JP6343475B2 - Value chain analysis apparatus and method - Google Patents

Description

  The present invention relates to a technique for automatically recognizing and analyzing a value chain from information on funds flow.

  When a bank provides a loan to a company, it analyzes the business details and business results of the company to grasp the risk at the time of executing the loan, and then determines whether or not the loan can be executed. A well-known technique for understanding risk is credit rating or scoring. Scoring quantifies the safety (ie, the risk level) of a company's loan by evaluating the company's financial statements using a predetermined evaluation function.

  However, the financial statements have a lot of room for interpretation, and the content of the interpretation does not necessarily accurately reflect the performance of the company. Furthermore, unlisted companies that do not undergo accounting audits tend to have a large discrepancy between the financial statements and actual performance. Therefore, Patent Document 1 proposes a technique that enables a credit rating in consideration of the difference between the financial statements and the actual situation of business performance. This technology inputs the additional information in addition to the financial statements submitted by the company, analyzes the actual financial structure, and corrects the deviation from the actual situation (ie, “decoration”) to determine the credit classification. However, this method is not sufficient for understanding business risk because it does not analyze the value chain.

  In addition, corporate activities are not closed within a single company. In addition, in the analysis of financial statements, etc., the business risk is only evaluated in the relationship between the target company and the company that has direct transactions in the form of suppliers and sales customers. However, in recent years, the expertise of corporations has advanced, and it has become possible to assemble final products by combining multiple functional modules. In other words, in order to complete the final product, not only direct business partners but also transactions with companies located further upstream are involved in a complicated manner. As a result, it is no longer possible to accurately grasp business risks simply by looking directly at business partners.

  For example, one part is incorporated into the final product along the following flow. (1) Incorporate parts into submodules, (2) Incorporate submodules into modules, (3) Incorporate modules into final products. All these integration tasks are handled by another company. In this way, when parts are procured hierarchically, the intention is to divide the risk by procuring modules and sub-modules from multiple companies, but the risk depends on a single component manufacturer at the sub-module stage. It may not be distributed. For this reason, value chain management is an extremely important issue.

  There is a technique described in Patent Document 2 as a method for credit rating by managing the entire value chain. With this technology, the entire value chain can be managed by manually registering the value chain in the system. However, this method has two problems. The problem is that companies that have not been recognized as value chains are leaked from management targets, and that it takes a lot of man-hours to apply to bank lenders.

  If there are 10 business partners for a single company, the number of companies in the five-level value chain is 100,000. It is not realistic to register all this manually. It is necessary to select and manage important companies, but control LSIs of hundreds of yen may be key parts of products of millions of yen. The importance cannot be grasped. Since bank lenders range from tens of thousands to hundreds of thousands, it is virtually impossible to manage the value chain manually. For this reason, a technology for automatically extracting which value chain is important and which company the value chain strongly depends on is required. In other words, when a company is represented as a node, a business-to-business transaction as an edge, and a value chain as a graph, this is a technique for determining which node in the graph is important.

  As a technique for ranking the importance of nodes in a graph, there is a technique described in Patent Document 3. This technology is used as a technology for ranking which sites (nodes) are important on the World Wide Web. In this technique, after giving an initial value of the weight to each node, a steady state is obtained by repeating the calculation of propagating the weight according to the structure of the link (edge), and a node having a larger weight is determined to be more important. Here, in order for the calculation to converge, all the nodes need to have outward edges (links to other sites). Such convergence calculation is equivalent to the problem of obtaining eigenvalues and eigenvectors. However, this convergence calculation does not converge if there is a node having no outward edge. For this reason, in the technique of Patent Document 3, after calculating a node weight by removing a node having no outward edge and calculating a weight of the node, the weight is propagated to a node having no outward edge, The weights of all nodes are calculated.

  By the way, in a business-to-business transaction, a company with a larger capital flow has a transaction with a plurality of banks. In order to apply this technique to the flow of funds between companies within a single bank, it is necessary to calculate the convergence by excluding companies that have transactions with multiple banks. However, if a company with a large flow of funds that has transactions with multiple banks is excluded, there is a problem that the ranking cannot be made accurately because the structure of the value chain changes.

  As a method for examining which companies in the value chain are important (fragile), the method described in Non-Patent Document 1 is known. This technology is used for systemic risk analysis and stress testing, and analyzes which part of interbank transactions is problematic by Monte Carlo simulation based on interbank OTC derivatives and cash retention status. To do. To apply this method to business-to-business transactions, it is necessary to obtain all information on accounts receivable and accounts payable. However, since it is difficult for banks to obtain such information, there is a problem that it cannot be applied.

JP 2002-92321 A JP-A-2005-44251 US Pat. No. 6,285,999

Viral V. Acharya, Lasse H. Pedersen, Thomas Philippon, and Matthew Richardson, Measuring Systemic Risk, (May 2010), AFA 2011 Denver Meetings Paper

With these previous cases, there are the following four problems in analyzing the value chain in order to understand the business risks of tens of thousands to hundreds of thousands of borrowers.
(1) It takes a lot of man-hours to manually recognize the value chain.
(2) Even an important company, such as a company that produces parts that do not work as substitutes, is overlooked by hand.
(3) Even if an attempt is made to extract an important company by calculation, it cannot be ranked correctly because the structure of the value chain is changed to exclude companies that have transactions with multiple banks.
(4) Even if the importance is examined by simulation, it cannot be calculated because the bank cannot grasp the accounts payable and accounts receivable of the borrower.

In order to solve these problems, it is necessary to solve the following technical problems.
(1) Recognize the value chain by calculation.
(2) Calculate the importance of the value chain.
(3) Estimate the company that is the root cause of the company's performance.

  In order to solve the above-described problems, the present invention employs, for example, the configurations described in the claims. The present specification includes a plurality of means for solving the above-described problems. To give an example, “a data linking unit that acquires information from a storage system that stores information on fund flows between companies, and the information” Value chain analyzer having a value chain recognizing unit for recognizing a value chain from a value chain and a value chain amplification factor calculating unit for calculating the amplification factor of funds of the entire value chain for each recognized value chain .

  According to the present invention, it is possible to automatically recognize a value chain from among fund flow information between companies, and it is possible to automatically calculate the importance of each value chain. This makes it possible to extract important value chains and estimate important nodes.

The figure which shows the structural example of a value chain analysis system. The schematic diagram explaining a money flow and a transition probability. The schematic diagram explaining a value chain. The flowchart explaining the recognition process of a value chain. The flowchart explaining the processing operation of an edge search subroutine. The flowchart explaining the calculation process of a value chain gain. The flowchart explaining the processing operation of a graph search subroutine. The flowchart explaining the calculation process of a node rank. The flowchart explaining main factor analysis processing. The figure which shows the example of an image of an interface screen.

  Hereinafter, embodiments of the present invention will be described with reference to the drawings. The embodiment of the present invention is not limited to the embodiments described later, and various modifications are possible within the scope of the technical idea. In the drawings, the same reference numerals indicate the same components.

[Example 1]
In this example, the company's business performance is obtained by recognizing the company's value chain by calculation from the inter-company's funds flow as understood by the bank, etc., and calculating the importance of the company using the recognized value chain. A value chain analysis device that can estimate the companies that affect the system will be described.

[System configuration]
FIG. 1 shows the overall configuration of the value chain analysis system used in this embodiment. The system includes a fund flow database 160 for accumulating fund flow information, a value chain analysis device 100 that acquires fund flow information from the database and analyzes a value chain, and a terminal 150 that is an input / output device for the device. Composed. The fund flow database 160 is a large-scale storage device and is composed of one or a plurality of storage device groups. The fund flow database 160 is connected to the value chain analyzer 100 via a network. The type and configuration of the network used for connection are arbitrary.

  Both the value chain analysis apparatus 100 and the terminal 150 are configured by a computer. However, the value chain analyzer 100 is required to have a high calculation capability. For this reason, the value chain analyzer 100 is configured based on a large computer. On the other hand, the terminal 150 includes a computer main body, a monitor, an input keyboard, and the like. Although only one terminal 150 is shown in FIG. 1, a plurality of terminals 150 can be used. The value chain analyzer 100 and the terminal 150 are also connected via a network.

  The processing function of the value chain analyzer 100 is realized through a program executed on the CPU. The processing function is roughly composed of four data linkage unit 101, value chain recognition unit 110, value chain amplification factor calculation unit 120, node rank calculation unit 130, and main factor analysis unit 140. Details of the processing operation executed in each processing unit will be described later.

[Fund flow and transition probability]
FIG. 2 schematically shows the relationship between the fund flow and the transition probability. In the figure, circled numbers are nodes corresponding to individual companies, and arrows are edges indicating business relationships between companies. FIG. 2 shows a graph 200 representing a fund flow between nodes, a graph 210 representing a forward transition matrix (probability) and a node amplification factor, and a graph 220 representing a reverse transition matrix (probability) and a node amplification factor. Represents. Details of FIG. 2 will be described later.

[Value chain]
FIG. 3 shows an example of a value chain assumed in this embodiment. In FIG. 3, 300 is a value chain, 301 is a node identified as being outside the range of the value chain due to stage number limitation, 302 is a node identified as being outside the range of the value chain due to contribution rate limitation, and 303 is not traceable from the starting point It shall indicate a node identified as out of range of the value chain. Details of FIG. 3 will be described later.

[Overview of processing operations]
Next, an outline of processing operations executed in the value chain analysis apparatus 100 will be described. First, the data linkage unit 101 reads information on the money flow from the money flow database 160 that accumulates information on the money flow. The information on the fund flow has at least information on the remittance source, remittance destination and remittance amount. 2 and 3, the remittance source and remittance destination are nodes, and the remittance amount from the remittance source to the remittance destination is the edge.

  The value chain recognizing unit 110 executes processing for recognizing the value chain from the read information of the money flow. First, the transition probability calculation unit 111 calculates the transition probability between nodes from the information of the fund flow. The threshold value setting unit 112 executes a process of setting a threshold value of the contribution rate of the money flow (defined as a product of transition probabilities for each edge on the movement route) when the money moves over a plurality of edges. The edge step number setting unit 113 executes processing for setting the maximum step number of edges to which funds move. The edge search unit 114 uses the set contribution rate and the number of edge steps as a recognition condition, and executes a process of recognizing a value chain from information on funds flow. Here, the threshold and the number of edge stages are set through the terminal 150.

  The value chain amplification factor calculation unit 120 calculates the amplification factor of funds in the value chain for each recognized value chain (that is, how much the funds input by the company that is the starting point of the value chain are amplified at each node constituting the value chain. Is executed). First, the node amplification factor calculation unit 121 calculates a fund amplification factor of a single node based on information on the inflow of funds to each node and the outflow of funds from the node. The node amplification factor calculation unit 122 calculates a node amplification factor element that is an amplification factor at each node of the value chain as viewed from the starting node of the value chain. The sum of the node amplification factor elements of all the nodes in the value chain is the value chain amplification factor. The graph search unit 123 causes the node amplification factor calculation unit 122 to be executed for all recognized value chains.

  The node rank calculation unit 130 executes a process for calculating the ranks of nodes constituting the entire graph. The above-described value chain amplification factor calculation unit 120 calculates the node amplification factor element of each node in the value chain as seen from the starting node of each value chain, but each node also belongs to a plurality of value chains. ing. Therefore, the node rank calculation unit 130 defines the rank of the node in the entire graph using all the node amplification factor elements calculated for all the value chains to which a certain node belongs.

  The amplification factor calculation unit 131 calculates the rank of a node as the sum of all node amplification factor elements calculated for all value chains to which a certain node belongs. The amplification amount calculation unit 132 calculates the sum of all the node amplification amount elements calculated by the product of the node amplification factor element of a certain value chain and the fund flow rate of the node that is the starting point of the corresponding value chain. In this embodiment, either the node rank calculated by the amplification factor calculation unit 131 or the node rank calculated by the amplification amount calculation unit 132 is used as the node rank.

  The main factor analysis unit 140 executes a process of extracting a company that is a root cause that greatly affects the performance of the entire graph. This is because the performance of a company is influenced not only by its own business strategy, but also by the performance of direct business partners and indirect business partners who conduct business through direct business partners. The extraction condition setting unit 141 executes processing for setting extraction conditions. The extraction condition is set through the terminal 150. The amplification factor totaling unit 142 performs processing for extracting a root cause company by paying attention to the node amplification factor element. The amplification amount totaling unit 143 executes processing for extracting a root cause company by paying attention to the node amplification amount element. Details of each process will be described later. Various conditions are set and calculation results are displayed through the terminal 150.

[Method of calculating transition probability and node amplification factor from funds flow]
Here, a method for calculating the transition probability and the node amplification factor from the fund flow will be described with reference to FIG. In the case of the graph 200, there is a fund flow f ₄₁ from node 4 to node 1, a fund flow f ₅₁ from node 5 to node 1, a fund flow f ₁₂ from node 1 to node 2, and a fund flow f ₁₃ from node 1 to node 3. . Since logistics exists in the opposite direction to the cash flow, to consider a value chain consisting of both suppliers and sellers, both the supplier who is the forward direction of the capital flow and the seller who is the opposite direction of the cash flow. I need to see it.

In this specification, symbols are defined as follows. Let f _{ij be} the fund flow from node i to node j. In this case, the forward transition probability and the backward transition probability from node i to node j are p _ij and rp _ij , respectively. Let g _i and rg _i be the forward gain and reverse gain of node i, respectively. These are defined as follows:
p _ij = f _ij Σ _k f _ik
rp _ij = f _ij Σ _k f _kj
g _i = Σ _k f _ik / Σf _ki
rg _i = Σ _k f _ki / Σf _ik

  A graph 210 in FIG. 2 shows the relationship between the forward transition probability and the node amplification factor for the fund flow (graph 200), and the graph 220 shows the relationship between the reverse transition probability and the node amplification factor for the fund flow (graph 200). Is shown. In the following, the forward direction and the reverse direction are not particularly specified, but the same calculation method can be applied to either direction.

[Value chain recognition example]
FIG. 3 shows an example of value chain recognition. FIG. 3 shows a value chain 300 starting from the node 1. The present embodiment is characterized in that the value chain is identified by “the number of edge stages” and “contribution rate” defined by the product of the transition probabilities of edges that give a movement path when moving a plurality of edges.

In FIG. 3, the threshold value for the maximum number of edge steps is set to 2 or less. In this case, since the node 6 (reference numeral 301) is the third-stage node when viewed from the node 1, it is treated as outside the range of the value chain 300. Further, in FIG. 3, the threshold value of the contribution rate and 5% p ₁₅ 8% and p ₅₈ is 10%, the node 8 (code contribution rate is calculated as 0.8% (= 8% × 10%) 302) is treated as outside the range of the value chain 300. In addition to the node 1, the node 5 has an edge from the node 9. However, since the node 9 has no route that can be reached from the node 1, the node 9 (reference numeral 303) is also treated as outside the range of the value chain 300.

As in the present embodiment, it is extremely important to use the contribution rate for recognition of the value chain 300. This is because even if the number of steps is smaller than the maximum number of edges, the contribution ratio is small, that is, there is an effect of removing an unimportant node from the value chain. For example, in a graph having 10 edges per node, if a value chain is identified only under the condition that the maximum number of edge stages is 5 or less, the number of nodes in the value chain is 100,000 (= 10 ⁵ ). If one value chain has 100,000 nodes, it takes a lot of time to process with a computer, and it becomes practically impossible for a human to interpret the value chain.

  On the other hand, if a condition that “excludes nodes with a contribution rate of less than 1%” is added, the number of nodes in the value chain will be several tens or less, and it will be possible to speed up the processing on the computer. , Human interpretation is also easier. However, the screening based on the contribution rate alone will make the important companies that make key parts out of the value chain, although the transaction amount is small. Therefore, in this embodiment, a mechanism for extracting such an important company as a value chain is also prepared. This mechanism will be described later.

[Value chain recognition processing]
An example of value chain recognition processing will be described based on FIGS. 4A and 4B. When the recognition process (step 400) starts, the value chain recognition unit 110 sets the number of edge steps and the threshold value of the contribution rate in step 401. The set value is input by a bank clerk or the like through the interface screen of the terminal 150.

  When the threshold of the number of edge steps and the contribution rate is set, the value chain recognition unit 110 proceeds to step 402 and calls the edge search subroutine 410. When the edge search subroutine 410 is started, the value chain recognizing unit 110 determines whether or not an edge to be processed exists in the information of the money flow acquired from the money flow database 160 (step 411). If there is an edge, the value chain recognition unit 110 proceeds to step 412. If not, the value chain recognition unit 110 ends the subroutine and returns to the main process.

  If there is an edge, the value chain recognition unit 110 acquires the edge transition probability (step 412). When the transition probability has not been calculated, the value chain recognition unit 110 calculates the transition probability and uses the value. In the edge search, there is a high possibility that the same edge will be traced multiple times. Therefore, if the calculated transition probability is stored in a storage area (not shown), it is stored if a value is required for the second and subsequent times. The amount of calculation can be reduced by reading the value. After the end of step 412, the value chain recognition unit 110 proceeds to step 413.

  In step 413, the value chain recognizing unit 110 uses the transition probability information acquired in step 412 to calculate the number of stages of each edge and the contribution rate when the edge is advanced. After the end of step 413, the value chain recognition unit 110 proceeds to step 414. In step 414, the value chain recognition unit 110 determines whether the contribution rate calculated in step 413 is greater than a threshold value. If it is determined that the value chain is greater than the threshold value, the value chain recognition unit 110 proceeds to step 415. Otherwise, the value chain recognition unit 110 ends the subroutine and returns to the main process.

  In step 415, the value chain recognition unit 110 determines whether the number of edge steps calculated in step 413 is equal to or less than a threshold value. The value chain recognizing unit 110 proceeds to step 402 when it is determined that the value is equal to or less than the threshold value, and otherwise ends the subroutine and returns to the main process. In step 402, the value chain recognition unit 110 recursively calls the edge search subroutine 410. Since the graph processing that moves from node to node via an edge has a recursive structure, the edge search subroutine is also a recursive processing. When the edge search subroutine ends, the value chain recognition unit 110 proceeds to step 411.

  The edge search subroutine 410 executes a recursive search for all nodes with a threshold of contribution rate and a threshold of the number of edge steps starting from a certain node. Since there is a high possibility of edge search for the same node a plurality of times, the processing can be speeded up by caching a recognized value chain (not shown).

[Value chain gain calculation processing]
Based on FIG. 5A and FIG. 5B, an example of value chain amplification factor calculation processing will be described. In this embodiment, the amplification factor G (i) of the value chain starting from the node i is defined by the following equation using the transition probability p _ik and the node amplification factor g _k .
G (i) = 1 + Σ _k (p _ik * g _k * G (k))
Here, G (k) is limited to the value chain starting from node i.

By expanding this mathematical formula, the node amplification factor element at each node can be calculated. If the amplification factor element of node j starting from node i is G _ij , the node amplification factor element in the value chain of FIG. 3 is given as follows. Here, since G ₁₁ is the starting point itself, the amplification factor was set to 1.
G ₁₁ = 1
G ₁₂ = p ₁₂ * g ₂
G ₁₃ = p ₁₂ * g ₂ * p ₂₃ * g ₃
G ₁₄ = p ₁₂ * g ₂ * p ₂₄ * g ₄
G ₁₅ = p ₁₅ * g ₅

  The definition of the value chain gain is not limited to the above-described definition. For example, there is a method including the node amplification factor at the starting point. In the present embodiment, the term value chain amplification factor is used because of the meaning of the mathematical expression, but any mathematical expression may be used as long as it is a defining expression that characterizes the value chain. For example, as described above, a method may be used in which both the transition probability and the node amplification factor are not used, but one is defined.

  Also in the value chain amplification factor calculation process (step 500), a recursive process similar to the value chain recognition process is executed. Here, it is assumed that a certain node is designated through the terminal 150, and a method for calculating the amplification factor of the value chain starting from the node will be described.

  When the calculation process (step 500) starts, the value chain amplification factor calculation unit 120 proceeds to step 501 and acquires a value chain graph of the designated node. After step 501 is completed, the value chain amplification factor calculation unit 120 proceeds to step 502 and calls the graph search subroutine 510. When the node amplification factor element is calculated in the graph search subroutine 510, the value chain amplification factor calculation unit 120 proceeds to step 503 and calculates the value chain amplification factor by the sum of the node amplification factor elements.

When the graph search subroutine 510 is started, the value chain amplification factor calculation unit 120 proceeds to step 511 and initializes the node amplification factor element. For example, in the case of the graph of FIG. 3, it is assumed that G ₁₁ = 1. After the end of step 511, the value chain amplification factor calculation unit 120 proceeds to step 512, and determines whether there is an edge to be processed. The value chain amplification factor calculation unit 120 proceeds to step 513 if there is an edge to be processed, and ends the subroutine and returns to the main processing if it does not exist.

  In step 513, the value chain amplification factor calculation unit 120 acquires the transition probability of the processing target edge. After completion of step 513, the value chain amplification factor calculation unit 120 proceeds to step 514. In step 514, the value chain amplification factor calculation unit 120 obtains the node amplification factor of the node ahead of the processing target edge. After step 514 is completed, the value chain amplification factor calculation unit 120 proceeds to step 502 and calls the graph search subroutine 510 again. By calling the graph search subroutine 510, the value chain amplification factor calculation unit 120 recursively executes a process of calculating node amplification factor elements after the node ahead of the processing target edge. move on. In step 515, the value chain amplification factor calculation unit 120 updates the node amplification factor element acquired in the subroutine 502 using the transition probabilities and node amplification factors acquired in step 513 and step 514, respectively. When step 515 ends, the value chain amplification factor calculation unit 120 returns to step 512.

[Node rank calculation]
The flow of funds between companies is performed using multiple banks. In the financial flow between companies that can be grasped by one bank, the cash flow graph is not closed due to the inflow of funds from other banks and the outflow of funds to other banks. Since the graph is not closed, the value does not converge in the iterative calculation. For this reason, it is not possible to use a method that uses the same convergence calculation as ranking sites on the World Wide Web to rank the importance of nodes in the graph.

  Therefore, in this embodiment, the node rank is calculated using the node amplification factor element. For example, when a node belongs to a plurality of value chains, the node amplification factor element is obtained every time the value chain amplification factor of each value chain is calculated. In the present embodiment, the node rank is calculated based on the sum of the amplification factor elements calculated for each of these value chains or the sum of the node amplification factor elements defined by the product of the flow rate of the starting node of the value chain to the node amplification factor element. Define Below, an example of a node rank calculation process is demonstrated based on FIG.

  When the calculation process (step 600) starts, the node rank calculation unit 130 proceeds to step 601. In step 601, the node rank calculation unit 130 determines whether there is a node to be processed. The node rank calculation unit 130 proceeds to step 602 if there is a node to be processed, and proceeds to step 610 otherwise. In step 602, the node rank calculation unit 130 acquires a value chain graph starting from the processing target node. After step 602 ends, the node rank calculation unit 130 proceeds to step 603. In step 603, the node rank calculation unit 130 obtains the funds flow rate to the processing target node. After step 603 ends, the node rank calculation unit 130 proceeds to step 604.

  In step 604, the node rank calculation unit 130 determines whether there is a node to be processed in the graph. If there is a node to be processed, the node rank calculation unit 130 proceeds to step 605. If not, the node rank calculation unit 130 returns to step 601 and continues the process using another node as the starting point of the value chain.

In step 605, the node rank calculation unit 130 acquires a node amplification factor element for the processing target node in the graph. After step 605 ends, the node rank calculation unit 130 proceeds to step 606. In step 606, the node rank calculation unit 130 stores the node amplification factor element acquired in step 605 in a list. For example, when the node 1 is included in the value chain starting from the node i, the node j, and the node k, the node rank calculation unit 130 searches for these three value chains and then performs the following for the node l: Get the list LRG _l .
LRG _l = [(i, G _il ), (j, G _jl ), (k, G _kl )]

  After the end of step 606, the node rank calculation unit 130 proceeds to step 607. In step 607, the node rank calculation unit 130 calculates a node amplification amount element from the fund flow rate acquired in step 603 and the node amplification factor element acquired in step 605. After step 607 ends, the node rank calculation unit 130 proceeds to step 608. In step 608, the node rank calculation unit 130 stores the node amplification amount element calculated in step 607 in the list as in step 606.

  After the end of step 608, the node rank calculation unit 130 returns to step 604 and continues the process for the remaining nodes to be processed in the graph. When the process is completed for all nodes in the graph, the process returns to step 601 to continue the process for the value chain starting from another node. When the processing is completed for all the nodes in step 601, the node rank calculation unit 130 proceeds to step 610. In step 610 and subsequent steps, the node rank calculation unit 130 totals the lists created in step 606 and step 608.

In step 610, the node rank calculation unit 130 determines whether there is a node to be processed. The node rank calculation unit 130 proceeds to step 611 when there is a node to be processed, and ends the process otherwise. In step 611, the node rank calculation unit 130 totals the list of node amplification factor elements created in step 606 for the nodes to be processed. For example, for node i, rank RG _i is calculated by the following sum.
RG _i = Σ _k G _ki

  After completing step 611 and proceeding to step 612, the node rank calculation unit 130 totalizes the list of node amplification amount elements created in step 608 for the nodes to be processed. The aggregation process is executed in the same manner as in step 611. After the end of step 612, the node rank calculation unit 130 returns to step 610 and continues the process for the remaining nodes to be processed in the graph.

  As described above, in this embodiment, instead of using convergence calculation, focusing on the value chain, the contribution rate and node amplification rate are used to define and reflect an index that reflects the value chain fund amplification effect. ing. This process is characterized in that the node ranking calculation is realized without using the convergence calculation by the node amplification factor as the initial value and the attenuation effect of the influence of the far node by the contribution rate.

  Depending on the size of the company and the structure of the value chain, the value chain fund amplification effect may be appropriate in terms of amplification rate or in terms of amplification amount. For this reason, in this embodiment, the node rank is calculated based on both the amplification factor and the amplification amount. For example, when introducing node ranks based on amplification factors, it is possible to extract an important company that produces key parts that are inexpensive but do not work as substitutes, even if they are far from the company that produces the final product. . In addition, companies with small transaction amounts may not be recognized as Batu Chain due to the threshold of contribution rate, but important companies will have a high node rank even if the transaction amount is small. Companies with high node ranks are re-registered in the value chain, or value chains starting from companies with high node ranks are used as a value chain that demands by reducing the contribution threshold and increasing the threshold for the number of edge steps. Special measures and other techniques eliminate the oversight of important companies that have small transaction amounts.

  In this embodiment, the node rank is calculated based on the node amplification factor element, but the node rank calculation method is not limited to this method. A method of calculating the node rank by defining another formula using the transition probability and node amplification factor that are the basis for calculating the node rank amplification factor, or either transition probability or node amplification factor A method using, or a method using another value or mathematical expression characterizing the value chain is also conceivable.

[Main factor analysis processing]
An example of the main factor analysis process will be described with reference to FIG. If the node rank is used, the rank among all the nodes in the graph can be found, and if the node amplification factor element list obtained during the node rank calculation is used, the node having a great influence on the noticed node can be found. On the other hand, here, a method of main factor analysis for extracting a node group having a large influence in the entire graph will be described. Since node ranks are assigned to all nodes, it is necessary to manually determine which nodes should be examined in descending order of rank. On the other hand, the method proposed here is effective in narrowing down the number of nodes to be noticed even when the number of nodes in the entire graph is large because only the nodes having a high influence degree are extracted. The main factor analysis unit 140 of the present embodiment extracts a starting node from a list of node amplification factor elements in a descending order according to a specified number and a condition equal to or more than a specified value, and determines a group of companies that are the root cause of corporate performance. Extract.

  When the analysis process (step 700) is started, the main factor analysis unit 140 sets node extraction conditions in step 701. Here, when the lists calculated in step 606 and step 608 are normalized so that the sum is 1 and rearranged in the descending order, the number of lower limit values and the maximum number of nodes extracted in the descending order are determined. specify. After step 701 ends, the main factor analysis unit 140 proceeds to step 702. In step 702, the main factor analysis unit 140 determines whether there is a node to be processed. If there is a node to be processed, the process proceeds to step 703; otherwise, the process proceeds to step 707.

In step 703, the main factor analysis unit 140 acquires a list of node amplification factor elements calculated in step 606 for the processing target node. The list stores values in the form of a pair of a node ID that is the starting point of the value chain and a node amplification factor element corresponding to the node ID. After the completion of step 703, the main factor analysis unit 140 proceeds to step 704. In step 704, the main factor analysis unit 140 stores the ID and amplification factor of the node that matches the condition specified in step 701. For example, if the list of node a is [(i, G _ia ), (j, G _ja ), (k, G _ka ), (l, G _la ), (m, G _ma )] When nodes i and j are to be extracted, _Gia and _Gja are stored in the newly created lists of nodes i and j, respectively.

  After step 704 ends, the main factor analysis unit 140 proceeds to step 705. In step 705, the main factor analysis unit 140 acquires the list of node amplification amount elements calculated in step 608 for the processing target node. After step 705 ends, the main factor analysis unit 140 proceeds to step 706. In step 706, the main factor analysis unit 140 stores the ID and amplification amount of the node that matches the condition specified in step 701. After the completion of step 706, the main factor analysis unit 140 returns to step 702. When the processing of all the nodes is completed in step 702, the main factor analysis unit 140 proceeds to step 707.

  In step 707, the main factor analysis unit 140 determines whether there is a node to be processed. The main factor analysis unit 140 proceeds to step 708 if there is a node to be processed, and ends the process if it does not exist. In step 708, the main factor analysis unit 140 aggregates the node amplification factor elements included in the list created in step 704 for the processing target node. After the completion of step 708, the main factor analysis unit 140 proceeds to step 709.

  In step 709, the main factor analysis unit 140 calculates a node amplification amount element included in the list created in step 706 for the processing target node. After step 709 ends, the main factor analysis unit 140 returns to step 707 and continues processing for the remaining nodes to be processed in the graph.

  With these processes, a score based on an amplification factor or an amplification amount can be calculated for a company that is the root cause of the performance of a node. Here, the higher the score, the greater the impact on the performance of many companies in the entire graph.

[Example of interface screen]
FIG. 8 shows an example of an interface screen image used in this embodiment. The interface screen 800 is displayed on the monitor of the terminal 150. The interface screen 800 is roughly divided into three areas: (1) an input area, (2) a node display area, and (3) a value chain drawing area. In the input area, a contribution rate threshold entry field 801, a maximum stage number entry field 802, a node number upper limit entry field 803, and a lower limit value entry field 804 are displayed. These input fields are used for inputting various values necessary for the processing operation described above.

  The node display area includes a node ID display area 811, a node rank display area 812 based on the amplification factor, a node rank display area 813 based on the amplification factor, a main factor analysis score display area 814 based on the amplification factor, and a main factor analysis score display region based on the amplification factor. 815 is displayed. These display areas are used to display the results of node rank calculation and principal factor analysis. After sorting these values, the user selects which node details to examine. Details of the selected node are displayed in the value chain drawing area.

  In the value chain drawing area, a value chain information display area 821 and an information display area 822 of a node group in the value chain are displayed. In these display areas, the details of the value chain starting from the node selected in the node display area are displayed. In addition to displaying the value chain as a graph, the amplification factor and amplification amount of the value chain, and detailed information of the nodes in the value chain (for example, node amplification factor element, node rank, main factor analysis score, etc.) are displayed.

[Summary]
Finally, the functional configuration characteristic of the value chain analysis apparatus 100 according to the present embodiment is organized. The following is listed for each feature.
(Characteristic 1) One of the characteristics is that the data linkage unit 101 acquires the information from the storage system that stores the information of the funds flow between the companies, the value chain recognition unit 110 that recognizes the value chain from the information, and the recognition A value chain amplification factor calculation unit 120 that calculates the amplification factor of funds in the value chain for each value chain that has been recognized, a node rank calculation unit 130 that calculates the rank of the nodes that constitute each recognized value chain, and a recognition The value chain analysis apparatus 100 includes a main factor analysis unit 140 that analyzes a node having a high influence on the performance of other nodes constituting the value chain. With this feature, the value chain analyzer 100 can (1) recognize the value chain by calculation, (2) calculate the importance of the value chain, and (3) estimate the company that is the root cause of the business performance of the company. Can do.

(Feature 2) One of the features is that the value chain recognition unit 110 in the value chain analysis device of feature 1 can recognize the value chain in two directions, ie, the forward direction and the reverse direction of the money flow. With this feature, the value chain analyzer 100 can recognize (1) the value chain by calculation and (2) calculate the importance of the value chain in any direction of the value chain (supplier and seller). (3) It is possible to estimate the company that is the root cause of the company's performance.

(Feature 3) One of the features is that the value chain recognizing unit 110 in the value chain analyzing apparatus of the feature 1 regards the flow of funds between companies as an edge, and calculates a transition probability of the edge; A threshold setting unit 112 for setting a contribution rate threshold defined as a product of transition probabilities of the transitioning edges when transitioning via a plurality of edges, and the number of edges that transition when transitioning via a plurality of edges An edge stage number setting unit 113 for setting a maximum value, and an edge search unit 114 for recognizing a node group that can transition under the condition that the contribution rate is larger than the threshold and the number of edges is equal to or less than the maximum value as a value chain. It is to prepare. With this feature, even when the number of nodes constituting one value chain is large, the amount of calculation processing can be suppressed to a practical level. In addition, the size of the extracted value chain can be suppressed to a scale that can be determined by the user.

(Feature 4) One of the features is that the value chain amplification factor calculation unit 120 in the value chain analysis device of Feature 1 regards each company as a node, and the flow of funds flowing into the node and the flow of money flowing out of the node. A node amplification factor calculation unit 121 that calculates the amplification factor of each node on the basis of the transition probability between the nodes calculated from the information of the fund flow and the node amplification factor based on the node amplification factor A node amplification factor element calculation unit 122 for calculating a node amplification factor element that is a fund amplification factor of a certain node in the graph as viewed from the node, and searching the graph to obtain the node amplification factor for each node of the graph. A graph search unit 123 that calculates the amplification rate of the value chain using the fund amplification rate calculated by the element calculation unit 122; It is be provided. By using this calculation method, it is possible to calculate a value for calculating an index reflecting the value chain fund amplification effect without using a convergence calculation.

(Feature 5) One of the features is that, in the value chain analysis device of feature 4, the node rank calculation unit 130 includes a value of the certain node as seen from the starting node of all the value chains for all the value chains including the certain node. An amplification factor calculation unit 131 for calculating the sum of the node amplification factor elements is provided. By using this calculation method, it is possible to calculate a rank that gives importance of a node in the value chain without using a convergence calculation.

(Feature 6) One of the features is that, in the value chain analysis device of feature 4, the node rank calculation unit 130 determines whether a certain node of all value chains including a certain node is viewed from the starting node of all the value chains. An amplification amount calculation unit 132 for calculating a sum of products of the node amplification factor element and the funding flow of the origin node; By using this calculation method, it is possible to calculate a rank that gives importance of a node in the value chain without using a convergence calculation.

(Feature 7) One of the features is that the main factor analysis unit 140 in the value chain analysis device of Feature 1 sets an extraction condition setting unit 141 that sets the upper limit of the number of nodes as an extraction condition, and all value chains including a certain node. The gain aggregating unit 142 collects the starting nodes of the node amplification factor elements that match the extraction condition in descending order of the node amplification factor elements of the certain node as viewed from the starting nodes of all the value chains. It is to provide. This feature facilitates the estimation of companies that have a large impact on the performance of many companies in the overall graph.

(Feature 8) One of the features is that the main factor analysis unit 140 in the value chain analysis device of feature 1 sets an extraction condition setting unit 141 that sets the upper limit of the number of nodes as an extraction condition, and all value chains including a certain node. , The node amplification factor element of the certain node seen from the origin node of all the value chains and the fund flow of the origin node meet the extraction condition in the order of the node amplification factor element in descending order. And an amplification amount totaling unit 143 that collects the starting node of the node amplification amount element to be performed. This feature facilitates the estimation of companies that have a large impact on the performance of many companies in the overall graph.

(Feature 9) One of the features is that the extraction condition setting unit 141 in the value chain analyzer of feature 7 includes a lower limit of the ratio in the extraction condition, and the amplification factor totaling unit 142 is configured to determine the amplification factor element of a certain node. Normalization is performed so that the sum is 1, and the condition that the extraction condition is matched includes that the normalized node amplification factor is larger than the lower limit of the ratio.

(Feature 10) One of the features is that the extraction condition setting unit 141 in the value chain analysis device of feature 8 includes a lower limit of the ratio in the extraction condition, and the amplification amount totaling unit 143 includes the amplification amount element of a certain node. Normalization is performed so that the sum is 1, and the condition that the extraction condition is matched includes that the normalized node amplification amount is larger than the lower limit of the ratio.

(Feature 11) One of the features is that, in the value chain analyzer of feature 1, the interface screen displays a contribution threshold value setting unit 801 defined as a product of edge transition probabilities, and the maximum number of edge transition stages. It includes an input field for a setting unit 802, a setting unit 803 for the upper limit of the number of nodes to be extracted as the main factor, and a setting unit 804 for the lower limit of the ratio to be extracted as the main factor.

[Other embodiments]
The present invention is not limited to the configuration of the embodiment described above, and includes various modifications. For example, in the above-described embodiments, in order to explain the present invention in an easy-to-understand manner, some embodiments are described in detail, and it is not always necessary to include all the configurations described. It is also possible to add other configurations to the embodiment, replace some configurations of the embodiments with other configurations, or delete some configurations of the embodiments.

  Moreover, you may implement | achieve some or all of each structure, a function, a process part, a process means, etc. which were mentioned above as an integrated circuit or other hardware, for example. In the above description, the case where the CPU interprets and executes the program corresponding to each function (that is, the case where the program is realized by software) has been described. However, the function can also be realized by hardware. Information such as programs, tables, and files required to implement each function is stored in storage devices such as memory, hard disk, and SSD (Solid State Drive), and storage media such as IC cards, SD cards, and DVDs. Can be stored. Control lines and information lines indicate what is considered necessary for the description, and do not represent all control lines and information lines necessary for the product. In practice, it can be considered that almost all components are connected to each other.

DESCRIPTION OF SYMBOLS 100 ... Value chain analyzer 101 ... Data cooperation part 110 ... Value chain recognition part 111 ... Transition probability calculation part 112 ... Threshold setting part 113 ... Edge stage number setting part 114 ... Edge search part 120 ... Value chain gain calculation part 121 ... Node Amplification factor calculation unit 122 ... Node amplification factor element calculation unit 123 ... Graph search unit 130 ... Node rank calculation unit 131 ... Amplification factor calculation unit 132 ... Amplification amount calculation unit 140 ... Main factor analysis unit 141 ... Extraction condition setting unit 142 ... Amplification Rate totalization unit 143 ... Amplification amount totalization unit 150 ... Terminal 160 ... Fund flow database 800 ... Interface screen 801 ... Contribution rate threshold entry field 802 ... Maximum stage number entry field 803 ... Node number upper limit entry field 804 ... Lower limit value entry field 811 ... Node ID display area 812 ... Node rank display area 813 based on amplification factor ... Amplification amount That node rank display area 814 ... display area of the nodes in the main factor analysis score display area 815 ... amplification amount by the main cause analysis score display area 821 ... value chain information display area 822 ... value chain by amplification factor

Claims (12)

A data linkage unit that obtains the information from a storage system that accumulates information on the flow of funds between companies;
A value chain recognition unit for recognizing a value chain from the information;
A value chain amplification factor calculation unit that calculates the amplification factor of funds for the entire value chain for each recognized value chain;
The value chain recognition unit
A transition probability calculation unit that regards the fund flow between companies as an edge and calculates the transition probability of the edge from the fund flow;
A threshold value setting unit that sets a threshold value of a contribution rate defined as a product of transition probabilities of each edge on the transition path when the cash flow transits via a plurality of edges;
An edge stage number setting unit for setting a maximum value of the plurality of edges;
An edge search unit that recognizes, as the value chain, a set of nodes in which the contribution rate calculated for the fund flow is greater than the threshold and the number of edges is transitionable under the maximum value condition;
Have
The value chain amplification factor calculation unit
A node amplification factor calculation unit that regards each company as a node and calculates the amplification factor of each node based on the money flow flowing into the node and the money flow flowing out of the node;
Based on the transition probability between nodes calculated from the information of the money flow and the node amplification factor, the node amplification which is the money amplification factor of a node in the graph viewed from the origin node in the graph constituting a certain value chain A node amplification factor calculation part of the graph for calculating the rate element;
A graph search unit that searches the graph and calculates an amplification factor of funds of the entire value chain using the node amplification factor element calculated by the node amplification factor calculation unit for each node of the graph;
Have
The sum of the node amplification factor elements of the certain node as viewed from the starting node of all the value chains for all the value chains including a certain node, and / or all the value chains of all the value chains including the certain node. A node rank calculation unit having a function of calculating the sum of the node amplification amount elements of the certain node as viewed from the starting node as a rank of the nodes constituting the value chain, and a function of displaying a calculation result on the interface screen Have
Value chain analyzer characterized by that . In the value chain analyzer of Claim 1,
The value chain recognizing unit recognizes the value chain in two directions, a forward direction and a reverse direction of money flow. In the value chain analyzer of Claim 1,
An extraction condition setting unit that sets the upper limit of the number of nodes as an extraction condition through the interface screen, and a node amplification factor element of the certain node as viewed from the starting node of all the value chains in all value chains including the certain node And a main factor analysis unit having an amplification factor totaling unit that collects and displays on the interface screen the starting nodes of the node amplification factor elements that meet the extraction condition in descending order of the node amplification factor element. Value chain analyzer. In the value chain analyzer of Claim 3 ,
The extraction condition setting unit includes a lower limit of the ratio as the extraction condition, and the amplification factor totaling unit is normalized so that the sum of amplification factor elements of a certain node is 1, and as a condition for matching the extraction condition, Including a normalized node gain greater than the lower limit of the ratio,
Value chain analyzer characterized by that. In the value chain analyzer of Claim 1,
Through the interface screen, an extraction condition setting unit that sets the upper limit of the number of nodes as an extraction condition, and in all value chains including a certain node, a node amplification factor element of the certain node viewed from the starting node of all the value chains, Amplification amount totaling unit that collects starting nodes of node amplification amount elements that match the extraction conditions in order from the largest node amplification amount element and displays them on the interface screen for the node amplification amount element calculated from the product of the funds flow; A value chain analysis device further comprising a main factor analysis unit having In the value chain analyzer of Claim 5 ,
The extraction condition setting unit includes a lower limit of the ratio as the extraction condition, and the amplification amount totaling unit is normalized so that the sum of amplification amount elements of a certain node is 1, and as a condition for matching the extraction condition, The amount of node amplification after normalization is greater than the lower limit of the ratio,
Value chain analyzer characterized by that. In the value chain analyzer of Claim 1,
On the interface screen, a setting area for the contribution threshold defined as the product of edge transition probabilities, a setting area for the maximum number of edge transition stages, a setting area for the maximum number of nodes to be extracted as the main factor, and extraction as the main factor A value chain analyzer characterized by having a function of displaying any one of the lower limit setting areas or a combination thereof. In the calculator,
A process of acquiring the information from a storage system that accumulates information on the flow of funds between companies;
Processing for recognizing a value chain from the information;
For each recognized value chain, a process for calculating the amplification factor of the entire value chain,
The process of recognizing the value chain is
A process of calculating the transition probability of the edge from the fund flow, considering the fund flow between companies as an edge,
A process of setting a contribution rate threshold defined as a product of transition probabilities of each edge on the transition path when the fund flow transitions via a plurality of edges;
A process of setting a maximum value of the plurality of edges;
Processing for recognizing as a value chain a set of nodes that can be transitioned under the condition that the contribution rate calculated for the funds flow is greater than the threshold and the number of edges is not more than the maximum value;
Have
The process of calculating the amplification factor of the value chain is as follows:
A process of regarding each company as a node and calculating the amplification factor of each node based on the money flow flowing into the node and the money flow flowing out of the node;
Based on the transition probability between nodes calculated from the information of the money flow and the node amplification factor, the node amplification which is the money amplification factor of a node in the graph viewed from the origin node in the graph constituting a certain value chain Processing to calculate the rate factor;
A process of searching the graph and calculating an amplification factor of funds of the entire value chain using the node amplification factor element calculated for each node of the graph;
Have
The sum of the node amplification factor elements of the certain node as viewed from the starting node of all the value chains for all the value chains including a certain node, and / or all the value chains of all the value chains including the certain node. The sum of the node amplification amount elements of the certain node as seen from the starting node is calculated as the rank of the node constituting the value chain, and the calculation result is displayed on the interface screen.
A value chain analysis method characterized by this . The value chain analysis method according to claim 8 ,
The value chain analysis method characterized in that the process of recognizing the value chain recognizes the value chain in two directions, a forward direction and a reverse direction of money flow. The value chain analysis method according to claim 8 ,
Processing to set the upper limit of the number of nodes as an extraction condition through the interface screen,
In all value chains including a certain node, with respect to the node amplification factor element of the certain node viewed from the starting node of all the value chains, the starting point of the node amplification factor element that matches the extraction condition in descending order of the node amplification factor element The value chain analysis method further comprising: collecting nodes and displaying them on the interface screen. The value chain analysis method according to claim 8 ,
Processing to set the upper limit of the number of nodes as an extraction condition through the interface screen,
In all value chains including a certain node, a node amplification factor element is calculated with respect to a node amplification factor element calculated from a product of the node amplification factor element of the certain node and the money flow viewed from the starting node of all the value chains. A value chain analysis method, further comprising: an amplification amount totaling unit that collects starting nodes of node amplification amount elements that match the extraction condition in descending order and displays them on the interface screen. The value chain analysis method according to claim 8 ,
On the interface screen, a setting area for the contribution threshold defined as the product of edge transition probabilities, a setting area for the maximum number of edge transition stages, a setting area for the maximum number of nodes to be extracted as the main factor, and extraction as the main factor A value chain analysis method characterized by displaying either one of the lower limit setting areas or a combination thereof.

Images