JP2017033191A - Estimation apparatus, estimation method, and estimation program - Google Patents

Abstract

PROBLEM TO BE SOLVED: To estimate the interest of a user in detail.SOLUTION: An estimation apparatus includes an acquisition unit and an estimation unit. The acquisition unit acquires an element of a model for estimating the interest of a user in a first field. The estimation unit estimates the interest of the user in a second field more specific than the frist field, on the basis of the element acquired by the acquisition unit.SELECTED DRAWING: Figure 3

Description

  The present invention relates to an estimation device, an estimation method, and an estimation program.

  In recent years, models for predicting information about users have been constructed. As a technique related to the construction of such a model, a technique using machine learning is known. For example, in such a technique, a user profile is determined from user event information including online activities.

Special table 2009-531782 gazette JP 2013-246757 A

  However, in the above-described conventional technology, the user's interest cannot be estimated in detail. For example, in the above-described conventional technology, the category user profile score is generated by applying the model parameters to the user event information. Therefore, when the quality of the user event information is low, the accuracy of the user profile score is low. For this reason, the above-described conventional technique cannot always estimate the user's interest in detail.

  The present application has been made in view of the above, and an object thereof is to provide an estimation device, an estimation method, and an estimation program that can estimate the user's interest in detail.

  The estimation apparatus according to the present application includes an acquisition unit that acquires an element of a model that predicts a user's interest in the first field, and a second that is more detailed than the first field based on the element acquired by the acquisition unit. And an estimation unit for estimating the user's interest in the field.

  According to one aspect of the embodiment, the user's interest can be estimated in detail.

FIG. 1 is an explanatory diagram illustrating an example of an estimation process performed by the estimation system according to the embodiment. FIG. 2 is an explanatory diagram for explaining an estimation process by the estimation device. FIG. 3 is a diagram illustrating a configuration example of the estimation apparatus according to the embodiment. FIG. 4 is a diagram illustrating an example of a user information storage unit according to the embodiment. FIG. 5 is a diagram illustrating an example of the model information storage unit according to the embodiment. FIG. 6 is a diagram illustrating an example of the interest information storage unit according to the embodiment. FIG. 7 is a diagram illustrating an example of the classification information storage unit according to the embodiment. FIG. 8 is a flowchart showing an estimation processing procedure by the estimation apparatus. FIG. 9 is a hardware configuration diagram illustrating an example of a computer that realizes the function of the estimation device.

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

[1. Embodiment)
[1-1. Estimation Process According to Embodiment]
First, an example of the estimation process according to the embodiment will be described with reference to FIG. FIG. 1 is an explanatory diagram illustrating an example of an estimation process performed by the estimation system 1 according to the embodiment. In the example of FIG. 1, an estimation process is performed in which the estimation apparatus 100 estimates the user's interest in more detail than “golf” using an element of a model that predicts the user's interest in “golf”.

  As illustrated in FIG. 1, the estimation system 1 includes terminal devices 10 </ b> A to 10 </ b> C and an estimation device 100. The terminal devices 10A to 10C and the estimation device 100 are connected to a network so that they can communicate with each other by wire or wirelessly. Hereinafter, the terminal devices 10 </ b> A to 10 </ b> C may be collectively referred to as “terminal device 10” when collectively referred to without distinction.

  The terminal device 10 is an information processing device such as a PC (Personal Computer), a tablet-type terminal, a smartphone, a mobile phone, or a PDA (Personal Digital Assistant). The estimation device 100 is a server device that estimates user interest. Specifically, the estimation apparatus 100 executes an estimation process for estimating a user's more detailed interest using an element of a model that predicts the user's interest. This point will be described in detail with reference to FIG.

  First, the terminal devices 10A to 10C transmit user information related to the users Us1 to Us3, respectively, to the estimation device 100 (step S1). For example, the terminal devices 10A to 10C transmit the search words used by the users Us1 to Us3 to the estimation device 100 as user information. As an example, the terminal device 10A transmits the search words “driver”, “iron”, “putter”, and “baseball player” used by the user Us1 for the search to the estimation device 100 as user information. Thereby, the estimation apparatus 100 receives user information from the terminal devices 10A to 10C.

Subsequently, the estimation apparatus 100 estimates whether or not the users Us1 to Us3 are interested in “golf” using a model that predicts the user's interest in “golf” (step S2). This point will be described in detail with reference to FIG. FIG. 2 is an explanatory diagram for explaining an estimation process performed by the estimation apparatus 100. As an example, the estimation apparatus 100 uses the model m1 of the linear regression model illustrated in FIG. 2 as a model for predicting the user's interest in “golf”, and determines whether the users Us1 to Us3 are interested in “golf”. Estimate. The model m1 that predicts the user's interest in “golf” is generated, for example, by machine learning a search word used by a user who likes golf. The elements x _{1 to} x _n of the model m1 are explanatory variables indicating that they are interested in “golf”. For example, the elements x ₁ of the model m1 corresponds to the "driver" of the golf club. Elements of the model m1 x ₂ corresponds to the "back swing" of Uchikata of golf. Element x ₂ of the model m1 corresponds to the "A player" of the golf player.

Here, the estimation apparatus 100 inputs information about the search words “driver”, “iron”, “putter”, and “baseball player” used by the user Us1 for the model m1 to make the user Us1 “golf”. Estimate whether you are interested. As an example, the estimation apparatus 100 assigns “1” to elements corresponding to the search words “driver”, “iron”, “putter”, and “baseball player” among the elements x _{1 to} x _n of the model m1. Based on the value of “y” calculated by substituting “0” for other elements, it is estimated whether or not the user Us1 is interested in “golf”. In this example, it is assumed that the estimating apparatus 100 estimates that the users Us1 to Us3 are interested in “golf”.

Subsequently, when estimating that the users Us <b> 1 to Us <b> 3 are interested in “golf”, the estimating apparatus 100 performs a process of estimating more detailed interest. Specifically, the estimating apparatus 100 _first acquires the elements x _{1 to} x _n of the model m1 that predicts the user's interest in “golf” (step S3).

Then, the estimating apparatus 100 estimates the interests of the users Us1 to Us3 in the field more detailed than “golf” based on the acquired elements x _{1 to} x _n (step S4). In the example of FIG. 2, the estimation apparatus 100 uses the elements x _{1 to} x _n of the model m 1 that predicts the user's interest in “golf”, and the golf “club”, “ The interests of the users Us1 to Us3 with respect to “how to hit” and “players” are estimated. Specifically, the estimation apparatus 100 determines “club” based on an element corresponding to “club”, an element corresponding to “how to hit”, and an element corresponding to “player” among the elements x _{1 to} x _n of the model m1. ”,“ How to hit ”, and“ player ”are estimated respectively for the user Us1. More specifically, the user's interest in “club”, “how to hit”, and “player” is estimated based on the elements corresponding to the user information of the user Us1 among the elements x _{1 to} x _n of the model m1. For example, the estimation apparatus 100 supports “club” based on the number of elements corresponding to “club” among the elements x _{1 to} x _n of the model m1 and the search word used by the user Us1 for the search. The score “4” of the element to be calculated is calculated. As an example, the estimating apparatus 100, "1 since the elements x ₁ and user Us1 search word and" driver "matches that used in the search to the score of the elements corresponding to the" club "corresponding to the" driver "of the club Is added. Then, the estimating apparatus 100 calculates the score “4” of the element corresponding to “club” by executing such calculation for all the search words included in the user information of the user Us1. Similarly, the estimating apparatus 100 calculates the score “1” of the element corresponding to “how to hit” using the user information of the user Us1. Further, the estimating apparatus 100 calculates the score “2” of the element corresponding to “player” using the user information of the user Us1. Then, the estimation apparatus 100 estimates the field corresponding to the element having a higher score as the field that the user is more interested in. In the example of FIG. 2, since the score of the element corresponding to “club” is the highest, the estimation apparatus 100 estimates “club” as a field in which the user Us1 is strongly interested.

  Similarly, the estimation apparatus 100 uses the user information of the user Us2 to score “1”, “4”, an element corresponding to “club”, an element corresponding to “how to hit”, an element corresponding to “player”, “3” is calculated, and “how to hit” corresponding to the element having the highest score is estimated as a field in which the user Us2 is strongly interested. In addition, the estimation apparatus 100 uses the user information of the user Us3, the element corresponding to “club”, the element corresponding to “how to hit”, and the scores “2”, “3”, “ 4 ”is calculated, and the“ player ”corresponding to the element with the highest score is estimated as a field in which the user Us3 is strongly interested. Thereby, the estimation apparatus 100 can estimate more detailed fields of interest “club”, “how to hit”, and “player” of the users Us1 to Us3 who are interested in “golf”. For this reason, the estimating apparatus 100 can enhance the advertising effect by, for example, distributing an advertisement relating to the golf club to the user Us1.

  As described above, the estimation apparatus 100 according to the embodiment acquires an element of a model that predicts a user's interest in the first field. In addition, the estimation apparatus 100 estimates the user's interest in the second field that is more detailed than the first field, based on the acquired element.

  Thereby, since the estimation apparatus 100 can estimate the user's interest in the second field, which is more detailed than the first field, the user's interest can be estimated in detail. For example, since the estimation apparatus 100 can estimate which field among golf fields a user who likes golf is interested in, it is possible to estimate the user's interest in more detail. As an example, the estimation apparatus 100 estimates in detail for each user whether the same golf enthusiast is interested in a golf club, whether to play golf, or interested in a golf player. be able to.

  In addition, since the estimation apparatus 100 can estimate the user's interest in the second field, which is more detailed than the first field, using the elements of the model that predicts the user's interest in the first field, more detailed information can be obtained. It is possible to reduce the burden of estimating the user's interest. For example, the estimation apparatus 100 estimates the user's interest in the second field without taking the trouble of generating the model, as compared with a case where a model for predicting the user's interest in the second field is newly generated. Therefore, the burden on estimating the user's interest can be reduced. In addition, the estimation apparatus 100 estimates the user's detailed interest without taking a great deal of time and effort to newly generate a model for predicting the user's interest in the second field, which is innumerable in more detail than the first field. be able to.

  In addition, the estimation apparatus 100 can estimate the user's interest in the second field in more detail than the first field using the elements of the model that predicts the user's interest in the first field. The user's interest in the field can be estimated with high accuracy. For example, since the estimation apparatus 100 uses a model element that predicts a user's interest in a first field that is a broader category than the second field, a model element that has a large amount of learning information and a statistically high accuracy is used. The user's interest can be estimated with high accuracy.

  In addition, the estimation apparatus 100 estimates the user's interest in the second field based on the elements corresponding to the second field among the elements of the model. Thereby, since the estimation apparatus 100 can estimate a user's detailed interest using the element relevant to a 2nd field | area, it can estimate a user's detailed interest with high precision.

  1 shows an example in which the estimation system 1 includes a plurality of terminal devices 10 and one estimation device 100, the estimation system 1 includes a plurality of terminal devices 10 and a plurality of devices. The estimation apparatus 100 may be included.

  FIG. 1 shows an example in which the model m1 for predicting the user's interest in “golf” is a linear regression model. However, the model m1 is not limited to the linear regression model, and may be other various models such as a logistic regression model. Good.

[1-2. Configuration of Estimation Device According to Embodiment]
Next, the configuration of the estimation apparatus 100 according to the embodiment will be described with reference to FIG. FIG. 3 is a diagram illustrating a configuration example of the estimation apparatus 100 according to the embodiment. As illustrated in FIG. 3, the estimation apparatus 100 includes a communication unit 110, a storage unit 120, and a control unit 130. The estimation device 100 includes an input unit (for example, a keyboard and a mouse) that receives various operations from an administrator who uses the estimation device 100, and a display unit (for example, a liquid crystal display) that displays various types of information. You may have.

(About the communication unit 110)
The communication unit 110 is realized by a NIC (Network Interface Card) or the like. Specifically, the communication unit 110 is connected to a network by wire or wireless, and transmits / receives information to / from the terminal device 10 via the network. As an example, the communication unit 110 performs communication for receiving user information from the terminal device 10.

(About the storage unit 120)
The storage unit 120 is realized by, for example, a semiconductor memory device such as a RAM (Random Access Memory) or a flash memory, or a storage device such as a hard disk or an optical disk. The storage unit 120 includes a user information storage unit 121, a model information storage unit 122, an interest information storage unit 123, and a classification information storage unit 124.

(User information storage unit 121)
The user information storage unit 121 stores information about the user. Specifically, the user information storage unit 121 stores, for each user, information related to a search word used for the search by the user. Here, FIG. 4 shows an example of the user information storage unit 121 according to the embodiment. As illustrated in FIG. 4, the user information storage unit 121 includes items such as “user ID” and “search word”.

  “User ID” indicates identification information for identifying a user. For example, a unique character string or the like given to the user is stored in the “user ID”. “Search word” indicates a keyword used by the user in the past for the search. For example, “search word” stores a character string or the like that is input to the search window of the search site by the user and used for the search.

  That is, FIG. 4 shows an example in which the user identified by the user ID “Us1” has searched in the past using keywords such as “driver”, “iron”, “putter”, and “baseball player”. .

(About the model information storage unit 122)
The model information storage unit 122 stores information related to a model that predicts a user's interest in the first field. Specifically, the model information storage unit 122 stores information regarding the elements forming the model for each first field. Here, FIG. 5 illustrates an example of the model information storage unit 122 according to the embodiment. As illustrated in FIG. 5, the model information storage unit 122 includes items such as “first field” and “element”.

  The “first field” indicates identification information for identifying the prediction target of the model. For example, in the “first field”, the name of the field of interest of the user is stored. “Element” indicates the weight of the element forming the model. For example, in “element”, the weight of a keyword related to the prediction target field is stored. The “elements” are classified and stored in a second field that is more detailed than the first field.

  That is, in FIG. 5, the models for predicting the user's interest in the first field “golf” are “driver”, “iron”, “putter”, “back swing”, “follow-through”, “torsion”, “ The example is formed by elements such as “A player”, “B player”, “C player”, and the like. Further, an example is shown in which the weight of the element “driver” is “1.2”. Further, the elements of “driver”, “iron”, and “putter” are classified into the field “club”. In addition, the elements of “back swing”, “follow-through”, and “twist” are classified into the field “how to hit”. In addition, the elements “A player”, “B player”, and “C player” are examples classified into the field “player”.

(About the interest information storage unit 123)
The interest information storage unit 123 stores information related to the user's interest. Specifically, the interest information storage unit 123 stores information related to interest in the second field for each user. Here, FIG. 6 shows an example of the interest information storage unit 123 according to the embodiment. As illustrated in FIG. 6, the interest information storage unit 123 includes items such as “user ID” and “second field”.

  “User ID” indicates identification information for identifying a user. For example, a unique character string or the like given to the user is stored in the “user ID”. The “second field” indicates a score of interest in a more detailed field than the first field. For example, a score of interest calculated based on an element corresponding to the second field is stored in the “second field”. As an example, the interest score is calculated based on the number of matches between the element corresponding to the second field and the search word used for the search by the user.

  That is, FIG. 6 shows an example in which the user identified by the user ID “Us1” has an interest score of “4” for “club”. The user identified by the user ID “Us1” shows an example in which the interest score for “how to hit” is “1”. Further, the user identified by the user ID “Us1” shows an example in which the interest score for “player” is “2”. That is, FIG. 6 shows an example in which the user identified by the user ID “Us1” has a stronger interest in the golf club than how to hit golf or the golf player.

(About the classification information storage unit 124)
The classification information storage unit 124 stores information regarding the classification to which the user belongs. Specifically, the classification information storage unit 124 stores, for each user, information related to a field in which the user is highly interested. Here, FIG. 7 illustrates an example of the classification information storage unit 124 according to the embodiment. As illustrated in FIG. 7, the classification information storage unit 124 includes items such as “user ID” and “category”.

  “User ID” indicates identification information for identifying a user. For example, a unique character string or the like given to the user is stored in the “user ID”. “Category” indicates a field in which the user is highly interested. For example, “category” stores a second field that is more detailed than the first field.

  That is, FIG. 7 shows an example in which the user identified by the user ID “Us1” is highly interested in “club”. In addition, the user identified by the user ID “Us2” shows an example of high interest in “how to hit”. Moreover, the user identified by the user ID “Us3” shows an example in which interest in the “player” is high.

(About the control unit 130)
Returning to FIG. 3, the control unit 130, for example, various programs (corresponding to an example of an estimation program) stored in a storage device inside the estimation device 100 by a CPU (Central Processing Unit), an MPU (Micro Processing Unit), or the like. Is implemented by using the RAM as a work area. The control unit 130 is realized by an integrated circuit such as an application specific integrated circuit (ASIC) or a field programmable gate array (FPGA).

  As shown in FIG. 3, the control unit 130 includes a reception unit 131, an acquisition unit 132, an estimation unit 133, and a classification unit 134, and realizes or executes functions and operations of information processing described below. . Note that the internal configuration of the control unit 130 is not limited to the configuration illustrated in FIG. 3, and may be another configuration as long as an estimation process described later is performed. Further, the connection relationship between the processing units included in the control unit 130 is not limited to the connection relationship illustrated in FIG. 3, and may be another connection relationship.

(About the receiver 131)
The receiving unit 131 receives user information about the user from the terminal device 10. Specifically, the receiving unit 131 receives, as user information, a search word used by the user for search from the terminal device 10. In addition, when receiving user information from the terminal device 10, the receiving unit 131 stores the received user information in the user information storage unit 121. For example, the receiving unit 131 stores the user information in the user information storage unit 121 in association with the user ID assigned to each user.

(About the acquisition unit 132)
The acquisition unit 132 acquires a model element that predicts a user's interest in the first field. Specifically, the acquisition unit 132 refers to the model information storage unit 122 and acquires a model element that predicts the user's interest in the first field. For example, the acquisition unit 132 acquires an element of a model that predicts a user's interest in a first field generated by learning user information by machine learning. As an example, the acquisition unit 132 is an explanatory variable such as a linear regression model or a logistic regression model that predicts a user's interest in the first field generated by learning a search word used by the user for search by machine learning. To get.

(About the estimation unit 133)
The estimation unit 133 estimates whether the user is interested in the first field using a model that predicts the user's interest in the first field. Specifically, the estimation unit 133 uses a model generated by machine learning of a search word used by a user who is interested in the first field, and the user is interested in the first field. It is estimated whether it has. For example, the estimation unit 133 estimates whether or not the user is interested in the first field by inputting information related to a search word used for the search to the model. As an example, the estimation unit 133 substitutes “1” for the element corresponding to the search word among the elements of the model, and based on the value of “y” calculated by substituting “0” for the other elements. To estimate whether the user is interested in the first field.

  Further, the estimation unit 133 estimates the user's interest in the second field that is more detailed than the first field, based on the elements acquired by the acquisition unit 132. Specifically, the estimation unit 133 estimates the user's interest in the second field based on an element corresponding to the user information of the user among the elements of the model that predicts the user's interest in the first field. In addition, the estimation unit 133 estimates the user's interest in the second field based on the elements corresponding to the second field among the model elements. Further, the estimation unit 133 estimates the user's interest in the second field based on an element indicating that the user is interested in the first field among the elements of the model. For example, the estimation unit 133 uses the user information indicating the search word used for the search by the user to calculate the score of the element corresponding to the second field that has a positive weight among the model elements. To do. As an example, the estimation unit 133 refers to the search word used for the search by the user, and is based on the number of the elements corresponding to the second field among the elements of the model and the search word used for the search by the user. To calculate the score of the element corresponding to the second field. Then, the estimation unit 133 estimates that the field corresponding to the element having the higher calculated score is the field in which the user is more interested. Note that the estimation unit 133 estimates the user's interest in the second field when the user is interested in the first field.

(About the classification unit 134)
The classification unit 134 classifies the user into the second field estimated by the estimation unit 133. For example, the classification unit 134 classifies the user based on the ratio of elements corresponding to the user information of the user among elements corresponding to the second field. As an example, the classification unit 134 classifies the user into the second field having the largest number of elements corresponding to the second field and the search word used for the search by the user. Then, the classification unit 134 stores the classified users in the classification information storage unit 124 in association with the second field.

[1-3. Estimation processing procedure according to embodiment]
Next, the procedure of the process by the estimation system 1 according to the embodiment will be described using FIG. FIG. 8 is a flowchart illustrating an estimation processing procedure by the estimation system 1 according to the embodiment.

  As illustrated in FIG. 8, first, the estimation device 100 receives user information from the terminal device 10 (step S101). For example, the estimation apparatus 100 receives, as user information, a search word used for a search by a user who is interested in the first field from the terminal apparatus 10. Then, the estimating apparatus 100 stores the received user information in the user information storage unit 121 (step S102).

  Subsequently, the estimation apparatus 100 estimates whether the user is interested in the first field using a model that predicts the user's interest in the first field (step S103). For example, the estimation apparatus 100 estimates whether or not the user is interested in the first field by inputting information related to the search word used by the user in the model that predicts the user's interest in the first field. To do.

  Subsequently, the estimating apparatus 100 acquires a model element that predicts the user's interest in the first field (step S104). For example, when it is estimated that the user is interested in the first field, the estimation apparatus 100 refers to the model information storage unit 122 and is generated by learning user information by machine learning. Get the elements of the model that predict the user's interest in the field.

  Subsequently, the estimation apparatus 100 estimates the user's interest in the second field, which is more detailed than the first field, based on the elements acquired by the acquisition unit 132 (Step S105). Specifically, the estimation apparatus 100 estimates the user's interest in the second field based on the elements corresponding to the second field among the elements of the model that predicts the user's interest in the first field. For example, when the user is interested in the first field, the estimating apparatus 100 estimates the user's interest in the second field.

  Thereafter, the estimating apparatus 100 classifies the user into the second field estimated by the estimating unit 133 (step S106). For example, the estimating apparatus 100 classifies users based on a ratio of elements corresponding to the user information of the user among elements corresponding to the second field. Then, the estimating apparatus 100 stores the classified user in the classification information storage unit 134 in association with the second field (step S107).

[1-4. Effects of the embodiment
As described above, the estimation apparatus 100 according to the embodiment includes the acquisition unit 132 and the estimation unit 133. The acquisition unit 132 acquires a model element that predicts a user's interest in the first field. Based on the element acquired by the acquisition unit 132, the estimation unit 133 estimates the user's interest in the second field that is more detailed than the first field.

  Thereby, since the estimation apparatus 100 can estimate the user's interest in the second field, which is more detailed than the first field, the user's interest can be estimated in detail. For example, since the estimation apparatus 100 can estimate which field among golf fields a user who likes golf is interested in, it is possible to estimate the user's interest in more detail. As an example, the estimation apparatus 100 estimates in detail for each user whether the same golf enthusiast is interested in a golf club, whether to play golf, or interested in a golf player. be able to.

  In addition, since the estimation apparatus 100 can estimate the user's interest in the second field, which is more detailed than the first field, using the elements of the model that predicts the user's interest in the first field, more detailed information can be obtained. It is possible to reduce the burden of estimating the user's interest. For example, the estimation apparatus 100 estimates the user's interest in the second field without taking the trouble of generating the model, as compared with a case where a model for predicting the user's interest in the second field is newly generated. Therefore, the burden on estimating the user's interest can be reduced. In addition, the estimation apparatus 100 estimates the user's detailed interest without taking a great deal of time and effort to newly generate a model for predicting the user's interest in the second field, which is innumerable in more detail than the first field. be able to.

  Moreover, in the estimation apparatus 100 according to the embodiment, the estimation unit 133 estimates the user's interest in the second field based on an element corresponding to the user information of the user among the model elements. Thereby, since the estimation apparatus 100 can estimate the user's detailed interest in association with information about the user, the user's detailed interest can be estimated with high accuracy.

  In the estimation apparatus 100 according to the embodiment, the estimation unit 133 estimates the user's interest in the second field based on the elements corresponding to the second field among the model elements. Thereby, since the estimation apparatus 100 can estimate a user's detailed interest using the element relevant to a 2nd field | area, it can estimate a user's detailed interest with high precision.

  In the estimation apparatus 100 according to the embodiment, the estimation unit 133 estimates the user's interest in the second field based on an element indicating that the user is interested in the first field among the model elements. To do. Thereby, the estimation apparatus 100 can estimate a user's detailed interest with high precision.

  In the estimation apparatus 100 according to the embodiment, the estimation unit 133 estimates the user's interest in the second field when the user is interested in the first field. Thereby, since the estimation apparatus 100 can estimate the interest in the detailed field of the user who is interested in the first field, the user's interest can be made clearer.

  In addition, the estimation apparatus 100 can estimate the user's interest in the second field in more detail than the first field using the elements of the model that predicts the user's interest in the first field. The user's interest in the field can be estimated with high accuracy. For example, since the estimation apparatus 100 uses a model element that predicts a user's interest in a first field that is a broader category than the second field, a model element that has a large amount of learning information and a statistically high accuracy is used. The user's interest can be estimated with high accuracy.

  In addition, the estimation apparatus 100 according to the embodiment further includes a classification unit 134 that classifies users into the second field estimated by the estimation unit 133. Thereby, since the estimation apparatus 100 can classify a user into a field in which the user is interested, the user can be classified into a suitable category. For example, the estimation apparatus 100 can perform an effective response to the user, such as distributing an advertisement suitable for the user, by classifying the user into a suitable category.

  In the estimation apparatus 100 according to the embodiment, the classification unit 134 classifies users based on the ratio of elements corresponding to the user information of the user among elements corresponding to the second field. Thereby, the estimation apparatus 100 can classify a user into a suitable category even when the user is interested in a plurality of fields.

[2. (Modification)
The estimation apparatus 100 according to the above-described embodiment may be implemented in various different forms other than the above-described embodiment. Therefore, in the following, another embodiment of the estimation device 100 will be described.

[2-1. (Bilinear)
In the above-described embodiment, an example has been described in which the estimation apparatus 100 estimates the user's interest in the second field, which is more detailed than the first field, based on the elements acquired by the acquisition unit 132. Here, the estimation apparatus 100 may estimate the user's interest based on an element group obtained by combining a plurality of elements.

  Specifically, the estimation unit 133 of the estimation apparatus 100 estimates the user's interest in the second field corresponding to the element group based on the element group obtained by combining a plurality of elements. For example, when targeting a user who is interested in a drive, the estimating apparatus 100 estimates the interest in the route based on an element group that combines scenery and traffic jam as a plurality of elements. Here, the user information of such a user indicates that he likes a route with a beautiful scenery but dislikes a route with a traffic jam. Further, it is assumed that the user information of the user prefers a route with a beautiful scenery even when there is a traffic jam. In this case, the estimation device 100 estimates a route with a beautiful scenery although there is a traffic jam as a more detailed interest of the user. Thus, for example, when the user searches for a route by car navigation, the estimation device 100 can propose a route with a beautiful scenery although there is a traffic jam that is a route suitable for the user.

  As described above, the estimation apparatus 100 according to the modification estimates the user's interest in the second field corresponding to the element group based on the element group obtained by combining a plurality of elements. Thereby, since the estimation apparatus 100 can estimate a user's interest based on the element group which combined the some element, it can estimate a user's interest with a finer granularity. For example, since the estimation apparatus 100 can estimate the user's interest based on a more expressive feature by combining a plurality of elements, it can accurately estimate the user's interest.

  In addition, the estimation apparatus 100 may estimate the user's interest in the second field based on an element group in which three or more elements are combined without being limited to two elements. Thereby, since the estimation apparatus 100 can estimate the user's interest with finer granularity, the user's interest can be estimated more accurately.

[2-2. (Model quality enhancement)
In the above embodiment, an example has been described in which the estimation apparatus 100 acquires a model element that predicts a user's interest in the first field. Here, the estimation apparatus 100 may acquire an element of a model whose quality is extended when a predetermined condition is satisfied.

  Specifically, the acquisition unit 132 of the estimation device 100 corrects the model by extending the learning information when the accuracy of the user's interest in the first field predicted by the model is less than a predetermined threshold, Get the corrected model elements. For example, the estimating apparatus 100 corrects the model by performing machine learning using learning information related to the first field and having different types, and acquires the corrected model elements.

  As described above, the estimation apparatus 100 according to the modification corrects the model by extending the learning information when the accuracy of the user's interest in the first field predicted by the model is less than the predetermined threshold, Get the model element. Thereby, since the estimation apparatus 100 can expand the quality of the model, the user's interest in the second field can be estimated with high accuracy. For example, since the estimation apparatus 100 can estimate the user's interest in the second field using high-quality model elements, the user's interest in the second field can be estimated with higher accuracy. .

[2-3. (Time series analysis)
In the above embodiment, an example in which the estimation apparatus 100 classifies users in the second field estimated by the estimation unit 133 has been described. Here, the estimation apparatus 100 may classify users using elements of a predetermined period.

  Specifically, the classification unit 134 of the estimation apparatus 100 classifies users using elements of a predetermined period among elements corresponding to the second field. For example, the estimating apparatus 100 classifies the user into the second field having the largest number of matches between the element corresponding to the second field and the search word used by the user in the search for a predetermined period. As an example, the estimation apparatus 100 classifies users as elements having a predetermined period corresponding to the second field in the most recent period.

  As described above, the estimation apparatus 100 according to the modification classifies users using elements of a predetermined period among elements corresponding to the second field. Thereby, since the estimation apparatus 100 can classify a user into a field in which the user is interested in a predetermined period, the user can be classified into a more accurate field. For example, the estimating apparatus 100 can classify users into fields that are still likely to be of interest by classifying users based on the elements of the most recent period. In addition, since the estimation apparatus 100 can compare a classification based on an element of a predetermined period with a classification based on an element of another period, the estimation apparatus 100 can grasp a time series change of a field of interest that the user has. Can do. For this reason, since the estimation apparatus 100 can estimate a field in which the user is most likely to be interested next, it is possible to execute a response such as advertisement distribution in which the transition of the user's interest is pre-read.

[2-4. (Use elements with negative weight)
In the above-described embodiment, an example is described in which the estimation apparatus 100 estimates the user's interest in the second field based on an element indicating that the user is interested in the first field among the elements of the model. did. Here, the estimation apparatus 100 may estimate the user's interest using an element having a negative weight among the elements of the model.

  Specifically, the estimation apparatus 100 estimates the user's interest based on an element indicating that the user is not interested in the first field among the elements of the model. More specifically, the estimation apparatus 100 estimates the user's interest using an element having a negative weight among the elements of the model. For example, when the user is interested in the first field, the estimation apparatus 100 uses an element having a negative weight among elements of a model that predicts the user's interest in the first field. , It is assumed that the user is interested in the first field, but not interested in the second field that is more detailed than the first field. As an example, when the user has an interest in “golf” and the weight of the element “tan” of the model that predicts the user's interest in “golf” is a negative value, It is presumed that such a user likes “golf” but dislikes a portion that “tans”.

  Thus, the estimation apparatus 100 estimates the user's interest based on an element indicating that the user is not interested in the first field among the elements of the model. Thereby, the estimation apparatus 100 can estimate a user's interest in detail. For this reason, since the estimation apparatus 100 can work on the user by using the gap between the comprehensive evaluation of the user's interest and the individual features of the user's interest, the estimation apparatus 100 can take a more appropriate response to the user. it can. For example, since the estimation apparatus 100 can distribute an advertisement related to sunscreen as well as an advertisement related to golf to a user who likes “golf” but dislikes a “sunburn” part, The advertising effect can be enhanced.

  In addition, when the user is not interested in the first field, the estimation apparatus 100 determines an element having a negative weight among the elements of the model that predicts the user's interest in the first field. By using it, the reason why the user is not interested in the first field may be estimated. For example, when the user is not interested in “golf”, the estimating apparatus 100 displays an element “sunburn” having a negative weight among the elements of the model. Estimated as a reason for not having.

  In another example, the estimation apparatus 100 estimates the user's interest by using an element having a negative weight among the elements of the model when there are only two conflicting estimation results. For example, it is assumed that there are only party A and party B. Also, if you support Party A, you will not support Party B. If you support Party B, you will not support Party A. In this case, the estimation apparatus 100 estimates the user's interest in the A party using an element having a negative weight among the elements of the model that predicts the user's interest in the B party. This is because an element having a negative weight among the elements of the model for predicting the user's interest in the B party is a reason for not supporting the B party, and is therefore considered to be a reason for supporting the A party.

  As described above, the estimation apparatus 100 according to the modified example estimates the user's interest based on an element indicating that the user is not interested in the first field among the elements of the model. Thereby, since the estimation apparatus 100 can estimate the user's interest based on the elements corresponding to the conflicting fields, the user's interest can be estimated with higher accuracy.

[2-5. (Advertisement delivery)
In the above-described embodiment, an example has been described in which the estimation apparatus 100 estimates the user's interest in the second field, which is more detailed than the first field, based on the elements acquired by the acquisition unit 132. Here, the estimation apparatus 100 may distribute an advertisement based on the user's interest in the estimated second field.

  Specifically, the estimating apparatus 100 distributes an advertisement having contents related to the second field in which the user is interested to the user. For example, when the user who is interested in the golf club browses the web page, the estimation apparatus 100 displays an advertisement related to the golf club in the advertisement frame of the web page.

  Thus, the estimation apparatus 100 according to the modified example distributes the advertisement based on the user's interest in the estimated second field. Thereby, since the estimation apparatus 100 which concerns on a modification can distribute an advertisement etc. based on a user's more detailed interest, it can improve an advertising effect.

[2-6. (Recommendation)
In the above-described embodiment, an example has been described in which the estimation apparatus 100 estimates the user's interest in the second field, which is more detailed than the first field, based on the elements acquired by the acquisition unit 132. Here, the estimation apparatus 100 may perform the recommendation based on the user's interest in the estimated second field.

  For example, when the user browses an online shopping site, the estimating apparatus 100 recommends a product having a content related to the second field in which the user is interested. As an example, when a user who is estimated to be interested in a golf club browses an online shopping site, the estimation device 100 displays a product image in which a link to a purchase page of a golf club is embedded as a site recommendation frame. To be displayed.

  As described above, the estimation apparatus 100 according to the modification performs recommendation based on the user's interest in the estimated second field. Thereby, since the estimation apparatus 100 can recommend goods etc. based on a user's more detailed interest, it can raise a user's goods purchase rate.

  In addition, the estimation apparatus 100 may recommend a field related to the field estimated to be of interest to the user. For example, when a user who is estimated to be interested in “golf club” browses an online shopping site, the estimation apparatus 100 recommends that an image embedded with a link to transition to a “golf wear” purchase page is recommended by the site. Display in the frame. This is because a user who is interested in “golf club” is likely to be interested in “golf wear”. Thereby, since the estimation apparatus 100 can recommend a product or the like of interest that the user himself / herself is not aware of, the product purchase rate of the user can be increased.

[2-7. (Estimation process)
In the above-described embodiment, an example has been described in which when the estimation apparatus 100 estimates that the user is interested in the first field, the user's interest in the second field is estimated. Here, the estimation apparatus 100 may not estimate whether the user is interested in the first field. Specifically, the estimation apparatus 100 may estimate the user's interest in the second field without estimating whether the user is interested in the first field.

[2-8. (Applicable)
In the above-described embodiment, the estimation apparatus 100 has been described with an example in which the user is classified into a field in which the user is estimated to be interested. Here, the estimation apparatus 100 may be applied not only to user classification but also to product classification. For example, the estimation apparatus 100 may be applied to clustering of products sold on an online shopping site. Thereby, since the estimation apparatus 100 can automatically classify the products, it is possible to reduce the trouble of setting the product classification.

[2-9. User information)
In the above embodiment, an example in which a search word used by a user for a search is adopted as an example of user information has been described. Here, the user information is not limited to the search word, and may be adopted from various types of information related to the user. For example, as the user information, feature words extracted from a purchase history in which a user has purchased a product at an online shopping site or a usage history using a service may be employed. In another example, the keyword extracted from the browsing history etc. of the web page which the user browsed in the past may be employ | adopted for user information. As an example, a feature word having a relatively high appearance frequency included in a news article browsed in the past by the user may be adopted as the user information.

[2-10. Effect of modification)
As described above, in the estimation apparatus 100 according to the modification, the classification unit 134 classifies users using elements of a predetermined period among elements corresponding to the second field. Thereby, since the estimation apparatus 100 according to the modification can classify the user into a field in which the user is interested in a predetermined period, the user can be classified into a more accurate field. For example, the estimating apparatus 100 can classify users into fields that are still likely to be of interest by classifying users based on the elements of the most recent period. In addition, since the estimation apparatus 100 can compare a classification based on an element of a predetermined period with a classification based on an element of another period, the estimation apparatus 100 can grasp a time series change of a field of interest that the user has. Can do. For this reason, since the estimation apparatus 100 can estimate a field in which the user is most likely to be interested next, it is possible to execute a response such as advertisement distribution in which the transition of the user's interest is pre-read.

  Moreover, in the estimation apparatus 100 according to the modification, the estimation unit 133 estimates the user's interest in the second field corresponding to the element group based on the element group obtained by combining a plurality of elements. Thereby, since the estimation apparatus 100 which concerns on a modification can estimate a user's interest based on the element group which combined the some element, it can estimate a user's interest with a finer granularity. For example, since the estimation apparatus 100 can estimate the user's interest based on a more expressive feature by combining a plurality of elements, it can accurately estimate the user's interest.

  Further, in the estimation apparatus 100 according to the modification, the acquisition unit 132 corrects the model by extending the learning information when the accuracy of the user's interest in the first field predicted by the model is less than a predetermined threshold. Then, the corrected model element is obtained. Thereby, since the estimation apparatus 100 according to the modified example can expand the quality of the model, it is possible to estimate the user's interest in the second field with high accuracy. For example, since the estimation apparatus 100 can estimate the user's interest in the second field using high-quality model elements, the user's interest in the second field can be estimated with higher accuracy. .

  Moreover, in the estimation apparatus 100 according to the modification, the estimation unit 133 estimates the user's interest based on an element indicating that the user is not interested in the first field among the elements of the model. Thereby, the estimation apparatus 100 can estimate a user's interest in detail.

[3. Others]
In addition, among the processes described in the above embodiment, all or part of the processes described as being automatically performed can be performed manually, or the processes described as being performed manually can be performed. All or a part can be automatically performed by a known method. In addition, the processing procedures, specific names, and information including various data and parameters shown in the document and drawings can be arbitrarily changed unless otherwise specified.

  Further, each component of each illustrated apparatus is functionally conceptual, and does not necessarily need to be physically configured as illustrated. In other words, the specific form of distribution / integration of each device is not limited to that shown in the figure, and all or a part thereof may be functionally or physically distributed or arbitrarily distributed in arbitrary units according to various loads or usage conditions. Can be integrated and configured.

  For example, the model information storage unit 122 illustrated in FIG. 3 may be stored in a storage server or the like without the estimation apparatus 100. In this case, the estimation apparatus 100 acquires a model element that predicts the user's interest in the first field by accessing the storage server.

  In addition, the estimation device 100 may be an estimation device that performs only recommendation processing without performing classification processing. In this case, the estimation device does not have the classification unit 134. Then, the classification device having the classification unit 134 classifies the users.

  Further, the estimation apparatus 100 according to the embodiment described above is realized by a computer 1000 having a configuration as shown in FIG. 9, for example. Hereinafter, the estimation apparatus 100 will be described as an example. FIG. 9 is a hardware configuration diagram illustrating an example of a computer 1000 that implements the functions of the estimation apparatus 100. The computer 1000 includes a CPU 1100, RAM 1200, ROM 1300, HDD 1400, communication interface (I / F) 1500, input / output interface (I / F) 1600, and media interface (I / F) 1700.

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

  The HDD 1400 stores programs executed by the CPU 1100, data used by the programs, and the like. The communication interface 1500 receives data from other devices via the 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.

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

  The media interface 1700 reads a program or data stored in the recording medium 1800 and provides it to the CPU 1100 via the RAM 1200. The CPU 1100 loads such a program from the recording medium 1800 onto the RAM 1200 via the media interface 1700, and executes the loaded program. The recording medium 1800 is, for example, an optical recording medium such as a DVD (Digital Versatile Disc) or PD (Phase change rewritable disk), a magneto-optical recording medium such as an MO (Magneto Optical disk), a tape medium, a magnetic recording medium, or a semiconductor memory. It is.

  For example, when the computer 1000 functions as the estimation device 100 according to the embodiment, the CPU 1100 of the computer 1000 implements the function of the control unit 130 by executing a program loaded on the RAM 1200. The HDD 1400 stores data in the storage unit 120. The CPU 1100 of the computer 1000 reads these programs from the recording medium 1800 and executes them. However, as another example, these programs may be acquired from other devices via the network N.

  As described above, some of the embodiments of the present application have been described in detail based on the drawings. It is possible to implement the present invention in other forms with improvements.

  Further, the estimation apparatus 100 described above may be realized by a plurality of server computers, and depending on functions, an external platform or the like may be realized by calling an API (Application Programming Interface), network computing, or the like. Can be changed flexibly.

  Further, “section (module, unit)” described in the claims can be read as “means”, “circuit”, and the like. For example, the receiving unit can be read as receiving means or a receiving circuit.

DESCRIPTION OF SYMBOLS 1 Recommendation system 10 Terminal apparatus 100 Estimation apparatus 121 User information storage part 122 Model information storage part 123 Interest information storage part 124 Classification information storage part 131 Reception part 132 Acquisition part 133 Estimation part 134 Classification part

Claims (13)

An acquisition unit that acquires an element of a model that predicts a user's interest in the first field;
An estimation device comprising: an estimation unit configured to estimate a user's interest in a second field that is more detailed than the first field based on an element acquired by the acquisition unit. The estimation unit includes
The estimation apparatus according to claim 1, wherein the user's interest in the second field is estimated based on an element corresponding to user information of the user among elements of the model. The estimation unit includes
The estimation apparatus according to claim 1, wherein a user's interest in the second field is estimated based on an element corresponding to the second field among elements of the model. The estimation unit includes
The estimation apparatus according to claim 1, wherein when the user has an interest in the first field, the user's interest in the second field is estimated. The estimation unit includes
The user's interest in the second field is estimated based on an element indicating that the user is interested in the first field among the elements of the model. The estimation apparatus as described in any one. The estimation unit includes
6. The user's interest is estimated based on an element indicating that the user is not interested in the first field among the elements of the model. The estimation apparatus described in 1. The estimation apparatus according to claim 1, further comprising a classification unit that classifies users into the second field estimated by the estimation unit. The classification unit includes:
The estimation apparatus according to claim 7, wherein the user is classified based on a ratio of elements corresponding to the user information of the user among elements corresponding to the second field. The classification unit includes:
The estimation apparatus according to claim 7 or 8, wherein the user is classified using an element of a predetermined period among elements corresponding to the second field. The estimation unit includes
The estimation apparatus according to claim 1, wherein the user's interest in the second field corresponding to the element group is estimated based on an element group obtained by combining a plurality of elements. The acquisition unit
When the accuracy of the user's interest in the first field predicted by the model is less than a predetermined threshold, the learning information is expanded to correct the model, and the corrected model element is obtained. The estimation device according to any one of claims 1 to 10. An estimation method executed by an estimation device,
Obtaining a model element that predicts the user's interest in the first field;
An estimation method comprising: an estimation step of estimating a user's interest in a second field that is more detailed than the first field based on the element acquired by the acquisition step. An acquisition procedure for acquiring an element of a model that predicts a user's interest in the first field;
An estimation program for causing a computer to execute an estimation procedure for estimating a user's interest in a second field that is more detailed than the first field based on elements acquired by the acquisition procedure.

Images