JP7040786B2 - Influence assessment system - Google Patents

Description

The present invention relates to an influence evaluation system for evaluating the influence of an entity such as a company or a person on other entities.

It is being considered to quantify and evaluate how a company or person has an influence on a third party.

One of the indicators of influence on a third party is the stockholding relationship. A company or person may influence the decision-making of a stock company by exercising the voting rights of the shares it holds at a general meeting of shareholders. The "Shapley-Shubik power index" of Non-Patent Document 1 below is known as a typical method for evaluating the influence that an entity such as a company or a person has on another entity.

The evaluation method of Non-Patent Document 1 is used as an index showing the direct influence between entities, for example, an index showing the influence based on the stockholding relationship. FIG. 17 schematically shows this. FIG. 17 shows the holding relationship of the shares of companies A to D and the index value of the influence according to Non-Patent Document 1 based on the holding relationship. FIG. 17A shows that Company B and Company C own 30% of the shares of Company A, and Company D holds 40% of the shares of Company A. Ordinary decisions at a stock company's general meeting of shareholders are made by ordinary resolution. In the case of Japan, ordinary resolutions are passed at a general meeting of shareholders at a joint-stock company unless otherwise provided in the articles of incorporation. This is a proposal that was attended and was approved by a majority of the voting rights of the shareholders who attended. Even if all shareholders who can exercise voting rights are present, if they hold a majority of the shares that can exercise voting rights alone, they will have a substantial influence on the decision-making of the company. Can be said to be possible.

In FIG. 17 (a), neither company B to company D holds a majority stake by itself. In addition, any combination that will be the majority of the shares of Company A can be established by partnering with other companies. Then, regarding the influence on company A, companies B to D are equal, and each can be evaluated as 1/3 each. This is schematically shown in FIG. 17 (b). PI is an index of influence. FIG. 17 (c) shows the relationship between the index of influence as an equation.

As described above, the evaluation method of Non-Patent Document 1 is a method of evaluating the influence only on the direct relationship between the two entities. Further, in Non-Patent Document 2 to Non-Patent Document 4, as in Non-Patent Document 1, only the index of influence regarding only the direct relationship is considered.

In both Non-Patent Document 1 and Non-Patent Document 4, it is generally not known in advance whether or not to exercise voting rights at a general meeting of shareholders. Therefore, all shareholders who can exercise voting rights attend and give voting rights. The influence on the entity is calculated on the premise of exercising.

Japanese Unexamined Patent Publication No. 2011-8375

Shapley, Shubik, “Shapley-Shubik power index”, American Political Science Review, 48, p.787-792, 1954 Manfred J. Holler, Hannu Nurmi, "Power, Voting, and Voting Power: 30 Years After", Springer Link, 2013 J.M.Gallardo, N. Jimenes, A. Jimenez-Losada, "A Shapley measure of power in hierarchies", ELSEVIER Inc., 2016 Norkhairul Hafiz Bajuri, Shanti Chakravarty, Noor Hazarina Hashim, "ANALYSIS OF CORPORATE CONTROL: CAN THE VOTING POWER INDEX OUTSHINE SHAREHOLDING SIZE?", AAMJAF, Vol.10, No.1, p75-94,2014

However, in the real world, there is a controlling relationship such as a parent company, a subsidiary, and a grandchild company, and the structure that influences the substance of other companies is more complicated. For example, as shown in FIG. 18, there may be a plurality of levels of relationships between entities, including cross-shareholding relationships between companies, or there may be a cyclical relationship such as cross-shareholdings. In the case of FIG. 18, it can be evaluated that companies B to D directly hold shares and have an influence on company A, but for company C, companies B and E further hold shares. is doing. If Company B holds 80% of Company C and Company E holds 20% of the shares, Company B holds a majority stake in Company C, so Company C holds a majority stake. Only Company B can exert influence. Then, for Company A, the majority of the 30% shares held directly by Company B and the 30% shares held through Company C, where Company B can exert influence, will be a substantial majority. Only Company B can exert influence over Company A.

As described above, when there is a hierarchical structure or a cyclical relationship, it is not possible to evaluate by the conventional evaluation methods of Non-Patent Document 1 to Non-Patent Document 4.

Further, Patent Document 1 is a system for analyzing the cause of a management problem and the like, and is not a system for evaluating the influence between entities.

Therefore, in view of the above problems, the present inventors index the influence that an entity has on other entities even when the relationship between the entities is complicated, for example, even when there are multiple hierarchies or cyclical relationships. Invented an influence assessment system that can be used.

The first invention is in network information in which a label indicating an entity is a node, the direction of influence of the entity is an edge, and information indicating the influence of the entity on other entities is used as influence information. , The upper node of a certain node is specified by using the edge, and the update label specifying processing unit that specifies the label to be updated by using the influence information is specified for the label in the node having the upper node. The influence is shown by using the label update processing unit that updates the node with the label and stores the updated label as a history, and the history of the updated label in the node after executing the label update process a predetermined number of times or more. It is an influence evaluation system that has an index calculation processing unit that calculates an index.

By using the present invention, it is possible to index the influence on other entities even when the relationship between the entities is complicated, for example, when there are a plurality of layers or a cyclic relationship.

In the above-described invention, the update label specifying processing unit can be configured like an influence evaluation system that specifies the label to be updated in the node having the upper node by the label propagation method.

In the case of a network with multiple layers and cyclical relationships, the influence of the entity can be propagated by specifying the label to be updated using the label propagation method as in the present invention, thereby propagating other factors. The influence on the entity can be indexed.

In the above-mentioned invention, the update label specifying processing unit can be configured like an influence evaluation system that specifies the label to be updated in the node having the upper node by random sampling.

It is not always certain which label of the higher-level nodes affects a node. Therefore, the simulation can be performed by using random sampling as in the present invention.

In the above-described invention, the update label specifying processing unit calculates the influence information of the label according to a predetermined condition for the label in the upper node, and when the total exceeds a predetermined threshold value, the effect information is calculated. It can be configured like an impact assessment system, specifying the label as an updated label.

Propagation of influence to other entities can be realized by performing a process as in the present invention.

In the above-described invention, the label update processing unit does not store the updated label as a history until the label update processing is executed a predetermined number of times, or from the history used for the calculation processing in the exponential calculation processing unit. It can be configured like an influence assessment system that excludes and stores.

When the number of executions is small, the influence of the entity does not extend from the top layer to the bottom layer of the network. Calculating the exponent using the update label while it is unaffected may affect the accuracy of the exponent. Therefore, it is advisable to configure the update label up to a predetermined number of times not to be used for calculating the index indicating influence.

In the above-described invention, the influence evaluation system has a network information input reception processing unit that receives input of the network information, and when the influence information is information indicating holding of shares, the network information input. When the entity has its own stock, the reception processing unit can be configured like an influence evaluation system that performs correction processing by the own stock for the network information that has received the input.

If the influence information is a shareholding, the entity company may hold its own shares. And if the treasury stock is not taken into consideration, the processing accuracy may be affected. Therefore, it is preferable to perform correction processing in consideration of the company's stock.

In the above-described invention, when the entity has its own stock, the network information input reception processing unit newly provides the node and the edge for the network information that has received the input, and the new edge has the above-mentioned. The label indicating the substance can be configured like an influence evaluation system in which the new edge performs a correction process of attaching an edge from the new node to the node.

If the company's stock influences the decision-making of the company, the above-mentioned correction process is to perform the correction process by providing a new node and edge for the own label (node) for the company's stock. be.

In the above-described invention, when the entity has its own stock, the network information input reception processing unit receives the input about the influence information of the label of the node having an edge on the entity in the network information. It can be configured like an influence evaluation system that performs correction processing to calculate influence information after deducting.

As the above-mentioned correction process, if the treasury stock does not affect the decision-making of the company and the initial influence information includes the treasury stock, the treasury stock can be deducted to correct the influence information. good.

In the above-mentioned invention, the exponential calculation processing unit can be configured like an influence evaluation system that calculates by performing an operation for weighting the scale of the entity when calculating the exponent indicating the influence. ..

The size of the entity ranges from extremely large to extremely small. Therefore, if the index is calculated simply based on its influence on other entities, the index will be higher as the relationship with other entities increases, regardless of the scale. Therefore, by weighting in consideration of the scale of the entity, the influence on the large-scale entity can be greatly evaluated. Therefore, it is possible to calculate an index of influence that is closer to the real world.

In the above invention, the influence evaluation system calculates the difference between the index showing the influence before the change of the entity in the network and the index showing the influence after the change of the entity in the network. It can be configured like a force assessment system.

By calculating the difference between the indexes indicating the influence before and after the change of the substance in the network in this way, it is possible to evaluate the act related to the change of the substance, for example, the decision of M & A. It is also possible to evaluate the butterfly effect on a third party.

The first invention can be realized by loading and executing the program of the present invention in a computer. That is, the computer is labeled with a label indicating the entity to the node, the direction of the influence of the entity is attached at the edge, and the information indicating the influence that the entity has on other entities is used as the influence information. In the network information, the upper node of a certain node is specified by using the edge, and the label in the node having the upper node is specified by the update label specifying processing unit, which specifies the label to be updated by using the influence information. The label update processing unit that updates the node with the label that has been updated and stores the updated label as a history. After executing the label update process a predetermined number of times or more, the influence is shown using the history of the updated label in the node. It is an influence evaluation program that functions as an index calculation processing unit that calculates the index.

By using the influence evaluation system of the present invention, it is possible to index the influence of an entity on other entities even when the relationship between the entities is complicated.

This is an example of a block diagram showing a system configuration of the influence evaluation system of the present invention. It is a figure which shows an example of the hardware composition of the computer which realizes the influence evaluation system of this invention. It is a flowchart which shows an example of the whole processing process of the influence evaluation system of this invention. It is a flowchart which shows an example of the processing process of the specific process of the update label in the node of the influence evaluation system of this invention. It is a figure which shows the example of the network information which expressed the relationship of the influence between the entities schematically. It is a figure which shows an example of the network information which the random number is attached to the label of each node schematically. It is a figure which shows the example of the network information of the state which specifies the label to be updated schematically. It is a figure which shows the example of the network information in the state where the label of a node is updated schematically. It is a figure which shows an example of the calculation process of a network influence index schematically. FIG. 5 is a diagram schematically showing an example of network information in a state where a random number is added to the label of each node in the first embodiment. FIG. 5 is a diagram schematically showing an example of network information in a state where the label of a node is updated in the first embodiment. In the second embodiment, it is a figure schematically showing an example of the process of adding the company stock E as a new entity as network information and correcting the network information. In the second embodiment, it is a figure schematically showing an example of the process of correcting the network information by deducting the treasury stock E as the network information. It is a figure which shows typically an example of the case where weighting is used in Example 3. FIG. FIG. 5 is a diagram schematically showing an example of network information in an initial state (before trading) in the fifth embodiment. FIG. 5 is a diagram schematically showing an example of network information after buying and selling in Example 5. It is a figure which shows the calculation of the index value based on the "Shapley-Shubik power index" which is the conventional method of evaluating influence schematically. It is a figure which shows typically an example of the structure which influences decision making in an actual model.

In the influence evaluation system 1 of the present invention, the relationship between the entities is networked, and the process is executed by using the label propagation method in order to numerically calculate it.

In the following description of the present specification, a case where the relationship between entities is networked as a stockholding relationship and the influence of one entity on another entity is evaluated in the network will be described. A typical example of an entity in this case is a company and / or a natural person, but the entity is not limited to these, and the entity is another entity such as a natural person, a company, or a group or organization, a fund, or a union that is a group of people. Anything that can constitute an entity that can be involved in the decision-making of an entity may be used.

The computer has an arithmetic unit 70 such as a CPU that executes arithmetic processing of a program, a storage device 71 such as a RAM or a hard disk for storing information, a display device 72 such as a display, and an input device 73 for inputting information. It has a communication device 74 that communicates various information such as processing results of the device 70 and information stored in the storage device 71. When the computer is provided with a touch panel display, the display device 72 and the input device 73 may be integrally configured. Touch panel displays are often used in portable communication terminals such as mobile phones, smartphones, and tablet computers, but are not limited to these.

The touch panel display is a device in which the functions of the display device 72 and the input device 73 are integrated in that input can be performed directly on the display with a predetermined input device (such as a pen for a touch panel) or a finger.

The influence evaluation system 1 may be realized by one computer, but its function may be realized by a plurality of computers. The computer in this case may be, for example, a cloud server.

Further, each processing unit in the influence evaluation system 1 of the present invention may form the same area physically or substantially, only the functions thereof are logically distinguished.

The influence evaluation system 1 has a network information input reception processing unit 10, a network information storage unit 11, an update label specifying processing unit 12, a label update processing unit 13, and an index calculation processing unit 14.

The network information input reception processing unit 10 receives input of information (network information) regarding a network structure expressing the relationship of influence between entities for which influence is to be evaluated using the present invention. As shown in Fig. 5 (a), the relationship of influence between entities is expressed by expressing the entity as a node (node) and the relationship as an edge (branch), and this edge is a directed graph in the direction of influence. Be expressed. When showing the influence of a stock holding relationship, a company or a natural person becomes a node, and a directed graph is attached as an edge from the node that is a shareholder to the node of the issuing company of the stock that it holds. In addition, as information showing the influence on other entities (influence information), the shareholding ratio is associated with the edge. The example of FIG. 5A shows the shareholding relationship in a network structure, in which Company B and Company C hold 30% each, Company E holds 40%, and Company C shares. It shows that Company B owns 50% and Company D owns 50%. The influence includes various influences on other entities, including the influence of risks.

FIG. 5B is a diagram schematically showing an example of embodying this as network information. In FIG. 5B, node identification information (node identification information) is attached to each node like N1 and N2, and the identification information (label) of an entity such as a company or a natural person in the node is the node identification information. Is assigned as the value of (including numerical values, alphanumeric characters, symbols, etc.). Further, each edge is expressed as E (X, Y), and is indicated by a node whose starting point is X and a node whose ending point is Y. Also, the shareholding ratio is substituted as the value of these edges. The method of expressing the network structure is not limited to this, and may be another method. In the case of FIG. 5 (b), for each node, N1 = A, N2 = E, N3 = B, N4 = C, N5 = D, E (N2, N1) = 40%, E (N3, N1). = 30%, E (N4, N1) = 30%, E (N3, N4) = 50%, E (N5, N4) = 50%.

The network information input reception processing unit 10 receives the input of the network information in the initial state, and stores the network information in the initial state in the network information storage unit 11 described later. The network information may be in a state in which the network structure shown in FIG. 5 or the like is appropriately embodied in processing by a computer.

The network information storage unit 11 stores the network information received by the network information input reception processing unit 10. As the network information, as the network information in the initial state, the node and its value (label), the influence information between the edge and the node, for example, the stockholding ratio, etc. are stored in association with each other. In addition, the values of the node and its label are stored in association with each other as a history.

The update label specifying processing unit 12 specifies the label of the propagating node in each node by the label propagating method. This identifies which of the entities in the node directly above the node affected the entity in the node. In reality, which entity is affected is executed by repeating, for example, random sampling.

Specifically, the update label specifying processing unit 12 executes the following processing, but is not limited to the following processing as long as the propagating label can be specified.

A random number is given to the label of the upper node that is directly connected to a certain node. Then, the labels of the upper nodes are sorted in ascending order based on the given random numbers. Then, in the order of the sorted labels, the influence information between the nodes, for example, the stockholding ratio is added. When the added value exceeds a predetermined threshold value, for example, 50%, the label exceeding the threshold value is specified as a label to be updated (propagated) (update label). It should be noted that the random numbers may not be sorted in ascending order, for example, in descending order, or in any other method, the order in which the influence information of the labels of each node to be added, which will be described later, can be randomly specified. Any method may be used. In addition, the influence information between nodes is added in the sorted order, and the influence information is calculated by any calculation method in the order in which the predetermined conditions are satisfied, so that the label exceeding the threshold value is specified. May be good.

The update label specifying processing unit 12 executes the update label specifying process for all the nodes having the nodes higher than itself.

For example, assume that the network information is shown in FIG. At this time, the label E of the node N2, the label B of the node N3, and the label C of the node N4 are above the label A of the node N1, and the random numbers are "0.52", "0.63", and ". "0.17" is given. In addition, another node having a node higher than itself is the label C of the node N4. Therefore, "0.58" and "0.21" are given as random numbers to the label B of the node N3 and the label D of the node N5, which are higher than the label C of the node N4, respectively. FIG. 6 schematically shows this state.

Then, the labels B, C, and E of the upper nodes of the node A are sorted in ascending order based on the random numbers assigned to them, and the labels C, E, and B are arranged in this order. Further, if the labels B and D of the upper nodes of the node C are sorted in ascending order in the same manner, the labels D and B are arranged in this order.

When the influence information is added in order for the label B, the label C, and the label E of the upper node of the label A, the label C is 30% and the label E is 40%, so that the threshold value exceeds 50% at this point. , Specify label E as the label to update node N1. Further, when the influence information is added in order for the upper label B and the label D of the label C, the label D is 50% and the label B is 50%, so that the threshold value exceeds 50% at this point. Label B is specified as the label for updating N4. FIG. 7 schematically shows this.

The label update processing unit 13 updates the label of each node with the update label specified by the update label specifying processing unit 12.

Assuming that the update label specified by the update label specifying processing unit 12 is FIG. 7, the label of the node is updated as shown in FIG. Then, the update label specifying processing unit 12 stores the updated label as a history. That is, the network information storage unit 11 stores the history of updating the label E for the node N1 and the label B for the node N4.

Then, in the state of this network (state of FIG. 8), the processing in the update label specifying processing unit 12 and the label update processing unit 13 is executed again. In this way, the processes of S110 to S150 are executed a predetermined number of times, for example, 10,000 times.

The index calculation processing unit 14 calculates the network influence index (NPI) according to the present invention based on the history of the label of the node stored in the network information storage unit 11. In the case of the network of FIG. 5, since the node N1 and the node N4 have higher-level nodes, the network influence index (NPI) is calculated in the node N1 and the node N4. For example, suppose that as a result of executing the processing 10,000 times, the history of the label at the node N1 is shown in FIG. 9 (a), and the history of the label at the node N4 is shown in FIG. 9 (b). At this time, the index calculation processing unit 14 calculates the network influence index as NPI (B → C) = 0.50 and NPI (D → C) = 0.50 (when the number of significant digits is the last two digits). ). Further, it is calculated as NPI (B → A) = 0.67, NPI (D → A) = 0.17, and NPI (E → A) = 0.16. This is shown in FIG. 9 (c). Then, when the network influence index is added for each entity, the entity A = 0, the entity B = 1.17 (= 0.50 + 0.67), the entity C = 0, and the entity D = 0.67 (0.50 + 0). .17), the network influence index for each entity can be calculated with the entity E = 0.16. This is shown in FIG. 9 (d).

Next, an example of the processing process of the influence evaluation system 1 of the present invention will be described with reference to the flowcharts of FIGS. 3 and 4. In this embodiment, it is assumed that the influence of the stock holding relationship on the decision-making is evaluated, and the network information in the initial state is shown in FIG. The threshold value when the influence information is added is set to 50%.

The operator inputs the network information in the initial state shown in FIG. 5 to be processed, and the network information input reception processing unit 10 accepts it (S100). The network information input reception processing unit 10 stores the received network information in the initial state in the network information storage unit 11.

Then, the update label specifying processing unit 12 executes the update label specifying process for all the nodes having the nodes higher than itself (S110).

First, the update label specifying processing unit 12 generates random numbers for the label E of the node N2 of the node above the node N1, the label B of the node N3, and the label C of the node N4, and for example, "0. 52 ”,“ 0.63 ”, and“ 0.17 ”are assigned, and random numbers are generated for the label B of the node N3 above the node N4 and the label D of the node N5, respectively. For example,“ 0. 58 ”and“ 0.21 ”are given (S200) (FIG. 6).

Next, the update label specifying processing unit 12 sorts the labels B, C, and E of the upper nodes N2 to N4 of the node N1 in ascending order based on the random numbers assigned to them (S210), and labels C and labels. The order is E and label B. Further, the labels B and D of the upper nodes of the node N4 are similarly sorted in ascending order (S210), and the order is label D and label B.

Then, as shown in FIG. 7, the influence information is added in the sorted order for the labels B, C, and E of the upper nodes N2 to N4 of the node N1 (S220), and the threshold value is obtained when the label E is added. Since it exceeds 50%, label E is specified as the update label of node N1 (S230). Further, the labels B and D of the upper nodes N3 and N5 of the node N4 are added based on the influence information in the sorted order (S220), and when the label B is added, the threshold value exceeds 50%, so that the node N4 Label B is specified as the update label of (S230).

As described above, when the update label specifying processing unit 12 executes the process of specifying the update label for all the nodes having higher-level nodes, the label update processing unit 13 updates the label of the node with the update label ( S120), the network information storage unit 11 stores the updated label as a history (S130). That is, the network information storage unit 11 stores that the node N1 has been updated to the label E and the node N4 has been updated to the label B. The network information in this state is shown in FIG.

Assuming that the iterative process is performed 10,000 times as a predetermined number of times, the process is executed for the first time, so the process after S110 is repeated again (S140). That is, in the state of FIG. 8, the processing after S110 is executed again.

That is, the update label specifying processing unit 12 executes the update label specifying process for all the nodes having nodes higher than itself (S110), as described above.

The update label specifying processing unit 12 generates random numbers for the label E of the node N2, the label B of the node N3, and the label B of the node N4, which are higher than the node N1, and the random number is "0.81". , "0.15" is given, and random numbers are generated for the label B of the node N3 above the node N4 and the label D of the node N5, respectively, and the random numbers are "0.08" and "0.98". (S200). The network information in this state is shown in FIG. Here, node N3 and node N4 are both labeled B to indicate that they are the same entity. Therefore, "0.15" is given to node N3 and node N4 as the same random number so that they can be processed as one entity in the following ascending order.

Next, the update label specifying processing unit 12 sorts the labels E and B of the upper nodes N2 to N4 of the node N1 in ascending order based on the random numbers assigned to them (S210), and the order of the label B and the label E. And. Further, the labels B and D of the upper nodes of the node N4 are similarly sorted in ascending order (S210), and the order is label B and label D.

Labels B and E of the upper nodes N2 to N4 of the node N1 are added based on the influence information in the sorted order (S220), and the influence information of the edge E (N3, N1) in the node N3 is 30%. Since the influence information of the edge E (N4, N1) in the node N4 is 30%, the influence information becomes 60% when the label B is added, and the threshold value exceeds 50% at this point, so that the node N1 is updated. Label B is specified as the label (S230). Further, the labels B and D of the upper nodes N3 and N5 of the node N4 are added based on the influence information in the sorted order (S220), and when the label D is added, the threshold value exceeds 50%, so that the node N4 Label D is specified as the update label of (S230).

As described above, when the update label specifying processing unit 12 executes the process of specifying the update label for all the nodes having higher-level nodes, the label update processing unit 13 updates the label of the node with the update label ( S150), the network information storage unit 11 stores the updated label as a history (S160). That is, the network information storage unit 11 stores that the node N1 has been updated to the label B and the node N4 has been updated to the label D. The network information in this state is shown in FIG.

Assuming that the iterative process is performed 10,000 times as a predetermined number of times, the process is executed for the second time, so the process after S110 is repeated again (S140). That is, in the state of FIG. 11, the processes from S110 to S130 are repeated.

In this way, after repeating the processes from S110 to S130 a predetermined number of times, the exponential calculation processing unit 14 uses the node as its label based on the history of the node's label stored in the network information storage unit 11. The network impact on each node by dividing the number of times by a predetermined value, for example, the number of iteration executions, the number of histories, or the number of histories minus the influence from the top layer to the bottom layer of the network described later. Calculate the force index (NPI). Then, by summing the network influence index (NPI) for each entity, the network influence index for each entity is calculated (S150).

For example, if the result of executing the processing 10,000 times is FIGS. 9A and 9B, and the number of significant digits is the last two digits, the index calculation processing unit 14 sets the network influence index as the network influence index. NPI (B → C) = 0.50 (= 4957/10000) and NPI (D → C) = 0.50 (= 5043/10000) are calculated. Further, NPI (B → A) = 0.67 (= 6661/10000), NPI (D → A) = 0.17 (= 1697/10000), NPI (E → A) = 0.16 (= 1642 /). 10000) is calculated. When the network influence index is added for each entity, the entity A = 0, the entity B = 1.17 (= 0.50 + 0.67), the entity C = 0, and the entity D = 0.67 (0.50 + 0). .17), the network influence index for each entity can be calculated with the entity E = 0.16.

In the above-mentioned processing of S110 to S130, the label in each node may be returned to the initial state based on the network information in the initial state at an arbitrary timing. This makes it possible to avoid the occurrence of the local minimum problem due to the initial value dependency.

To return the label of each node to the initial state, for example, a random number p is generated before the processing of S110, and if the random number p satisfies the condition regarding the initialization reference value, the label of each node is returned to the initial state. You may try to return to. For example, when the random number p is 0 ≦ p ≦ 1, if 0 ≦ p ≦ 0.005 is set as the condition of the initialization reference value for returning to the initial state, and when the random number p is 0 ≦ p ≦ 0.005, the initial state. Refer to the network information of, return each node to the label in the initial state, and specify it as the update label. On the other hand, when the random number p> 0.005, the process after S200 may be executed to specify the update label.

The initialization reference value can be set arbitrarily.

In addition, in network information, it takes a predetermined number of trials for the influence of an entity to propagate from the top layer to the bottom layer of the network. Therefore, although the processes of S110 to S130 are executed from the first time to a predetermined number of times, for example, about 19 times, the updated label may not be stored as a history. Alternatively, even if it is stored, it may be excluded from the history of the label used for calculating the network influence index. The predetermined number of times can be set sequentially based on the size of the number of layers of the network.

In Example 1, the shareholding ratio is used as it is as the influence information, but the company's own stock may be considered. That is, there are cases where the issuing company holds shares that can be exercised with voting rights (so-called treasury stock). In Japan, treasury stock cannot exercise voting rights and does not affect decision-making, but if treasury stock is included in the shareholding ratio that shows influence information, that is taken into consideration. You may. For example, in the network information of FIG. 5, if the entity F is 50% of the issued shares of the issued shares and the remaining 50% is the treasury stock, the treasury stock is used as the network information. The network information may be corrected so that E is newly added as an entity. FIG. 12 schematically shows this. That is, originally, as shown in FIG. 12A, the network information is provided with node, edge, and influence information, but the upper node of the entity E is only the entity F, and the total of the influence information is also 100%. It is not. Therefore, as shown in FIG. 12 (b), a node N7 indicating the company's stock was added, and an edge E (N7, N2) from the node N7 to the node N2 (the company's node) was added to network the network information. The information input reception processing unit 10 may receive the input and execute the processing of the first embodiment.

In addition to the above, the treasury stock may be processed as follows. That is, with regard to the shareholding ratio of the shares showing the influence information, the influence information may be calculated by the shareholding ratio of the issued shares excluding the treasury stock that can exercise voting rights.

That is, the shareholding ratio of a certain entity is calculated by dividing the number of shares held by the entity by the number of issued shares that can exercise voting rights minus the number of own shares, and holding the shares of the entity. The ratio may be calculated. For example, in the case of FIG. 12 (a) described above, of the issued shares, if the entity F is 50% of the shares that can exercise the voting rights of the entity E and the remaining 50% is the company's shares, the shares of the entity F are held. The ratio may be calculated as 100% (= number of shares of E held by F / (total number of issued shares-number of own shares) x 100). FIG. 13 shows an example of the network information processed in this way. In FIG. 13, the correction of the shareholding ratio is calculated using the number of shares, but it may be calculated using the shareholding ratio.

In addition, companies that issue class shares that influence the company's decisions, such as so-called golden shares (shares that can exercise the right to refuse specific resolutions at a general meeting of shareholders), shares with multiple voting rights, and non-voting shares. In the case of, influence information is not determined solely based on the shareholding ratio. Therefore, after considering the class stock and correcting the influence information, the network information may be set as the initial state and input to the network information input reception processing unit 10.

In the first and second embodiments, when the update label specifying processing unit 12 specifies the label to be updated, the random numbers attached to the labels of each node are sorted, and the influence information of each label is added in that order. Therefore, the label that exceeds a predetermined threshold value is specified as a label to be updated, but this threshold value may be set to another threshold value instead of 50%. For example, it may be set to 2/3 (67%, 66.7%, etc.) and 3/4 (75%).

In the case of Japan, important matters of the company such as transfer of an important part of the company's business, amendment of the articles of incorporation, reduction of the amount of capital, etc. are not ordinary resolutions but special resolutions, special resolutions, etc. It is also possible to set the above threshold value as a standard of influence on the company based on it.

When the influence evaluation system 1 of the present invention is used, the influence is evaluated regardless of the scale of the entity. Therefore, the network influence index may be calculated higher for an entity that holds a large number of shares of a very small company than if it holds a small number of shares of a very large company.

Therefore, when calculating the network influence index in the index calculation processing unit 14, even if the individual influence index for each entity is added after weighting with information indicating the size of the entity such as net assets and market capitalization. good. For example, when the index calculation processing unit 14 calculates the network influence index NPI for each node and then calculates the network influence index NPI for each entity, the NPI for a company with a small entity is added as it is. The NPI for a company with a large substance may be multiplied by a predetermined coefficient as a weight and the value may be added.

FIG. 14 schematically shows this. For example, as shown in FIG. 9 (c) of Example 1, it is a network influence index for each node, and since the scale of the entity C is larger than that of other entities, the influence on the entity C is doubled and weighted. If you do. FIG. 14A schematically shows this.

In this case, the network influence index for each entity is entity A = 0, entity B = 1.67 (= 0.50 × 2 + 0.67), entity C = 0, entity D = 1.17 (0.50 ×). The network influence index for each entity can be calculated with 2 + 0.17) and entity E = 0.16. FIG. 14 (b) schematically shows this. By weighting according to the size of the entity in this way, the network influence index for large-scale entities can be evaluated with a larger numerical value, and not only the number of entities that have influence but also the scale is taken into consideration. Since the network influence index can be calculated, it is possible to calculate an index that is more in line with the actual influence.

This weighting can be set as appropriate according to the scale of the entity itself, the scale of the country to which the entity belongs, and the weighting can be set in a plurality of stages according to the scale.

Furthermore, weighting can be performed according to the purpose of calculating the network influence index. For example, when making an evaluation based on ESG (Environment, Social, Governance), a predetermined standard is whether each entity in each of the above-mentioned examples is considered for the environment, social responsibility, and corporate governance. Weighting is increased for entities that are not considered for ESG, and decreased for entities that are given consideration for ESG, and the network influence index for entities (company) that are at risk for ESG. Can be calculated.

As a result, for example, it is possible to avoid unexpected debt generation and reputation risk for M & A and investee companies.

Of course, the weighting may be increased for the entity in which consideration is given to ESG, and the weighting may be decreased for the entity in which consideration is not given to ESG, so that the positiveness toward ESG may be evaluated.

In addition, each of the above-mentioned examples can be used for evaluation of corporate value at the time of M & A or the like.

For example, it is assumed that the network information in the initial state is shown in FIG. Such network information is read into the influence evaluation system 1 and the network influence index for each entity is calculated. As a result, it is assumed that the network influence index for each entity is entity A = 1.33, entity B = 0.33, and entity C = 0.33.

At this time, when the entity B purchases all the shares of the entity α from the entity A (that is, B purchases the company α from A), the present invention can be used to evaluate the corporate value of the entity α.

First, as shown in FIG. 16, the network information after the entity B purchases the entity α is read, and the network influence index for each entity is calculated. As a result, it is assumed that the network influence index for each entity is entity A = 0, entity B = 2, and entity C = 0.

Then, the value of the entity associated with the sale can be calculated by the difference between the network influence index before and after the sale. In the case of FIGS. 15 and 16, the value of the entity α associated with the sale is 1.33 (= | 0-1.33 |) for the entity A and 1.67 (= | 2-0.33) for the entity B. |), It can be evaluated as 0.33 (= | 0-0.33 |) for the entity C.

Therefore, if the entity α is bought and sold at an amount corresponding to the network influence index between 1.33 and 1.66, it can be seen that the price is appropriate for the entity A and the entity B. Furthermore, the influence (butterfly effect) on a third party (entity C) before and after the sale can be calculated.

That is, as in the present embodiment, by using the influence evaluation system 1 of the present invention, the network influence index before and after the change of the substance is calculated, and the difference is compared to change the substance. In addition to the value of the accompanying entity, the impact on third parties (butterfly effect) can also be evaluated.

The same as in each of the above embodiments can be applied to the calculation of the stockholding relationship, that is, the influence on decision making in various fields other than the influence on the decision making of the company. For example, it can be applied to the influence on decision making in the country. In this case, when some kind of legislation is made in the Diet, the number of seats of each political party in the Diet indicates the influence information of the political party (entity) in the Diet. And, for example, the support group (entity) of the political party has influence on the political party, and the influence is the amount of political contribution, the number of members of the group, the number of members of the Diet from that group, etc. Can be quantified and set as influence information. Furthermore, as the influence on the support group, it can be set as influence information by evaluating the number of people constituting the faction (entity) within the support group, and the influence can be evaluated.

In this way, the connection between entities such as people and companies in society can be expressed by a network structure, and by quantifying the influence of each entity with a predetermined index and setting it as influence information, one entity becomes another entity. It is possible to evaluate what kind of influence it has on.

By using the influence evaluation system 1 of the present invention, it is possible to index the influence of an entity on other entities even when the relationship between the entities is complicated.

1: Impact evaluation system 10: Network information input reception processing unit 11: Network information storage unit 12: Update label specification processing unit 13: Label update processing unit 14: Index calculation processing unit 70: Arithmetic unit 71: Storage device 72: Display Device 73: Input device 74: Communication device

Claims (11)

In network information where the label indicating an entity is on the node, the direction of influence of the entity is on the edge, and the information indicating the influence of the entity on other entities is the influence information, the upper level of a node. An update label specifying processing unit that identifies a node using the edge and specifies a label to be updated using the influence information for a label in a node having the upper node.
A label update processing unit that updates the node with the specified label and stores the updated label as a history.
An index calculation processing unit that calculates an index indicating influence using the history of updated labels in the node after executing the label update processing a predetermined number of times or more.
An influence assessment system characterized by having. The update label specifying processing unit is
The label to be updated in the node having the upper node is specified by the label propagation method.
The influence evaluation system according to claim 1. The update label specifying processing unit is
The label to be updated in the node having the upper node is specified by random sampling.
The influence assessment system according to claim 1 or 2, wherein the influence assessment system is characterized in that. The update label specifying processing unit is
For the label in the upper node, the influence information of the label is calculated according to a predetermined condition, and when the total exceeds a predetermined threshold value, the label is specified as a label to be updated.
The influence evaluation system according to any one of claims 1 to 3, wherein the influence evaluation system is characterized in that. The label update processing unit is
The updated label is not stored as a history or is excluded from the history used for the calculation process in the index calculation processing unit and stored until the label update process is executed a predetermined number of times.
The influence evaluation system according to any one of claims 1 to 4, wherein the influence evaluation system is characterized in that. The influence evaluation system is
It has a network information input reception processing unit that accepts the input of the network information.
If the influence information is information indicating the holding of shares,
The network information input reception processing unit is
When the entity has its own stock, the network information that received the input is corrected by the own stock.
The influence evaluation system according to any one of claims 1 to 5, wherein the influence evaluation system is characterized in that. The network information input reception processing unit is
When the entity has its own stock, the node and the edge are newly provided for the network information that has received the input.
A label indicating the substance is attached to the new edge, and the new edge is subjected to a correction process of attaching an edge from the new node to the node.
The influence evaluation system according to claim 6, wherein the influence evaluation system is characterized in that. The network information input reception processing unit is
When the entity has its own stock, correction processing is performed to calculate the influence information obtained by deducting the treasury stock from the influence information of the label of the node having an edge to the entity in the network information that received the input.
The influence evaluation system according to claim 6, wherein the influence evaluation system is characterized in that. The exponential calculation processing unit
When calculating the index indicating the influence, the calculation is performed by weighting the scale of the entity.
The influence evaluation system according to any one of claims 1 to 8, wherein the influence evaluation system is characterized in that. The influence evaluation system is
The difference between the index showing the influence before the change of the entity in the network and the index showing the influence after the change of the entity in the network is calculated.
The influence evaluation system according to any one of claims 1 to 9, wherein the influence evaluation system is characterized in that. Computer,
In network information, a label indicating an entity is attached to a node, the direction of the influence of the entity is attached at an edge, and information indicating the influence of the entity on other entities is used as influence information. Update label specifying processing unit, which identifies the upper node of the node using the edge, and specifies the label to be updated using the influence information for the label in the node having the upper node.
A label update processing unit that updates the node with the specified label and stores the updated label as a history.
An index calculation processing unit that calculates an index indicating influence using the history of updated labels in the node after executing the label update processing a predetermined number of times or more.
An influence assessment program characterized by functioning as.

Images