JP6440661B2 - Determination device, determination method, and determination program - Google Patents

Description

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

  In recent years, with the rapid spread of the Internet, a lot of information has been transmitted and received. Therefore, for example, an attempt has been made to utilize information transmitted and received by the user for business.

  For example, Patent Literature 1 discloses a technique for calculating a person's creditworthiness by a calculation process using information obtained from a social network (SNS).

JP 2014-206792 A

  However, with the above-described conventional technology, it is not always possible to determine a user's social credibility with high accuracy. For example, in the above-described conventional technology, the information processing apparatus determines the user's creditworthiness in credit based on the user's profile on the SNS, comments made by the user on the SNS, user friendships on the SNS, and the like. Generate the credit model shown. Then, the information processing apparatus calculates a risk score using the generated credit model. However, the user often pretends to be a fictitious person on the SNS, and the above-described conventional technology does not always determine the user's social credibility with high accuracy.

  The present application has been made in view of the above, and an object thereof is to provide a determination device, a determination method, and a determination program that can determine a user's social credibility with high accuracy.

  The determination apparatus according to the present application is based on an acquisition unit that acquires behavior information of a predetermined user in electronic commerce, and the relationship between an arbitrary user behavior information in electronic commerce and the reliability of the arbitrary user. And a determination unit that determines the reliability of the predetermined user from which the behavior information is acquired.

  According to one aspect of the embodiment, there is an effect that the social credibility of the user can be determined with high accuracy.

FIG. 1 is a diagram illustrating an example of a determination process according to the embodiment. FIG. 2 is a diagram illustrating a configuration example of the determination apparatus according to the embodiment. FIG. 3 is a diagram illustrating an example of the behavior tendency storage unit according to the embodiment. FIG. 4 is a flowchart illustrating a determination processing procedure by the determination apparatus according to the embodiment. FIG. 5 is a hardware configuration diagram illustrating an example of a computer that realizes the function of the determination apparatus.

  Hereinafter, a mode for carrying out a determination device, a determination method, and a determination program according to the present application (hereinafter referred to as “embodiment”) will be described in detail with reference to the drawings. In addition, the determination apparatus, the determination method, and the determination program concerning this application are not limited by this embodiment.

[1. Determination process〕
First, an example of the determination process according to the embodiment will be described with reference to FIG. FIG. 1 is a diagram illustrating an example of a determination process according to the embodiment. The determination process according to the embodiment is performed by the determination apparatus 100 illustrated in FIG.

  The determination system 1 shown in FIG. 1 includes a shopping server 10 and a determination device 100. The shopping server 10 and the determination apparatus 100 are connected via a network so as to be communicable by wire or wirelessly. Note that the determination system 1 illustrated in FIG. 1 may include a plurality of shopping servers 10 and a plurality of determination devices 100.

  First, the shopping server 10 is a server device that performs electronic commerce with a user. Specifically, the shopping server 10 is a server device that provides products that are exhibited by a plurality of stores so that they can be purchased online. For example, the user accesses the shopping server 10 using his / her terminal device, searches for a purchase desired product, and places an order for the searched purchase desired product.

  In addition, for example, the user performs an evaluation action for evaluating a product purchased after purchasing the product. An example of such evaluation behavior is the provision of an evaluation score indicating the degree of satisfaction with the product (for example, 3 out of 5), or the impression and evaluation comment on the product, that is, the posting of a review comment.

  The determination apparatus 100 according to the embodiment acquires behavior information of a predetermined user in electronic commerce, and acquires behavior information based on a relationship between behavior information of an arbitrary user in electronic commerce and the reliability of the arbitrary user. A predetermined user's credit rating is determined.

  Specifically, the determination apparatus 100 generates a learning model by performing machine learning on the correlation between the behavior information of an arbitrary user and the reliability of the arbitrary user. And the determination apparatus 100 determines the reliability of a predetermined user by applying the action information acquired from the predetermined user to the learning model.

  In the present embodiment, the determination device 100 is assumed to be a server device used by a predetermined financial institution (for example, a bank, a card company, a loan company, etc.). Below, an example of the determination process performed by the determination apparatus 100 is demonstrated.

  First, the determination apparatus 100 includes a behavior tendency storage unit 120. The behavior tendency storage unit 120 is a storage unit that associates and stores behavior information on the Internet for users who have had financial transactions so far and an index indicating creditworthiness in the financial transaction. In the example of FIG. 1, the behavior tendency storage unit 120 stores behavior information on a product by a user who has purchased a product at the shopping server 10 and a credit loss rate in a conventional financial transaction in association with each other. Specifically, the behavior trend storage unit 120 includes behavior information such as how many evaluation points each user has given to the same product among the products purchased by each user, and financial transactions so far. The credit loss rate of each user is stored in association with each other.

  In such a case, the determination apparatus 100 refers to the behavior tendency storage unit 120, and obtains a correlation between the behavior information such as how much the credit rating rate is given by the user to whom the evaluation point is given and the credit rating. Machine learning. Thereby, the determination apparatus 100 generates a trustworthiness model (an example of a learning model) for determining the trustworthiness of a predetermined user, that is, a trustworthiness determination target user (step S1).

  For example, the determination apparatus 100 learns a credit model that estimates the user's creditworthiness by causing the model to learn the correlation between the evaluation score given by the user and the credit loss rate of the user. For example, the determination apparatus 100 uses a SVM (Support Vector Machine) or DNN (Deep Neural Network) technology to create a trustworthiness model that performs clustering of users according to credit loss rates based on evaluation points given by users. create. As a result of such processing, the determination apparatus 100 can create a trustworthiness model for estimating the trustworthiness of the user from the evaluation points given by the user whose trustworthiness is unknown.

  For example, the determination apparatus 100 calculates an average value of the evaluation points given by each user stored in the behavior tendency storage unit 120, a user who gives an evaluation score lower than the calculated average value, and the user's Machine learning the correlation with the credit loss rate. For example, it is assumed that the determination apparatus 100 obtains a correlation such as “users who give an evaluation score lower than the average value tend to have a high rate of credit loss” by machine learning.

  On the other hand, it is assumed that the determination apparatus 100 obtains a correlation such as “users who are given an evaluation score higher than the average value tend to have a low credit loss rate” by machine learning.

  In the case of the above example, for example, the determination apparatus 100 is a credit model for calculating a credit rating in a financial transaction low for a user who has been given a score lower than the average value, and has a score higher than the average value. A creditworthiness model for calculating a high creditworthiness in the financial transaction is generated for the assigning user. FIG. 1 illustrates an example in which the determination apparatus 100 generates “credit model M1” as such a credit model.

  In such a state, it is assumed that the determination apparatus 100 receives a borrowing request (an example of a financial service) from the user U10 (step S2). In such a case, the determination apparatus 100 determines the user U10 in the financial transaction (borrowing) using the trustworthiness model M1 generated in step S1.

  Here, it is assumed that the user U10 has already performed an evaluation action such as giving an evaluation score by purchasing a product with the shopping server 10 so far. Therefore, the determination apparatus 100 accesses the shopping server 10 and acquires action information of the user U10 (step S3). FIG. 1 shows an example in which the determination apparatus 100 acquires “evaluation score 1” as the action information of the user U10. In addition, for example, when the user U10 has given evaluation points to a plurality of products so far, the determination apparatus 100 acquires the average value of the evaluation points for each product as the behavior information of the user U10. May be.

  Next, the determination apparatus 100 calculates a credit score by applying the acquired action information “grant evaluation score 1” to the credit model M1, and determines the credit of the user U10 based on the calculated credit score (step S4). For example, it is assumed that the determination apparatus 100 calculates a credit score “1” as illustrated in FIG. For example, if the maximum value of the credit score calculated from the credit model M1 is “5”, and this credit score “1” is lower than the threshold in the credit determination, the determination device 100 determines the credit of the user U10 in the financial transaction. Is determined to be “low”.

  For example, it can be said that dealing with the user U10 having a low degree of trust has a high risk for the financial institution managing the determination apparatus 100. Therefore, as illustrated in FIG. 1, the determination device 100 notifies the terminal device of the user U10 of information indicating “transaction impossible” (step S5).

  Although not shown, when the determination device 100 determines that the credit rating of the user U10 is “high”, the determination device 100 notifies the terminal device of the user U10 of information indicating “transaction permission”.

  As described above, the determination apparatus 100 according to the embodiment acquires behavior information of a predetermined user in electronic commerce, and based on the relationship between the behavior information of an arbitrary user in electronic commerce and the reliability of the arbitrary user. The reliability of a predetermined user who has acquired the information is determined.

  For example, since electronic commerce such as online shopping has high anonymity, it can be said that a user's person is likely to appear in the user's behavior in electronic commerce. The determination apparatus 100 according to the embodiment generates a credit model by machine learning the correlation between the person and the credit rating (for example, the credit loss rate) in the financial transaction. And the determination apparatus 100 can determine the reliability with high precision about the user whose reliability is unknown by using a reliability model.

[2. (Configuration of judgment device)
Next, the determination apparatus 100 according to the embodiment will be described with reference to FIG. FIG. 2 is a diagram illustrating a configuration example of the determination apparatus 100 according to the embodiment. As illustrated in FIG. 2, the determination apparatus 100 according to the embodiment includes a communication unit 110, a behavior tendency storage unit 120, and a control unit 130.

(About the communication unit 110)
The communication unit 110 is realized by, for example, a NIC (Network Interface Card). The communication unit 110 is connected to the network by wire or wireless, and transmits and receives information to and from the shopping server 10, for example.

(About the behavior tendency storage unit 120)
The behavior tendency 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 behavior tendency storage unit 120 is a storage unit that stores behavior information on the Internet for users who have had financial transactions so far in association with an index indicating creditworthiness in the financial transaction.

  Here, FIG. 3 shows an example of the behavior tendency storage unit 120 according to the embodiment. In the example of FIG. 3, the behavior tendency storage unit 120 includes items such as “user ID”, “evaluation point”, “comment length”, “comment rate”, and “lender collapse rate”.

  The “user ID” is a user who has made a financial transaction so far with the determination apparatus 100 or a financial institution that manages the determination apparatus 100, and is a predetermined evaluation action for the purchased product purchased at the shopping server 10. The identification information which identifies the user who performed is shown.

  “Evaluation point”, “comment length”, and “comment rate” are examples of user behavior information. “Evaluation score” is information indicating an evaluation action such as giving an evaluation score to a purchased product. In the example of FIG. 3, it is shown that the user identified by the user ID “U1” has performed an evaluation action such as giving “evaluation point 2” to his / her purchased product.

  “Comment length” is information indicating an evaluation action such as posting a review comment for a purchased product. In the example of FIG. 3, it is shown that the user identified by the user ID “U1” has performed an evaluation action such as posting a review comment having a comment length of “90 characters” on his / her purchased product.

  The “comment rate” is information indicating an evaluation behavior such as how much a corresponding user has posted a review comment for a previously purchased product. In the example of FIG. 3, the ratio of the user identified by the user ID “U1” who has posted the review comment so far with respect to his / her purchased product is “3%”. It should be noted that the purchased merchandise targeted in the behavior tendency storage unit 120 is the same for all users. Or the purchase goods made into object in the action tendency memory | storage part 120 may be a thing of the same classification or the same category with all the users.

  The “loan rate” indicates the ratio of the user's credit transaction out of the determination apparatus 100 or the financial transaction of the user with the financial institution managing the determination apparatus 100. The example of FIG. 3 shows an example in which the credit loss rate of the user identified by the user ID “U1” is “75%”.

(About the control unit 130)
Returning to FIG. 2, the control unit 130 executes various programs stored in the storage device inside the determination apparatus 100 using the RAM as a work area by a CPU (Central Processing Unit), an MPU (Micro Processing Unit), or the like. It is realized by. 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 illustrated in FIG. 2, the control unit 130 includes a model generation unit 131, a provision unit 132, an acquisition unit 133, and a determination unit 134, and implements or executes information processing functions and operations described below. To do. The internal configuration of the control unit 130 is not limited to the configuration illustrated in FIG. 2, and may be another configuration as long as the information processing described later is performed. In addition, the connection relationship between the processing units included in the control unit 130 is not limited to the connection relationship illustrated in FIG. 2, and may be another connection relationship.

(About the model generation unit 131)
The model generation unit 131 generates a trustworthiness model by machine learning of the correlation between the behavior information of any user and the trustworthiness of any user. This point will be described with reference to the example of FIG.

  For example, the model generation unit 131 refers to the behavior tendency storage unit 120 and performs machine learning on the correlation between the behavior information such as how much the credit rating rate is given by the user to whom the evaluation point is given and the credit rating. .

  For example, the model generation unit 131 calculates an average value of evaluation scores given by each user stored in the behavior tendency storage unit 120, and a user who gives an evaluation score lower than the calculated average value and the user Machine learning of the correlation with the default rate of credit loss. In addition, for example, the model generation unit 131 performs machine learning on the correlation between a user who is assigned an evaluation score higher than the calculated average value and the credit loss rate of the user.

  Here, it is assumed that the model generation unit 131 obtains a correlation such as “users who give an evaluation score lower than the average value tend to have a high rate of credit loss” by machine learning. On the other hand, it is assumed that the model generation unit 131 obtains a correlation such as “users who are given an evaluation score higher than the average value tend to have a low credit loss rate” by machine learning. In such a case, the model generation unit 131 is a credit model for calculating a low credit score in a financial transaction to a user who is assigned a score lower than the average value, and the model generator 131 assigns a score higher than the average value. The credit model M1a for calculating a high creditworthiness in the financial transaction is generated for the user who is present.

  In addition, the model generation unit 131 refers to the behavior tendency storage unit 120 and calculates the correlation between the behavior information such as how much the user who posts the review comment has how much the credit loss rate and the credit rating. Machine learning. For example, the model generation unit 131 calculates the average value of the comment length posted by each user stored in the behavior tendency storage unit 120, and the user who has posted a review comment shorter than the calculated average value, Machine learning of the correlation with the user's credit loss rate. For example, the model generation unit 131 performs machine learning on the correlation between a user who has posted a review comment longer than the calculated average value and the credit loss rate of the user.

  Here, it is assumed that the model generation unit 131 obtains a correlation such as “a user who has posted a review comment shorter than the average value tends to have a high credit loss rate” by machine learning. On the other hand, it is assumed that the model generation unit 131 obtains a correlation such as “users who have posted review comments longer than the average value tend to have a low credit loss rate” by machine learning. In such a case, the model generation unit 131 is a credit model for calculating a low credit rating in a financial transaction to a user who has posted a review comment shorter than the average value, and posts a review comment longer than the average value. For users who are present, a credit level model M1b for generating a high credit level in financial transactions is generated.

  In addition, the model generation unit 131 refers to the behavior tendency storage unit 120 and performs machine learning on the correlation between the behavior information such as how much the user with a comment rate is and how much the credit loss rate is, and the credit rating. For example, the model generation unit 131 calculates an average value of comment rates posted by each user stored in the behavior tendency storage unit 120, a user with a comment rate lower than the calculated average value, and a credit loss rate of the user Machine learning the correlation between the two. For example, the model generation unit 131 performs machine learning on the correlation between a user with a comment rate higher than the calculated average value and the credit loss rate of the user.

  Here, it is assumed that the model generation unit 131 obtains a correlation such as “users with a comment rate lower than the average value tend to have a high loan loss rate” by machine learning. On the other hand, it is assumed that the model generation unit 131 obtains a correlation such as “users with a comment rate higher than the average value tend to have a low loan loss rate” by machine learning. In such a case, the model generation unit 131 is a credit model for calculating a low credit rating in a financial transaction for a user with a comment rate lower than the average value, and a user with a comment rate higher than the average value is in a financial transaction. A trustworthiness model M1c for calculating a trustworthiness is generated.

  And the model production | generation part 131 produces | generates the reliability model M11 as one reliability model which integrated the reliability models M1a-M1c produced | generated as mentioned above, for example.

(About the providing unit 132)
The providing unit 132 provides various financial services to the user. Further, the providing unit 132 receives a request from the user to receive a financial service.

(About the acquisition unit 133)
The acquisition unit 133 acquires behavior information of a predetermined user in electronic commerce. For example, the acquisition unit 133 uses, as behavior information, an evaluation behavior for a product, an evaluation behavior for a product seller, an evaluation behavior for a product purchaser, a behavior for posting a comment, a behavior for canceling a transaction in product sales, or a product The information regarding at least one action is acquired among each action performed until purchasing.

(About determination unit 134)
The determination unit 134 determines the reliability of a predetermined user whose behavior information is acquired by the acquisition unit 133 based on the relationship between the behavior information of an arbitrary user in electronic commerce and the reliability of the arbitrary user. Specifically, the determination unit 134 determines the creditworthiness of a predetermined user by applying the behavior information acquired by the acquisition unit 133 to the creditworthiness model. Further, for example, the determination unit 134 determines the reliability used when providing the financial service.

  Here, the process of determining the credit rating of the user U10 using the trustworthiness model M11 generated by the model generating unit 131 as described above will be described. First, acquisition part 133 accesses shopping server 10, and acquires action information on user U10.

  For example, it is assumed that the acquisition unit 133 acquires “evaluation score 1” as information related to the evaluation behavior for the purchased product. In addition, it is assumed that the acquisition unit 133 acquires “post a review comment with a comment length of 70 characters” as information regarding the action of posting a comment. Further, it is assumed that the acquisition unit 133 acquires “comment rate 10%”.

  Then, the determination unit 134 calculates the credit score by applying the three behavior information acquired by the acquisition unit 133 to the credit model M11. For example, it is assumed that the determination unit 134 calculates the credit score “1” in the credit model M11 having the maximum credit score of “5”. If the credit score “1” is lower than the threshold value in the credit determination, the determination unit 134 determines that the credit of the user U10 in the financial transaction is “low”.

  Thus, when it is determined that the credit rating is “low”, the providing unit 132 notifies the terminal device of the user U10 of information indicating “transaction impossible”. On the other hand, when the reliability is determined to be “high”, the providing unit 132 notifies the terminal device of the user U10 of information indicating “transaction permission”.

[3. Processing procedure)
Next, a procedure of determination processing executed by the determination apparatus 100 according to the embodiment will be described with reference to FIG. FIG. 4 is a flowchart illustrating a determination processing procedure performed by the determination apparatus 100 according to the embodiment. Here, it is assumed that the determination apparatus 100 has already generated a trustworthiness model.

  First, the providing unit 132 determines whether a financial service request (for example, a borrowing request) has been received from a user (step S101). When not receiving (step S101; No), the provision part 132 waits until it receives. On the other hand, if the obtaining unit 133 determines that the request for the financial service has been received by the providing unit 132 (step S101; Yes), the obtaining unit 133 obtains behavior information on the electronic commerce of the requesting user (step S102). .

  Next, the determination unit 134 calculates a reliability score by applying the behavior information acquired by the acquisition unit 133 to the reliability model (step S103). Next, the determination unit 134 determines whether or not the calculated usage score is greater than or equal to a threshold (step S104). If the determination unit 134 determines that the credit score is equal to or greater than the threshold value (step S104; Yes), the determination unit 134 transmits a notification permitting service provision to the user terminal device (step S105). On the other hand, if the determination unit 134 determines that the credit score is lower than the threshold (step S104; No), the determination unit 134 transmits a notification that the service cannot be provided to the user terminal device (step S106).

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

[4-1. Model generation (1)]
In the embodiment described above, the model generation unit 131 generates the learning model by performing machine learning on the correlation between the behavior information of the arbitrary user and the creditworthiness of the arbitrary user. However, the model generation unit 131 may generate such a learning model for each piece of user attribute information. In such a case, the determination unit 134 determines the reliability of the predetermined user using a learning model corresponding to the attribute information of the predetermined user. This point will be described with reference to the example of FIG. Here, an example in which “sex” is used as user attribute information is shown. In other words, the model generation unit 131 generates a learning model for each “gender” of any user based on the behavior information and credit rating of the arbitrary user.

  First, the behavior tendency storage unit 120 illustrated in FIG. 1 further includes an item “gender”. First, the model generation unit 131 divides the users stored in the behavior tendency storage unit 120 into the sexes “male” and “female”.

  In such a state, the model generation unit 131 machine-learns the correlation between the behavior information such as how much the credit rating rate is given by the user who gives the evaluation score for “male” and the credit rating. . Similarly, in the case of “female”, the model generation unit 131 performs machine learning on the correlation between the behavior information such as how much the credit rating rate is given by the user to whom the evaluation point is given and the credit rating.

  Here, it is assumed that the model generation unit 131 obtains a correlation such as “male users who are given an evaluation score lower than the average value tend to have a high credit loss rate” by machine learning. On the other hand, it is assumed that the model generation unit 131 obtains a correlation such as “male users who are given an evaluation score higher than the average value tend to have a low credit loss rate” by machine learning. In such a case, the model generation unit 131 is a credit model for calculating a credit rating in a financial transaction low to a male user who has been assigned a score lower than the average value, and the model generation unit 131 gives a score higher than the average value. For a male user who is, a credit model M1-M is generated for calculating a high credit rating in financial transactions.

  Although description is omitted, based on the same idea, the model generation unit 131 is a credit model for calculating low creditworthiness in a financial transaction for a female user who has been given a lower evaluation score than the average value, For a female user who is given a higher evaluation score than the average value, a credit rating model M1-W for generating a high credit rating in the financial transaction is generated.

  Here, for example, it is assumed that the user U10 who has requested borrowing is “male”. In such a case, the determination unit 133 determines the reliability of the user U10 by applying the behavior information of the user U10 acquired by the acquisition unit 133 to the male reliability model M1-M.

  For example, depending on the attribute information of the user, a specific correlation may be established between the behavior information and the credit rating. In accordance with such a situation, the determination apparatus 100 according to the embodiment generates a trustworthiness model corresponding to user attribute information. And since the determination apparatus 100 can calculate the credit score of the user who is the target of credit determination using the credit model corresponding to the attribute information of the user who is the target of credit determination, the determination device 100 can determine the credit more accurately. .

  In this modification, an example in which the user's “gender” is used as the user attribute information is shown, but it is not necessary to be limited to “gender”. For example, if a correlation indicating a specific tendency is obtained for each “age” of the user as attribute information of the user by machine learning, the model generation unit 131 generates a credit model for each “age”. Also good.

[4-2. Model generation (2)]
In addition, the model generation unit 131 may generate a learning model for each user's health state based on the behavior information and reliability of any user. In such a case, the determination unit 134 determines the reliability of the predetermined user using a learning model corresponding to the health state of the predetermined user.

  If it demonstrates using the example of FIG. 1, the action tendency memory | storage part 120 will further have an item "health state." For example, “good” or “bad” may be input as the health state, and “disease name” or the like may be input as a more detailed classification.

  In addition, the determination device 100 collects, for example, behavior information of users who have had financial transactions so far from the shopping server 10 and accumulates them in the behavior tendency storage unit 120. At this time, the health state of each user And enter the determined health status. Moreover, not only this example but the determination apparatus 100 may determine a health state about each user accumulate | stored in the action tendency memory | storage part 120 at arbitrary timings.

  Moreover, the determination apparatus 100 can determine each user's health state using arbitrary methods. For example, the determination apparatus 100 may determine the health state based on the user's purchased product, or may determine the health state based on the posted content.

  In addition, when the determination apparatus 100 generates the trustworthiness model, the user stored in the behavior tendency storage unit 120 is classified for each health state, and the correlation between the behavior information and the credit loss rate for each health state. To learn. And the determination apparatus 100 produces | generates the learning model which shows the correlation, when a specific correlation is obtained between action information and a loan fall rate in a specific health state.

  For example, depending on the health condition of the user, a specific correlation may be established between the behavior information and the credit rating. In accordance with such a situation, the determination apparatus 100 according to the embodiment generates a trustworthiness model corresponding to the user's health state. And since the determination apparatus 100 can determine the creditworthiness of the user who is a creditworthiness determination target using the trustworthiness model corresponding to the health condition of the user who is the creditworthiness determination target, it can calculate the creditworthiness with higher accuracy.

[4-3. Model generation (3)]
Further, the model generation unit 131 may generate a trustworthiness model based on mutual rating performed between the seller and the purchaser. Here, although the example using the shopping server 10 which provides online shopping was shown in FIG. 1, it demonstrates taking the case of the auction server X which performs an auction process.

  The auction server X has a mechanism that allows sellers and successful bidders to evaluate each other (provide evaluation points). In such a case, the model generation unit 131 acquires a combination of the exhibitor, the successful bidder, and their evaluation points from the auction server X. And the model production | generation part 131 specifies collapse of evaluation balance about each combination. For example, if the maximum value of the evaluation score is “5” and the middle is “3”, the model generation unit 131 gives the seller who gives the highest evaluation score “5” to the successful bidder and the seller. The combination with the successful bidder who gives the minimum evaluation score “1” is specified.

  Then, the model generation unit 131 performs machine learning on the tendency of the credit loss rate of the seller who has performed the action of giving the maximum evaluation score “5” to the successful bidder who has given the minimum evaluation score “1”. In addition, the model generation unit 131 performs machine learning on the tendency of the credit loss rate of the successful bidder who performed the action such as assigning the minimum evaluation score “1” to the exhibitor who gave the maximum evaluation score “5” in the pattern A. To do.

  Then, for example, the model generation unit 131 calculates a high credit rating of the seller who has performed an action such as giving the highest evaluation score “5” to the successful bidder who has given the lowest evaluation score “1”. A learning model for generating a low credit rating of a successful bidder who has performed an action such as giving a minimum evaluation score of “1” to an exhibitor who has given “5” is generated.

  Also, in such a case, the determination unit 135 determines behavioral information such as how the credit target user has been rated in the auction so far (for example, an average value of evaluation points for the seller when the credit target user is a successful bidder). Alternatively, the creditworthiness of the creditworthiness target user is determined by applying the average value of evaluation points for the winning bidder when the creditworthiness target user is an exhibitor to the creditworthiness model generated as described above.

[4-4. Model generation (4)]
In addition, the model generation unit 131 may generate a trustworthiness model based on information regarding actions performed until a product is purchased. The information regarding the action performed until the product is purchased is, for example, the number of times the product has been searched before the product is purchased.

  In this case, for example, in the example of the figure, the behavior tendency storage unit 120 stores, for each user, the number of searches performed until a specific product is purchased and the user's credit loss rate in association with each other. . The specific product here is not limited in price, type, or the like, but is the same product for all users. Here, the model generation unit 131 calculates the average value of the number of searches for each user stored in the behavior tendency storage unit 120, and between the number of search times less than the calculated average value and the credit loss rate of the user. Machine learning of correlation in In addition, for example, the model generation unit 131 performs machine learning on the correlation between a user whose search count is greater than the calculated average value and the credit loss rate of the user.

  Here, it is assumed that the model generation unit 131 obtains a correlation such as “users with a smaller number of searches than the average value tend to have a high loan loss rate” by machine learning. On the other hand, it is assumed that the model generation unit 131 obtains a correlation such as “users with a greater number of searches than the average value tend to have a low credit loss rate” by machine learning. In such a case, the model generation unit 131 is a credit model for calculating a low credit rating in a financial transaction for a user with a search count less than the average value. Generate a trustworthiness model to calculate high trustworthiness.

  Further, in such a case, the determination unit 135 applies the number of searches related to the specific product performed until the credit target user purchases the specific product to the credit target by using the learning model generated as described above. Determine the user's trustworthiness.

[4-5. Model generation (5)]
Moreover, in the said embodiment, the model production | generation part 131 showed the example which produces | generates a reliability model using the information regarding the evaluation action with respect to purchased goods, the action which posts a comment, or the action performed until the goods were purchased. . However, the model generation unit 131 may generate a trustworthiness model using behavior information other than these.

  For example, the model generation unit 131 uses information regarding an evaluation behavior for a product seller, an evaluation behavior for a product purchaser, or behavior for canceling a transaction in product sales (cancellation rate at which the user canceled the transaction). A confidence model may be generated.

[4-6. Model generation (6)]
Moreover, in the said embodiment, the model production | generation part 131 may produce | generate a trustworthiness model by learning the correlation between the user's action information regarding resale, and trustworthiness. The user behavior information related to resale is, for example, the number of resales, the number of products resold at a time, resale balance such as how many products are purchased at a time and how much is resold at a time. As an example, the model generation unit 131 may obtain a correlation such that the higher the number of resales, the lower the credit loss rate, by machine learning. In such a case, for example, the model generation unit 131 generates a trustworthiness model that calculates a higher trustworthiness score as the number of resales increases.

  Moreover, the determination part 135 determines the creditworthiness of the user who is a creditworthiness judgment object by applying the user's action information regarding resale of the creditworthiness judgment object user to the creditworthiness model generated as described above.

  In addition, resale may be performed in units of corporations other than when performed by individual users. For example, a certain company makes a large number of successful bids at a predetermined auction site and sells the successful bid items at a predetermined shopping site. Therefore, the model generation unit 131 may generate a credit model dedicated to a corporation by learning a correlation between behavior information regarding resale of the corporation (company) and the credit quality of the corporation.

  In this case, for example, if there is a financial transaction request from a company, the determination unit 135 determines the creditworthiness of the requesting company by applying the behavior information related to the resale of the company to the creditworthiness model dedicated to the corporation.

[5. Hardware configuration)
In addition, the determination apparatus 100 according to the above-described embodiment is realized by a computer 1000 configured as shown in FIG. FIG. 5 is a hardware configuration diagram illustrating an example of a computer 1000 that realizes the function of the determination 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 communication network 50 and sends the data to the CPU 1100, and transmits the data generated by the CPU 1100 to other devices via the communication network 50.

  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. Etc.

  For example, when the computer 1000 functions as the determination apparatus 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 behavior tendency storage unit 120. The CPU 1100 of the computer 1000 reads these programs from the recording medium 1800 and executes them, but as another example, these programs may be acquired from other devices via the communication network 50.

[6. Others]
Of the processes described in the above embodiment, all or part of the processes described as being automatically performed can be performed manually, or all of the processes described as being performed manually 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.

  Moreover, each embodiment mentioned above can be combined suitably in the range which does not contradict a process content.

[7. effect〕
The determination apparatus 100 according to the embodiment includes an acquisition unit 133 and a determination unit 134. The acquisition unit 133 acquires behavior information of a predetermined user in electronic commerce. The determination unit 134 determines the reliability of a predetermined user whose behavior information has been acquired by the acquisition unit, based on the relationship between the behavior information of an arbitrary user in electronic commerce and the reliability of the arbitrary user.

  Thereby, the determination apparatus 100 concerning embodiment can determine a user's social credibility with high precision.

  In addition, the determination apparatus 100 according to the embodiment includes a model generation unit 131. The model generation unit 131 generates a learning model by machine learning of the correlation between the behavior information of an arbitrary user and the reliability of the arbitrary user. And the determination part 134 determines the reliability of a predetermined user by applying the action information acquired by the acquisition part 133 to a learning model.

  Thus, the determination apparatus 100 according to the embodiment increases the user's social credibility by using the learning model generated by machine learning of the correlation between the user behavior information and the user's credibility. The accuracy can be determined.

  In addition, the model generation unit 131 generates a learning model for each arbitrary user attribute information based on the arbitrary user behavior information and trustworthiness, and the determination unit 134 determines the learning model corresponding to the predetermined user attribute information. Is used to determine the trustworthiness of a given user.

  Thereby, the determination apparatus 100 according to the embodiment can determine the reliability more accurately.

  Moreover, the model production | generation part 131 produces | generates a learning model based on arbitrary user's action information and creditworthiness for every arbitrary user's health state. Further, the determination unit 134 determines the reliability of the predetermined user using a learning model corresponding to the health state of the predetermined user.

  Thereby, the determination apparatus 100 according to the embodiment can determine the reliability more accurately.

  Further, the acquisition unit 131 includes, as behavior information, an evaluation behavior for a product, an evaluation behavior for a product seller, an evaluation behavior for a product purchaser, a behavior for posting a comment, a behavior for canceling a transaction in product sales, The information regarding at least one action is acquired among each action performed until purchasing. And the determination part 134 determines the reliability of a predetermined user based on the relationship between the information regarding the action in at least one arbitrary user among each action, and the reliability of an arbitrary user.

  Thus, since the determination apparatus 100 concerning embodiment determines a user's trustworthiness based on the action information of various users in electronic commerce, it can determine a user's trustworthiness with high precision.

  The determination unit 134 also determines the reliability used when providing the financial service. Thereby, the determination apparatus 100 according to the embodiment can reduce the risk received from the user of the trading partner.

  The embodiments of the present application have been described in detail with reference to several drawings. However, these are merely examples, and various modifications, including the aspects described in the disclosure section of the invention, based on the knowledge of those skilled in the art, It is possible to implement the present invention in other forms with improvements.

  In addition, the “section (module, unit)” described above can be read as “means” or “circuit”. For example, the acquisition unit can be read as acquisition means or an acquisition circuit.

DESCRIPTION OF SYMBOLS 1 Determination system 10 Shopping server 100 Determination apparatus 120 Behavior tendency memory | storage part 131 Model production | generation part 132 Provision part 133 Acquisition part 134 Determination part

Claims (9)

An acquisition unit that acquires an evaluation score given to a predetermined target by the predetermined user as behavior information of the predetermined user in electronic commerce;
As the action information for any users in electronic commerce, the evaluation point where the arbitrary user has given to the predetermined target, based on the relationship between credit of the given user, action information acquired by the acquisition unit And a determination unit that determines the reliability of the predetermined user. A generator that generates a learning model by machine learning a correlation between the behavior information of the arbitrary user and the creditworthiness of the arbitrary user;
The determination device according to claim 1, wherein the determination unit determines the credit rating of the predetermined user by applying the behavior information acquired by the acquisition unit to the learning model. The generation unit generates the learning model based on the behavior information and trustworthiness of the arbitrary user for each attribute information of the arbitrary user,
The determination apparatus according to claim 2, wherein the determination unit determines the reliability of the predetermined user using a learning model corresponding to the attribute information of the predetermined user. The generation unit generates the learning model based on the behavior information and the credit rating of the arbitrary user for each arbitrary user's health state,
The determination apparatus according to claim 2, wherein the determination unit determines the reliability of the predetermined user using a learning model corresponding to the health state of the predetermined user. The acquisition unit, as the behavior information, evaluation point the predetermined users granted in pairs in the product, the predetermined evaluation points that the user has granted against the seller of the product, the predetermined user of the product determination apparatus according to any one of claims 1 to 4, wherein the obtaining at least one behavior information among the evaluation points were granted against the buyer. The acquisition unit includes at least one of the behaviors of the behavior of posting a comment that evaluates a predetermined target as the behavior information, the behavior of canceling the transaction in the merchandise sales, or the behavior performed until the merchandise is purchased Action information about one action,
The determination unit, as the behavior information of the arbitrary user, based on the relationship between the behavior information of the arbitrary user corresponding to the behavior information acquired by the acquisition unit and the reliability of the arbitrary user, The reliability of the predetermined user whose behavior information is acquired by the unit is determined
The determination apparatus according to claim 1, wherein: The said determination part determines the reliability used when providing a financial service. The determination apparatus as described in any one of Claims 1-6 characterized by the above-mentioned. A determination method executed by a determination device,
An acquisition step of acquiring an evaluation score given to a predetermined target by the predetermined user as behavior information of the predetermined user in electronic commerce;
As behavior information of an arbitrary user in electronic commerce, behavior information is acquired by the acquisition step based on a relationship between an evaluation score given by the arbitrary user to the predetermined target and a credit rating of the arbitrary user. And a determination unit that determines the reliability of the predetermined user. An acquisition procedure for acquiring an evaluation score given to a predetermined target by the predetermined user as behavior information of the predetermined user in electronic commerce;
As behavior information of an arbitrary user in electronic commerce, behavior information is acquired by the acquisition procedure based on a relationship between an evaluation score given by the arbitrary user to the predetermined target and a credit rating of the arbitrary user. A determination program that causes a computer to execute a determination unit that determines the reliability of the predetermined user.

Images