated to satisfy the predetermined conditions.

[0123] The number of conversions in the ad group is the total number of conversions corresponding to all ad presentation types in the ad group, but may also be the number of conversions for each ad presentation type in the ad group. The number of conversions for each ad presentation type in the ad group is, for example, the number of conversions in a predetermined period TA, but is not limited to this example. The determination unit 32 can determine whether the number of conversions for each ad presentation type is equal to or greater than a threshold.

[0124] 3.3.4. Calculation unit 33 The calculation unit 33 calculates an attribution score Sc, which is a score related to the relative contribution of each of the plurality of advertisements to conversion. For example, the calculation unit 33 calculates the attribution score Sc for each of the plurality of advertisements included in the advertisement group determined by the determination unit 32 that the number of conversions is equal to or greater than a threshold.

[0125] The calculation unit 33 calculates an attribution score Sc for each of the plurality of advertisements using a first prediction model that predicts conversion based on information about advertisements provided to a user U. For example, the calculation unit 33 calculates the attribution score Sc for each of the plurality of advertisements based on provided advertisement information (e.g., first provided advertisement information) including information about conversion for each combination or each order of providing one or more advertisements among the plurality of advertisements.

[0126] The calculation unit 33 stores attribution score information including the calculated attribution score Sc of each advertisement for each advertisement provision type in the attribution score storage unit 23.

[0127] For example, the calculation unit 33 performs a calculation process to calculate the attribution score Sc for each advertisement A, B, and C by inputting information on the advertisement combination or information on the advertisement presentation order into the first prediction model and obtaining the predicted number of conversions output from the first prediction model.

[0128] For example, assume that the first prediction model is an RNN such as LSTM or GRU. In this case, the calculation unit 33 inputs information about the advertisement combination or information about the advertisement presentation order to the first prediction model, and calculates output values ​​of a softmax function in the first prediction model at this time as the attribution scores Sc of advertisements A, B, and C.

[0129] The output value of the softmax function is obtained by applying the softmax function to the output of hierarchical attention, and is expressed as the probability distribution of advertisements A, B, and C. Hierarchical attention is an example of an attention mechanism.

[0130] The calculation unit 33 can also calculate a value obtained using the output value of the softmax function as the attribution score Sc for each of the multiple advertisements. The value obtained using the output value of the softmax function is, for example, a value obtained by multiplying the output value of the softmax function by a weight w shown in the following formula (11). w=(1-TreatCV / CtrlCV) ···(11)

[0131] TreatCV is a control conversion, and TreatCV is a treat conversion. If the advertisements to be provided are advertisements A, B, and C, a control conversion is, for example, a conversion that occurs when advertisements A, B, and C are not included in the advertisements to be provided, and a treat conversion is, for example, a conversion that occurs when advertisements A, B, and C are included in the advertisements to be delivered.

[0132] In addition, the calculation unit 33 can estimate the CtrlCV for each advertisement presentation order by controlling the delivery of advertisements A, B, and C using, for example, an advertisement evaluation method (e.g., A / B testing), which is a method for evaluating the performance of advertisements.

[0133] In the advertising evaluation method, for example, users who send advertising requests are divided into Treat users and Ctrl users, and delivery control is performed in which one of advertisements A, B, or C is delivered to Treat users, and none of advertisements A, B, or C is delivered to Ctrl users.

[0134] The calculation unit 33 can also treat, for example, the total number of users U to whom none of advertisements A, B, or C was provided during a predetermined period TA prior to conversion from the provided advertising information acquired by the acquisition unit 30 as CtrlCV.

[0135] Furthermore, it is assumed that the first prediction model is a classification model such as a linear regression model or GBDT. In this case, the calculation unit 33 inputs information about the advertisement combination into the first prediction model and performs a process of obtaining pCV from the first prediction model for each advertisement combination. Then, the information processing device 1 performs a process of calculating an attribution score Sc of each advertisement A, B, and C for each advertisement combination using the pCV for each advertisement combination.

[0136] For example, assume that the advertisement combination is a combination of advertisements A, B, and C. In this case, the calculation unit 33 performs the calculations of the above formulas (2) to (4). The calculation unit 33 performs the calculations of the above formulas (5) to (8) based on the calculation results of the above formulas (2) to (4).

[0137] The calculation unit 33 similarly calculates the attribution scores Sc(A), Sc(B), and Sc(C) of advertisements A, B, and C for advertisements A, B, and C for other advertisement combinations than the combination of advertisements A, B, and C. In this way, the calculation unit 33 performs a calculation process to calculate the attribution scores Sc(A), Sc(B), and Sc(C) of advertisements A, B, and C for each advertisement combination.

[0138] In this way, the calculation unit 33 can calculate an attribution score Sc, which is a score indicating the relative contribution of each of the multiple advertisements, based on the conversion predicted by the first prediction model when multiple advertisements are provided to user U and the conversion predicted by the first prediction model when only some of the multiple advertisements are provided to user U.

[0139] The calculation unit 33 can also calculate the attribution score Sc of each advertisement for each advertisement presentation order. For example, the calculation unit 33 calculates the number of conversions for each advertisement presentation order. For example, the calculation unit 33 calculates the number of conversions when the advertisement presentation order is advertisement A, advertisement B, advertisement C by tallying up the number of conversions when the advertisement presentation order is advertisement A, advertisement B, advertisement C.

[0140] Then, the calculation unit 33 generates learning information including information on the order of presenting advertisements and information indicating the number of conversions for each order of presenting advertisements, and generates a first prediction model using the learning information.

[0141] Furthermore, the feature of the first prediction model may include user attribute information in addition to the information on the type of advertisement provision. In this case, the user attribute information is information on the attributes of the user, such as the age and gender of the user, but is not limited to such examples and may include, for example, occupation, place of residence, annual income, etc.

[0142] [3.3.5. Generation unit 34] The generation unit 34 generates a second prediction model that predicts the CVR of the advertisement to be predicted, based on the attribution score Sc of each of the plurality of advertisements acquired by the acquisition unit 30.

[0143] For example, the generation unit 34 generates the second prediction model shown in the above formula (9). The feature value shown in the above formula (9) is the target feature value described above, and is, for example, a feature value obtained from one or more pieces of information selected from the attribute information of the user U, the behavior history of the user U, and the advertisement display environment information.

[0144] The attribute information of user U is information on the demographic attributes of user U, but may include information on psychographic attributes of user U instead of or in addition to the information on the demographic attributes of user U. The behavioral history of user U includes, for example, information on the behavioral history of user U in online services. The advertisement display environment information includes, for example, information such as the category of the website on which the advertisement is provided, the type of application (application of user terminal 2) on which the advertisement is provided, the OS (Operating System) of user terminal 2, and the display position of the advertisement provided.

[0145] The generation unit 34 generates a second prediction model based on second provided advertisement information acquired by the acquisition unit 30, the second provided advertisement information including target feature amounts and information indicating the presence or absence of conversion, and the attribution score Sc of each of the plurality of advertisements acquired by the acquisition unit 30. As described above, the target feature amounts are feature amounts obtained from one or more pieces of information selected from the user attribute information, the user behavior history, and the advertisement display environment information.

[0146] For example, assume that advertisements A, B, and C are provided to user U within a predetermined period TB before the user converts. In this case, the second provided advertisement information includes target feature amounts when advertisements A, B, and C are provided within the predetermined period TB, and information indicating whether user U, who was provided with advertisements A, B, and C within the predetermined period TB, converted, for each advertisement A, B, and C or for each combination of advertisements A, B, and C. The predetermined period TB is the same period as the predetermined period TA, but may be a period different from the predetermined period TA.

[0147] If the second provided advertising information includes information indicating whether or not a conversion has occurred for each of advertisements A, B, and C, the information indicating whether or not a conversion has occurred will include information indicating whether a conversion has occurred in the case of the last advertisement provided before the conversion occurred, and information indicating whether or not a conversion has occurred in other cases, but is not limited to such examples.

[0148] The generation unit 34 generates learning information including target features when advertisements are provided to a user U during a predetermined period TB and the attribution score Sc of each advertisement, and including multiple pieces of information as training information in which the attribution score Sc of each advertisement is used as a label.

[0149] The generation unit 34 generates a second prediction model by machine learning using learning information including the generated plurality of pieces of teacher information. This allows the generation unit 34 to generate a prediction model that reflects the contribution to conversion provided to the user up to the conversion, and can generate a prediction model that appropriately predicts the conversion rate of each advertisement.

[0150] The generation unit 34 generates a CVR prediction model trained using the attribution score Sc as a second prediction model for a group of advertisements determined by the determination unit 32 to have a number of conversions equal to or greater than a threshold.

[0151] Furthermore, the generation unit 34 generates a CVR prediction model that has learned the last touch tendency as a second prediction model for a group of advertisements determined by the determination unit 32 that the number of advertisements that resulted in conversions is less than a threshold. The CVR prediction model that has learned the last touch tendency is a prediction model that determines that the advertisement that corresponds to the conversion is the advertisement that is last provided before the conversion among multiple advertisements.

[0152] For example, if during a given period TB, user U is presented with advertisements A, B, and C in that order and then converts, there will be no conversions corresponding to advertisements A and B, and only a conversion corresponding to advertisement C will be treated as occurring.

[0153] In this way, the generation unit 34 can generate the second prediction model by using the advertisement that was last presented before the conversion among the multiple advertisements included in the group of advertisements that the determination unit 32 has determined to have resulted in less than a threshold number of conversions as the advertisement corresponding to the conversion.

[0154] In addition, the generation unit 34 can generate a second prediction model using the attribution score Sc as a label for advertising types for which the determination unit 32 determines that the number of conversions is greater than or equal to a threshold, and using conversion as a label for advertising types for which the determination unit 32 determines that the number of conversions is not greater than a threshold.

[0155] [3.3.6. Decision Unit 35] The determination unit 35 determines the bid amount of the advertisement to be predicted based on the pCVR (predicted value of conversion rate) obtained using the second prediction model generated by the generation unit 34.

[0156] For example, when the reception unit 31 receives an advertisement request, the determination unit 35 can calculate the pCVRs of advertisements A, B, and C using the second prediction model generated by the generation unit 34.

[0157] The determining unit 35, for example, inputs target features corresponding to the advertisement request received by the receiving unit 31 into the second prediction model, and obtains the pCVR of each of advertisements A, B, and C output from the second prediction model, thereby calculating the pCVR of advertisements A, B, and C. The target features corresponding to the advertisement request include, for example, features obtained from attribute information and behavior history of user U of user terminal 2 that sent the advertisement request.

[0158] For example, in the example shown in Fig. 1(b), the pCVR of advertisement A is "0.05", the pCVR of advertisement B is "0.1", and the pCVR of advertisement C is "0.2". In this way, the information processing device 1 can calculate a pCVR that reflects the contribution to conversion provided to the user up to the conversion.

[0159] The determination unit 35 uses the calculated pCVR to calculate the bid amount (Bid) for each of the advertisements A, B, and C. The determination unit 35 calculates the bid amount (Bid) for each of the advertisements A, B, and C, for example, using the above formula (10).

[0160] [3.3.7.Providing Department 36] The providing unit 36 ​​provides various types of information to the user U and the advertiser O. The providing unit 36 ​​can provide the advertiser O with the information on the attribution score Sc of each advertisement calculated by the calculation unit 33 by transmitting the information on the attribution score Sc of each advertisement to the advertiser terminal 3.

[0161] In addition, when an advertising request is accepted by the accepting unit 31, the providing unit 36 ​​performs automatic bidding processing based on the bid amount for each advertisement determined by the determining unit 35, and can transmit the advertisement selected by the automatic bidding processing to the user terminal 2 that is the source of the advertisement request.

[0162] In the automatic bidding process, automatic bidding is performed in which the bid with the second highest bid price, called second price action, is won, but automatic bidding in which the highest bid price, called first price auction, is won may also be performed.Furthermore, in the automatic bidding process, automatic bidding other than second price action or first price auction may also be performed.

[0163] [4. Processing Procedure] Next, a procedure of information processing by the processing unit 12 of the information processing device 1 according to the embodiment will be described. Fig. 7 is a flowchart showing an example of information processing by the processing unit 12 of the information processing device 1 according to the embodiment, which is executed for each advertisement group, for example.

[0164] 7, the processing unit 12 of the information processing device 1 determines whether it is time to generate a model (step S10). The model generation timing may be, for example, the timing when a model generation request is received from the user terminal 2, the timing when a predetermined period has arrived, or the timing when the number of new conversions exceeds a threshold, but is not limited to these examples.

[0165] When the processing unit 12 determines that it is time to generate a model (step S10: Yes), it determines whether the number of conversions is equal to or greater than a threshold (step S11). When the processing unit 12 determines that the number of conversions is equal to or greater than a threshold (step S11: Yes), it generates a first prediction model based on the first provided advertisement information and the like (step S12).

[0166] Then, the processing unit 12 calculates an attribution score Sc for each advertisement using the first prediction model generated in step S12 (step S13). After that, the processing unit 12 generates a second prediction model based on the attribution score Sc for each advertisement calculated in step S13 and the second provided advertisement information (step S14).

[0167] When the processing of step S14 is completed, when it is determined that the time for model generation has not yet arrived (step S10: No), or when it is determined that the number of conversions is not greater than or equal to the threshold (step S11: No), the processing unit 12 determines whether or not an advertising request has been received (step S15).

[0168] When the processing unit 12 determines that the advertisement request has been received (step S15: Yes), it determines the bid amount for each advertisement (step S16), and determines advertisements to be provided by automatic bidding processing based on the bid amounts determined in step S16 (step S17).Then, the processing unit 12 provides the advertisements determined in step S17 (step S18).

[0169] When the processing of step S18 is completed or when it is determined that an advertisement request has not been received (step S15: No), the processing unit 12 determines whether or not the operation end time has arrived (step S19). The processing unit 12 determines that the operation end time has arrived when, for example, the power of the information processing device 1 is turned off.

[0170] If the processing unit 12 determines that the operation end time has not yet arrived (step S19: No), it proceeds to step S10, and if it determines that the operation end time has arrived (step S19: Yes), it terminates the processing shown in Figure 7.

[0171] [5. Modifications] In the above-described example, the calculation unit 33 calculates the attribution score Sc for each advertisement using the first prediction model, but the example is not limited to this, and the calculation unit 33 can also calculate the attribution score Sc for each advertisement without using a prediction model.

[0172] For example, the calculation unit 33 tallies the number of conversions for each advertisement type. Then, the calculation unit 33 can calculate the attribution score Sc for each advertisement based on the number of conversions for each advertisement type. In this case, the calculation unit 33 uses CV(·) indicating the number of conversions instead of f(·) indicating the first prediction model. For example, assume that the advertisement types are advertisements A, B, and C in this order.

[0173] In this case, the calculation unit 33 uses, for example, CV(φ), CV(A), CV(A, B), and CV(A, B, C) instead of f(φ), f(A), f(A, B), and f(A, B, C) in the above formulas (2) to (5).The calculation unit 33 then calculates the attribution scores Sc(A), Sc(B), and Sc(C) of advertisements A, B, and C by performing the calculations of the above formulas (2) to (8).

[0174] The calculation unit 33 can also tally the number of conversions for each combination of the advertisement provision type and user information including feature amounts obtained from the attribute information and history information of the user U. In this case, the calculation of CV(·) includes feature amounts obtained from the history information of the user U, and the calculation unit 33 calculates the attribution score Sc of each advertisement based on the number of conversions for each combination of the user information and the advertisement provision type.

[0175] [6. Hardware Configuration] The information processing device 1 according to the embodiment described above is realized by, for example, a computer 80 configured as shown in Fig. 8. Fig. 8 is a hardware configuration diagram showing an example of the computer 80 that realizes the functions of the information processing device 1 according to the embodiment. The computer 80 has a CPU 81, a RAM 82, a ROM (Read Only Memory) 83, an HDD (Hard Disk Drive) 84, a communication interface (I / F) 85, an input / output interface (I / F) 86, and a media interface (I / F) 87.

[0176] The CPU 81 operates and controls each part based on programs stored in the ROM 83 or the HDD 84. The ROM 83 stores a boot program executed by the CPU 81 when the computer 80 starts up, programs that depend on the hardware of the computer 80, and the like.

[0177] The HDD 84 stores programs executed by the CPU 81, data used by such programs, etc. The communication interface 85 receives data from other devices via the network N (see FIG. 2) and sends it to the CPU 81, and transmits data generated by the CPU 81 to other devices via the network N.

[0178] The CPU 81 controls output devices such as a display and a printer, and input devices such as a keyboard or a mouse, via the input / output interface 86. The CPU 81 acquires data from the input devices via the input / output interface 86. The CPU 81 also outputs generated data to the output devices via the input / output interface 86.

[0179] The media interface 87 reads a program or data stored in a recording medium 88 and provides it to the CPU 81 via the RAM 82. The CPU 81 loads the program or data from the recording medium 88 onto the RAM 82 via the media interface 87 and executes the loaded program. The recording medium 88 is, for example, an optical recording medium such as a DVD (Digital Versatile Disc) or a PD (Phase Change Rewritable Disc), a magneto-optical recording medium such as an MO (Magneto-Optical disk), a tape medium, a magnetic recording medium, or a semiconductor memory.

[0180] For example, when the computer 80 functions as the information processing device 1 according to the embodiment, the CPU 81 of the computer 80 executes programs loaded onto the RAM 82 to realize the functions of the processing unit 12. In addition, the HDD 84 stores data in the storage unit 11. The CPU 81 of the computer 80 reads and executes these programs from a recording medium 88, but as another example, the CPU 81 may obtain these programs from another device via the network N.

[0181] [7. Other] Furthermore, among the processes described in the above embodiments, all or part of the processes described as being performed automatically can be performed manually, or all or part of the processes described as being performed manually can be performed automatically using known methods. In addition, the information including the processing procedures, specific names, various data, and parameters shown in the above documents and drawings can be changed as desired unless otherwise specified. For example, the various information shown in each drawing is not limited to the information shown in the drawings.

[0182] Furthermore, the components of each device shown in the figure are conceptual functional components and do not necessarily have to be physically configured as shown in the figure. In other words, the specific form of distribution and integration of each device is not limited to that shown in the figure, and all or part of them can be functionally or physically distributed and integrated in any unit depending on various loads, usage conditions, etc.

[0183] For example, the information processing device 1 described above may be realized by a terminal device and a server computer, or may be realized by multiple server computers. Furthermore, depending on the function, the configuration can be flexibly changed, such as by calling an external platform using an API (Application Programming Interface) or network computing.

[0184] Furthermore, the above-described embodiments and modifications can be combined as appropriate within the scope of not causing any contradiction in the processing content.

[0185] [8. Effects] As described above, the information processing device 1 according to the embodiment includes the acquisition unit 30 and the generation unit 34. The acquisition unit 30 acquires a score related to the relative contribution to conversion of each of a plurality of advertisements provided to the user U until the user U reaches conversion. The generation unit 34 generates a prediction model that predicts the conversion rate of the advertisement to be predicted, based on the score of each of the plurality of advertisements acquired by the acquisition unit 30. This allows the information processing device 1 to generate a prediction model that appropriately predicts the conversion rate of the advertisement.

[0186] The information processing device 1 also includes a determination unit 35 that determines the bid amount for the advertisement to be predicted based on the predicted value of the conversion rate obtained using the prediction model generated by the generation unit 34. This allows the information processing device 1 to appropriately determine the bid amount for the advertisement.

[0187] The information processing device 1 also includes a calculation unit 33 that calculates a score related to the relative contribution of each of the plurality of advertisements to conversion, based on provided advertisement information including information related to conversion for each combination or each order of provision of one or more advertisements among the plurality of advertisements, and the acquisition unit 30 acquires the score of each of the plurality of advertisements calculated by the calculation unit 33. This enables the information processing device 1 to generate a prediction model that appropriately predicts the conversion rate of advertisements.

[0188] Furthermore, the information processing device 1 includes a generation unit 34 that includes a determination unit 32 that determines whether a group of advertisements provided to the user U before conversion is an advertisement group for which the number of conversions is equal to or greater than a threshold, and a calculation unit 33 that calculates scores related to the relative contribution to conversion of a plurality of advertisements included in the advertisement group determined by the determination unit 32 to be equal to or greater than the threshold. This enables the information processing device 1 to generate a prediction model that appropriately predicts the conversion rate of advertisements.

[0189] Furthermore, the generation unit 34 generates a prediction model by setting the advertisement that was last presented before the conversion as the advertisement corresponding to the conversion among the advertisements included in the advertisement group determined by the determination unit 32 that the number of conversions is less than the threshold value. This enables the information processing device 1 to generate a prediction model that appropriately predicts the conversion rate of the advertisement.

[0190] Furthermore, the calculation unit 33 calculates a score indicating the relative contribution of each of a plurality of advertisements using a model that predicts conversion based on advertisement information provided to the user U. This enables the information processing device 1 to generate a prediction model that appropriately predicts the conversion rate of advertisements.

[0191] Furthermore, the calculation unit 33 calculates a score indicating the relative contribution of each of the multiple advertisements, based on the conversion predicted by the model when multiple advertisements are provided to the user U and the conversion predicted by the model when some of the multiple advertisements are provided to the user U. This enables the information processing device 1 to generate a prediction model that appropriately predicts the conversion rate of the advertisements.

[0192] The calculation unit 33 also calculates the output value of the softmax function used in the model or a value obtained using the output value as a score indicating the relative contribution of each of the multiple advertisements. This allows the information processing device 1 to generate a prediction model that appropriately predicts the conversion rate of the advertisement.

[0193] The above describes the embodiments of the present application in detail based on the drawings, but this is merely an example, and the present invention can be implemented in other forms that include the embodiments described in the Disclosure of the Invention section and that have been modified and improved in various ways based on the knowledge of those skilled in the art.

[0194] Furthermore, the above-mentioned "section, module, unit" can be read as "means" or "circuit," etc. For example, an acquisition unit can be read as an acquisition means or an acquisition circuit. [Explanation of symbols]

[0195] 1. Information processing equipment 2. User terminal 3. Advertiser terminal 10. Communications Department 11 Storage section 12 Processing section 20 User information storage unit 21 Advertising information storage unit 22 Ad serving history memory unit 23 Attribution score storage 30 Acquisition Department 31 Reception 32 Judgment section 33 Calculation section 34 Generation part 35 Decision Section 36 Providing Department 100 Information Processing Systems

Claims

1. An acquisition unit that acquires a score relating to a relative contribution of each of a plurality of advertisements provided to a user until the user reaches a conversion; a generation unit that generates a prediction model that predicts a conversion rate of an advertisement to be predicted based on the score of each of the plurality of advertisements acquired by the acquisition unit.

23. An information processing apparatus comprising:

2. A determination unit that determines a bid amount for the advertisement to be predicted based on a predicted value of the conversion rate obtained by using the prediction model generated by the generation unit.

2. The information processing apparatus according to claim 1,

3. a calculation unit that calculates a score related to a relative contribution of each of the plurality of advertisements to the conversion based on provided advertisement information including information related to the conversion for each combination or each provision order of one or more advertisements among the plurality of advertisements, The acquisition unit is The score of each of the plurality of advertisements calculated by the calculation unit is obtained.

3. The information processing apparatus according to claim 1, wherein the information processing apparatus is a computer.

4. The generation unit is a determination unit that determines whether or not a group of advertisements provided to the user before the conversion is a group of advertisements that have led to the conversion in a number equal to or greater than a threshold value; The calculation unit is Calculating a score related to a relative contribution to the conversion of a plurality of advertisements included in an advertisement group determined by the determination unit to have a number of advertisements that have led to the conversion equal to or greater than a threshold value 4. The information processing apparatus according to claim 3.

5. The generation unit is The prediction model is generated by determining, as an advertisement corresponding to the conversion, an advertisement that was last provided before the conversion among a plurality of advertisements included in an advertisement group determined by the determination unit to have a number of conversions that is less than a threshold value.

5. The information processing apparatus according to claim 4.

6. The calculation unit is Calculating a score indicating the relative contribution of each of the plurality of advertisements using a model that predicts conversions based on information about the advertisements provided to the user.

4. The information processing apparatus according to claim 3.

7. The calculation unit is Calculating a score indicating a relative contribution of each of the plurality of advertisements based on a conversion predicted by the model when the plurality of advertisements are provided to the user and a conversion predicted by the model when a portion of the plurality of advertisements are provided to the user.

7. The information processing apparatus according to claim 6,

8. The calculation unit is An output value of a softmax function used in the model or a value obtained by using the output value is calculated as a score indicating a relative contribution of each of the plurality of advertisements.

7. The information processing apparatus according to claim 6,

9. 1. A computer-implemented information processing method, comprising: acquiring a score relating to a relative contribution of each of a plurality of advertisements provided to a user to the conversion; and generating a prediction model for predicting a conversion rate of an advertisement to be predicted based on the score of each of the plurality of advertisements acquired by the acquisition step.

23. An information processing method comprising:

10. An acquisition step of acquiring a score relating to a relative contribution to a conversion of each of a plurality of advertisements provided to the user until the user achieves the conversion; a generating step of generating a prediction model for predicting a conversion rate of an advertisement to be predicted, based on the score of each of the plurality of advertisements acquired by the acquiring step, 2. An information processing program comprising:

Citation Information

Patent Citations

  • Information processing device, information processing method and information processing program

    JP2019220052A

Cited By

  • Behavior prediction device and behavior prediction method

    JP7920479B1