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

The information processing device uses machine learning to enhance advertising sales by predicting advertiser preferences and product matches, addressing the inefficiencies in current advertising distribution systems.

JP7577699B2Active Publication Date: 2024-11-05LY CORP
View PDF 4 Cites 0 Cited by

Patent Information

Application Number
JP2022005498
Authority / Receiving Office
JP · JP
Patent Type
Patents
Current Assignee / Owner
Filing Date
2022-01-18
Publication Date
2024-11-05
Estimated Expiration
2041-04-23

AI Technical Summary

Technical Problem

Existing advertising distribution systems lack effective support for advertisers in purchasing advertising products, as they rely heavily on sales know-how and experience rather than data-driven methods.

Method used

An information processing device that utilizes machine learning models to estimate advertising costs and match target advertisers with suitable advertising products, incorporating advertiser and product vectors, market gaps, and sales results to predict purchasing behavior.

Benefits of technology

Enhances the sales of advertising products by accurately predicting which products advertisers are likely to purchase, allowing for targeted and efficient sales activities.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure 0007577699000001
    Figure 0007577699000001
  • Figure 0007577699000002
    Figure 0007577699000002
  • Figure 0007577699000003
    Figure 0007577699000003
Patent Text Reader

Abstract

An information processing device, an information processing method, and an information processing program are provided that support the sale of advertised products to advertisers. [Solution] In an information processing device (10) that outputs information on advertised products estimated from information on target advertisers, a purchase amount estimation unit (123) estimates expected advertising costs, which are the expected amounts of advertising costs to be paid by the target advertiser, from the information on the target advertiser using a purchase amount estimation model trained using information on advertisers whose advertising costs satisfy predetermined conditions as correct answer data. A gap calculation unit (124) calculates gap information indicating the difference between the expected advertising costs estimated by the purchase amount estimation unit (123) and the actual advertising costs, which are the actual amounts of advertising costs ordered from the target advertiser.
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, advertisement distribution via the Internet has become popular. For example, advertisement content for a company, product, etc. is displayed in an advertisement space set in an advertisement medium such as a web page, and when the advertisement content is clicked, the user is redirected to the advertiser's web page (see Patent Document 1).

[0003] Advertisements in ad distribution can take various forms, such as the advertising medium on which the advertisement is displayed, the position and size of the advertisement on the advertising medium, whether it is a still image or a video, and whether it is linked to user operations, and an advertiser purchases an advertisement of any form from an advertising distributor or the like while taking into consideration the advertising cost, and displays it on the advertising medium. Hereinafter, an advertisement of any form sold to an advertiser is referred to as an "advertising product." [Prior art documents] [Patent documents]

[0004] [Patent Document 1] JP 2002-203119 A Summary of the Invention [Problem to be solved by the invention]

[0005] The above-mentioned conventional techniques have room for further improvement in terms of supporting the sale of advertising products to advertisers.

[0006] The present application has been made in consideration of the above, and has an object to support sales of advertising products to advertisers. [Means for solving the problem]

[0007] The information processing device according to the present application includes a purchase amount estimation unit that estimates expected advertising expenses, which are the expected amount of advertising expenses to be paid by a target advertiser, from information on the target advertiser using a purchase amount estimation model trained using information on advertisers whose advertising expenses satisfy specified conditions as correct answer data, and a gap calculation unit that calculates gap information indicating the difference between the expected advertising expenses estimated by the purchase amount estimation unit and the actual advertising expenses, which are the actual amounts of advertising expenses received from the target advertiser. Effect of the Invention

[0008] According to one aspect of the embodiment, it is possible to support the sale of advertising products to advertisers. [Brief description of the drawings]

[0009] [Figure 1] FIG. 1 is a diagram illustrating an example of information processing according to an embodiment. [Diagram 2] FIG. 2 is a diagram illustrating an example of the configuration of the information processing device according to the embodiment. [Diagram 3] FIG. 3 is a diagram illustrating an example of information stored in an advertiser vector storage unit according to the embodiment. [Figure 4] FIG. 4 is a diagram illustrating an example of information stored in a product vector storage unit according to the embodiment. [Diagram 5] FIG. 5 is a diagram illustrating an example of information stored in a market gap storage unit according to the embodiment. [Figure 6] FIG. 6 is a diagram illustrating an example of information stored in a sales result storage unit according to the embodiment. [Figure 7] FIG. 7 is a flowchart showing an example of the flow of the target setting process according to the embodiment. [Figure 8] FIG. 8 is a flowchart showing an example of the flow of a product information output process according to the embodiment. [Figure 9] FIG. 9 is a hardware configuration diagram showing an example of a computer that realizes the functions of the information processing device according to the embodiment. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

[0010] Hereinafter, the information processing device, the information processing method, and the information processing program according to the present application 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 the embodiments.

[0011] [Embodiment] [1. 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.

[0012] 1, an information processing device 10 according to the embodiment is an information processing device that matches target advertisers with advertised products. The information processing device 10 is realized by, for example, a server device or a cloud system.

[0013] Here, the target advertiser is, for example, an object set as a sales destination of the advertising product. In other words, the target advertiser from the viewpoint of the seller of the advertising product corresponds to, for example, an object to sell the advertising product. The target advertiser is, for example, an enterprise (a company, a corporation, etc.), but is not limited to an enterprise and may be a national government, a prefecture, a city, a town, a village, or other such local government.

[0014] An advertising product is an advertisement of any aspect sold to an advertiser (including a target advertiser). In addition, any aspect of an advertisement is, for example, an advertising medium on which the advertisement is displayed, the position and size of the advertisement on the advertising medium, whether it is a still image or a video, and whether it is linked to a user operation. The concept of an "advertising product" is also the advertising product itself, and includes customizable parts of the advertising product such as functions and settings, and in the case of search advertising, the intent of a search query, etc. For example, advertising products with different customization contents such as functions and settings may be treated as different products. The intent of a search query may be vectorized using a technology such as DQNN (Deep Q Neural Network). In addition, an advertising product may be represented in a hierarchical structure, such as the product itself and the functions and settings of the advertising product.

[0015] Conventionally, when advertising vendors such as advertising distributors sell advertising products to target advertisers, they narrow down advertising products that the target advertisers are likely to purchase from among many existing advertising products based on the sales know-how and experience they have cultivated. The information processing device 10 supports the sale of advertising products by automatically matching such target advertisers with advertising products using a machine learning model (result estimation model).

[0016] Specifically, when the information processing device 10 receives information about a target advertiser (step S01), it estimates the advertised product that the target advertiser will purchase using a result estimation model (step S02). Then, the information processing device 10 outputs information about the estimated advertised product (step S03).

[0017] Here, the result estimation model is a model that learns the relationship between information about an advertiser and an advertised product purchased by the advertiser. A method for generating the result estimation model will be described below.

[0018] The information processing device 10 has an advertiser vector storage unit 132 that stores an advertiser vector as information related to an advertiser. The advertiser vector is a characteristic vector of an advertiser.

[0019] The advertiser vector is generated using an advertiser vector model, which will be described later. The advertiser vector model is a model that has been trained to output an advertiser vector similar to the advertiser vectors of other advertisers with similar information when various information on an advertiser, such as various financial figures such as sales, profit margins, and advertising expenses, text explaining business details and management issues, the industry to which the advertiser belongs, capital structure, market, the contents of information posted on the Internet, search data (jump source query), advertising expenses, and advertising media, is input. Vectors being similar means, for example, that the cosine distances are similar, that is, that the cosine similarity is high. A specific example of a method for generating an advertiser vector will be described later. Note that "advertising expenses" may be read as "advertising sales."

[0020] The information processing device 10 also includes a product vector storage unit 134 that stores a product vector. The product vector is a feature vector of an advertised product.

[0021] The product vector is generated using a product vector model, which will be described later. The product vector model is a model that is trained to output a product vector similar to the product vectors of other advertised products having similar information when various information including purchase patterns related to the advertised product, such as information on the advertiser who made the purchase (advertiser vector), purchase amount, purchase period (distribution period), search data (jump source query), CTR (Click Through Rate), CVR (Conversion Rate), and advertising medium, is input. For example, the product vector model generates a product vector from information on the advertised product so that the product vectors of advertised products purchased by the same advertiser are similar to each other. A specific example of a method for generating a product vector will be described later.

[0022] The information processing device 10 also has a market gap storage unit 136 that stores a market gap (an example of gap information). The market gap is the difference between the expected amount of advertising expenses to be paid by an advertiser (expected advertising expenses) and the actual amount of advertising expenses currently received from the advertiser (actual advertising expenses).

[0023] For example, the information processing device 10 may calculate a prospective advertising cost based on the advertiser vector stored in the advertiser vector storage unit 132, and calculate the difference between the calculated prospective advertising cost and the actual advertising cost as the market gap. Furthermore, the information processing device 10 may calculate the prospective advertising cost using a learning model (purchase amount estimation model) that is trained to output the purchase amount of each advertised product when various information on an advertiser (particularly, an advertiser whose advertising sales have increased through sales) is input. A specific example of a method for calculating the market gap will be described later.

[0024] Furthermore, the information processing device 10 has a sales result storage unit 138 that stores sales result information indicating the content and results of sales activities for each advertiser. The sales result information may include, as the content of the sales activities, information such as whether or not sales activities were conducted for the advertiser, the type of sales activities conducted, the specific content of the sales activities conducted, and the proposed advertising product. The sales result information may also include, as the result of the sales activities, information such as an increase or decrease in sales due to sales, advertising expenses after the sales, and whether or not the proposed advertising product was sold (whether or not orders or upselling were successful). Note that the sales result information may include not only the results of human sales activities, but also, for example, the results when the advertising product is purchased via an advertising submission tool or the like.

[0025] Then, the information processing device 10 performs machine learning using the advertiser vector, the product vector, and the market gap as inputs and the sales result information as output, thereby generating a result estimation model.

[0026] The information processing device 10 acquires the advertiser vector of the target advertiser from the advertiser vector storage unit 132, and acquires the market gap of the target advertiser from the market gap storage unit 136. The information processing device 10 also acquires an arbitrary product vector from the product vector storage unit 134. The information processing device 10 also inputs each piece of information of the acquired advertiser vector, market gap, and product vector into a result estimation model. Then, the information processing device 10 estimates the advertised product to be purchased by the target advertiser using the sales result information output from the result estimation model.

[0027] For example, the information processing device 10 can estimate the advertised product that is most likely to be purchased or the advertised product with the largest purchase amount by fixing the advertiser vector and market gap and replacing the product vector among the various information input into the result estimation model.

[0028] Furthermore, by calculating the market gap, the information processing device 10 can find out advertisers whose expected advertising costs are significantly different from their actual advertising costs. For example, the information processing device 10 can find out advertisers whose expected advertising costs are 3 billion yen but whose actual advertising costs are only 1 billion yen. By conducting sales activities targeting advertisers whose expected advertising costs are significantly different from their actual advertising costs, advertising sellers can efficiently increase advertising orders. Furthermore, by taking into account the content of sales activities, it can be estimated, for example, whether advertising orders will increase by conducting human sales activities or not, and efficient sales activities can be conducted.

[0029] In this way, the information processing device 10 according to the embodiment can support the sale of advertised products to advertisers.

[0030] 2. Example of the configuration of an information processing device Next, a configuration example of the information processing device 10 will be described with reference to Fig. 2. Fig. 2 is a diagram showing a configuration example of the information processing device 10 according to an embodiment. As shown in Fig. 2, the information processing device 10 has a communication unit 110, a control unit 120, and a storage unit 130.

[0031] The communication unit 110 is realized by, for example, a network interface card (NIC) etc. The communication unit 110 transmits and receives information to and from an external device via a network such as various wireless communication networks, such as 4G (Generation), 5G, LTE (Long Term Evolution), Wifi (registered trademark), or wireless LAN (Local Area Network), or various wired communication networks.

[0032] The storage unit 130 is realized by, for example, a semiconductor memory element such as a RAM or a flash memory, or a storage device such as a hard disk or an optical disk. The storage unit 130 also has an advertiser vector model storage unit 131, an advertiser vector storage unit 132, a product vector model storage unit 133, a product vector storage unit 134, a purchase amount estimation model storage unit 135, a market gap storage unit 136, a target information storage unit 137, a sales result storage unit 138, and a result estimation model storage unit 139.

[0033] The advertiser vector model storage unit 131 stores an advertiser vector model. The advertiser vector model is a learning model that converts information about an advertiser into an advertiser vector.

[0034] For example, when various information about an advertiser, such as various expenses including advertising and promotion costs, capital structure, market, content of information posted on the Internet, search data (jump source query), advertised products purchased, purchase amount of the advertised products, advertising cost which is the total amount of the advertised products purchased, and advertising media, is input into the model, the information processing device 10 learns the model so as to output an advertiser vector similar to the advertiser vectors of other advertisers with similar information.

[0035] Here, "similar" advertisers do not simply mean that their markets are similar, but that the content and results of their advertising sales are similar.

[0036] For example, the information processing device 10 calculates a similarity score indicating the similarity between advertisers, including the details of sales activities for a certain advertiser (whether sales activities were conducted or not, the type of sales activities conducted, the specific details of the sales activities conducted, the advertised product proposed, etc.) and their results (for example, advertised product A was proposed and purchased, but advertised product B was proposed and not purchased, etc.), the medium in which the advertisement is placed (television, radio, newspapers, magazines, the Internet, transportation, etc.), advertising costs, etc. Then, the information processing device 10 performs learning of an advertiser vector model that outputs advertiser vectors from various information regarding advertisers, such that the higher the calculated similarity score, the higher the similarity of the advertiser vectors.

[0037] In addition, in the learning stage of the advertiser vector model, the information processing device 10 may use data obtained by vectorizing the information about the advertiser using a technology such as W2V (Word2Vec) instead of the information about the advertiser itself (i.e., text data) as input data. For example, the information processing device 10 may use a conversion model such as W2V that converts words with similar meanings into similar vectors to vectorize the words input as the information about the advertiser, and may use vectors obtained by combining the vectors or composite vectors that are a composite of the vectors as input data to learn the advertiser vector model.

[0038] Furthermore, the information processing device 10 may use, for example, corporate information published in the Kaisha Shikiho and the contents of various analyst reports as information about advertisers. For example, the information processing device 10 may vectorize the corporate information published in the Kaisha Shikiho using a technology such as W2V, and may perform learning of an advertiser vector model so as to output similar advertiser vectors for advertisers whose vector data are similar.

[0039] The advertiser vector storage unit 132 stores the advertiser vector. Fig. 3 is a diagram showing an example of information stored in the advertiser vector storage unit 132 according to the embodiment. As shown in Fig. 3, the advertiser vector storage unit 132 includes information such as "advertiser ID" and "advertiser vector".

[0040] The "Advertiser ID" field stores identification information for identifying an advertiser. Note that the "Advertiser ID" may also include identification information for a company that is not yet an advertiser, i.e., a company that is a candidate for a new advertiser.

[0041] The "Advertiser Vector" field stores an advertiser vector corresponding to an advertiser identified by an advertiser ID. In Fig. 3, an example is shown in which the advertiser vector is two-dimensional vector data "10,24", but the advertiser vector may be multi-dimensional (N-dimensional) vector data and is not limited to two dimensions.

[0042] The product vector model storage unit 133 stores a product vector model. The product vector model is a learning model that converts information about an advertisement product into a product vector.

[0043] For example, the information processing device 10 identifies advertised products purchased by the same advertiser. Then, the information processing device 10 performs learning of a product vector model in which various information related to the advertised products is input and product vectors are output so that vectors of advertised products purchased by the same advertiser are similar.

[0044] As an example, the information processing device 10 calculates a similarity score between advertisement products of information related to the advertisement products, such as information on the advertiser who made the purchase (advertiser vector), purchase amount, purchase period (distribution period), search data (jump source query), CTR, CVR, and advertisement medium.The information processing device 10 may then perform learning of a product vector model that outputs a product vector from information related to the advertisement products, such that the higher the calculated similarity score, the higher the similarity of the product vectors.

[0045] The product vector storage unit 134 stores the product vector. Fig. 4 is a diagram showing an example of information stored in the product vector storage unit 134 according to the embodiment. As shown in Fig. 4, the product vector storage unit 134 includes information such as "advertised product ID" and "product vector".

[0046] The "advertised product ID" field stores identification information for identifying an advertised product. The "product vector" field stores a product vector corresponding to an advertised product identified by the advertised product ID. FIG. 4 shows an example in which the product vector is two-dimensional vector data "33, 15", but the product vector may be multi-dimensional (N-dimensional) vector data and is not limited to two dimensions.

[0047] The purchase amount estimation model storage unit 135 stores a purchase amount estimation model. The purchase amount estimation model is a model that learns the relationship between information about an advertiser and advertised products that the advertiser has purchased in the past and the purchase amount thereof. Specifically, the purchase amount estimation model is a learning model that is trained to output the purchase amount of each advertised product when information about an advertiser (advertiser vector) is input.

[0048] For example, the information processing device 10 identifies an advertiser whose advertising sales have increased due to sales activities. The information processing device 10 may identify an advertiser whose advertising sales have increased due to sales activities, for example, based on information stored in the sales result storage unit 138 described later. Then, the information processing device 10 trains the purchase amount estimation model using various information related to the identified advertiser as correct answer data. Specifically, the information processing device 10 trains the purchase amount estimation model using a set of information related to the advertiser whose advertising sales have increased due to sales activities (excluding the advertised product and purchase amount purchased by the advertiser) and the advertised product and purchase amount purchased by the advertiser as correct answer data. That is, the information processing device 10 trains the purchase amount estimation model using a technique such as backpropagation so that the purchase amount of the advertised product purchased by the advertiser is output when information related to the advertiser whose advertising sales have increased due to sales activities is input to the purchase amount estimation model.

[0049] In addition, in the learning stage of the purchase amount estimation model, the information processing device 10 may learn the purchase amount estimation model by limiting the information on the advertisers who have previously purchased the advertised product of a specific advertising seller. Furthermore, without being limited to this, the information processing device 10 may use, as input data, not only the information on the advertisers whose advertising sales have increased due to sales activities, but also information on other advertisers, i.e., advertisers whose advertising sales are unknown whether or not they will increase due to sales activities. In this case, the information processing device 10 may weight the information on the advertisers whose advertising sales have increased due to sales activities so that the learning result reflects the information. In other words, the information processing device 10 may learn the purchase amount estimation model by using information on a plurality of advertisers adjusted so that the weighting of the information on the advertisers whose advertising sales have increased due to sales activities is heavier than the information on the advertisers whose advertising sales are unknown whether or not they will increase due to sales activities.

[0050] In addition, the information used as the correct answer data is not limited to information on advertisers whose advertising sales have increased through sales activities. For example, the information processing device 10 may train the purchase amount estimation model using information on an advertiser whose advertising expenses exceed a predetermined threshold as correct answer data. In addition, the information processing device 10 may train the purchase amount estimation model using information on an advertiser whose advertising expenses are higher than the average advertising expenses of multiple advertisers belonging to the same business field as correct answer data. In addition, the information processing device 10 may train the purchase amount estimation model using information on an advertiser whose advertising sales of each product exceed a predetermined threshold as correct answer data. For example, the information processing device 10 may train the purchase amount estimation model using information on an advertiser whose purchase amount of at least one of the purchased advertised products exceeds a threshold as correct answer data.

[0051] The market gap storage unit 136 stores the market gap. The market gap is information indicating the difference between the expected advertising cost, which is the expected amount of advertising cost to be paid by an advertiser, and the actual advertising cost, which is the advertising cost currently ordered from the advertiser (the total purchase amount of each advertised product). The expected advertising cost is calculated by the gap calculation unit 124, which will be described later, using the purchase amount estimation model described above. This point will be described later.

[0052] Fig. 5 is a diagram showing an example of information stored in the market gap storage unit 136 according to the embodiment. As shown in Fig. 5, the market gap storage unit 136 includes information such as "advertiser ID" and "market gap". In the "advertiser ID" item, identification information for identifying an advertiser is stored. In the "market gap" item, a market gap corresponding to an advertiser identified by an advertiser ID is stored.

[0053] The target information storage unit 137 stores, for example, identification information of advertisers determined as target advertisers by the determination unit 125 described later.

[0054] The sales result storage unit 138 stores sales result information indicating the results of sales activities for each advertiser. Fig. 6 is a diagram showing an example of information stored in the sales result storage unit 138 according to the embodiment.

[0055] As shown in FIG. 6, the sales result storage unit 138 stores items such as "advertiser ID", "proposed advertising product", "whether or not an order has been received", "purchase amount", "increase / decrease in advertising sales", and "amount of increase / decrease". The "advertiser ID" item stores identification information for identifying an advertiser. The "proposed advertising product" item stores identification information of an advertising product proposed to an advertiser identified by an advertiser ID. The "whether or not an order has been received" item stores information on whether or not an order has been received for the proposed advertising product. The "purchase amount" item stores the purchase amount of the proposed advertising product. The "increase / decrease in advertising sales" item stores information indicating whether or not advertising sales by an advertiser identified by an advertiser ID will increase. The "amount of increase / decrease" item stores the amount of increase / decrease in advertising sales of the entire advertiser identified by an advertiser ID. Note that the "amount of increase / decrease" may be expressed as the difference between advertising sales for the previous term and the estimated amount of advertising sales for the current term (for example, the amount of advertising sales at the current time in the current term converted into advertising sales for the current term) based on advertising sales for each period in a quarter. In addition, the "increase or decrease" may be shown, for example, as the difference between the average annual advertising sales (e.g., the average amount for the past five fiscal years) and the estimated amount of advertising sales for the current fiscal year (e.g., the amount of advertising sales at this point in time for the current fiscal year converted into annual advertising sales for the current fiscal year).

[0056] The result estimation model storage unit 139 stores the result estimation model. The result estimation model is a learning model that is trained to output sales result information when an advertiser vector, a product vector, and a market gap are input. For example, the information processing device 10 takes the advertiser vector stored in the advertiser vector storage unit 132, the product vector stored in the product vector storage unit 134, and the market gap stored in the market gap storage unit 136 as inputs, and creates a set of the sales result information stored in the sales result storage unit 138 as output for each advertiser as correct answer data. Then, the information processing device 10 trains the result estimation model using the created correct answer data. That is, the information processing device 10 uses a technique such as backpropagation to train the result estimation model so that the sales result information or information corresponding to the sales result information is output when an advertiser vector, a product vector, and a market gap are input.

[0057] The result estimation model generated in this manner can, for example, when an advertiser vector, a product vector, and a market gap are input, output information indicating whether or not the advertiser of the input advertiser vector will purchase the advertised product corresponding to the input product vector.

[0058] The control unit 120 is, for example, a controller, and is realized by a CPU (Central Processing Unit), MPU (Micro Processing Unit), etc., executing various programs stored in a storage device inside the information processing device 10 using a RAM as a working area. The control unit 120 is also a controller, and is realized, for example, by an integrated circuit such as an ASIC (Application Specific Integrated Circuit) or an FPGA (Field Programmable Gate Array).

[0059] As shown in FIG. 2, the control unit 120 has an advertiser vector generation unit 121, a product vector generation unit 122, a purchase amount estimation unit 123, a gap calculation unit 124, a determination unit 125, a result estimation unit 126, and an output unit 127.

[0060] The advertiser vector generation unit 121 generates an advertiser vector for each advertiser using the advertiser vector model stored in the advertiser vector model storage unit 131. Specifically, the advertiser vector generation unit 121 inputs information about the advertiser to the advertiser vector model. Then, the advertiser vector generation unit 121 acquires, as the advertiser vector, vector data output by the advertiser vector model when the information about the advertiser is input.

[0061] The advertiser vector generating unit 121 stores the acquired advertiser vector in the advertiser vector storage unit 132 in association with the advertiser ID.

[0062] The product vector generation unit 122 generates a product vector for each advertised product using the product vector model stored in the product vector model storage unit 133. Specifically, the product vector generation unit 122 inputs information about the advertised product to the product vector model. Then, the product vector generation unit 122 acquires, as a product vector, vector data output by the product vector model when the information about the advertised product is input.

[0063] The product vector generating unit 122 stores the acquired product vector in the product vector storage unit 134 in association with the advertisement product ID.

[0064] In addition, the advertising products to be vectorized are not limited to those sold by one advertising sales source. In other words, the product vector generation unit 122 may convert information about advertising products sold by multiple advertising sales sources into product vectors using a product vector model.

[0065] The purchase amount estimation unit 123 uses a purchase amount estimation model to estimate a potential advertising cost of an advertiser from information about the advertiser.

[0066] For example, the purchase amount estimation unit 123 inputs information about the advertiser (e.g., an advertiser vector) to the purchase amount estimation model. The information about the advertiser may include information about the advertised products purchased by the advertiser (e.g., identification information and purchase amount of the advertised products, etc.). The information about the advertiser may not include information about the advertised products purchased by the advertiser. Next, the purchase amount estimation unit 123 acquires data output from the purchase amount estimation model when the information about the advertiser is input, that is, the purchase amount of each advertised product. Specifically, the purchase amount estimation model outputs the purchase amount of each advertised product that is estimated to be purchased by a hypothetical advertiser similar to the advertiser whose information is input to the purchase amount estimation model.

[0067] Then, the purchase amount estimation unit 123 estimates the total purchase amount of each advertised product output from the purchase amount estimation model as the ideal advertising cost, that is, the expected advertising cost, of the advertiser who inputted information into the purchase amount estimation model.

[0068] The gap calculation unit 124 calculates the difference between the expected advertising cost estimated by the purchase amount estimation unit 123 and the total amount of advertising products currently ordered from advertisers who have input information into the purchase amount estimation model, i.e., the actual advertising cost, as a market gap. Note that the actual advertising cost may be stored in the storage unit 130 in association with an advertiser ID, for example, or may be acquired from an external device via the communication unit 110. The gap calculation unit 124 associates the calculated market gap with identification information of the advertiser who has input information into the purchase amount estimation model and stores it in the market gap storage unit 136.

[0069] The determination unit 125 compares the market gap stored in the market gap storage unit 136 with a threshold value. If the market gap exceeds the threshold value, the determination unit 125 determines that there is a discrepancy between the expected advertising cost and the actual advertising cost for the advertiser identified by the identification information associated with the market gap.

[0070] The result estimation unit 126 sets, as a target advertiser, an advertiser whose expected advertising cost and actual advertising cost are determined to be deviated from each other by the determination unit 125. Then, the result estimation unit 126 uses the result estimation model stored in the result estimation model storage unit 139 to estimate the advertised product that the target advertiser will purchase.

[0071] Specifically, the result estimation unit 126 acquires the advertiser vector of the target advertiser from the advertiser vector storage unit 132, and acquires the market gap of the target advertiser from the market gap storage unit 136. In addition, the result estimation unit 126 acquires an arbitrary product vector from the product vector storage unit 134. Next, the result estimation unit 126 inputs each piece of information of the acquired advertiser vector, market gap, and product vector into a result estimation model. Then, the result estimation unit 126 acquires information on sales results output from the result estimation model. Specifically, the result estimation model outputs information on whether or not the advertiser indicated by the advertiser vector (i.e., the target advertiser) will purchase the advertised product indicated by the advertised product vector.

[0072] The result estimation unit 126 fixes the advertiser vector and the market gap and replaces the product vector among various information input to the result estimation model, thereby classifying, for example, all the product vectors stored in the product vector storage unit 134 into advertised products that the target advertiser will purchase and advertised products that the target advertiser will not purchase. Then, the result estimation unit 126 passes information on the advertised products that the target advertiser will purchase to the output unit 127.

[0073] The result estimation unit 126 may narrow down the information of the advertising products to be passed to the output unit 127 from among the information of the advertising products to be purchased by the target advertiser according to a predetermined condition. For example, the result estimation unit 126 may narrow down the information of the advertising products to be passed to the output unit 127 based on the purchase amount. That is, the result estimation unit 126 may output to the output unit 127 the information of the advertising products whose rankings are within a threshold value (for example, within 5th place) when the advertising products estimated to be purchased by the target advertiser by the result estimation model are sorted in order of purchase amount.

[0074] Furthermore, the result estimation unit 126 may narrow down the information on the advertised products to be passed to the output unit 127 based on the purchase status by other advertisers. For example, the result estimation unit 126 may output to the output unit 127 information on advertised products whose rankings are within a threshold (for example, within 5th place) when the advertised products estimated to be purchased by the target advertiser by the result estimation model are sorted in order of sales volume (sales volume).

[0075] The output unit 127 outputs the information on the advertisement product received from the result estimation unit 126. For example, the output unit 127 may output the information on the advertisement product received from the result estimation unit 126 to an external device via the communication unit 110 and a network. The external device serving as the output destination is, for example, a terminal device used by a sales representative of an advertisement sales agency. The external device serving as the output destination may also be a terminal device owned by a target advertiser.

[0076] 3. Processing Procedure Next, an example of the flow of processing executed by the information processing device 10 according to the embodiment will be described with reference to Fig. 7 and Fig. 8. First, an example of target setting processing will be described with reference to Fig. 7. Fig. 7 is a flowchart showing an example of the flow of the target setting processing according to the embodiment.

[0077] 7, the information processing device 10 inputs information about the advertiser into a purchase amount estimation model (step S101). Next, the information processing device 10 calculates the sum of the purchase amounts of the advertised products output from the purchase amount estimation model as an expected advertising cost (step S102). Then, the information processing device 10 calculates the difference between the expected advertising cost and the actual advertising cost as a market gap (step S103).

[0078] Next, the information processing device 10 determines whether the market gap exceeds the threshold value (step S104). In this process, if it is determined that the market gap exceeds the threshold value (step S104, Yes), the information processing device 10 determines that the expected advertising cost and the actual advertising cost are diverging (step S105). Then, the information processing device 10 sets the advertiser whose expected advertising cost and the actual advertising cost are determined to be diverging as the target advertiser (step S106).

[0079] When the process of step S106 is completed, or in step S104, if the market gap does not exceed the threshold value (step S104, No), the information processing device 10 ends the target setting process.

[0080] Next, an example of the product information output process will be described with reference to Fig. 8. Fig. 8 is a flowchart showing an example of the flow of the product information output process according to the embodiment.

[0081] 8, the information processing device 10 inputs the advertiser vector of the target advertiser, the market gap of the target advertiser, and an arbitrary product vector into a result estimation model (step S201). Then, the information processing device 10 obtains the output of the result estimation model, that is, information indicating whether the advertised product indicated by the input product vector is purchased by the target advertiser (step S202).

[0082] Next, the information processing device 10 determines whether the processes of steps S201 and S202 have been executed for all the advertising products (step S203). In this process, if there are advertising products for which the processes of steps S201 and S202 have not yet been executed (step S203, No), the information processing device 10 returns the process to step S201 and executes the processes of steps S201 and S202 for the unprocessed advertising products.

[0083] On the other hand, in step S203, if it is determined that the processes in steps S201 and S202 have been executed for all the advertised products (step S203, Yes), the information processing device 10 outputs information on the advertised products purchased by the target advertiser (step S204).

[0084] It should be noted that "all advertising products" in step S203 refers to, for example, all advertising products whose product vectors are stored in the product vector storage unit 134, but is not limited to this and may refer to, for example, all advertising products specified by a user (e.g., a sales representative).

[0085] 4. Modifications The above-described information processing device 10 may be implemented in various different forms other than the above embodiment, so other embodiments of the information processing device 10 will be described below.

[0086] In the learning stage of the result estimation model, the information processing device 10 may input the contents of the sales activities (whether or not the sales activities were performed, the type of the sales activities performed, the specific contents of the sales activities performed, the proposed advertised product, etc.) as input data to the result estimation model. For example, the information processing device 10 may learn the result estimation model using the advertiser vector, the product vector, the market gap, and information on the sales activities as input data, and information on the sales results (whether or not an order was received, the amount of increase or decrease, etc.) as output data. In this case, the information processing device 10 may input the advertiser vector and market gap, the product vector, and the contents of the sales activities (for example, the type of sales activity) of the target advertiser to the result estimation model in the utilization stage of the result estimation model. Thereby, for example, when an advertised product of the input product vector is proposed to the target advertiser with the input type of sales activity, information on whether or not the target advertiser will purchase the advertised product can be obtained.

[0087] In the above-described embodiment, an example has been described in which the information output from the result estimation model is information on whether or not the target advertiser will purchase the advertised product (i.e., binary information of Yes / No). Without being limited to this, the result estimation model may be trained to output the likelihood that the target advertiser will purchase the advertised product (i.e., the likelihood of Yes) when the advertiser vector, the product vector, and the market gap are input. In this case, the information processing device 10 may output the advertised product that is most likely to be purchased by the target advertiser.

[0088] The result estimation model may be trained to output information indicating whether advertising sales by the target advertiser will increase or not when the advertiser vector, the market gap, and the product vector are input. The result estimation model may be trained to output the increase or decrease in advertising sales by the target advertiser when the advertiser vector, the market gap, and the product vector are input. In this case, the result estimation model may be trained to output the increase or decrease in advertising sales by the target advertiser as is, or may output the increase or decrease, which is a continuous value, by replacing it with a discrete value through discretization. For example, when the increase or decrease is 8.3 million yen, the result estimation model may be trained to output information of "5 million yen or more and less than 10 million yen" as a discrete value. The result estimation model may be trained to output an estimated value of the order amount of the advertising product when the advertiser vector, the market gap, and the product vector are input.

[0089] In the above embodiment, an example has been described in which the advertiser vector, the product vector, and the market gap are input data to the result estimation model in the learning stage of the result estimation model, but the market gap does not necessarily have to be included in the input data. That is, the result estimation model may be learned using the advertiser vector and the product vector as input data. That is, when the advertiser vector and the product vector are input, the result estimation model may be learned to output sales result information indicating whether or not the advertised product corresponding to the product vector (the advertised product proposed to the advertiser) has been purchased.

[0090] When the information stored in the sales result storage unit 138 is updated, the information processing device 10 may update various learning models (e.g., a purchase amount estimation model, a result estimation model, etc.) based on the updated information.

[0091] In the learning stage of the advertiser vector model, the information processing device 10 may generate an advertiser vector model by performing pre-learning to capture the structure of language from a huge amount of text without a teacher using, for example, BERT (Bidirectional Encoder Representations from Transformers), and then performing readjustment using information about advertisers as learning data. In addition, instead of pre-learning BERT with natural language, the model structure of BERT may be used to perform pre-learning using BERT with multimodal data, for example, information in a plurality of different formats (information transmission formats) related to advertisers, such as images (company logos, etc.), voices, text (business overviews, management issues, etc.), and flags (advertising tag usage status of competitors), as input data. That is, the advertiser vector may be generated by performing learning using an approach that solves a fill-in-the-blank problem for data in a plurality of different formats. Note that, not limited to BERT, a model having a Transformer network configuration other than BERT may be used.

[0092] The information about the advertiser may include information about the group to which the advertiser belongs. A group here refers to a collection of advertisers (such as a holding company) that have a capital relationship, such as a relationship between a subsidiary and a parent company or a relationship between subsidiaries of the same parent company. In this case, the result prediction model can output information on whether or not to purchase the advertised product, taking into account the tendency of the group to which the target advertiser belongs (for example, whether the group to which the target advertiser belongs is proactive or passive about advertising, etc.).

[0093] The information processing device 10 may estimate an advertiser vector that will obtain a desired sales result from the desired sales result, using a model that has learned the association between information about an advertiser and the results of sales to the advertiser. In this case, the information processing device 10 may, for example, estimate an advertiser vector that increases advertising sales through sales activities using the above model, and identify an advertiser that has an advertiser vector similar to the estimated advertiser vector as a target advertiser.

[0094] In the learning stage of the advertiser vector model, the information processing device 10 may convert information about advertisers into a one-hot vector indicating the result of clustering by type, and use the concatenated vector as input data to learn the advertiser vector model.

[0095] In the above-described embodiment, an example was described in which various information on advertisers whose advertising sales have increased due to sales activities is used as correct answer data in the learning stage of the purchase amount estimation model. Not limited to this, the information processing device 10 may learn the purchase amount estimation model using information on advertisers whose advertising sales due to sales activities exceed the predicted value as correct answer data. Furthermore, the information processing device 10 may cluster information on advertisers by type, extract the top n% advertisers with the highest advertising sales increase rate from the clustered group of advertisers, and learn the purchase amount estimation model using the information on the extracted advertisers as correct answer data. The extraction condition for the advertisers is not limited to the increase rate of advertising sales, and may be, for example, the top n% with the highest advertising sales. Furthermore, the extraction condition for the advertisers may be the top n% with the highest sales amount of the advertised product relative to the scale of the advertiser's sales revenue or advertising and promotion expenses. Furthermore, the extraction condition for the advertisers may be the top n% with the highest sales volume of the product that is the subject of the advertisement (the product that the advertiser wants to sell), etc. In this way, the information processing device 10 may train the purchase amount estimation model using information on advertisers whose sales results satisfy a predetermined success condition (for example, increased advertising sales, etc.) as correct answer data. Also, the information processing device 10 may train the purchase amount estimation model using information on one or more advertisers that rank highly when advertisers whose sales results satisfy a predetermined success condition (for example, increased advertising sales, etc.) are ranked according to the success condition as correct answer data.

[0096] In the above embodiment, an example of setting an advertiser whose market gap exceeds a threshold value as a target advertiser has been described. However, the information processing device 10 may calculate market gaps for multiple advertisers, and set an advertiser whose ranking (e.g., within 5th place) when the multiple advertisers are sorted in descending order of the calculated market gaps as a target advertiser.

[0097] Furthermore, in the learning stage of the advertiser vector model, the product vector model, and the result estimation model, the information processing device 10 may discretize numerical information among the data to be input to the models before inputting it to the models. For example, in the learning stage of the advertiser vector model, the information processing device 10 may discretize numerical values, such as sales, profit margins, and advertising and promotion expenses, among the information about the advertiser, before inputting them to the advertiser vector model. In addition, in the learning stage of the result estimation model, the information processing device 10 may discretize the "increase or decrease amount" included in the sales result information before inputting it to the result estimation model. In addition, when discretizing numerical values, the number of digits may be adjusted to a certain extent by, for example, a log scale, before inputting them to the model.

[0098] In the learning stage of the advertiser vector model, the information processing device 10 may learn the advertiser vector using a technique such as backpropagation so as to output the advertiser vector when various information regarding the advertisers and information indicating the similarity between the advertisers (for example, information associating a set of advertisers with the similarity) are input.

[0099] 5. Effects As described above, the information processing device 10 according to the embodiment includes a result estimation unit 126 that estimates the advertised product to be purchased by the target advertiser from the information on the target advertiser using a result estimation model that learns the relationship between advertiser information on the advertiser (advertiser vector, as an example) and the advertised product purchased by the advertiser, and an output unit 127 that outputs the information on the advertised product estimated by the result estimation unit 126. As a result, for example, a sales representative of an advertisement sales agency can narrow down the advertised products that the target advertiser is likely to purchase as advertised products to be proposed to the target advertiser based on the information output from the information processing device 10. In this way, the information processing device 10 according to the embodiment can support the sale of advertised products to advertisers.

[0100] The result estimation unit 126 may use the result estimation model to estimate the advertising product that the target advertiser will purchase from gap information indicating the difference between the expected advertising cost, which is the expected amount of advertising cost to be paid by the target advertiser, and the actual advertising cost, which is the actual amount of advertising cost ordered from the target advertiser, and information on the target advertiser. This makes it possible to more accurately estimate the advertising product that the target advertiser will purchase.

[0101] The information processing device 10 according to the embodiment may include a purchase amount estimation unit 123 that estimates expected advertising costs from information about a target advertiser using a purchase amount estimation model trained to output the purchase amount of each advertised product when advertiser information is input, and a gap calculation unit 124 that calculates gap information using the expected advertising costs and actual advertising costs estimated by the purchase amount estimation unit 123. This makes it easier to calculate the gap information compared to, for example, a case where a sales representative calculates the gap information by himself. Therefore, the information processing device 10 according to the embodiment can further support sales of advertised products to advertisers.

[0102] The information processing device 10 according to the embodiment may include an advertiser vector generation unit 121 that generates an advertiser vector from advertiser information using an advertiser vector model that has been trained to output a similar advertiser vector when similar advertiser information is input. In this case, the result estimation unit 126 may use the advertiser vector of the target advertiser generated by the advertiser vector generation unit 121 as information on the target advertiser. In this way, by abstracting (meta-izing) information on the advertiser, it is possible to appropriately estimate the advertised product that the target advertiser will purchase.

[0103] The advertiser vector model may be a model trained so that the higher the similarity score indicating the similarity between advertisers, including the content and results of sales activities for the advertisers, the higher the similarity of the advertiser vectors. This makes it possible to abstract (meta-ize) information about advertisers, including information about sales activities, and therefore to estimate the advertising products that the target advertiser will purchase, taking into account information about sales activities.

[0104] The information processing device 10 according to the embodiment may include a product vector generation unit 122 that generates an advertiser vector from information on an advertised product using a product vector model that has been trained to output a similar product vector when information on an advertised product with a similar purchase pattern is input. In this case, the result estimation unit 126 may use the advertiser vector generated by the product vector generation unit 122 as input data for the result estimation model. In this way, by abstracting (meta-izing) information on the advertised product, it is possible to appropriately estimate the advertised product that the target advertiser will purchase.

[0105] The product vector model may be a model that is trained so that the product vectors of advertising products purchased by the same advertiser are similar to each other. As a result, for example, by inputting information about a target advertiser and the product vector into the result estimation model, it is possible to estimate an advertising product purchased by an advertiser similar to the target advertiser (or an advertising product similar to the advertising product) as an advertising product to be purchased by the target advertiser.

[0106] The result estimation unit 126 may estimate the advertising product that the target advertiser will purchase from the information on the target advertiser and the information indicating the content of the sales activities by using the result estimation model. This makes it possible to estimate the advertising product that the target advertiser will purchase while taking into account the content of the sales activities.

[0107] The information processing device 10 according to the embodiment includes a purchase amount estimation unit 123 that estimates an expected advertising cost, which is an expected amount of advertising cost to be paid by a target advertiser, from information about the target advertiser using a purchase amount estimation model trained using information about an advertiser whose advertising cost satisfies a predetermined condition as correct answer data, and a determination unit 125 that determines whether or not the expected advertising cost and the actual advertising cost of the target advertiser are different from each other, based on the expected advertising cost estimated by the purchase amount estimation unit 123 and the actual advertising cost, which is the actual amount of advertising cost received from the target advertiser. As a result, for example, by selling an advertising product to a target advertiser whose expected advertising cost and actual advertising cost are different from each other, effective sales activities can be performed. Therefore, the information processing device 10 according to the embodiment can support the sale of advertising products.

[0108] The information processing device 10 according to the embodiment may include a gap calculation unit 124 that calculates gap information indicating a difference between the expected advertising cost and the actual advertising cost, using the expected advertising cost and the actual advertising cost estimated by the purchase amount estimation unit 123. In this case, the determination unit 125 may determine that the expected advertising cost and the actual advertising cost are diverging when the gap information calculated by the gap calculation unit 124 exceeds a threshold value. It is possible to appropriately find target advertisers whose expected advertising cost and actual advertising cost are diverging.

[0109] The purchase amount estimation model may be a model trained using information on advertisers whose advertising costs have increased as a result of sales activities as correct answer data. This can increase the reliability of the information output from the purchase amount estimation model, and can increase the estimation accuracy of the expected advertising costs.

[0110] The purchase amount estimation model may be a model trained using information on advertisers whose advertising costs exceed a threshold as correct answer data. This can increase the reliability of the information output from the purchase amount estimation model, and can increase the accuracy of estimating the expected advertising costs.

[0111] The purchase amount estimation model may be a model trained using information on advertisers whose advertising costs are higher than the average advertising costs of multiple advertisers in the same business field as correct answer data. This can increase the reliability of the information output from the purchase amount estimation model and improve the estimation accuracy of the expected advertising costs.

[0112] The purchase amount estimation model may be a model trained using information of advertisers whose purchase amount for at least one of the advertised products exceeds a threshold as correct answer data. This can increase the reliability of the information output from the purchase amount estimation model, and can increase the estimation accuracy of the expected advertising cost.

[0113] The purchase amount estimation model may be a model trained using a set of information on the advertiser and information on the advertised product purchased by the advertiser, the information including at least the purchase amount, as correct answer data. This can increase the reliability of the information output from the purchase amount estimation model, and can increase the estimation accuracy of the expected advertising cost.

[0114] The purchase amount estimation model may be a model trained using information of a plurality of advertisers adjusted so that information of an advertiser whose first condition advertising expense satisfies the second condition is weighted more heavily than information of an advertiser whose first condition advertising expense does not satisfy the second condition as correct answer data. This can increase the reliability of the information output from the purchase amount estimation model and improve the estimation accuracy of the expected advertising expense.

[0115] [6. Hardware Configuration] The information processing device 10 according to the embodiment described above is realized by a computer 1000 having a configuration as shown in Fig. 9. Fig. 9 is a hardware configuration diagram showing an example of a computer that realizes the functions of the information processing device 10 according to the embodiment. The computer 1000 has a CPU 1100, a RAM 1200, a ROM 1300, a HDD 1400, a communication interface (I / F) 1500, an input / output interface (I / F) 1600, and a media interface (I / F) 1700.

[0116] The CPU 1100 operates and controls each unit based on a program stored in the ROM 1300 or the HDD 1400. The ROM 1300 stores a boot program executed by the CPU 1100 when the computer 1000 is started up, programs that depend on the hardware of the computer 1000, and the like.

[0117] The HDD 1400 stores programs executed by the CPU 1100, data used by such programs, etc. The communication interface 1500 receives data from other devices via a network (communication network) N and sends the data to the CPU 1100, and transmits data generated by the CPU 1100 to other devices via the network N.

[0118] The CPU 1100 controls output devices such as a display and a printer, and input devices such as a keyboard and a mouse (in FIG. 9, the output devices and the input devices are collectively referred to as "input / output devices") via an input / output interface 1600. The CPU 1100 acquires data from the input devices via the input / output interface 1600. The CPU 1100 also outputs generated data to the output devices via the input / output interface 1600.

[0119] The media interface 1700 reads a program or data stored in the recording medium 1800 and provides it to the CPU 1100 via the RAM 1200. The CPU 1100 loads the 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 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.

[0120] For example, when the computer 1000 functions as the information processing device 10 according to the embodiment, the CPU 1100 of the computer 1000 executes programs loaded onto the RAM 1200 to realize the functions of the control unit 120. The CPU 1100 of the computer 1000 reads and executes these programs from the recording medium 1800, but as another example, the CPU 1100 may obtain these programs from another device via the network N.

[0121] Although some of the embodiments of the present application have been described in detail above with reference to the drawings, these are merely examples, and the present invention can be embodied in other forms that incorporate various modifications and improvements based on the knowledge of those skilled in the art, including the aspects described in the Disclosure of the Invention section.

[0122] [7. Other] Furthermore, among the processes described in the above embodiments and modifications, 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 by a known method. In addition, the information including the processing procedures, specific names, various data and parameters shown in the above documents and drawings can be changed arbitrarily unless otherwise specified. For example, the various information shown in each drawing is not limited to the illustrated information.

[0123] In addition, each component of each device shown in the figure is a functional concept, and does 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 according to various loads, usage conditions, etc. For example, the advertiser vector generation unit 121, the product vector generation unit 122, the purchase amount estimation unit 123, the gap calculation unit 124, the determination unit 125, the result estimation unit 126, and the output unit 127 may be connected via a network as external devices of the information processing device 10. In addition, the advertiser vector generation unit 121, the product vector generation unit 122, the gap calculation unit 124, the determination unit 125, the result estimation unit 126, and the output unit 127 may be included in separate devices, and may be connected to a network to cooperate with each other to realize the functions of the information processing device 10. That is, the information processing device 10 may be realized by multiple server computers, and depending on the function, the configuration can be flexibly changed by calling an external platform, etc. using an API (Application Programming Interface) or network computing, etc.

[0124] Furthermore, the above-described embodiments and modifications can be appropriately combined as long as the processing contents are not contradictory.

[0125] In addition, the above-mentioned "section, module, unit" can be read as "means" or "circuit", etc. For example, the advertiser vector generating unit 121 can be read as an advertiser vector generating means or an advertiser vector generating circuit. [Explanation of symbols]

[0126] 10 ; Information processing device 110; Communications Department 120; Control section 121; Advertiser Vector Generation Unit 122: Product vector generation unit 123; Purchase amount estimation section 124; Gap calculation section 125; Judgment section 126;Result estimation part 127; Output section 130;Memory part 131; Advertiser vector model storage section 132; Advertiser vector storage unit 133; Product vector model storage section 134: Product vector storage unit 135; Purchase amount estimation model memory section 136; Market Gap Memory Section 137; Target information storage section 138; Sales results memory section 139; Result estimation model memory section

Claims

1. a purchase amount estimation unit that estimates an expected advertising cost, which is an expected amount of advertising costs to be paid by a target advertiser, from information about the target advertiser using a purchase amount estimation model trained using information about an advertiser whose advertising cost satisfies a predetermined condition as correct answer data; a gap calculation unit that calculates gap information indicating a difference between the expected advertising cost estimated by the purchase amount estimation unit and an actual advertising cost that is an actual amount of advertising cost ordered from the target advertiser; An information processing device comprising:

2. The purchase amount estimation model is This model is trained using information about the advertiser whose advertising expenses increased as a result of sales activities as correct answer data.

2. The information processing apparatus according to claim 1,

3. The purchase amount estimation model is The model is trained using information of the advertiser whose advertising cost exceeds a threshold as correct answer data.

2. The information processing apparatus according to claim 1,

4. The purchase amount estimation model is The model is trained using information on the advertiser whose advertising cost is higher than the average advertising cost of the multiple advertisers belonging to the same business field as correct answer data.

2. The information processing apparatus according to claim 1,

5. The purchase amount estimation model is The model is trained using information about the advertiser whose purchase amount for at least one of the advertised products exceeds a threshold as correct answer data.

2. The information processing apparatus according to claim 1,

6. The purchase amount estimation model is The model is trained using information about one or more advertisers that rank highly when the advertisers whose sales activity results meet a certain success condition are ranked according to the success condition as correct answer data.

2. The information processing apparatus according to claim 1,

7. The purchase amount estimation model is The model is trained using a set of information on the advertiser and information on the advertised product purchased by the advertiser, the information including at least the purchase amount, as correct answer data.

7. The information processing apparatus according to claim 1,

8. The purchase amount estimation model is The model is trained using information of a plurality of advertisers as correct answer data, the information of the advertiser whose advertising cost of the first condition satisfies the second condition being weighted more heavily than the information of the advertiser whose advertising cost of the first condition does not satisfy the second condition.

8. The information processing device according to claim 1,

9. 1. A computer-implemented information processing method, comprising: a purchase amount estimation step of estimating an expected advertising cost, which is an expected amount of advertising cost to be paid by a target advertiser, from information on the target advertiser using a purchase amount estimation model trained using information on the advertiser whose advertising cost satisfies a predetermined condition as accurate data; a calculation step of calculating gap information indicating a difference between the expected advertising cost estimated by the purchase amount estimation step and an actual advertising cost which is an actual amount of advertising cost ordered from the target advertiser; 13. An information processing method comprising:

10. a purchase amount estimation step of estimating an expected advertising cost, which is an expected amount of advertising cost to be paid by a target advertiser, from information on the target advertiser using a purchase amount estimation model trained using information on the advertiser whose advertising cost satisfies a predetermined condition as accurate data; a calculation step of calculating gap information indicating a difference between the expected advertising cost estimated by the purchase amount estimation step and an actual advertising cost which is an actual amount of advertising cost ordered from the target advertiser; An information processing program for causing a computer to execute the above.

Citation Information

Patent Citations

  • Device and method for distributing advertisement and content

    JP2002203119A

  • Advertising cost computing method

    JP2010044795A

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

    JP2022167665A

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

    JP2022167666A