Information processing apparatus, information processing system, and information processing method

Abstract

To allow for maintaining the safety of a commercial transaction and efficiently selecting a transaction partner by solving the problem of insufficient safety in selecting the transaction partner and inefficiency in the transaction because confirming the reliability of an account participating in transactions or electronic information exchanges over a network depends on posted information of which the personal identification and reference are unclear.SOLUTION: An information processing apparatus comprises: a collection unit which collects account information and transaction histories of a plurality of accounts for performing transactions over a network; and an account evaluation unit which generates account evaluation values of the respective accounts on the basis of the relationships between the plurality of accounts in the transaction histories. An information processing method includes collecting account information and transaction histories of the plurality of accounts performing transactions over the network and generating account evaluation values of the respective accounts on the basis of the relationships between the plurality of accounts in the transaction histories.SELECTED DRAWING: Figure 3

Description

[Technical field]

[0001] The present disclosure relates to an information processing device, an information processing system, and an information processing method for generating account evaluation values ​​for multiple accounts that conduct transactions on a network. [Background technology]

[0002] Account KYC (Know Your Customer) is known as a means of verifying the trustworthiness of a trading partner in online commercial transactions. [Prior art documents] [Patent documents]

[0003] [Patent Document 1] US Patent Application Publication No. 2016 / 0071108 [Patent Document 2] Special Publication No. 2020-524855 [Patent Document 3] Special Publication No. 2019-517084 Summary of the Invention [Problem to be solved by the invention]

[0004] In online commerce, there are limited ways to verify the trustworthiness of trading partners, leading to many reports of crimes such as fraud. The current method of verification using account KYC (Know Your Customer) relies on personal certificates, which means that people from countries where personal certificates cannot be trusted are cut off from transactions. There are also issues with protecting personal information. Furthermore, ratings that rely solely on user posts allow for unfair transactions in exchange for high ratings, and arbitrary ratings by raters with no usage history, which leads to problems with unclear standards and low reliability.

[0005] Patent Document 1 monitors the exchange of cryptocurrency and legal currency in cryptocurrency transactions and strengthens the automatic settlement system to prevent money laundering. Patent Document 2 monitors the actions of privileged users such as administrator users and judges security risks. Patent Document 3 scores entities based on the trust level of users.

[0006] In order to confirm the reliability of accounts participating in online commercial transactions, it is desirable to assign a highly reliable evaluation value that does not depend on personal authentication or the like. [Means for solving the problem]

[0007] An information processing device according to one embodiment of the present disclosure includes a collection unit that collects account information and transaction history of multiple accounts conducting transactions on a network, and an account evaluation unit that generates an account evaluation value for each account based on the relationships between the multiple accounts in the transaction history. Effect of the Invention

[0008] According to the present disclosure, in order to confirm the reliability of accounts participating in commercial transactions on a network, it is possible to assign a highly reliable evaluation value that does not rely on personal authentication or the like.

[0009] Note that the effects described herein are not necessarily limited to those described herein, and may be any of the effects described in this disclosure. [Brief description of the drawings]

[0010] [Figure 1] 1 illustrates an information processing system according to an embodiment of the present disclosure. [Diagram 2] This shows the functional configuration of the information system. [Diagram 3] 3 shows an operation flow of the information processing device. [Figure 4] Indicates the account database. [Diagram 5] A transaction history database is shown. [Figure 6] 2 shows a configuration of an initial evaluation value generating unit. [Figure 7] 2 shows the configuration of a clustering unit. [Figure 8] The cluster database is shown. [Figure 9] The configuration of the propagation section is shown. [Figure 10] 1 is an example of graph data showing trading relationships between multiple accounts. [Figure 11] 1 shows the configuration of a grouping unit. [Figure 12] 1 shows the configuration of a group evaluation unit. [Figure 13] The group database is shown. [Figure 14] 1 shows the configuration of a role analysis unit. [Figure 15] The configuration of the time series analysis unit is shown below. [Figure 16] The structure of the time-series feature database is shown below. [Figure 17] 1 shows a schematic diagram of Example 1. [Figure 18] 1 shows a schematic diagram of Example 2. [Figure 19] 3 shows a schematic diagram of Example 3. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

[0011] Hereinafter, embodiments of the present disclosure will be described with reference to the drawings.

[0012] 1. Information Processing System

[0013] FIG. 1 illustrates an information processing system according to an embodiment of the present disclosure.

[0014] The information processing system 1 includes an information processing device 10, a storage device 20, and a web server 30, which are communicatively connected via a network or the like. The information processing device 10 functions as a back-end computing server and executes arithmetic processing. The storage device 20 is a large-capacity non-volatile storage device that functions as a server-side database server. The information processing device 10 and the storage device 20 may be independently connected via a network or the like (as shown), or may be constructed as an integrated information processing device (not shown). The web server 30 is a front-end web server that provides data generated by the information processing device 10 through arithmetic processing and stored in the storage device 20 in response to a request from a terminal device of a client, which is a third party.

[0015] 2. Functional configuration of the information processing system

[0016] FIG. 2 shows the functional configuration of the information system.

[0017] The information processing device 10 operates as a collection unit 110, an account evaluation unit 100, and a providing unit 190 by the processor of the control circuit loading various information processing programs recorded in the auxiliary storage device into the main storage device and executing them. The account evaluation unit 100 has an initial evaluation value generation unit 120, a clustering unit 130, a propagation unit 140, a grouping unit 150, a group evaluation unit 160, a role analysis unit 170, and a time series analysis unit 180.

[0018] The storage device 20 stores an account database 210 , a transaction history database 220 , a cluster database 230 , and a group database 240 .

[0019] 3. Operation of information processing device

[0020] FIG. 3 shows an operation flow of the information processing device.

[0021] The collection unit 110 periodically collects account information and transaction history of multiple accounts that conduct transactions on the network. In this specification, a "transaction" on the network is intended to have a broad meaning that includes not only a "commercial transaction" involving money, but also an "exchange" of electronic information on SNS and the like. The accounts may be personal accounts or corporate accounts. The account information includes various information such as criminal history, evaluation posting data, rumors on SNS and the Web, past evaluation values, tag data, and financial information (especially in the case of corporate accounts). The collection unit 110 accumulates the account information in the account database 210 and accumulates the transaction history in the transaction history database 220.

[0022] FIG. 4 shows the account database.

[0023] A plurality of account data 210A are registered in the account database 210. Each of the account data 210A includes an account serial ID 211, an account name 212, a criminal history 213, a tag 214, a role 215, an initial evaluation value 216, an account evaluation value 217, a cluster ID 218, and a group ID 219, which are registered in association with each other.

[0024] The account serial ID 211 may be a serial number that uniquely identifies one transaction history within the account database 210. The account name 212 may be an address that is included in the collected account information and is uniquely assigned to the account by a private or public community, such as a blockchain platform network for building distributed applications and smart contracts. The criminal history 213 may be included in the collected account information and expressed as a binary value of True or False. The tag 214 is an attribute that is previously assigned to the account, such as "Official" or "Bank," a negative attribute that is later assigned to the account, such as "Fraud," or a positive attribute that indicates that the account is conducting honest transactions.

[0025] In addition to the account name 212 and the criminal history 213, other information included in the collected account information (rating posting data, rumours on SNS or the web, past rating values, tag data, etc.) may be further registered in the account database 210. Role 215, initial rating value 216, account rating value 217, cluster ID 218, and group ID 219 will be explained later.

[0026] The account database 210 may be implemented in a form that holds an update history of the registered data. In this case, the update history is used by the time series analysis unit 180.

[0027] FIG. 5 shows the transaction history database.

[0028] A plurality of transaction history data 220A are registered in the transaction history database 220. Each transaction history data 220A is registered in such a manner that a transaction serial ID 221, a transaction source account name 222, a transaction destination account name 223, a transaction amount 224, and a transaction date and time 225 are associated with each other.

[0029] For one transaction, a transaction serial ID 221, a source account name 222, a destination account name 223, a transaction amount 224, and a transaction date and time 225 are registered in association with each other. One transaction history does not mean, for example, a series of transactions including multiple nodes from an initial seller to an ultimate purchaser, but means a transaction between one source account and one destination account (two nodes that directly trade).

[0030] The transaction serial ID 221 may be a serial number that uniquely identifies one transaction history within the transaction history database 220. The transaction source account name 222 and the transaction destination account name 223 are any of the account names 212 registered in the account database 210.

[0031] The account evaluation unit 100 judges whether the collection unit 110 has collected new account information and transaction history and whether the account database 210 and / or the transaction history database 220 have been updated (step S1A). If the account database 210 and / or the transaction history database 220 have been updated (step S1B, Yes), the account evaluation unit 100 proceeds to a process for updating the account database 210 (step S1C). On the other hand, if the account database 210 and / or the transaction history database 220 have not been updated (step S1B, No), the account evaluation unit 100 goes into a standby state until the collection unit 110 next periodically collects new account information and / or transaction history (step S1D).

[0032] In the process for the account evaluation unit 100 to update the account database 210, each module in the account evaluation unit 100 (initial evaluation value generation unit 120, clustering unit 130, propagation unit 140, grouping unit 150, group evaluation unit 160, and role analysis unit 170) executes processes (dashed line frames surrounding steps S2 to S7) independently and in parallel. The operation of each module in the account evaluation unit 100 will be described below.

[0033] FIG. 6 shows the configuration of the initial evaluation value generating unit.

[0034] As a preliminary step of step S2, the initial evaluation model 122 used by the initial evaluation value generation unit 120 is generated in advance by the machine learning unit 121, for example, by the following method. The machine learning unit 121 reads out a large number of pieces of account information AC on the network and a large number of pieces of account data 210A and transaction history data 220A registered in the account database 210 and the transaction history database 220 (step S21), and trains a machine learning model (step S22). The account data 210A in which at least the initial evaluation value 216, preferably the account evaluation value 217, is registered in the account database 210 can be used as training data. The machine learning unit 121 trains the machine learning model using the account information AC for a plurality of accounts as training data, and generates the initial evaluation model 122.

[0035] Assume that in step S1A (described above), the collection unit 110 collected account information ACx and updated the account database 210 and / or the transaction history database 220 based on the account information ACx.

[0036] In step S2, the initial evaluation value generation unit 120 reads the account data 210A of the newly registered account information ACx and the transaction history data 220A of the account from the account database 210 and the transaction history database 220 (step S23). The initial evaluation value generation unit 120 inputs the read account data 210A and transaction history data 220A to the initial evaluation model 122, and infers the initial evaluation value 216 of this account (step S24). The initial evaluation value generation unit 120 writes the initial evaluation value 216 output by the initial evaluation model 122 into the account database 210.

[0037] FIG. 7 shows the configuration of the clustering unit.

[0038] In step S3, the clustering unit 130 reads out a large number of account information AC on the network and a large number of account data 210A and transaction history data 220A registered in the account database 210 and the transaction history database 220, preprocesses and shapes the data, and inputs it to the clustering model (step S31). The clustering model determines the characteristics of each account from the input account data 210A and transaction history data 220A, etc., performs clustering analysis based on the account characteristics, and classifies multiple accounts into multiple clusters 1 to K for each account with similar characteristics (step S32). In other words, the clustering model classifies multiple accounts that behave similarly into one of clusters 1 to K by machine learning. The clustering unit 130 registers cluster IDs 218 indicating clusters 1 to K to which each account belongs in the account database 210 in association with the account.

[0039] Examples of account characteristics include the following characteristics (1) to (12). Characteristics (1) to (7) are examples when the transaction is a commercial transaction. Characteristics (8) to (12) are examples when the transaction is an electronic information exchange such as an SNS.

[0040] (1) Frequency of deposits and withdrawals (average interval between deposits / withdrawals (minutes)) (2) Total number of deposits and total number of payments (3) Total number of accounts that received funds and total number of accounts that withdrew funds (4) The minimum, maximum, and average deposit amounts, and the minimum, maximum, and average withdrawal amounts in the past (5) Total amount of deposits and withdrawals (6) Current amount of money (7) The length of time that your account has been active (8) Number of responses (9) Response frequency (average response interval (min) (10) Total number of accounts that responded (11) Response data volume (12) Number of followings and followers

[0041] The clustering unit 130 calculates a cluster evaluation value 233 for each of clusters 1 to K based on the characteristics of multiple account data 210A in which the same cluster ID 218 is registered (i.e., belonging to the same clusters 1 to K) by machine learning (step S33). The clustering unit 130 registers the cluster evaluation value 233 in the cluster database 230.

[0042] FIG. 8 shows the cluster database.

[0043] A plurality of cluster data 230A are registered in the cluster database 230. In each cluster data 230A, a cluster serial ID 231, a cluster ID 232, a cluster evaluation value 233, etc. are registered in association with each other.

[0044] The cluster serial ID 231 may be a serial number that uniquely identifies one cluster within the cluster database 230. The cluster ID 232 is a cluster name that uniquely identifies one cluster, and is the same as the cluster ID 218 in the account database 210. The cluster evaluation value 233 is a score calculated by the clustering unit .

[0045] As a result, a cluster ID 218 is registered for each account in the account database 210, and a cluster ID 232 that is the same as the cluster ID 218 is registered in the cluster database 230, and is associated with a cluster evaluation value 233. In this way, since the cluster database 230 is a sub-database of the account database 210, as a modified example, it may be in the form of a table integrated with the account database 210.

[0046] The clustering unit 130 updates the account evaluation value 217 of each account based on the cluster evaluation values ​​of the clusters 1 to K to which each account belongs. That is, the clustering unit 130 reads out the cluster evaluation value 233 associated with the cluster ID 232 that is the same as the cluster ID 218 associated with each account name 212 from the cluster database 230 in the account database 210, and updates the account evaluation value 217 by reflecting the cluster evaluation value 233 in the account evaluation value 217 through machine learning.

[0047] Fig. 9 shows the configuration of the propagation unit. Fig. 10 shows an example of graph data showing the transaction relationships of multiple accounts.

[0048] In step S4, the propagation unit 140 estimates the relationship between the multiple accounts having a trading relationship based on the transaction history, and propagates the account evaluation values ​​of the multiple accounts having the trading relationship to each other according to the strength of the relationship, thereby updating the account evaluation value of the account. That is, when updating the account evaluation value 217 of a specific account, the propagation unit 140 directly influences not only the personal factors of the account itself, but also the account evaluation values ​​217 of other accounts having a trading relationship.

[0049] Specifically, the propagation unit 140 reads out transaction history data 220A in which a certain account AC (account A in FIG. 10) is the source account name 222 or the destination account name 223 from the transaction history database 220. The propagation unit 140 extracts the source account name 222 or the destination account name 223 included in the read transaction history data 220A. In other words, the propagation unit 140 extracts all primary nearby accounts (accounts B, F, and G in FIG. 10) that have a direct (i.e., primary) transaction relationship with account A (step S41).

[0050] As shown in the relationship between accounts A and B, the number of directed edges connecting a pair of accounts A and B is not limited to one because there exists a concept of time and a relationship between deposits and withdrawals. That is, there may be multiple directed edges from account A to account B, and a directed edge from account A to account B and a directed edge from account B to account A may coexist. Also, for example, the deposit direction may be referred to as the +xth order neighborhood, and the withdrawal direction may be referred to as the -xth order neighborhood, etc., and may be distinguished. However, in this embodiment, both the deposit direction and the withdrawal direction are referred to as the xth order neighborhood, and no distinction is made.

[0051] The propagation unit 140 reads out the account data 210A of each of the extracted first-order nearby accounts B, F, and G from the account database 210 (step S42). The propagation unit 140 estimates the strength of the relationship between the account AC and each of the first-order nearby accounts B, F, and G from the criminal history 213 and tag 214 included in the read account data 210A and the transaction history data 220A (obtained in step S41) by, for example, machine learning.

[0052] The propagation unit 140 updates the account evaluation value 217 of the account AC registered in the account database 210 by propagating the account evaluation value 217 of the account AC to the account evaluation values ​​217 of the primary neighborhood accounts B, F, and G in accordance with the strength of the relationship between the account AC and each of the primary neighborhood accounts B, F, and G. Conversely, the propagation unit 140 updates the account evaluation values ​​217 of the accounts B, F, and G registered in the account database 210 by propagating the account evaluation values ​​217 of the primary neighborhood accounts B, F, and G to the account evaluation value 217 of the account AC in accordance with the strength of the relationship between the account AC and each of the primary neighborhood accounts B, F, and G (step S43).

[0053] In this way, the propagation unit 140 estimates the strength of the relationship (depth of the relationship) from the transaction history and determines the degree to which the evaluation value is reflected. An account that has a direct transaction relationship with an account with a high account evaluation value 217 will receive a higher account evaluation value 217. Conversely, an account that has a direct transaction relationship with an account with a low account evaluation value 217 will receive a low account evaluation value 217. In this way, the trust between accounts that have a transaction relationship, i.e., the account evaluation value 217, is propagated.

[0054] FIG. 11 shows the configuration of the grouping unit.

[0055] In step S5, the grouping unit 150 generates groups including multiple accounts with trading relationships based on the transaction history, and associates each account with a group ID indicating the group to which the account belongs. For example, the grouping unit 150 performs a link analysis between the accounts, analyzes the strength of the connections between the accounts, and generates a group to which each account belongs.

[0056] Specifically, the grouping unit 150 reads out transaction history data 220A in which a certain account AC (account A in FIG. 10) is the source account name 222 or the destination account name 223 from the transaction history database 220. The grouping unit 150 extracts the source account name 222 or the destination account name 223 included in the read transaction history data 220A. In other words, the grouping unit 150 extracts all primary neighboring accounts (accounts B, F, and G in FIG. 10) that have a direct (i.e., primary) transaction relationship with account A (step S51). The grouping unit 150 repeats the extraction of the x-th neighboring account for account A a predetermined number of times (D times) (step S52, No). For example, in the second search, the grouping unit 150 extracts the secondary neighboring account C for account A via the primary neighboring account B.

[0057] The grouping unit 150 determines that the search stop condition is met when the extraction of the xth order neighboring account for account A is repeated a predetermined number of times (D times), or when the extraction of the xth order neighboring account for account A is repeated less than a predetermined number of times (D times) (step S52, Yes). Examples of the search stop condition include a weak association to be considered as a neighbor, no more unsearched neighbors (all neighbors have been searched), and the like. "A weak association to be considered as a neighbor" means, for example, that account C in FIG. 10 is extracted as a second order neighboring account of account A, but the quality and strength of the association with account A are relatively low (weak) compared to other xth order neighboring accounts with account A. In this case, account C is separated from group 1 to which account A belongs, depending on the quality and strength of the association with account A.

[0058] The grouping unit 150 generates a group Y including the account A and all the first to Dth neighboring accounts for the account A. The grouping unit 150 writes a group ID 219 indicating the group Y to the account data 210A of all the accounts in the account database 210.

[0059] FIG. 12 shows the configuration of the group evaluation unit.

[0060] In step S6, the group evaluation unit 160 generates a group evaluation value for the group based on the account evaluation values ​​217 of the multiple accounts in the group. Specifically, the group evaluation unit 160 extracts the account data 210A of all accounts 1 to N in which a group ID 219 indicating a specific group X is registered from the account database 210 (step S61). The group evaluation unit 160 estimates and integrates the account evaluation values ​​217 and tags 214 of the account data 210A of all accounts 1 to N, and the transaction history data 220A including the account names 212 of the accounts 1 to N registered in the transaction history database 220, for example, by machine learning, and generates a group evaluation value 243 for the group X to which the accounts 1 to N belong (step S62). The group evaluation unit 160 registers the group evaluation value 243 in the group database 240.

[0061] FIG. 13 shows the group database.

[0062] A plurality of group data 240A are registered in the group database 240. In each group data 240A, a group serial ID 241, a group ID 242, a group evaluation value 243, etc. are registered in association with each other.

[0063] The group serial ID 241 may be a serial number that uniquely identifies one group within the group database 240. The group ID 242 is a group name that uniquely identifies one group, and is the same as the group ID 219 in the account database 210. The group evaluation value 243 is a score calculated by the group evaluation unit 160.

[0064] As a result, a group ID 219 is registered for each account in the account database 210, and a group ID 242 that is the same as the group ID 219 is registered in the group database 240, and is associated with a group evaluation value 243. In this way, since the group database 240 is a sub-database of the account database 210, as a modified example, it may be in the form of a table integrated with the account database 210.

[0065] The group evaluation unit 160 updates the account evaluation value 217 of each account based on the group evaluation value of the group X to which each of the accounts 1 to N belongs. That is, the group evaluation unit 160 reads out the group evaluation value 243 associated with the same group ID 242 as the group ID 219 associated with each account name 212 from the group database 240 in the account database 210, and updates the account evaluation value 217 by reflecting the group evaluation value 243 by, for example, machine learning.

[0066] In this way, the group evaluation unit 160 integrates the account evaluation values ​​of the accounts belonging to each group to calculate the group evaluation value 243. This allows the reliability and evaluation of the entire group (the account evaluation values ​​of all accounts included in the group to which the account belongs) to be reflected in the account evaluation value of each individual account. In other words, when updating the account evaluation value 217 of a specific account, the group evaluation unit 160 directly influences the group evaluation value 243, which depends not only on the personal elements of the account itself but also on the account evaluation values ​​217 of other accounts belonging to the same group.

[0067] FIG. 14 shows the configuration of the role analysis unit.

[0068] In step S7, the role analysis unit 170 analyzes the role of each of the multiple accounts in the group, associates role information indicating the role of the account with each account, and updates the account evaluation value of the account based on the role of each account. Specifically, the role analysis unit 170 extracts the account data 210A of all accounts 1 to N in which a group ID 219 indicating a specific group X is registered from the account database 210 (step S71). The role analysis unit 170 analyzes the role of each of the multiple accounts 1 to N based on the account evaluation value 217 and tag 214 of the account data 210A of all accounts 1 to N and the transaction history data 220A including the account names 212 of the accounts 1 to N registered in the transaction history database 220 (in other words, classifies the multiple accounts 1 to N into multiple roles). The role analysis unit 170 associates the account data 210A of each account 1 to N with the role 215 indicating the role of the account 1 to N, and registers it in the account database 210 (step S72).

[0069] The role analysis unit 170 updates the account evaluation value 217 of each account based on the role 215 of each account 1 to N. That is, when updating the account evaluation value 217 of a particular account, the role analysis unit 170 directly influences not only the personal factors of the account itself, but also the role 215 of the account itself, which becomes clear only from the relationship with other accounts belonging to the same group.

[0070] Furthermore, the role analysis unit 170 may update the tag 214 of each account based on the role 215 of each account 1 to N. In this way, the role analysis unit 170 analyzes the role of the account within each group and associates it with the tag 214. This makes it possible to clarify the characteristics and roles of the accounts within the group and reflect them in the tag 214.

[0071] The role analysis unit 170 judges an account having a specific account evaluation value in a group having a specific group evaluation value as an actualized positive / negative account having a positive or negative role. The role analysis unit 170 judges an account that has a direct or indirect trading relationship with multiple positive / negative accounts as a hub account having a hub role. The role analysis unit 170 judges the role of a new account that has the same trading relationship as any existing account's trading relationship with the hub account and trades directly or indirectly with the hub account as a potential positive / negative account.

[0072] For example, the role analysis unit 170 selects a group having a specific group evaluation value 243 (lower than a predetermined threshold) as group X. The role analysis unit 170 may determine that an account having a specific account evaluation value 217 (lower than a predetermined threshold) in group X is an exposed negative account having a role 215 (negative role) of "fraud". The role analysis unit 170 may determine that an account that has a direct or indirect business relationship with multiple negative accounts is a hub account having a role 215 of hub. Furthermore, the role analysis unit 170 may also determine that a hub account is an exposed negative account.

[0073] Furthermore, when a new account appears, the role analysis unit 170 can estimate the potential role of the new account. For example, the role analysis unit 170 may determine the role of a new account that has the same trading relationship with the hub account as that of any existing manifested negative account and trades directly or indirectly with the hub account as a potential negative account. The role analysis unit 170 may update the account evaluation value 217 of the hub account based on the account evaluation value 217 of the manifested negative account. The role analysis unit 170 may update the account evaluation value 217 of a new potential negative account based on the account evaluation values ​​217 of multiple existing manifested negative accounts.

[0074] The above describes an example of the behavior of a negative account with a negative role among positive / negative accounts. On the other hand, the same explanation can be given for a positive account with a positive role, such as conducting honest transactions.

[0075] For example, the role analysis unit 170 selects a group having a specific (higher than a predetermined threshold) group evaluation value 243 as group X. The role analysis unit 170 may determine that multiple accounts in group X having a specific (higher than a predetermined threshold) account evaluation value 217 are manifested positive accounts having a positive role 215. The role analysis unit 170 may determine that an account that has a direct or indirect trading relationship with multiple positive accounts is a hub account having a hub role 215. Furthermore, the role analysis unit 170 may also determine that the hub account is also a manifested positive account having a positive role 215, such as conducting honest transactions.

[0076] Furthermore, for example, the role analysis unit 170 may determine the role of a new account that has the same trading relationship with the hub account as that of any existing manifested positive account and trades directly or indirectly with the hub account as a potential positive account. The role analysis unit 170 may update the account evaluation value 217 of the hub account based on the account evaluation value 217 of the manifested positive account. The role analysis unit 170 may update the account evaluation value 217 of a new potential positive account based on the account evaluation values ​​217 of multiple existing manifested positive accounts.

[0077] FIG. 15 shows the configuration of the time series analysis unit.

[0078] In step S8, the time series analysis unit 180 reads out the account data 210A of the account information ACx of each account from the account database 210 (step S81). The time series analysis unit 180 performs time series analysis using, for example, machine learning based on the update history of data such as the criminal history 213, the tag 214, the role 215, the initial evaluation value 216, the account evaluation value 217, the cluster ID 218, and the group ID 219 registered in the account database 210 (step S82), and calculates the time series feature 253 for each account. That is, the time series analysis unit 180 analyzes the change history of the account evaluation value 217 and the information associated with the account from the past to the present in a time series manner. The time series analysis unit 180 registers it in the time series feature database 250. The time series feature 253 may be, for example, a vector (a sequence of real numbers), a real number (score), a matrix (real numbers in the form of n×m), or the like.

[0079] FIG. 16 shows the time-series feature database.

[0080] A plurality of time-series feature data 250A are registered in the time-series feature database 250. In each time-series feature data 250A, a time-series feature serial ID 251, an account serial ID 252, a time-series feature 253, etc. are registered in association with each other.

[0081] The time-series feature serial ID 251 may be a serial number that uniquely identifies one account within the time-series feature database 250. The account serial ID 252 is the same as the account serial ID 211 in the account database 210. The time-series feature 253 is a feature calculated by the time-series analysis unit 180.

[0082] As a result, an account serial ID 211 is registered for each account in the account database 210, and an account serial ID 252 that is the same as the account serial ID 211 is registered in the time-series feature database 250, and is associated with a time-series feature 253. In this way, since the time-series feature database 250 is a sub-database of the account database 210, as a modified example, it may be in the form of a table integrated with the account database 210.

[0083] The providing unit 190 provides all or a part of the account information 210A registered in the account database 210 (at least the account evaluation value 217) to a third party via the web server 30 in response to a request from a terminal device of a client which is a third party (FIG. 1). The providing unit 190 can provide more accurate account information 210A as a result of updating and complementing the initial input information based on the results of the characteristic clustering analysis, the propagation of the evaluation value, the account linkage analysis, the group evaluation, the account role analysis, and the tag data. The providing unit 190 may further provide the group data 240A (particularly, the group evaluation value 243) of the group database 240 to the third party via the web server 30.

[0084] 4. Working Example

[0085] Below, several examples will be described as use cases of the above embodiment.

[0086] FIG. 17 shows a schematic diagram of the first embodiment.

[0087] Example 1 is an example in which the technology of this embodiment is used for crime prevention based on the analysis results.

[0088] Accounts A, B, C, and D are determined to belong to the same group based on an analysis of deposits and withdrawals. Within the group, account A (A1, A2) is a "fraudulent" account that is clearly involved in a crime, and is an obvious negative account with a negative role. Because this group includes negative account A, the group evaluation value 243 is influenced by the account evaluation value 217 of account A, which is an obvious negative account, and is a low value. Because group evaluation value 243 is lower than a predetermined threshold, the group is determined to be a "fraudulent" group with a negative role. Accounts B, C, and D that belong to this "fraudulent" group can be presumed to be related to the fraudulent account A, and can be determined to be highly likely to be accounts with a negative role, just like account A.

[0089] On the other hand, it is determined that Account 0, Official, does not belong to this group. By combining the analyses, the following tags 214 and / or roles 215 are assigned to each account.

[0090] Within the group, Account A is a "fraudulent" account that is clearly involved in a crime and is a negative account that has become apparent. Account B is an account whose role is simply to pass on deposited amounts to other accounts. Account C is a hub account that receives deposits from multiple accounts A and B and has the function of withdrawing funds to multiple accounts A3 and D1. Account D is an account that acts as a contact point for accounts at financial institutions.

[0091] Account 0 is a normal user account and is the account of the "victim." Account Official is an account that is publicly announced as an official service (financial institution, securities account, etc.).

[0092] It is possible to identify individuals by identifying Account D, which is responsible for collecting money from Accounts C and other accounts in the "fraud" group and withdrawing it to Account Official (such as a securities account).

[0093] By identifying a hub account C1 that receives withdrawals from each negative account A1, A2 that is deemed to be "fraudulent," it is possible to identify a new account A3 (a potential negative account) that is deemed to be potentially "fraudulent" and that receives withdrawals from hub account C1.

[0094] In this way, by estimating the newly created account A3 as a potential "fraud" account (potential negative account) and as belonging to the "fraud" group, new criminal acts can be prevented. By identifying the Official financial account that receives transactions directly from Account D, which serves as the window for accounts at financial institutions, it is possible to increase the possibility of recovering funds stolen during crimes and prevent criminal acts.

[0095] FIG. 18 illustrates the second embodiment.

[0096] Example 2 is an example in which the technology of this embodiment is applied to business flow analysis and service provision.

[0097] Account A is a manufacturer. Account B is a wholesaler. Account C is a primary distributor. Account D is a general user. Account E is a secondary (second-hand) distributor.

[0098] For example, when analyzing the incoming and outgoing data for authenticity verification of brand-name products, the distribution flow is analyzed and each role is extracted. This allows the manufacturing company A to understand the situation of the secondhand dealer E, whose sales situation is difficult to grasp from the manufacturing company A's point of view. It also becomes easier to understand the route through which general users D purchase products, making it easier to provide services. For example, among general users D, it is possible to analyze user D1 who resells new products to secondary (secondhand) dealer E1 after purchasing them for the first time, user D3 who holds new products that he purchased first for a long time, and user D6 who only purchases secondhand products from secondary (secondhand) dealer E1. It is also possible to distinguish these users D1, D3, and D6 as having different characteristics by performing clustering analysis based on characteristics (step S3).

[0099] FIG. 19 illustrates the third embodiment.

[0100] Example 3 is an example in which the technology of this embodiment is used for selecting business partners and providing services through account analysis.

[0101] Account A1 has transaction records with multiple accounts (not in the same group). Account A2 has transaction records only with accounts in the same group. Account A3 has no transaction records.

[0102] For example, in an auction service, multiple accounts in the same group may repeatedly bid and drive up product prices (wash trading, leverage, etc.). For example, in the case of posted evaluations from users participating in an auction, if accounts A1, A2, and A3 have the same evaluation value, it is difficult to distinguish between accounts A1, A2, and A3.

[0103] On the other hand, by analyzing accounts A1, A2, and A3 based on this embodiment, it becomes possible to determine that multiple accounts that repeatedly make bids belong to the same group. Differences arise in the evaluation values ​​of account A2, which only trades within the group, account A3, which has no trading history, and account A1, which does not belong to a group but has trading history. This makes it possible to hold a more fair auction. Furthermore, when providing services such as an auction service, it is possible to select an appropriate trading partner from among accounts A1, A2, and A3.

[0104] 5. Conclusion

[0105] In transactions such as commercial transactions and electronic information exchanges on the Internet, there is a lack of means to confirm the reliability of trading partners, and many frauds and crimes have been reported. In contrast, the present embodiment provides a more reliable method that does not rely on limited information such as personal certificates and corporate financial information. Specifically, by combining methods such as collection of account information, utilization of machine learning, clustering analysis, propagation of evaluation values, chain analysis, grouping and role analysis, it is possible to more accurately evaluate the reliability of trading partners by evaluating the roles in the commercial flow and the relationships between each account based on transaction information with accounts that have a trading relationship, which is separated from personal information. This embodiment enables many people to safely participate in the commercial flow and is useful for efficient operation of the commercial flow.

[0106] Specifically, this embodiment provides the following advantageous effects.

[0107] It is possible to calculate an individual's account evaluation value without relying on personal information such as account KYC (Know Your Customer). In order to confirm the reliability of accounts participating in online commercial transactions and electronic information exchange transactions, it is possible to assign a highly reliable evaluation value without relying on limited information such as personal certificates, corporate financial information, and information and ratings posted on restaurant rating sites and price rating sites with unclear standards. This maintains the safety of transactions and makes it possible to efficiently select trading partners.

[0108] By inferring relationships between accounts, it is possible to identify accounts that are effectively operating as one unit.

[0109] Even if the accounts are registered as owned by different people through KYC, if they are determined to be closely related, they will be considered to be in the same group and will be strongly influenced by each other's evaluation values. This can prevent collusion between multiple accounts.

[0110] If there is a relationship (transaction), it will be reflected in the evaluation value. Also, transactions within a group are not affected, so transactions such as match pumps (self-made transactions) by multiple accounts in collusion are excluded.

[0111] Because it does not rely on KYC, people from countries or regions where personal identification is underdeveloped or credibility is low can also become suppliers and buyers of services. It is possible to create a system where an account's evaluation value within the community will increase as long as you continue to make honest transactions regardless of the country or region you belong to.

[0112] By combining the settlement function based on the account evaluation value, a cooling-off mechanism can be created. A mechanism can be created that calculates the risk value of a transaction based on the evaluation value, and matches the amount to be returned in the event of a transaction not being concluded with the person who will receive interest income by substituting that amount. This ensures that each party has an incentive to act, ensuring the safety of the transaction.

[0113] By using this service based on the transaction history of the public chain and applying the evaluation value of this service to the private chain, the safety and efficiency of transactions within the private chain can be improved.

[0114] A decentralized internet (known as Web 3.0) is being explored to replace the centralized internet (known as Web 2.0). In Web 3.0, data ownership will be decentralized, allowing individuals who are the original owners of their data to manage their own data. In peer-to-peer transactions, which are a representative example of this, the functions previously held by the distribution industry will be broken down into payment, supply and demand matching, and reliability. Of these, this invention can ensure reliability. This invention can provide an environment in peer-to-peer transactions where even first-time traders can trade with peace of mind.

[0115] Although the embodiments and modified examples of the present technology have been described above, the present technology is not limited to the above-described embodiments, and it goes without saying that various modifications can be made within the scope of the gist of the present technology. [Explanation of symbols]

[0116] 1. Information Processing Systems 10. Information processing device 20 Storage device 30 Web Server 100 Account Evaluation Department 110 Collection Department 120 Initial evaluation value generation unit 121 Machine Learning Department 122 Initial Evaluation Model 130 Clustering Department 140 Propagation Department 150 Grouping Section 160 Group Evaluation Department 170 Role Analysis Department 180 Time Series Analysis Section 190 Provision Department 210 Account Database 220 Transaction History Database 230 Cluster Database 240 Group Database 250 Time Series Feature Database

Claims

1. A collection unit that collects account information and transaction history of a plurality of accounts that conduct transactions on the network; an account evaluation unit that generates an account evaluation value for each account based on a relationship between the multiple accounts in the transaction history; An information processing device comprising:

2. 2. The information processing device according to claim 1, The account evaluation unit is Inferring relationships between multiple accounts that have a trading relationship based on the transaction history; The account evaluation values ​​of the multiple accounts having a trading relationship are mutually propagated according to the strength of the relationship, and the account evaluation value of the relevant account is updated. Has a propagation part Information processing device.

3. 2. The information processing device according to claim 1, The account evaluation unit is generating a group including a plurality of accounts having a trading relationship based on the transaction history; A group ID indicating the group to which the account belongs is associated with each account. Has a grouping section Information processing device.

4. 4. The information processing device according to claim 3, The account evaluation unit is A group evaluation value of the group is generated based on the account evaluation values ​​of the multiple accounts in the group. Has a group evaluation department Information processing device.

5. 4. The information processing device according to claim 3, The account evaluation unit is Analyzing the role of each of the plurality of accounts within the group; Role information indicating the role of each account is associated with the account. Has a role analysis section Information processing device.

6. 6. The information processing device according to claim 5, The role analysis unit updates the account evaluation value of each account based on the role of the account. Information processing device.

7. 7. The information processing device according to claim 5, The role analysis unit is In a group with a specific group evaluation value, an account with a specific account evaluation value is determined to be a positive / negative account having a positive or negative role, An account that has direct or indirect trading relationships with multiple positive / negative accounts is deemed to be a hub account that has the role of a hub. Determine the role of a new account that has the same trading relationship with the hub account as any existing positive / negative account and that trades directly or indirectly with the hub account as a potential positive / negative account. Information processing device.

8. The information processing device according to claim 7, The role analysis unit updates the account evaluation value of the hub account based on the account evaluation values ​​of the positive / negative accounts with which the hub account has a direct or indirect business relationship. Information processing device.

9. The information processing device according to claim 7, The role analysis unit updates the account evaluation value of the potential positive / negative account based on the account evaluation values ​​of the existing positive / negative accounts and the hub account. Information processing device.

10. 2. The information processing device according to claim 1, The account evaluation unit is classifying any accounts into particular clusters based on the transaction history and characteristics of the plurality of accounts; A cluster ID indicating the cluster to which each account belongs is associated with the account. Has a clustering section Information processing device.

11. The information processing device according to claim 10, The clustering unit includes: Calculating a cluster evaluation value for each cluster based on characteristics of the plurality of accounts included in the cluster; Update the account rating of each account based on the cluster rating of the cluster to which the account belongs. Information processing device.

12. 2. The information processing device according to claim 1, a providing section for providing at least the account evaluation values ​​of the plurality of accounts to a third party; The information processing device further comprises:

13. 2. The information processing device according to claim 1, The account evaluation unit is A time series analysis unit that analyzes the change history of the account evaluation value and the information associated with the account from the past to the present in a time series manner. Information processing device.

14. A storage device; a collection unit that collects account information and transaction history of a plurality of accounts that conduct transactions on the network and stores the collected information in the storage device; an account evaluation unit that generates an account evaluation value for each account based on the relationship between the multiple accounts in the transaction history and stores the account evaluation value in the storage device; An information processing device having An information processing system comprising:

15. Collect account information and transaction history of multiple accounts conducting transactions on the network, Generate an account evaluation value for each account based on the relationships between the multiple accounts in the transaction history. Information processing methods.

Images