Customer analysis device, customer analysis system, and customer analysis method

Abstract

To provide a technique for extracting a group related to a suspicious transaction by using fund movement information.SOLUTION: The present invention is directed to a customer analysis apparatus by way of an aspect of the present disclosure. The customer analysis apparatus has an input unit for acquiring an analysis request indicating a target customer, a related customer group extracting unit for extracting a related customer group having relation with the target customer in fund movement information on fund movement between customers, a center customer specifying unit for specifying a center customer in the related customer group, and an output unit for outputting an analysis result indicating the related customer group and the center customer.SELECTED DRAWING: Figure 2

Description

[Technical Field]

[0001] The present disclosure relates to a customer analysis device, a customer analysis system, and a customer analysis method. [Background technology]

[0002] In recent years, advanced information processing technologies such as AI (Artificial Intelligence) have been actively introduced into the financial industry, spurring the evolution of financial technology, often referred to as FinTech.

[0003] For example, FinTech not only makes it possible to provide various financial services to customers via the Internet, but also enables financial institutions to obtain desired analytical results from transaction information, market information, etc. [Prior art documents] [Patent documents]

[0004] [Patent Document 1] Special Publication No. 2022-508106 [Patent Document 2] Special Publication No. 2022-518370 Summary of the Invention [Problem to be solved by the invention]

[0005] In financial crimes such as fraud, and organized crimes such as underground banking transactions involving cryptocurrencies, fraudulent acts tend to be committed by groups rather than by individual customers. For this reason, there is a need for technology that can use fund transfer information to identify groups involved in "suspicious transactions" such as criminal activities.

[0006] One objective of the present disclosure is to provide a technique for extracting groups related to suspicious transactions using fund transfer information. [Means for solving the problem]

[0007] One aspect of the present disclosure relates to a customer analysis device having an input unit that acquires an analysis request indicating a target customer, a related customer group extraction unit that extracts a related customer group that is related to the target customer in fund transfer information between customers, a core customer identification unit that identifies a core customer in the related customer group, and an output unit that outputs analysis results that indicate the related customer group and the core customer. [Effects of the Invention]

[0008] According to the present disclosure, funds transfer information can be used to extract clusters related to suspicious transactions. [Brief explanation of the drawings]

[0009] [Figure 1] FIG. 1 is a schematic diagram illustrating a customer analysis system according to an embodiment of the present disclosure. [Figure 2] FIG. 10 is a conceptual diagram illustrating an analysis result of fund transfer information according to an embodiment of the present disclosure. [Figure 3] FIG. 2 is a block diagram showing a hardware configuration of a customer analysis device according to an embodiment of the present disclosure. [Figure 4] FIG. 2 is a block diagram illustrating a functional configuration of a customer analysis device according to an embodiment of the present disclosure. [Figure 5] FIG. 10 is a diagram illustrating an example of a customer analysis process according to an embodiment of the present disclosure. [Figure 6] FIG. 10 is a diagram illustrating an example of a customer analysis process according to an embodiment of the present disclosure. [Figure 7] FIG. 10 is a diagram illustrating an example of a customer analysis process according to an embodiment of the present disclosure. [Figure 8] FIG. 10 is a diagram illustrating an example of a customer analysis process according to an embodiment of the present disclosure. [Figure 9] 10 is a flowchart illustrating a customer analysis process according to an embodiment of the present disclosure. DETAILED DESCRIPTION OF THE INVENTION

[0010] Hereinafter, embodiments of the present disclosure will be described with reference to the drawings.

[0011] In the following embodiment, a customer analysis system is disclosed that extracts a group of related customers having a relationship with a given target customer and a central customer in the related customer group from fund transfer information.

[0012] [Summary] 1, in a customer analysis system 10 according to an embodiment of the present disclosure, upon receiving an analysis request indicating a target customer from a terminal 60, the customer analysis device 100 acquires inter-customer fund transfer information indicating fund transfers such as deposits and withdrawals between customers from a database device 50, and extracts a group of related customers who are related to the target customer (for example, related based on the amount of deposits and withdrawals). Furthermore, the customer analysis device 100 identifies a central customer who is presumed to play a central role or a leader in the extracted group of related customers, and provides the terminal 60 with an analysis result indicating the group of related customers and the central customer.

[0013] Specifically, as shown in FIG. 2A, the inter-customer fund transfer information may be graph information expressed in a graph format defined by nodes representing customers (e.g., CIF (Customer Information File), etc.) and edges indicating the amounts of deposits and withdrawals between customers (e.g., remittance amounts, net deposits and withdrawals, etc.). Inter-customer fund transfer information may typically be composed of transaction histories between customers stored in financial institutions. When a target customer is specified by the terminal 60, the customer analysis device 100 extracts a group of related customers who have a relationship with the target customer in accordance with predetermined extraction conditions. The relationship between customers may be quantified, for example, by the size of the amount of remittances between customers within the aggregation period, the frequency of transactions, etc. Furthermore, the predetermined extraction conditions include, but are not limited to, the following: i) The relevant customer group includes customer information of the target customer; ii) The transaction volume within the relevant customer group is large; and iii) The transaction amounts between related and unrelated customer groups are small; For example, any suitable known clustering method, such as local clustering, may be used to determine a group of related customers that meet these extraction conditions.

[0014] Furthermore, the customer analysis device 100 may calculate a centrality index for each customer in the extracted related customer group, associate the calculated centrality index with each node, and output graph information. For example, the centrality index may be any appropriate known index representing centrality, such as PageRank or Personalized PageRank. For information on local clustering and Personalized PageRank, see, for example, "Local Graph Partitioning using PageRank Vectors" by Reid Andersen et al. (https: / / ieeexplore.ieee.org / document / 4031383).

[0015] After acquiring the related customer group and the centrality index in this manner, the customer analysis device 100 may provide the terminal 60 with inter-customer fund transfer information in which the related customer group and the centrality index are associated, as shown in Fig. 2B. This enables the user of the terminal 60 to check the predicted related customer group and central customer for the specified target customer.

[0016] Here, the customer analysis device 100 is realized by a computing device such as a server or a personal computer, and may have a hardware configuration such as that shown in Fig. 3. That is, the customer analysis device 100 has a storage device 101, a processor 102, a user interface (UI) device 103, and a communication device 104, which are interconnected via a bus B.

[0017] The programs or instructions that realize the various functions and processes described below in the customer analysis device 100 may be downloaded from any external device via a network, or may be provided from a removable storage medium such as a CD-ROM (Compact Disk-Read Only Memory) or flash memory.

[0018] Storage device 101 may be implemented by random access memory, flash memory, a hard disk drive, or the like, and stores installed programs or instructions as well as files, data, etc. used in executing the programs or instructions. Storage device 101 may also include a non-transitory storage medium.

[0019] The processor 102 may be realized by one or more CPUs (Central Processing Units), GPUs (Graphics Processing Units), processing circuitry, etc., which may be composed of one or more processor cores, and performs various functions and processes of the customer analysis device 100 described below in accordance with programs, instructions, data such as parameters required to execute the programs or instructions, etc. stored in the memory device 101.

[0020] The interface device 103 realizes an interface between the database device 50 and the terminal 60 and the customer analysis device 100. For example, a user operates a GUI (Graphical User Interface) displayed on the display or touch panel of the terminal 60 using a keyboard, mouse, etc., to send and receive various information, data, instructions, etc. to and from the customer analysis device 100 via the interface device 103.

[0021] The communication device 104 is realized by various communication circuits that execute communication processes with external devices, the Internet, a communication network such as a LAN (Local Area Network), and the like.

[0022] However, the above-described hardware configuration is merely an example, and the customer analysis device 100 according to the present disclosure may be realized by any other appropriate hardware configuration.

[0023] [Customer analysis device] Next, a customer analysis device 100 according to an embodiment of the present disclosure will be described with reference to Figures 4 to 8. Figure 4 is a block diagram showing the functional configuration of the customer analysis device 100 according to an embodiment of the present disclosure.

[0024] As shown in FIG. 4, the customer analysis device 100 includes an input unit 110, a related customer group extraction unit 120, a core customer identification unit 130, and an output unit 140.

[0025] The input unit 110 acquires an analysis request indicating a target customer. Specifically, the input unit 110 receives an analysis request including customer information (e.g., CIF) indicating the target customer from the terminal 60, and passes the received customer information to the related customer group extraction unit 120. In detecting a criminal group such as organized crime, the target customer may be one of the members of the criminal group.

[0026] The related customer group extraction unit 120 extracts related customer groups that are related to the target customer from the fund transfer information between customers. The fund transfer information may typically represent various fund transfers between customers during a specific period, such as deposit and withdrawal transactions within a specific financial institution, deposit and withdrawal transactions between multiple financial institutions, and / or deposit and withdrawal transactions between a specific financial institution and a foreign financial institution. Such fund transfer information may be represented in the form of a weighted directed graph defined by nodes representing customers and weighted directed edges representing transaction amounts between customers.

[0027] For example, fund transfer information can be represented as a weighted directed graph as shown in FIG. 5. In the illustrated example, customers are represented by node A, node B, ..., node N, and transaction amounts between customers are represented by weighted directed edges. For example, a transaction between customers A and B indicates that customer A transferred 10 million yen to customer B, and a transaction between customers A and C indicates that customer C transferred 20 million yen to customer A. Note that the weight associated with an edge is not limited to the transaction amount, and may be the frequency of transactions, the number of transactions, etc.

[0028] For example, if customer A is designated as the target customer, the related customer group extraction unit 120 extracts the following: i) The relevant customer group includes customer information of the target customer; ii) The transaction volume within the relevant customer group is large; and iii) The transaction amounts between related and unrelated customer groups are small; A group of customers who satisfy the three search conditions i) to iii) above may be extracted.

[0029] Specifically, it has been mathematically proven that the related customer group extraction unit 120 can extract an approximate node group that satisfies search conditions i) to iii) according to a known local clustering method (e.g., https: / / ieeexplore.ieee.org / document / 4031383). The related customer group extraction unit 120 first calculates a price-weighted personalized PageRank for each node according to the local clustering method. Note that the direction of directed edges is ignored here. Personalized PageRank is a derivative of PageRank, and compared to PageRank, it is biased so that the centrality index is higher around a specified starting point (in this example, the target customer).

[0030] For example, the related customer group extraction unit 120 can calculate the price-weighted personalized page rank by the following procedure: First, the related customer group extraction unit 120 treats the directed graph in FIG. 5 as an undirected graph and executes the following procedure. 1) Create the following virtual directed graph: For any two nodes X and Y connected by an edge in the undirected version of the graph in Figure 5, the weight of the edge from node X to node Y in the virtual directed graph is set as the quotient of the weight of the undirected version of the graph in Figure 5 divided by the amount-weighted degree of node X (the sum of the absolute values ​​of the incoming and outgoing amounts). For example, the weight of the edge from node A to node E is 3 / (10 + 20 + 5 + 3) ≒ 0.0789, and the weight of the edge from node E to node A is 3 / (3 + 10 + 1 + 5) ≒ 0.1579. 2) The initial value of the provisional centrality of node A is set to 1, and the initial value of the provisional centrality of other nodes is set to 0. Here, the provisional centrality is a provisional centrality index generated in the calculation process of the monetary weighted personalized page rank. 3) Repeat steps a) to d) below to update the provisional centrality value of each node. 3a) Replace the provisional centrality of each node with the sum of the products of the weights of the edges flowing into that node in the virtual directed graph and the provisional centrality of the original node (the provisional centrality of the node from which that edge flows out). For example, after the first execution of this process, the provisional centrality of node E is 0.0789×1+0.2500×0+0.1250×0+0.2273×0=0.0789. Here, 0.2500, 0.1250, and 0.2273 are the weights of the edges from nodes C, D, and F to node E in the virtual directed graph, respectively, calculated in 1). 3b) Multiply the tentative centrality of every node by a decay factor parameter. For example, the decay factor parameter may be set as 0.85. 3c) Add 1 to the provisional centrality of node A and subtract the decay factor parameter. 3d) For each node, calculate the difference between the previous provisional centrality and the current provisional centrality. Note that "previous" in the first processing refers to the initial value 2. For all nodes, if the calculated difference is within a predetermined tolerance (e.g., 0.0000001), use the current provisional centrality as the monetary-weighted personalized page rank; if not, return to 3a).

[0031] From the above calculations, the monetary weighted personalized page ranks of nodes A to N are 0.2788, 0.1145, 0.1765, 0.0378, 0.0696, 0.0484, 0.1105, 0.0657, 0.0298, 0.0269, 0.0092, 0.0182, 0.0091, and 0.0049, respectively.

[0032] Next, the related customer group extraction unit 120 divides the amount-weighted Personalized PageRank by the amount-weighted degree (i.e., the transaction amount of each node, which is the sum of the absolute values ​​of the input and output amounts). For example, in the example shown in FIG. 6, for node A, the quotient obtained by dividing the amount-weighted Personalized PageRank by the amount-weighted degree of 38 (= 10 + 5 + 3 + 20) is 0.220, as shown in the figure. Similarly, the quotient for each node is the value shown in the figure. Here, to prevent errors caused by calculating large values, the amount-weighted degree may be calculated by dividing the weights of all edges by the maximum value (30 million yen in the example of FIG. 5) in advance. For example, for node A, the amount-weighted degree may be 38 / 30 ≒ 1.2667, and the quotient obtained by dividing the amount-weighted Personalized PageRank by the amount-weighted degree may be 0.2788 / 1.2667 ≒ 0.220.

[0033] After calculating a quotient by dividing the price-weighted personalized page rank by the price-weighted degree for each node, the related customer group extraction unit 120 ranks the nodes in descending order of the quotient value. As shown in Fig. 6, the related customer group extraction unit 120 ranks the nodes in the order of node A, node D, node C, ...

[0034] The related customer group extraction unit 120 then calculates the amount-weighted conductance for the ranked nodes, such as the first-ranked node A, the first- and second-ranked nodes A and D, the first- to third-ranked nodes A, D, and C, and the first- to fourth-ranked nodes A, D, C, and E. The calculation results can be plotted as shown in FIG. 7. For example, when the number of extracted nodes is four, the amount-weighted conductances for the first- to fourth-ranked nodes A, D, C, and E are indicated by the arrows. Specifically, the amount-weighted conductance is calculated by dividing the sum of the weights of the edges on the boundary of the extracted node group by the sum of the amount-weighted degrees (the sum of the absolute values ​​of the incoming and outgoing amounts) of the nodes belonging to the extracted node group. For example, when the extracted node group consists of the first- to fourth-ranked nodes A, D, C, and E, the amount-weighted conductance is (10 + 2 + 5 + 10) / (38 + 8 + 40 + 19) ≈ 0.257.

[0035] After calculating the amount-weighted conductance for each number of extracted nodes, the related customer group extraction unit 120 identifies the number of extracted nodes that minimizes the amount-weighted conductance and determines the node group included in the identified number of extracted nodes as the related customer group. In the example shown in FIG. 7, the amount-weighted conductance for the extracted node number of 8 is minimized, and the related customer group extraction unit 120 determines the node groups A, D, C, E, B, H, F, and G ranked 1st to 8th as the related customer group. This is because it is known that the more conditions ii) and iii) are satisfied, the smaller the amount-weighted conductance. After extracting the related customer group, the related customer group extraction unit 120 notifies the core customer identification unit 130 of the extracted related customer group.

[0036] The center customer identification unit 130 identifies center customers in the related customer group. Specifically, the center customer identification unit 130 calculates a price-weighted PageRank for each node in the extracted related customer group, taking into account the edge direction. The price-weighted PageRank for each node in the related customer group is the value shown in FIG. 8. That is, the center customer identification unit 130 determines the centrality index of node A to be 1.022, the centrality index of node B to be 1.343, the centrality index of node C to be 0.953, the centrality index of node D to be 0.723, the centrality index of node E to be 0.832, the centrality index of node F to be 1.391, the centrality index of node G to be 0.892, and the centrality index of node H to be 0.842. For example, node F, which has the largest centrality index, may be identified as the center customer.

[0037] Specifically, the central customer identification unit 130 calculates the price-weighted PageRank taking into account the direction of edges according to the following procedure. In this example, the related customer group is considered to be the entire graph, that is, the price-weighted PageRank is calculated on a graph consisting only of A to H and edges whose ends are contained within these. This is done to identify central people within the related customer group. However, if it is desired to evaluate centrality while also taking into account customers who are not included in the related customer group, the price-weighted PageRank may be calculated on the entire graph, not limited to the related customer group. 1) Create a virtual directed graph as follows: 1a) For nodes that do not have outgoing edges (e.g., nodes B and F), draw edges to all nodes other than the node itself and set the weight to 1. Note that the weight is not necessarily limited to 1 and may be any positive value. The purpose of setting this weight is to prevent division by zero in 1b) below. 1b) For any two nodes X and Y connected by an edge, taking into account the direction, the weight of the edge from node X to node Y is the quotient obtained by dividing the original weight by the amount-weighted outdegree of node X (the sum of the outdegrees of node X). For example, the weight of the edge from node A to node E is 3 / (10+5+3) ≒ 0.1667, there is no edge from node E to node A, and the weight of the edge from node B to node A is 1 / 7 ≒ 0.1429. 2) The initial value of the provisional centrality of all nodes is set to 1. Here, the provisional centrality is a provisional centrality index generated in the calculation process of the monetary weighted PageRank. 3) Repeat steps a) to d) below to update the provisional centrality value of each node. When the update is complete, the monetary weighted PageRank is output. 3a) Replace the provisional centrality of each node with the sum of the product of the weight of the edge flowing into that node in the virtual directed graph and the provisional centrality of the original node, i.e., the provisional centrality of the node from which that edge flows out. For example, after the first pass, the provisional centrality of node E is 0.1667 x 1 + 0.3333 x 1 = 0.5000, where 0.3333 is the weight of the edge from node D to node E in the virtual directed graph calculated in 1b. 3b) Multiply the tentative centrality of every node by a decay factor parameter. For example, the decay factor parameter may be set to 0.85. 3c) Add 1 to the pseudo-centrality of every node and subtract the decay factor parameter. 3d) For each node, calculate the difference between the previous provisional centrality and the current provisional centrality. Note that "previous" in the first processing refers to the initial value 2. For all nodes, if the calculated difference is within a predetermined tolerance (e.g., 0.0000001), use the current provisional centrality as the monetary-weighted PageRank; if not, return to 3a).

[0038] Note that the central customer according to the present disclosure is not limited to a single node but may be multiple nodes. The central customer identification unit 130 may also classify each node into multiple centrality levels based on the centrality index value. After identifying the central customer in this manner, the central customer identification unit 130 notifies the output unit 140 of the related customer group and the central customer. Here, a value-weighted PageRank that takes into account the edge direction increases the value of a node to which more funds "inflow" through the remittance network. However, a value-weighted PageRank that ignores the edge direction does not distinguish between inflow and outflow, and therefore increases the value of a node to which more funds flow in or out through the remittance network. When calculating the centrality index, it is preferable to calculate the index taking into account the edge direction in order to evaluate to whom funds are concentrated. Meanwhile, the reason for ignoring the edge direction when calculating the value-weighted Personalized PageRank in local clustering is that when a remittance is made between two nodes, the existence of a "relationship" between the two people, regardless of the direction, is reflected in the related person search.

[0039] The output unit 140 outputs the analysis results indicating the related customer group and the central customer. For example, the output unit 140 may associate the node group belonging to the related customer group with the centrality index of each node of the node group and display the fund transfer information in a graph format on the terminal 60.

[0040] Specifically, the output unit 140 may increase the size of the node or color it according to the magnitude of the centrality index. The output unit 140 may also display / hide nodes of related customer groups according to the attributes of the nodes. For example, the output unit 140 may display only the bank's own customers among the related customer group, and hide transfer / foreign exchange remittance counterparties to other banks. Alternatively, the output unit 140 may display only customers among the related customer group who have conducted transactions equal to or greater than a set transaction amount, and hide customers who have conducted transactions less than the set transaction amount.

[0041] [Customer analysis processing] Next, a customer analysis process according to an embodiment of the present disclosure will be described with reference to Fig. 9. The customer analysis process is executed by the above-described customer analysis device 100, and more specifically, may be realized by one or more processors 102 of the customer analysis device 100 executing one or more programs or instructions stored in one or more storage devices 101. Fig. 9 is a flowchart showing the customer analysis process according to an embodiment of the present disclosure.

[0042] 9, in step S101, the customer analysis device 100 acquires an analysis request indicating a target customer. Specifically, the customer analysis device 100 receives the analysis request indicating the target customer from the terminal 60. The analysis request may also include a search range such as a target period and a customer group to be searched.

[0043] In step S102, the customer analysis device 100 extracts a group of related customers who are related to the target customer. Specifically, the customer analysis device 100 may acquire information on fund transfers between customers corresponding to the search range from the database device 50, and extract a group of related customers who are related to the specified target customer from the fund transfer information. For example, the relationship may be the transaction amount and / or transaction frequency between customers, and the search conditions may be i) The relevant customer group includes customer information of the target customer; ii) The transaction volume within the relevant customer group is large; and iii) The transaction amounts between related and unrelated customer groups are small; As described above, the customer analysis device 100 extracts a related customer group according to a known local clustering method. Specifically, the customer analysis device 100 may determine, as the related customer group, a node group for which the conductance calculated based on the personalized page rank is the smallest.

[0044] In step S103, the customer analysis device 100 identifies a central customer in the related customer group. Specifically, the customer analysis device 100 may use the page rank of each node in the related customer group as a centrality index, and identify the central customer based on the centrality index.

[0045] In step S104, the customer analysis device 100 outputs an analysis result indicating the related customer group and the central customer. Specifically, the customer analysis device 100 may transmit to the terminal 60 screen information that associates a centrality index with each node of the related customer group and displays the related customer group in a graph format.

[0046] According to the above-described embodiment, the customer analysis device 100 can extract a group of related customers that are related to the target customer and determine the centrality index of each customer in the related customer group. This makes it possible to extract groups involved in fraudulent transactions from inter-customer fund transfer information and identify central figures in the groups, which may enable effective detection of criminal groups.

[0047] Although the examples of the present disclosure have been described in detail above, the present disclosure is not limited to the specific embodiments described above, and various modifications and variations are possible within the scope of the gist of the present disclosure as set forth in the claims. [Explanation of symbols]

[0048] 50 Database device 60 devices 100 customer analysis equipment 110 Input section 120 Related Customer Group Extraction Unit 130 Core Customer Identification Department 140 Output section

Claims

1. an input unit that receives an analysis request indicating a target customer from a terminal; a related customer group extraction unit that extracts related customer groups that have a relationship with the target customer based on transaction amounts with the target customer during a predetermined period from fund transfer information expressed as a weighted directed graph with each customer as a node and transaction amounts between customers as weighted directed edges, in accordance with predetermined extraction conditions related to the customer information of the target customer, transaction amounts within the related customer group, and transaction amounts between the related customer group and unrelated customer groups; a central customer identification unit that identifies a central customer who plays a central role in the related customer group based on the amount-weighted page rank of each node in the related customer group; an output unit that outputs an analysis result indicating the related customer group; A customer analysis device having:

2. 2. The customer analysis device according to claim 1, wherein the related customer group extraction unit calculates a price-weighted Personalized PageRank for each node according to a local clustering technique, and extracts a node group having the smallest conductance calculated based on the price-weighted Personalized PageRank as the related customer group.

3. The customer analysis device according to claim 1 , wherein the central customer identification unit identifies the central customer based on a centrality index of the customer in the related customer group.

4. The customer analysis device according to claim 3 , wherein the centrality index is determined based on a price-weighted PageRank.

5. The customer analysis device according to claim 1 , wherein the analysis result displays the related customer group in a graph format.

6. a database device for storing fund transfer information expressed as a weighted directed graph in which each customer is a node and transaction amounts between customers are weighted directed edges; A customer analysis device; and The customer analysis device an input unit that receives an analysis request indicating a target customer from a terminal; a related customer group extraction unit that extracts, from the inter-customer fund transfer information, a related customer group that has a relationship with the target customer based on the transaction amount with the target customer during a predetermined period, in accordance with predetermined extraction conditions related to the inclusion of customer information of the target customer, the transaction amount within the related customer group, and the transaction amount between the related customer group and a non-related customer group; a central customer identification unit that identifies a central customer who plays a central role in the related customer group based on the amount-weighted page rank of each node in the related customer group; an output unit that outputs an analysis result indicating the related customer group; A customer analysis system with

7. receiving an analysis request indicating a target customer from a terminal; A step of extracting a group of related customers that have a relationship with the target customer based on transaction amounts with the target customer during a predetermined period from fund transfer information expressed as a weighted directed graph with each customer as a node and transaction amounts between customers as weighted directed edges, in accordance with predetermined extraction conditions related to the customer information of the target customer, transaction amounts within the related customer group, and transaction amounts between the related customer group and unrelated customer groups; Identifying a central customer who plays a central role in the related customer group based on the amount-weighted page rank of each node of the related customer group; outputting an analysis result indicating the related customer group; A customer analysis method performed by a computer.

Images