Decision support device, decision support method, and decision support program

Abstract

This decision support device executes: acquisition processing for acquiring a feature amount related to a measure to be taken by a group, a feature amount related to an action of the group, and a feature amount related to the group; first prediction processing for inputting the feature amount related to the measure, the feature amount related to the action, and the feature amount related to the group to a machine learning model, the feature amounts being acquired by the acquisition processing, and outputting a first prediction result of prediction of the action in cooperation among members constituting the group, the machine learning model predicting whether to perform the action in cooperation among members constituting the group when the feature amount related to the action and the feature amount related to the group are input depending upon whether the measure has been implemented; and outputting processing for outputting the first prediction result obtained by the first prediction processing.

Description

Technical Field

[0001] The present invention relates to a decision-making support device for processing information, a decision-making support method, and a decision-making support program.

Background Art

[0002] The following Patent Document 1 discloses an organizational improvement activity support device that improves the continuity of organizational improvement activities without the intervention of experts on-site. This organizational improvement activity support device includes a measure suitability acquisition means for acquiring a measure suitability that quantitatively indicates the degree of suitability of the measures implemented by the target organization for the characteristics and improvement themes of the target organization, a measure agreement degree acquisition means for acquiring a measure agreement degree that quantitatively indicates the degree of agreement of the members of the target organization with respect to the measures, a measure activity degree acquisition means for acquiring a measure activity degree that quantitatively indicates the degree of implementation of the measures in the target organization, an effect point calculation means for calculating an effect point that quantitatively indicates the degree of effect of the organizational improvement activities in the target organization based on the measure suitability, the measure agreement degree, and the measure activity degree, and an information presentation means for presenting at least information regarding the activity status in the self-organization to the facilitator and / or members of the target organization based on the effect point.

Prior Art Documents

Patent Documents

[0003]

Patent Document 1

Summary of the Invention

Problems to be Solved by the Invention

[0004] The measures presented based on the characteristics of the organization are not specific measures, and manpower is required to consider specific measures.

[0005] An object of the present invention is to automatically present the cooperation rate among members constituting a group when measures for promoting actions are implemented for the group.

Means for Solving the Problems

[0006] A decision support device representing the first aspect of the invention disclosed in this application is a decision support device having a processor that executes a program and a storage device that stores the program, wherein the processor performs an acquisition process to acquire features relating to measures that a group should take, features relating to the actions of the group, and features relating to the group, and when the features relating to the actions and features relating to the group are input depending on whether the measures are implemented, the processor inputs the features relating to the measures, features relating to the actions, and features relating to the group acquired by the acquisition process into a machine learning model that predicts whether the members constituting the group will cooperate to take the measures Do you The system is characterized by performing a first prediction process that outputs a first prediction result, and an output process that outputs the first prediction result obtained by the first prediction process.

[0007] A decision support device comprising a second aspect of the invention disclosed herein is a decision support device having a processor for executing a program and a storage device for storing the program, and capable of communicating with a plurality of computers, wherein the decision support device holds master information in which recommended actions for a group are ranked according to the characteristics of the group, each of the plurality of computers holds slave information in which recommended actions for a group are ranked according to the characteristics of the group, and the processor has specific characteristics of the group, specific recommended actions for the group, and first answers to questions made to each member constituting the group before the implementation of measures to be taken by the group. The method is characterized by the following steps: receiving from each of the plurality of computers the implementation results, which include a first score for the group based on the group and a second score for the group based on the second answers to the questions given to each of the members after the implementation of the measures; calculating a correction value for the ranking of specific recommended actions for the group based on specific characteristics of the group, based on the first score and the second score received by the receiving process; correcting the ranking of the specific recommended actions based on the correction value calculated by the calculation process; and transmitting the corrected results from the correcting process to the plurality of computers.

[0008] A decision support device comprising a third aspect of the invention disclosed herein is a decision support device having a processor for executing a program and a storage device for storing the program, and capable of communicating with a plurality of computers, wherein the decision support device holds master information in which recommended actions for a group are ranked according to the characteristics of the group, and each of the plurality of computers holds a machine learning master model that predicts whether the members constituting the group will cooperate to take the action when a feature quantity relating to the group's actions and a feature quantity relating to the group are input, depending on whether or not the measures to be taken by the group are implemented, and each of the plurality of computers, depending on whether or not the measures to be taken are implemented, The processor is characterized by having a machine learning slave model that, upon input of features relating to an action and features relating to the group, predicts whether the members constituting the group will cooperate in performing the action, and by performing a receiving process that receives information regarding the learning of the machine learning master model in the machine learning slave model from each of the plurality of computers, a learning process that learns the machine learning master model based on the learning information received from each of the plurality of computers by the receiving process, and a transmission process that transmits the learning parameters of the machine learning master model learned by the learning process to the plurality of computers. [Effects of the Invention]

[0009] According to a typical embodiment of the present invention, when measures to promote behavior are implemented in a group, the cooperation rate among the members constituting the group can be automatically displayed. Other issues, configurations, and effects will be clarified by the following description of the embodiments. [Brief explanation of the drawing]

[0010] [Figure 1] Figure 1 is a block diagram showing an example of the hardware configuration of a decision support device. [Figure 2] Figure 2 is an explanatory diagram showing an example of a member database. [Figure 3] Figure 3 is an explanatory diagram showing an example of a response database. [Figure 4]Figure 4 is an explanatory diagram showing an example of a group characteristics table. [Figure 5] Figure 5 is an explanatory diagram showing an example of a recommended action table. [Figure 6] Figure 6 is a flowchart showing an example of an information processing procedure by a decision support device. [Figure 7] Figure 7 is an explanatory diagram showing an example of the results of the group analysis performed by the group characteristics analysis process (step S602). [Figure 8] Figure 8 is an explanatory diagram showing an example of the results of a recommended behavior analysis. [Figure 9] Figure 9 is an explanatory diagram showing an example of a group data input screen. [Figure 10] Figure 10 is an explanatory diagram showing an example of the promotion measures analysis process (step S605). [Figure 11] Figure 11 is an explanatory diagram showing an example of the first data transformation table in data transformation (step S1001). [Figure 12] Figure 12 is an explanatory diagram showing an example of a second data transformation table in data transformation (step S1001). [Figure 13] Figure 13 is an explanatory diagram showing an example of a reference feature 1011 output by data transformation (step S1001). [Figure 14] Figure 14 is an explanatory diagram showing an example of a policy feature table. [Figure 15] Figure 15 is an explanatory diagram showing an example of addition (step S1002). [Figure 16] Figure 16 is an explanatory diagram showing an example of cooperation rate prediction (step S1003). [Figure 17] Figure 17 is an explanatory diagram showing an example of an input screen for the results of promotion measures. [Figure 18] Figure 18 is an explanatory diagram illustrating an example of a decision support system. [Figure 19] Figure 19 is an explanatory diagram showing an example of the group of survey results shown in Figure 18. [Figure 20] Figure 20 is an explanatory diagram showing an example of updated data. [Figure 21] FIG. 21 is an explanatory diagram showing an example of a recommended action master table. [Figure 22] FIG. 22 is a sequence diagram for updating the recommended action table. [Figure 23] FIG. 23 is an explanatory diagram showing an example of the promotion measure implementation result group shown in FIG. 18. [Figure 24] FIG. 24 is a sequence diagram for updating the machine learning model.

MODE FOR CARRYING OUT THE INVENTION

EXAMPLE

[0011] <Hardware Configuration Example of Decision Support Device> FIG. 1 is a block diagram showing a hardware configuration example of a decision support device. The decision support device 100 includes a processor 101, a storage device 102, an input device 103, an output device 104, and a communication interface (communication IF) 105. The processor 101, the storage device 102, the input device 103, the output device 104, and the communication IF 105 are connected by a bus 106. The processor 101 controls the decision support device 100. The storage device 102 serves as a working area for the processor 101. The storage device 102 is a non-temporary or temporary recording medium that stores various programs and data. Examples of the storage device 102 include a ROM (Read Only Memory), a RAM (Random Access Memory), a HDD (Hard Disk Drive), and a flash memory. The input device 103 inputs data. Examples of the input device 103 include a keyboard, a mouse, a touch panel, a numeric keypad, a scanner, a microphone, and a sensor. The output device 104 outputs data. Examples of the output device 104 include a display, a printer, and a speaker. The communication IF 105 connects to a network and transmits and receives data.

[0012] <Member DB> Figure 2 is an explanatory diagram showing an example of a member database. Member DB200 is a database that stores the personal information of members. Members are people who belong to an organization such as a company, government agency, or school, or people who receive services from such an organization (members). When multiple members come together, some kind of decision is made by the group composed of those multiple members.

[0013] Member DB200 includes the fields Member ID 201, Age 202, Gender 203, and Nationality 204. It may also contain personal information other than Age 202, Gender 203, and Nationality 204. Each combination of fields in the same row constitutes an entry that identifies the personal information of a single member. Note that the value z of xxIDyyy (xx is a string, yyy is a sign) may be represented as xx#z. For example, a member with Member ID 201 value "001" is represented as Member#001.

[0014] <Answer DB> Figure 3 is an explanatory diagram showing an example of a response database. The response database 300 is a database that stores the answers to questions included in a questionnaire administered to each member of a group. Figure 3 shows an example of answers to questions from 35 members belonging to a particular group.

[0015] The answer database 300 has the fields Member ID 201 and A-type questions 301A to F-type questions 301F. If A-type questions 301A to F-type questions 301F are not distinguished, they are written as X-type question 301X. A-type question 301A is a question about autonomy. B-type question 301B is a question about adaptability. C-type question 301C is a question about creativity. D-type question 301D is a question about equality. E-type question 301E is a question about cooperation. F-type question 301F is a question about belonging.

[0016] Questions 301A through 301F in the A-series each contain four questions (1) through (4). Questions (1) through (4) in the X-series question 301X are similar questions, intended to increase the reliability of the answers. For each of questions (1) through (4), one of the options "1" through "5" is selected as the score.

[0017] Options "1" through "5" represent scores on a 5-point scale. For example, option "1" means "not at all applicable," option "2" means "somewhat not applicable," option "3" means "applies," option "4" means "applies well," and option "5" means "applies very much."

[0018] Note that in Figure 3, six A-type questions 301A to F-type questions 301F are used as an example, but the number of questions is not limited to six. Also, X-type question 301X includes four questions (1) to (4), but the number of questions is not limited to four. Furthermore, the number of choices is not limited to a 5-point scale. In addition, the choices may be reaction times to the answers.

[0019] Furthermore, the response database 300 includes the question score 302 for question 301X of the X series. The question score 302 is the average of 35 × 4 scores for questions (1) to (4) from all 35 members belonging to that group. For example, in the case of question 301A of the A series, it indicates that the average of 35 × 4 scores for questions (1) to (4) from the 35 members is "4.40".

[0020] Furthermore, the response DB300 includes a total score of 303. The total score of 303 is the average of 35 × 4 × 6 scores for questions (1) to (4) in the A-type questions 301A to F-type questions 301F for all 35 members belonging to that group. For example, in the case of A-type questions 301A to F-type questions 301F, it indicates that the average of 35 × 4 × 6 scores for questions (1) to (4) for the 35 members is "3.78".

[0021] Furthermore, while the mean was used for the question score 302 and the overall score 303, statistical values ​​such as the median or the mean after excluding outliers may also be used.

[0022] <Group Characteristics Table> Figure 4 is an explanatory diagram showing an example of a group characteristics table 400. The group characteristics table 400 has the following fields: characteristic ID 401, comparison result with reference value 402, group type 403, and features and challenges 404. A combination of values ​​for each field in the same row constitutes an entry representing one group characteristic.

[0023] Characteristic ID 401 is identification information that uniquely identifies the group characteristics. Comparison result 402 with the baseline value shows the comparison result between each question score 302 of A-type question 301A to F-type question 301F and the baseline value. In the example of a 5-point scale, the score range is 1 to 5, so the baseline question score 302 is set to "3.00", and if it is higher than the baseline value, it is marked "+", and if it is lower, it is marked "-". The combination of "+" and "-" for A-type question 301A to F-type question 301F in comparison result 402 with the baseline value uniquely identifies the group type 403.

[0024] Group type 403 is a string indicating the type of group characteristics. Features and challenges 404 are strings indicating the features and challenges of the group that corresponds to group type 403.

[0025] In Figure 3, the score for question 302 of type A question 301A is "4.40", the score for question 302 of type B question 301B is "3.23", the score for question 302 of type C question 301C is "4.05", ..., and the score for question 302 of type F question 301F is "1.55". If the score for question 302 of type D question 301D is above the baseline value of "3.00", and the score for question 302 of type E question 301E is below the baseline value of "3.00", then the group shown in Figure 3 corresponds to the group characteristics of "G61".

[0026] Note that while Figure 4 uses 64 types of group characteristics, it is not limited to 64 types. Also, although the comparison result 402 with the reference value is represented using binary values ​​"+" and "-", the comparison result 402 with the reference value may be represented using multiple values ​​based on the difference from the reference value. Alternatively, the question scores 302 for questions A-type 301A to F-type 301F may be used as vectors, and the reference value scores for questions A-type 301A to F-type 301F may be used as reference value vectors, and the comparison result 402 with the reference value may be represented by the distance between the two vectors.

[0027] Furthermore, the strings shown in Group Type 403 and Characteristics and Issues 404 are not limited to these, and other expressions may be used.

[0028] <Recommended Action Table> Figure 5 is an explanatory diagram showing an example of a recommended behavior table. The recommended behavior table 500 is a table that identifies the behaviors recommended for groups that correspond to group characteristics. The recommended behavior table 500 includes a recommended behavior ID 501, a recommended behavior 502, and a group characteristic-based ranking 503.

[0029] Recommended Action ID 501 is an identifier that uniquely identifies Recommended Action 502. Recommended Action 502 includes actions recommended to leaders within a group and actions recommended to members within a group. Recommended Action ID 501 that uniquely identifies Recommended Action 502 recommended to leaders is appended with "L", while Recommended Action ID 501 that uniquely identifies Recommended Action 502 recommended to members is appended with "M".

[0030] Recommended Action 502 is a string of characters indicating an action recommended for a group. Recommended Action 502 includes Recommended Action 502 recommended for leaders and Recommended Action 502 recommended for members. For example, Recommended Action #R01L is "Sharing the meaning of the group," and Recommended Action #R01M is "Participating in regular meetings."

[0031] The group characteristic ranking 503 shows the ranking of recommended behaviors 502 for each group characteristic. In this example, since 20 recommended behaviors are specified, each recommended behavior 502 is ranked from 1st to 20th for each group characteristic.

[0032] The member database 200, response database 300, group characteristics table 400, and recommended behavior table 500 are specifically stored in, for example, the memory device 102, but may also be stored in other computers accessible via the communication interface 105 through networks such as the Internet, LAN (Local Area Network), or WAN (Wide Area Network).

[0033] <Information Processing Procedure> Figure 6 is a flowchart illustrating an example of the information processing procedure performed by the decision support device 100. The decision support device 100 performs the response data acquisition process (step S601) and the group characteristics analysis process (step S602).

[0034] In the response data acquisition process (step S601), the decision support device 100 reads the question score 302 and the overall score 303 from the response DB 300 as response data. Alternatively, the decision support device 100 may read the scores of questions (1) to (4) of the X-series question 301X for each member from the response DB 300 and calculate the question score 302 and the overall score 303 as response data.

[0035] [Group Characteristics Analysis Process (Step S602)] In the group characteristics analysis process (step S602), the decision support device 100 analyzes group characteristics based on the response data acquired in the response data acquisition process (step S601). Specifically, for example, the decision support device 100 refers to the group characteristics table 400, compares the question score 302 with a reference value, and identifies the comparison result 402 with the reference value. This identifies the characteristic ID 401, group type 403, and features and challenges 404 of the group (hereinafter referred to as the target group) composed of members who provided the response data (question score 302) as analysis information.

[0036] Figure 7 is an explanatory diagram showing an example of the group analysis results obtained by the group characteristics analysis process (step S602). The group analysis results 700 can be displayed, for example, on a display which is an example of an output device 104. The group analysis results 700 include a radar chart 701, analysis information 702, an analysis button 703, and an exit button 704.

[0037] Radar chart 701 is a diagram that visualizes the score 302 for each question from A-type question 301A to F-type question 301F. The first chart, 711, consists of six black dots representing the score 302 for each question from A-type question 301A to F-type question 301F, and line segments connecting adjacent black dots. The second chart, 711 1 2 indicates the baseline value (for example, question score: 3.00). See Chart 7 (2nd chart). 1 Item 2 may be a chart of another group.

[0038] Analysis information 702 is information that visualizes the characteristics ID 401, group type 403, and features and challenges 404 of the target group.

[0039] The analysis button 703 and the exit button 704 are user interface elements that can be pressed by a user (for example, the leader of the target group) by operating the input device 103. When the analysis button 703 is pressed (step S603: Yes), the recommended behavior analysis process (step S604) is executed. When the exit button 704 is pressed (step S603: No), the series of processes ends.

[0040] Furthermore, the group characteristics analysis process (step S602) can also be implemented using machine learning, where the response data is used as features and group type 403 is used as the ground truth data, rather than being rule-based as described above.

[0041] In Figure 6, the decision support device 100 performs recommended behavior analysis processing (step S604), promotion measure analysis processing (step S605), and measure implementation processing (step S606).

[0042] [Recommended Behavioral Analysis Process (Step S604)] In the recommended behavior analysis process (step S604), the decision support device 100 analyzes the recommended behaviors 502. Specifically, for example, the decision support device 100 obtains the characteristic ID 401 and the string of the issue within the features and issues 404 from the analysis information 702. The decision support device 100 also obtains the top N recommended behaviors 502 in the group characteristic ranking 503 from the recommended behavior table 500, which are the recommended behaviors 502 for the group characteristics identified in the group characteristic analysis process (step S602).

[0043] Figure 8 is an explanatory diagram showing an example of recommended behavior analysis results. The recommended behavior analysis results 800 include a task display area 801, a higher recommended behavior display area 802, and a selection button 803. The recommended behavior analysis results 800 can be displayed on a display, for example, which is an example of an output device 104.

[0044] The issue display area 801 is the area that displays the characteristic ID 401 of the target group, its features, and the issue string from issue 404. In this example, since the acquired characteristic ID 401 is "G61", "Low sense of belonging, therefore prone to dropping out" is displayed.

[0045] The top recommended behavior display area 802 is the area that displays the top N recommended behaviors 502 in the group characteristic-based ranking 503 obtained from the recommended behavior table 500. In this example, N=4. If characteristic ID 401 is "G61", then the recommended behaviors #R02 (1st place), #R07 (2nd place), #R20 (3rd place), and #R13 (4th place) will be displayed.

[0046] The leader's recommended actions 821 are recommended actions #R02L, #R07L, #R20L, and #R13L, and the member's recommended actions 822 are recommended actions #R02M, #R07M, #R20M, and #R13M. The member's recommended actions 822 have a checkbox 823. The checkbox 823 is a user interface that can be selected by the user (for example, the leader of the target group) by operating the input device 103.

[0047] The selection button 803 is a user interface that can be pressed by a user (for example, the leader of the target group) using the input device 103. When the selection button 803 is pressed while one or more member-side recommended actions 822 are selected in the checkbox 823, the decision support device 100 saves the member-side recommended actions 822 selected in the checkbox 823 as the recommended action selection result 804. In this example, the recommended action selection result 804 is "proactive information sharing" (recommended action ID 501 is R02M).

[0048] [Group Data Input Screen] Figure 9 is an explanatory diagram showing an example of a group data input screen. The group data input screen 900 is displayed when the selection button 803 is pressed. The group data input screen 900 displays the recommended action selection result 804 as the selected recommended action. The group data input screen 900 has a group data input area 901 and an analysis button 902.

[0049] The input area 901 for group data includes, for example, input fields related to group size, average age, percentage of women, educational background, nationality, place of activity, years of activity, organizational structure, closeness, and cooperativeness.

[0050] The group size indicates the range to which the target group's members belong. The decision support device 100 identifies the number of group members from the member database 200 based on the response data (question score 302), identifies the range to which those members belong, and displays it as the group size. The group size can also be selected by the user (for example, the leader of the target group) by operating the input device 103. In this example, since the group has 35 members, the group size is displayed as "30-50".

[0051] The average age is the average of the ages of each member of the target group. The decision support device 100 identifies the ages of the group members from the member database 200 in the response data (question score 302), calculates the average value of the ages, and displays it as the average age.

[0052] The female ratio is the proportion of the group in which gender 203 is female. The decision support device 100 identifies the number of members in the group whose gender 203 is female from the member DB 200, calculates the female ratio by dividing it by the total number of members in the group, and displays it.

[0053] Educational history refers to a general overview of an individual's educational background within a given group. Educational History This is selected by the user (for example, the leader of the target group) through the operation of input device 103.

[0054] Nationality is a qualification that indicates that a member of the target group belongs to a specific country and is a citizen of that country. It indicates the distribution of nationalities for the group as a whole, rather than the nationality 204 of individual members. The decision support device 100 may read the nationality 204 of each member and select from several categories such as "Japan only" or "including foreigners," or it may be selected by the user (for example, the leader of the target group) by operating the input device 103.

[0055] The activity location, length of activity, organizational structure, closeness, and level of cooperation are selected by the user (for example, the leader of the target group) through the operation of input device 103.

[0056] The analysis button 902 is a user interface that can be pressed by a user (for example, the leader of the target group) by operating the input device 103. When the analysis button 902 is pressed, the input data entered in the group data input area 901 (group size, average age, percentage of women, educational history, nationality, activity location, years of activity, organizational structure, closeness, cooperativeness) is saved as group data 910.

[0057] Furthermore, the recommended behavior analysis process (step S604) can also be implemented using machine learning, with group type 403 as a feature and recommended behavior 502 as ground truth data, rather than the rule-based approach described above.

[0058] [Promotion Measure Analysis Processing (Step S605)] Returning to Figure 6, in the promotion measure analysis process (step S605), the decision support device 100 analyzes the promotion measures. Promotion measures are the measures that the target group should take in order to promote the recommended behavior selected in the recommended behavior selection result 804.

[0059] Figure 10 is an explanatory diagram showing an example of the promotion measure analysis process (step S605). In the promotion measure analysis process (step S605), the decision support device 100 converts the recommended action selection result 804 and the group data 910 into data.

[0060] In the promotion measure analysis process (step S605), the decision support device 100 inputs the group data 910 and the recommended action selection result 804 to perform data transformation (step S1001) and outputs the reference feature 1011. The decision support device 100 then performs addition (step S1002) of the reference feature 1011 and the measure feature table 1012. Next, the decision support device 100 inputs the result of the addition in step S1002 into the machine learning model 1013 to perform cooperation rate prediction (step S1003). After this, the decision support device 100 aggregates the cooperation rate, which is the prediction result of cooperation rate prediction (step S1003) (step S1004). The data transformation (step S1001), addition (step S1002), cooperation rate prediction (step S1003), and result aggregation (step S1004) will be explained sequentially below.

[0061] • Data conversion (Step S1001) Figure 11 is an explanatory diagram showing an example of a first data transformation table in data transformation (step S1001). The first data transformation table 1100 is a table for transforming the group data 910 into feature quantities (groups) F51 to F75. The first data transformation table 1100 defines the data items to be transformed 1101, the feature quantities (groups) 1102, and the transformation algorithm 1103.

[0062] The data items to be converted 1101 are the data items (group size, average age, percentage of women, educational history, nationality, place of activity, years of activity, organizational structure, closeness, and cooperativeness) within the group data 910 that are subject to data conversion in the data conversion (step S1001).

[0063] Feature (group) 1102 is a feature related to a group and serves as the index for the data item 1101 to be transformed in the same row. For example, F51 represents a feature related to the number of people in the group. In Figure 11, there are 25 types of feature (group) 1102 (F51 to F25), but the number of types is not limited to 25.

[0064] The conversion algorithm 1103 is a function that converts the values ​​of the data items 1101 to be converted into features (groups) 1102. For example, in the case of F51, the set range for the number of people in the group is "30 to 50", so the "center of the set range" is "40". Therefore, F51 is calculated by the following formula (1).

[0065] F51 = log(center of the set range) ÷ log(1000) = log(40) ÷ log(1000) =0.534 ···(1)

[0066] Furthermore, for features F72-F75 (cooperativeness), one of the features F72-F75 will be set to "1", and the rest will be set to "0". In the example in Figure 9, since cooperativeness is set to "individualism", F72 is set to "1", and F73-F75 are set to "0". In this way, the decision support device 100 obtains group-related features F51-F75 from the group data 910.

[0067] Figure 12 is an explanatory diagram showing an example of a second data transformation table in data transformation (step S1001). The second data transformation table 1200 is a table for transforming the recommended action selection result 804 into features (situation).

[0068] The second data conversion table 1200 has the fields Selection Recommended Action ID 1201 and Feature Quantities (Recommended Actions) 1202 (F26~F50). Selection Recommended Action ID 1201 is identified by the Recommended Action Selection Result 804. Feature Quantities (Recommended Actions) 1202 are the feature quantities F26~F50 related to the recommended action. Feature quantities F26~F50 take different values ​​for each Feature Quantity (Recommended Action) 1202. In this example, the Recommended Action Selection Result 804 is the value of Selection Recommended Action ID 1201, "R02M" (see Figures 8 and 9). Therefore, the decision support device 100 obtains the feature quantities F26~F50 for the row with "R02M".

[0069] In Figure 12, there are 25 types of features (recommended behaviors) (F26-F50), but the number of types is not limited to 25.

[0070] Figure 13 is an explanatory diagram showing an example of a reference feature 1011 output by data transformation (step S1001). The reference feature 1011 has the following fields: reference ID 1300, feature (policy) 1301, feature (recommended behavior) 1202, and feature (group) 1102.

[0071] The criterion ID 1300 is identification information that uniquely identifies the criterion feature 1011. The feature (measure) 1301 consists of features F01 to F25 related to the measure. While the feature F01 to F25 related to the measure are set for each measure, the criterion feature 1011 is set to 0 because it serves as the standard for all measures.

[0072] The feature vector (recommended behavior) 1202 is set to feature vectors F26 to F50, which have been transformed using the second data transformation table 1200 in Figure 12. The feature vector (group) 1202 is set to feature vectors F51 to F75, which have been transformed using the first data transformation table 1100 in Figure 11.

[0073] • Addition (Step S1002) Figure 14 is an explanatory diagram showing an example of a policy feature table 1012. The policy feature table 1012 is a table that defines the features related to the policies that a group should adopt for each policy. The policy feature table 1012 has a policy ID 1400, a feature (policy) 1401, a feature (recommended behavior) 1402, and a feature (group) 1403.

[0074] Policy ID 1400 is identification information that uniquely identifies a policy. In this example, 25 types of policies (T01 to T25) are defined. In Figure 14, there are 25 types of policy ID 1400 (T01 to T25), but the number of types is not limited to 25.

[0075] Feature (Measure) 1401 is a feature related to the measure that uniquely characterizes the measure identified by measure ID 1400. In this example, features whose last digit matches the last digit of measure ID 1400 are set to "1", and features whose last digits do not match are set to "0". For example, if measure ID 1400 is "T01", the value of feature F01 is set to "1", and the values ​​of features F2 to F25 are set to "0". In other words, feature (Measure) 1401, when viewed as a whole for measures T01 to T25, is represented as a diagonal matrix where the diagonal elements are "1". Note that although a diagonal matrix is ​​used in this example, it does not have to be a diagonal matrix if the values ​​of features F1 to F25 differ for each measure.

[0076] In the policy feature table 1012, feature (recommended behavior) 1402 and feature (group) 1403 are set to "0" in order to represent feature (policy) 1401.

[0077] Figure 15 is an explanatory diagram showing an example of addition (step S1002). In addition (step S1002), the reference feature 1011 and the policy feature table 1012 are added together. Specifically, for example, feature (policy) 1301 and feature (policy) 1401 are added together, feature (recommended behavior) 1202 and feature (recommended behavior) 1402 are added together, and feature (group) 1102 and feature (group) 1403 are added together. As a result, the decision support device 100 obtains the addition result 1500.

[0078] The summation result 1500 has feature (policy) 1401, feature (recommended behavior) 1402, and feature (group) 1403. In other words, it shows features F1 to F75 in which feature (recommended behavior) 1402 and feature (group) 1403 are taken into consideration in the policy.

[0079] • Prediction of cooperation rate (step S1003) and aggregation of results (step S1004) Figure 16 is an explanatory diagram showing an example of cooperation rate prediction (step S1003). In cooperation rate prediction (step S1003), the summation result 1500 and the reference feature 1011 are input to the machine learning model 1013. The machine learning model 1013 outputs a first prediction result 1601 when the summation result 1500 is input, and outputs a second prediction result 1602 when the reference feature 1011 is input. The first prediction result 1601 is the set of cooperation rates 1613 for each measure. The second prediction result 1602 is the cooperation rate 1613 when no measure is applied.

[0080] The cooperation rate 1613 is the probability that, if the target group implements the measure (or if the measure is not implemented in the case of the second prediction result 1602), the members of that target group will cooperate to perform the recommended choice behavior #R02M "active information sharing". For example, if none of the measures are implemented in the target group, the cooperation rate 1613 for members of that target group to cooperate to perform the recommended choice behavior #R02M "active information sharing" is 54.8%. On the other hand, if measure #T02 "discuss the meaning of recommended behavior" (measure 1612, ranked 1st) is implemented in the target group, the cooperation rate 1613 for members of that target group to cooperate to perform the recommended choice behavior #R02M "active information sharing" is 70.3%.

[0081] Machine learning model 1013 is a model that has been trained in advance using a training dataset consisting of sample data equivalent to 1500 summation results (or just the reference feature 1011) and its ground truth data. The ground truth data may be labels indicating "1" for cooperative actions and "0" for non-cooperative actions, or they may be probability values ​​between 0 and 1 (inclusive).

[0082] The decision support device 100 retrains the machine learning model 1013 by backpropagating errors using the value of a loss function based on the difference between the cooperation rate 1613 shown by the first prediction result 1601 and the second prediction result 1602 and the actual cooperation rate when the measures are implemented (which may be subjective to the leader of the group in question).

[0083] Furthermore, in the results aggregation (step S1004), the decision support device 100 displays the promotion measures analysis results screen 1600 based on the first prediction result 1601 and the second prediction result 1602. The promotion measures analysis results screen 1600 includes the recommended action selection result 804 and the promotion measures analysis result 1610.

[0084] The promotion measures analysis result 1610 has the following fields: rank 1611, promotion measures 1612, cooperation rate 1613, and improvement 1614. Rank 1611 indicates the position of promotion measures 1612 sorted in descending order of cooperation rate 1613 in the second prediction result 1602. A higher cooperation rate 1613 results in a higher rank. Note that rank 1611 is not set for no measures, i.e., the second prediction result 1602.

[0085] Promotion measure 1612 is the string associated with measure ID 1400. Note that for no measure, i.e., second prediction result 1602, it will be displayed as "No measure (B00)". "B00" corresponds to base ID 1300.

[0086] The cooperation rate of 1613 is the first prediction result 1601 and the second prediction result 1602 output from the machine learning model 1013. The improvement 1614 is the absolute value of the difference obtained by subtracting the cooperation rate of 1613 (no action taken) of the second prediction result 1602 from the cooperation rate 1613 (ranked 1611) of the first prediction result 1601. If the difference is positive, the absolute value of the difference and an upward arrow are displayed; if the difference is negative, the absolute value of the difference and an upward arrow are displayed.

[0087] For example, if the promotion measure 1612, "Discuss the meaning of recommended behavior," which is measure #T02 ranked 1st with a rank of 1611, is applied to the group, the cooperation rate 1613 (predicted value) is predicted to be 70.3%, which is a 15.5% increase compared to the cooperation rate 1613 (54.8%) without any measures.

[0088] [Implementation process for measures (Step S606)] Returning to Figure 6, the decision support device 100 executes the policy implementation process (step S606). In the policy implementation process (step S606), the decision support device 100 receives input of the results of the promotion measures implemented by the user (for example, the leader of the target group) and stores the input data in the storage device 102.

[0089] Figure 17 is an explanatory diagram showing an example of an input screen for the results of promotion measures. The input screen 1700 for the results of promotion measures includes a characteristic ID 401, a promotion measure implementation result 1701, and a confirmation button 1710. Characteristic ID 401 indicates the characteristic ID of the target group. In this example, it is G61.

[0090] The results of the promotion measures 1701 consist of recommended actions 1702, promotion measures 1703, and implementation results 1704. Recommended actions 1702, promotion measures 1703, and implementation results 1705 are displayed separately for the leader side (L) and the member side (M). The leader side (L) recommended actions 1702, promotion measures 1703, and implementation results 1704 are appended with "L" to the end, while the member side (M) recommended actions 1702, promotion measures 1703, and implementation results 1704 are appended with "M" to the end.

[0091] Recommended action 1702L represents the leader's recommended action 821, which corresponds to the member's recommended action selection result 804 in Figure 8. In this example, recommended action #R02L, "Conversation with the member," is displayed as recommended action 1702L.

[0092] Recommended Action 1702 M This shows the recommended action #R02M, which is the result of the member's recommended action selection 804 in Figure 8. In this example, "Proactive information sharing" for recommended action #R02M is displayed as recommended action 1702M.

[0093] Promotion measure 1703L is not displayed. In this example, the group data 910 is generated based on the member's recommended action selection result 804, and not based on recommended action 1702L.

[0094] Promotion measure 1703M represents promotion measure 1612 selected from the promotion measure analysis results 1610. Promotion measure 1703M is a pull-down user interface that allows a user (for example, a leader of the target group) to select promotion measure 1612 by operating the input device 103. In the example in Figure 17, measure #T02, "Discuss the meaning of the recommended actions," is selected. In addition to promotion measure 1612, the selection items for promotion measure 1703M also include "Not implemented."

[0095] The implementation result 1704L indicates whether or not the promotion measures were implemented. The implementation result 1704L is a pull-down user interface that allows the user (for example, the leader of the target group) to select "Not implemented" or "Implemented" by operating the input device 103. In the example in Figure 17, "Implemented" is selected.

[0096] The implementation results 1704M include the pre-implementation cooperation rate 1741M and the post-implementation cooperation rate 1742M. The pre-implementation cooperation rate 1741M is the cooperation rate before the implementation of the promotion measure 1612 selected in promotion measure 1703M, and is input information entered by the user (for example, the leader of the target group) through the operation of the input device 103.

[0097] The post-implementation cooperation rate 1742M is the cooperation rate after the implementation of the promotion measure 1612 selected in promotion measure 1703M, and is input information entered by the user (for example, the leader of the target group) through the operation of the input device 103. If promotion measure 1703M is "not implemented", the post-implementation cooperation rate 1742M is also automatically set to "not implemented".

[0098] The confirmation button 1710 is a user interface that can be pressed by a user (for example, the leader of the target group) by operating the input device 103. When the confirmation button 1710 is pressed, the decision support device 100 saves the promotion measure implementation result 1720 to the storage device 102. The promotion measure implementation result 1720 includes the input data from the promotion measure implementation result 1701 (promotion measure 1703M, pre-implementation cooperation rate 1741M, post-implementation cooperation rate 1742M), the summation result 1500, and the reference feature 1011. This completes the measure implementation process (step S606).

[0099] Returning to Figure 6, the decision support device 100 receives an input confirming the improvement effect from a user (for example, the leader of the target group) by operating the input device 103 (step S607). If the input confirming the improvement effect is received (step S607: Yes), the process returns to step S601. On the other hand, if the input confirming the improvement effect is not received and the process ends (step S607: No), the series of processes ends.

[0100] If the group type 403 of the target group is known, the decision support device 100 may start execution from the recommended behavior analysis process (step S604).

[0101] Thus, according to Example 1, when the promotion measure 1612 is implemented in the target group, the cooperation rate 1613 among the members constituting the subgroup can be automatically presented. Furthermore, this increases the likelihood that the recommended behaviors indicated in the recommended behavior selection result 804 will be successful when the promotion measure 1612 is implemented in the target group, thereby improving the characteristics of the target group. [Examples]

[0102] Next, Example 2 will be described. Example 2 describes a decision support system for managing multiple decision support devices 100 according to Example 1. In Example 2, the focus will be on the differences from Example 1, so common parts with Example 1 will be given the same reference numerals and their descriptions will be omitted.

[0103] Figure 18 is an explanatory diagram showing an example of an information processing system. The information processing system includes a management device 1801 and a plurality of decision support devices 100. The management device 1801 and the plurality of decision support devices 100 are connected to each other via a network 1802 such as the Internet, LAN (Local Area Network), or WAN (Wide Area Network). The management device 1801 is a decision support device implemented with a hardware configuration such as that shown in Figure 1. The management device 1801 may be any of the plurality of decision support devices 100.

[0104] The management device 1801 includes a survey results group 1811, a recommended behavior master table 1812, a promotion measures implementation results group 1813, and a machine learning master model 1814. Specifically, the survey results group 1811, the recommended behavior master table 1812, the promotion measures implementation results group 1813, and the machine learning master model 1814 are stored, for example, in the storage device 102 of the management device 1801, but may also be stored in another computer accessible to the management device 1801 via the network 1802 through the communication IF 105 of the management device 1801.

[0105] The results of the survey (1811) will be described later in Figure 19. The results of the promotion measures (1813) will be described later in Figure 23. The recommended behavior master table (1812) is a table with the same structure as the recommended behavior table (500). The machine learning master model (1814) is a model with the same structure as the machine learning model (1013).

[0106] The recommended action master table 1812 and the machine learning master model 1814 are updated by the management device 1801. The management device 1801 distributes the updated recommended action master data in the recommended action master table 1812 and the learning parameters constituting the machine learning master model 1814 to each decision support device 100. The decision support device 100 receives the recommended action master data in the recommended action master table 1812 and the learning parameters constituting the machine learning master model 1814 from the management device 1801 and updates the recommended action table 500 and the machine learning model 1013, which are slave information.

[0107] <Update to Recommended Actions Table 500> Next, we will explain an example of updating the recommended action table 500.

[0108] Figure 19 is an explanatory diagram showing an example of the questionnaire results group 1811 shown in Figure 18. The questionnaire results group 1811 contains the questionnaire results from each of the multiple decision support devices 100. The questionnaire results group 1811 has the following fields: computer ID 1900, characteristic ID 401, recommended selection action ID 1201, and overall score 303.

[0109] Computer ID 1811 is identification information that uniquely identifies the computer, i.e., the decision support device 100. Therefore, one entry represents one survey result. For example, entry 1911 is a survey result obtained from the decision support device 100 on computer #C001.

[0110] The overall score of 303 has two subfields: pre-measure 1901 and post-measure 1902. Pre-measure 1901 is the overall score of 303 from a questionnaire administered to the target group before the promotion measure 1612, corresponding to the recommended choice behavior ID 1201, was implemented in the target group identified by the decision support device 100 with computer ID 1811 (see Figure 3).

[0111] Post-measure 1902 is the overall score of 303 for a questionnaire administered to the target group (the same questionnaire as pre-measure 1901) after the promotion measure 1612 corresponding to the recommended choice behavior ID 1201 was implemented in the target group identified by the decision support device 100 with computer ID 1811 (see Figure 3).

[0112] For example, entry 1911 indicates that, in the decision support device 100 of computer #C001, the overall score 303 of responses to a questionnaire conducted before the policy was implemented 1901, when the recommended selection action #R02M corresponds to a target group having characteristic #G61, is "3.78", and the overall score 303 of responses to the same questionnaire conducted after the policy was implemented 1902 is "3.99". Note that if the questionnaire was not conducted after the policy was implemented 1902, "Not implemented" will be stored in the entry for after the policy was implemented 1902, as shown in entry 1912.

[0113] Figure 20 is an explanatory diagram showing an example of update data 2000. Update data 2000 is data generated by the management device 1801 by referring to the questionnaire implementation results group 1811. Update data 2000 has two fields: correction target 2001 and correction value 2002. Correction target 2001 is the combination of characteristic ID 401 and recommended selection action ID 1201 that are subject to correction by the correction value 2002. For example, in the case of entry 1911 in Figure 19, since characteristic ID 401 is "G61" and recommended selection action ID 1201 is "R02M", characteristic ID 401 and recommended selection action ID 1201 are concatenated to become "G61R02M".

[0114] In the group type 403 of the target group identified by characteristic ID 401, the group characteristic rank 503 of the recommended action ID 1201 (recommended action ID 501) is identified as the target for correction 2001. For example, in the case of entry 1911 in Figure 19, characteristic ID 401 is "G61" and recommended action ID 1201 is "R02M". Therefore, in Figure 5, the group characteristic rank 503 is identified as "1".

[0115] The adjustment value 2002 is data used to adjust the group characteristic rankings 503 identified in the adjustment target 2001. The adjustment value 2002 consists of a positive or negative sign and a step size. The sign is determined by comparing the overall score 303 of pre-measure 1901 and the overall score 303 of post-measure 1902. If pre-measure 1901 < post-measure 1902, the sign is determined to be negative. If pre-measure 1901 > post-measure 1902, the sign is determined to be positive. If pre-measure 1901 = post-measure 1902, or if post-measure 1902 has not been implemented, the step size is determined to be 0.

[0116] The step size is a pre-set value, and in the example in Figure 20, it is set to "0.1". The step size is not limited to "0.1". Also, the step size value may differ for each correction target 2001.

[0117] Figure 21 is an explanatory diagram showing an example of the recommended behavior master table 1812. For example, if the correction target 2001 is "G61R02M", the correction value 2002 is "-0.1". In the recommended behavior master table 1812, the value "1" is identified as the group characteristic rank 503 at the intersection of the column where characteristic ID 401 is "G61" and the row where recommended behavior ID 501 is "R02M".

[0118] The control device 1801 adds the correction value 2002, "-0.1", which is the value when the correction target 2001 is "G61R02M", to the identified value "1", and calculates "0.9". This "0.9" becomes the value of the identified group characteristic rank 503 in the recommended behavior master table 1812. The control device 1801 performs this process for each survey result.

[0119] <Update sequence for Recommended Action Table 500> Figure 22 is an update sequence diagram of the recommended action table 500. Each of the multiple decision support devices 100 transmits the survey results to the management device 1801 (step S2201). The management device 1801 receives the survey results from each of the multiple decision support devices 100 and stores them as a group of survey results 1811 (step S2202).

[0120] The management device 1801 calculates a correction value 2002 by comparing the total score 303 before the measure 1901 and after the measure 1902 for each survey result, and generates updated data 2000 (step S2203).

[0121] Based on the update data 2000 from step S2203, the management device 1801 corrects the ranking of the correction target 2001 for each survey result, as shown in Figure 21, and updates the recommended action master table 1812 (step S2204).

[0122] The management device 1801 distributes the recommended action master data in the updated recommended action master table 1812 from step S2204 to each decision support device 100 (step S2205). The recommended action master data is the matrix data of the group characteristic rank 503 in the updated recommended action master table 1812 from step S2204. The recommended action master data distributed may also be the values ​​of the correction target 2001 and the updated group characteristic rank 503.

[0123] Each decision support device 100 updates the recommended action table 500 with the received recommended action master data (step S2206). As a result, the recommended action table 500 in each decision support device 100 matches the recommended action master table 1812. This completes the update of the recommended action table 500.

[0124] <Update to machine learning model 1013> Next, we will explain an example of updating machine learning model 1013.

[0125] Figure 23 is an explanatory diagram showing an example of the group of promotion measures implementation results 1813 shown in Figure 18. The group of promotion measures implementation results 1813 has the following fields: computer ID 1900, promotion measure ID 2301, feature quantity 2302, and cooperation rate result 2303. Two rows of entries with the same computer ID 1900 correspond to the promotion measures implementation results 1720 from the decision support device 100 identified by that computer ID 1900.

[0126] Promotion measure ID 2301 includes standard ID 1300 (no measure) and measure ID 1400 of promotion measure 1703M.

[0127] Feature vector 2302 includes the values ​​of features F01-F75 for baseline ID 1300 (no measures) and the values ​​of features F01-F75 for promotion measures 1703M. For example, the value of features F01-F75 for baseline ID 1300 (no measures) is baseline feature 1011. For example, the value of features F01-F75 for promotion measures 1703M is the sum result 1500.

[0128] The cooperation rate result 2303 includes the pre-implementation cooperation rate of 1741M for standard ID 1300 (no measures) and the post-implementation cooperation rate of 1742M for promotion measures 1703M. In cooperation rate result 2303, "Not entered" indicates that the implementation result 1704M has not yet been entered in the decision support device 100 identified by computer ID 1900.

[0129] The two entries with the same computer ID 1900 represent the results of the promotion measures 1720, which are information regarding the training of the machine learning master model 1814 in the machine learning model 1013 (hereinafter referred to as training information), that is, information used for retraining the machine learning master model 1814.

[0130] <Update sequence for machine learning model 1013> Figure 24 is an update sequence diagram of the machine learning model 1013. Each of the multiple decision support devices 100 transmits learning information to the management device 1801 (step S2401). If each decision support device 100 has already been retrained based on the results of the promotion measures 1720, each decision support device 100 may transmit the learning parameters of its machine learning model 1013 as learning information.

[0131] The management device 1801 receives the results 1720 of the promotion measures implemented from each of the multiple decision support devices 100 and stores them as a group of promotion measures implemented results 1813 (step S2402).

[0132] The control device 1801 retrains the machine learning master model 1814 using the features F01 to F75 of the promotion measure implementation results group 1813 (step S2403). Specifically, for example, the control device 1801 uses feature 2302 as training data and cooperation rate results 2303 as ground truth data, inputs the training data into the machine learning master model 1814, and predicts the cooperation rate. The control device 1801 then retrains the machine learning master model 1814 by backpropagation using the value of the loss function based on the difference between the predicted cooperation rate and the ground truth data. The control device 1801 does not use the promotion measure implementation results 1720, for which the cooperation rate result 2303 is "not input", for retraining.

[0133] Furthermore, the management device 1801 may perform relearning (step S2403) when a predetermined number or more of the promotion measures implementation results 1720, in which the cooperation rate result 2303 is not "not entered," have been accumulated. This eliminates the need to wait for the "not entered" state of the cooperation rate result 2303 to be resolved before relearning (step S2403) can be performed, thereby shortening the update interval of the machine learning model 1013.

[0134] Furthermore, if the learning information is the learning parameters of the machine learning model 1013, the management device 1801 may average the learning parameters from each decision support device 100 through federated learning and update the machine learning master model 1814 with the averaged learning parameters.

[0135] Then, the management device 1801 distributes the updated learning parameters of the machine learning master model 1814 to each decision support device 100 (step S2404).

[0136] Each decision support device 100 updates the machine learning model 1013 with learning parameters from the management device 1801 (step S2106). As a result, the machine learning model 1013 in each decision support device 100 matches the machine learning master model 1814. This completes the update of the machine learning model 1013.

[0137] Thus, according to Embodiment 2, by having the management device 1801 centrally manage multiple decision support devices 100, the variability in the probability of success of the recommended actions indicated by the recommended action selection results 804 among the multiple decision support devices 100 is suppressed, and improvements can be made to the characteristics of the target group in each of the multiple decision support devices 100.

[0138] It should be noted that the present invention is not limited to the embodiments described above, but includes various modifications and equivalent configurations within the spirit of the attached claims. For example, the embodiments described above are described in detail to make the present invention easier to understand, and the present invention is not necessarily limited to having all of the described configurations. Furthermore, some of the configurations of one embodiment may be replaced with those of another embodiment. Furthermore, some of the configurations of one embodiment may be added to those of another embodiment. Furthermore, some of the configurations of each embodiment may be added, deleted, or replaced with other configurations.

[0139] Furthermore, each of the aforementioned configurations, functions, processing units, and processing means may be implemented in hardware, for example, by designing them as integrated circuits, or they may be implemented in software by having a processor interpret and execute programs that realize each function.

[0140] Information such as programs, tables, and files that implement each function can be stored in memory, hard disks, SSDs (Solid State Drives), or on recording media such as IC (Integrated Circuit) cards, SD cards, and DVDs (Digital Versatile Discs).

[0141] Furthermore, the control lines and information lines shown are those deemed necessary for explanation purposes and do not necessarily represent all control lines and information lines required for implementation. In reality, it can be assumed that almost all components are interconnected.

Claims

1. A decision support device having a processor for executing a program and a storage device for storing the program, The aforementioned processor, A data acquisition process that acquires features related to the measures that the group should take, features related to the group's behavior, and features related to the group itself. Depending on whether the aforementioned measures are implemented, when features related to the aforementioned behavior and features related to the aforementioned group are input, a first prediction process inputs the features related to the aforementioned measures, features related to the aforementioned behavior, and features related to the aforementioned group obtained by the acquisition process into a machine learning model that predicts whether the members constituting the group will cooperate in performing the aforementioned behavior, and outputs a first prediction result that predicts whether the members constituting the group will cooperate in performing the aforementioned behavior. An output process that outputs the first prediction result obtained by the first prediction process, A decision support device characterized by performing the following actions.

2. A decision support device according to claim 1, The aforementioned processor, The machine learning model is given features related to the behavior and features related to the group, and a second prediction process is performed to output a second prediction result that predicts whether the members constituting the group will cooperate in performing the behavior. In the output processing, the processor outputs the second prediction result obtained by the second prediction processing. A decision-making support device characterized by the following features.

3. A decision support device according to claim 1, In the acquisition process, the processor accepts the selection of a specific action from one or more recommended actions, and acquires feature quantities related to the specific action of the group. In the first prediction process, the processor inputs features relating to the policy, features relating to the specific action, and features relating to the group into the machine learning model and outputs a first prediction result that predicts whether the members constituting the group will cooperate to perform the specific action. A decision-making support device characterized by the following features.

4. A decision support device according to claim 1, The aforementioned processor, Based on the characteristics of the group, a recommended behavior analysis process is performed to identify one or more recommended behaviors. In the acquisition process, the processor accepts the selection of a specific behavior from the one or more recommended behaviors identified by the recommended behavior analysis process, and acquires feature quantities related to the specific behavior of the group. A decision-making support device characterized by the following features.

5. A decision support device according to claim 4, The aforementioned processor, Based on the answers given to the questions made to each of the aforementioned members, a group characteristics analysis process is performed to analyze the characteristics of the group. In the recommended behavior analysis process, the processor identifies one or more recommended behaviors based on the characteristics of the group analyzed by the group characteristics analysis process. A decision-making support device characterized by the following features.

6. A decision support device according to claim 1, The aforementioned processor, A learning process that learns the machine learning model based on the first prediction result and input information showing the results of the actions taken by the members in cooperation when the measures are implemented, A decision support device characterized by performing the following actions.

7. A decision support device having a processor for executing a program and a storage device for storing the program, and capable of communicating with multiple computers, The decision support device maintains master information in which recommended actions for a group are ranked according to the characteristics of that group, Each of the aforementioned computers maintains slave information in which recommended actions for the group are ranked according to the characteristics of the group, The aforementioned processor, A receiving process that receives implementation results from each of the multiple computers, including specific characteristics of the group, specific recommended actions for the group, a first score for the group based on first answers to questions given to each member of the group before the implementation of measures to be taken by the group, and a second score for the group based on second answers to the same questions given to each member after the implementation of the measures. Based on the first score and the second score received by the reception process, A calculation process for calculating a correction value for the ranking of specific recommended actions for the group based on specific characteristics of the group, A correction process that corrects the ranking of the specific recommended actions based on the correction value calculated by the calculation process, A transmission process that transmits the correction result obtained by the correction process to the multiple computers, A decision support device characterized by performing the following actions.

8. A decision support device having a processor for executing a program and a storage device for storing the program, and capable of communicating with multiple computers, The decision support device maintains master information in which recommended actions for a group are ranked according to the characteristics of that group, Each of the aforementioned computers, upon receiving feature quantities related to the group's behavior and feature quantities related to the group, depending on whether or not the measures to be taken by the group are implemented, maintains a machine learning master model that predicts whether the members constituting the group will cooperate in taking the action. Each of the aforementioned computers, depending on whether the measure is implemented or not, receives features related to the behavior and features related to the group as input, and maintains a machine learning slave model that predicts whether the members constituting the group will cooperate in performing the behavior. The aforementioned processor, A receiving process that receives information regarding the learning of the machine learning master model in the machine learning slave model from each of the multiple computers, A learning process that learns the machine learning master model based on the learning information received from each of the multiple computers through the receiving process, A transmission process that transmits the learning parameters of the machine learning master model learned by the learning process to the multiple computers, A decision support device characterized by performing the following actions.

9. A decision support device according to claim 8, The learning information includes features relating to the policy, features relating to the behavior, features relating to the group, and first input information indicating the results of the behavior performed by the members in cooperation when the policy is implemented. In the learning process, the processor learns the machine learning master model based on the first prediction result output as a result of inputting the features related to the policy, the features related to the behavior, and the features related to the group, which are received from each of the multiple computers by the receiving process, into the machine learning master model, and the first input information. A decision-making support device characterized by the following features.

10. A decision support device according to claim 9, The learning information includes second input information showing the results of the actions taken by the members in cooperation when the measures are not implemented. In the learning process described above, the processor learns the machine learning master model based on the second prediction result output as a result of inputting the features related to the behavior and the features related to the group into the machine learning master model, and the second input information. A decision-making support device characterized by the following features.

11. A decision support device according to claim 8, The learning information is the learning parameters of the machine learning slave model, In the learning process, the processor learns the machine learning master model based on the learning parameters received from each of the multiple computers through the receiving process. A decision-making support device characterized by the following features.

12. A decision support device according to claim 8, The aforementioned processor, When information regarding the learning process is received from a predetermined number or more of the aforementioned computers, the learning process is executed. A decision-making support device characterized by the following features.

13. A decision support method performed by a decision support device having a processor for executing a program and a storage device for storing the program, The aforementioned processor, A data acquisition process that acquires features related to the measures that the group should take, features related to the group's behavior, and features related to the group itself. Depending on whether the aforementioned measures are implemented, when features related to the aforementioned behavior and features related to the aforementioned group are input, a first prediction process inputs the features related to the aforementioned measures, features related to the aforementioned behavior, and features related to the aforementioned group obtained by the acquisition process into a machine learning model that predicts whether the members constituting the group will cooperate in performing the aforementioned behavior, and outputs a first prediction result that predicts whether the members constituting the group will cooperate in performing the aforementioned behavior. An output process that outputs the first prediction result obtained by the first prediction process, A decision support method characterized by performing the following actions.

14. In the processor, A data acquisition process that acquires features related to the measures that the group should take, features related to the group's behavior, and features related to the group itself. Depending on whether the aforementioned measures are implemented, when features related to the aforementioned behavior and features related to the aforementioned group are input, a first prediction process inputs the features related to the aforementioned measures, features related to the aforementioned behavior, and features related to the aforementioned group obtained by the acquisition process into a machine learning model that predicts whether the members constituting the group will cooperate in performing the aforementioned behavior, and outputs a first prediction result that predicts whether the members constituting the group will cooperate in performing the aforementioned behavior. An output process that outputs the first prediction result obtained by the first prediction process, A decision support program characterized by causing the program to execute.

Images