JP7603284B2 - NETWORK ANALYSIS METHOD, NETWORK ANALYSIS SYSTEM AND SERVER - Google Patents

Description

The present invention relates to network analysis technology in populations.

In recent years, with the advancement of sensing devices and video analysis technology, the use of sensing technology has also been attracting attention in the field of sports. In particular, professional sports teams and Olympic national teams are seeking quantitative evaluation of the performance of individual and group athletes using IoT in order to improve their competitive ability and win games.

In assessing individual performance, there is increasing use of sensing technologies such as video analysis and wearable sensors that can record and analyze the amount of movement, position, motion, vital signs, etc. during competition.

When evaluating group performance, it is important to analyze the network that represents the relationships between individuals during the competition. For example, Patent Document 1 provides technology that analyzes and visualizes the network of a group, even when individuals in the group do not face each other but move in a related manner.

JP 2018-81406 A

When applying conventional technology to sports data analysis, reductionist statistical analysis is the norm. In sports such as soccer, rugby, and basketball, where the game situation changes drastically and is difficult to reproduce, it is difficult to effectively use the results of conventional analysis.

Conventional technology focuses on analyzing data about players related to the ball or those at the center of the game, making it difficult to visualize players who are making effective plays.

For example, in soccer, if a player on the right wing is dribbling the ball up the field, a player who runs into the open space on the left wing can be considered to be making an effective play even if he does not receive a pass. Also, an offensive player who keeps the defense engaged even without the ball can be considered to be making an effective play.

In addition, conventional technology mainly focuses on analyzing past data, making it difficult to evaluate players or teams in real time.

The present invention was made in consideration of the above problems, and aims to analyze the network between individuals in a group and visualize individuals that perform effective movements in relation to the content of the competition.

The present invention is a network analysis method in which a computer including a processor and a memory analyzes a network between individuals in a population, and includes a sensing data collection step in which the computer collects time-series information on the movements of individuals exercising in a population as sensor data, a feature calculation step in which the computer calculates features related to the movements of each individual from the sensor data, a network analysis step in which the computer analyzes the relationships between individuals for combinations of individuals in the population based on the features and generates a network between individuals in a time series from the analysis results of the relationships between the individuals, and a clique evaluation step in which the computer extracts a group consisting of multiple individuals having mutual relationships from among the individuals having relationships in the network as a clique, and analyzes the clique in a time series.

Therefore, the present invention makes it possible to visualize individuals that perform effective actions by analyzing the network between individuals in a population and extracting multiple cliques.

Details of at least one implementation of the subject matter disclosed herein are set forth in the accompanying drawings and the description below. Other features, aspects, and advantages of the disclosed subject matter will become apparent from the following disclosure, drawings, and claims.

FIG. 1 illustrates a first embodiment of the present invention, showing an example of a network analysis system applied to the field of sports. FIG. 1 is a block diagram illustrating an example of a configuration of a network analysis system according to a first embodiment of the present invention. 4 is a flowchart illustrating an outline of a process performed by the server according to the first embodiment of the present invention. FIG. 4 illustrates an example of sensor data according to the first embodiment of the present invention. FIG. 11 illustrates an example of a screen displayed by an environmental information input unit according to the first embodiment of the present invention. FIG. 2 illustrates an example of environmental information according to the first embodiment of the present invention. 5 is a flowchart illustrating an example of processing performed by the feature amount calculation unit according to the first embodiment of the present invention. FIG. 10 illustrates an example of feature amount information according to the first embodiment of the present invention. 11 is a flowchart illustrating an example of processing performed by the network analysis unit according to the first embodiment of the present invention. FIG. 2 illustrates an example of synchronous network time series data according to the first embodiment of the present invention. 5 is a flowchart illustrating an example of a process performed by the network evaluation unit according to the first embodiment of the present invention. 10 is a flowchart illustrating details of a degree centrality calculation process performed by the network evaluation unit according to the first embodiment of the present invention. 11 is a flowchart illustrating details of a closeness centrality calculation process performed by the network evaluation unit according to the first embodiment of the present invention. 11 is a flowchart illustrating the details of a calculation process of betweenness centrality by a network evaluation unit according to the first embodiment of the present invention. 10 is a flowchart illustrating an example of a process for calculating an individual that is a center of a population by a network evaluation unit according to the first embodiment of the present invention. 11 is a flowchart illustrating an example of a process for calculating a group activity level by a network evaluation unit according to the first embodiment of the present invention. FIG. 11 illustrates an example of an individual evaluation table according to the first embodiment of the present invention. FIG. 11 illustrates an example of a group evaluation table according to the first embodiment of the present invention. FIG. 2 illustrates an example of synchronous network time series data according to the first embodiment of the present invention. FIG. 2 illustrates the first embodiment of the present invention and is a diagram illustrating an example of a relationship between sensor data and time-series data of a synchronous network. FIG. 1 illustrates a first embodiment of the present invention and shows an example of extracting a clique from a synchronization network. 11 is a flowchart illustrating an example of a process performed by a clique evaluation unit according to the first embodiment of the present invention. FIG. 13 illustrates the types of two cliques according to the first embodiment of the present invention. FIG. 11 illustrates an example of two-clique information according to the first embodiment of the present invention. 13 is a graph illustrating an example of time series data by type for two cliques according to the first embodiment of the present invention. 13 is a graph illustrating an example of individual time series data of two cliques according to the first embodiment of the present invention. FIG. 1 illustrates the types of three cliques according to the first embodiment of the present invention. FIG. 11 illustrates an example of 3-clique information according to the first embodiment of the present invention. 13 is a graph illustrating an example of time series data by type for three cliques according to the first embodiment of the present invention. 13 is a graph illustrating another example of time series data by type of three cliques according to the first embodiment of the present invention. 13 is a graph illustrating an example of individual time series data of three cliques according to the first embodiment of the present invention. FIG. 11 is a block diagram illustrating an example of a configuration of a network analysis system according to a second embodiment of the present invention. 10 is a flowchart illustrating an outline of a process performed by a server according to a second embodiment of the present invention.

The following describes an embodiment of the present invention with reference to the attached drawings.

As a first preferred embodiment of the present invention, an example of a network analysis system between players in a group sport is described below with reference to the drawings.

Figure 1 is a diagram showing an example in which the network analysis system of the first embodiment is applied to a group sport. The network analysis system of the first embodiment is a computer system that analyzes relationships based on the movements of individuals (players) that make up a group, creates a network (connections between individuals), and visualizes the evaluation of the individuals.

Note that, although the following describes a system for visualizing human networks, the system may be applied to any system that analyzes and visualizes networks between moving individuals, such as animals. In the following, the individuals that make up the network will be referred to as users.

The network visualization system includes a sensing device 1, a network 4, and a server 5.

An example of an athletic activity for which a network is displayed in Example 1 is soccer. However, the type of athletic activity for which a network is visualized may be any sport in which two or more individuals participate, and the network analysis system of Example 1 may be applied to group sports such as rugby, basketball, volleyball, goalball, baseball, tennis, handball, lacrosse, track and field, and speed skating.

The network analysis system of Example 1 is not limited to networks between individuals, and may be used to display networks between human bodies and objects, or between objects, by measuring the movements of objects such as balls, bats, rackets, and other tools and implements with the sensing device 1.

In Example 1, one or more sensing devices 1 are attached to the body of an individual 30 performing physical activity. The individual 30 in Example 1 shows an example in which a wristwatch-type wearable device equipped with a three-axis acceleration sensor is directly attached to the human body as the sensing device 1. The individual 30 is an individual performing physical activity on the field 100, but the location where the physical activity is performed is not limited to the field 100 and may be any location.

In addition, by specifying the location where the physical activity will be performed, even if the individual wearing the sensing device 1 is in multiple locations, it is possible to display a network of only the individuals in the specified location. In this embodiment, the individual 30 is an individual that is displayed in the network and is also an individual that is measured by the sensing device 1.

The sensing device 1 in this embodiment is equipped with at least one of various sensors that measure the movement (exercise) of an individual. The various sensors include, but are not limited to, an acceleration sensor, a gyro sensor, a pedometer, a heart rate monitor, a position sensor, an impact sensor, a magnetometer, a temperature sensor, a humidity sensor, a wind sensor, a sound sensor, a barometric pressure sensor, a positioning sensor, an infrared sensor, and a position information sensor. The results of measurements by these various sensors may be transmitted to the server 5 as sensing data.

The individual 30 may also wear the sensing device 1 so that it is physically connected to a body part of the individual 30, such as the head, neck, shoulders, back, arms, wrists, hands, fingers, waist, hips, legs, ankles, feet, heels, and toes. If there is one or more pieces of clothing, footwear, or athletic protective gear between the sensing device 1 and the body of the individual 30, the individual 30 may wear the sensing device 1 by various removable or non-removable connecting means such as straps, adhesives, pockets, and clips, in a state where the sensing device 1 is integrated with the clothing, footwear, or athletic protective gear used for the athletic activity. The sensing device 1 may also be attached to or embedded in an object such as an implement or tool, such as a ball, bat, or racket.

In addition, the sensing device 1 is not limited to a wearable sensor, but may be any device capable of measuring the movement of an individual or object, such as a device that measures the movement of an individual or object through image analysis, a laser radar that can detect the position of an object by reflecting laser light, a force plate that can measure pressure or impact, or other stationary sensors such as a sonic sensor, a magnetic sensor, an RGB camera, a depth sensor, or a multi-array microphone.

The sensing data regarding the movements measured by the sensing device 1 is transmitted to the server 5 via the network 4, where it is subjected to analysis processing and the like, and is displayed on the display of the input/output device 56.

In this embodiment, the server 5 analyzes the sensing data received from the sensing device 1 in real time to generate synchronous network time series data (described later) that indicate the connections or relationships between the individuals 30, and displays the connections or relationships between the individuals 30 on the display of the input/output device 56 as clique evaluation information (described later) that classifies the relationships between the individuals 30 in real time.

By referring to the clique evaluation information, it is possible to grasp in real time which individuals 30 have effective moves in an ongoing competition, etc.

Figure 2 is a block diagram showing an example of the configuration of a network analysis system. The server 5 communicates with the sensing device 1 via the network 4. The server 5 is a computer having a CPU 51, a memory 54, an auxiliary memory device (storage device) 55, a communication unit 52, and an input/output device 56. In the following description, a network without a symbol is used to indicate collaboration between individuals, and is distinguished from the network 4 that connects computers.

The server 5 realizes the functions of the analysis unit 10 and the environmental information input unit 53 by executing a program loaded from the auxiliary storage device 55 to the memory 54 with the CPU 51. The memory 54 of the server 5 can be configured with a ROM, which is a non-volatile storage element, or a RAM, which is a volatile storage element.

The ROM stores programs (e.g., BIOS) and the like. The RAM is a high-speed, volatile storage element such as DRAM (Dynamic Random Access Memory), and temporarily stores the programs stored in the auxiliary storage device 55 and data used when the programs are executed.

The database 20 is stored in the auxiliary storage device 55. The auxiliary storage device 55 is, for example, a large-capacity, non-volatile storage device such as a magnetic storage device (HDD) or a flash memory (SSD). The auxiliary storage device 55 also stores the programs executed by the CPU 51 and data used by the programs. That is, the programs are read from the auxiliary storage device 55, loaded into the memory 54, and executed by the CPU 51.

The server 5 has an information collection unit (not shown) that stores the sensing data received from the sensing device 1 in the sensor data 22 of the database 20, and collects the sensing data transmitted at a predetermined interval.

The analysis unit 10 is a functional unit that calculates time-series acceleration sensor information and position information based on the sensor data 22, calculates a network between individuals based on the movements and trajectories of the individuals detected from the acceleration sensor information and position information, and evaluates the network and cliques. When the connections or relationships between individuals in a competition or the like are expressed as edges, a group of vertices (or nodes) that include one or more edges is considered to be a clique. In other words, a group made up of multiple individuals among the individuals that have relationships within the network is considered to be a clique.

The analysis unit 10 mainly has functional units of a feature calculation unit 11, a network analysis unit 12, a network evaluation unit 13, a clique evaluation unit 15, and a display unit 14. Note that the location of an individual may be measured using location information detected by a GNSS 35 or the like in addition to the sensing device 1 attached to the individual. Also, the association of the time-series location information with the sensor data 22 may be performed by a computer connected to the server 5.

The display unit 14 and the environmental information input unit 53 output the data recorded in the database 20 to a computer or device connected via the network 4, or to an input/output device 56, thereby making the data available to a user (a user of the analysis system) who evaluates the group network. The input/output device 56 is composed of a keyboard, a mouse, a display, etc.

The CPU 51 operates as a functional unit that provides a specific function by processing according to the program of each functional unit. For example, the CPU 51 functions as the analysis unit 10 by processing according to an analysis program. The same applies to other programs. Furthermore, the CPU 51 also operates as a functional unit that provides each function of the multiple processes executed by each program. A computer and a computer system are devices and systems that include these functional units.

The communication unit 52 is a network interface that controls communication with other devices according to a specific protocol. The communication unit 52 can communicate with other devices via the network 4 and send and receive data.

The analysis unit 10 and the environmental information input unit 53 may be implemented by a program, or may be implemented by a physical integrated circuit. In particular, the analysis unit 10 may be implemented by multiple programs or multiple integrated circuits for executing the functional units included in the analysis unit 10. In addition, the feature calculation unit 11, the network analysis unit 12, the network evaluation unit 13, and the display unit 14 may each be implemented by multiple programs or multiple integrated circuits for each process that they execute.

The programs executed by the CPU 51 are provided to the server 5 via removable media (CD-ROM, flash memory, etc.) or a network, and are stored in the auxiliary storage device 55, which is a non-transitory storage medium. For this reason, the server 5 should have an interface for reading data from removable media.

The server 5 is a computer system that is physically configured on one computer, or logically or physically configured on multiple computers, and the above-mentioned programs may be executed in different threads on the same computer, or may be executed on virtual computers constructed on multiple physical computers.

The database 20 stores sensor data 22, feature information 21, environmental information 23, synchronous network time series data 24, network evaluation information 25, and clique evaluation information 40.

The sensing data measured by the sensing device 1 for each individual is stored as sensor data 22. In this embodiment, the sensor data 22 indicates the movement measurement results for each individual.

The feature information 21 indicates features related to the movement of each individual calculated by executing the feature calculation unit 11 on the sensor data 22. Details will be described later.

The environmental information 23 is input through the environmental information input unit 53. The environmental information 23 indicates information related to athletic activity, and includes information for identifying the sensor data 22, the feature amount information 21, and the synchronous network time series data 24 based on the athletic activity.

The environmental information 23 indicates, for example, the content of the athletic activity, the time and place at which the athletic activity was performed, and the participants (individuals) in the athletic activity. Details will be described later. Note that, when analyzing and displaying a network targeting all sensor data 22, the environmental information 23 and the environmental information input unit 53 are not necessarily required.

The synchronous network time series data 24 indicates information about the network between individuals calculated by executing the network analysis unit 12 on the feature information 21. Details will be described later.

The network evaluation information 25 is evaluation information regarding the network of a population calculated by executing the network evaluation unit 13 on the synchronous network time series data 24, and includes an individual evaluation table 250 and a population evaluation table 258. Details will be described later.

The clique evaluation information 40 is data that classifies the connections or relationships between individuals calculated from the synchronous network time series data 24 by clique type, and expresses the frequency of occurrence of each clique type in a time series, as described below.

The sensing device 1 is a device that measures the movement of a target individual for which a network is to be analyzed. In this embodiment, a wearable device worn by the target individual is used as an example. Note that the sensing device 1 may measure any content as sensing data, as long as the value changes as a result of the user's movement.

The sensing device 1 includes an acceleration sensor 31, a Global Navigation Satellite System (GNSS) 35, a memory 32, a microcomputer (MCU in the figure) 33, and a communication unit 34.

The acceleration sensor 31 measures the user's acceleration, for example, at a rate of about 20 to 1000 times per second. The GNSS 35 can be configured with a location information system such as the Global Positioning System (GPS) or the Quasi-Zenith Satellite System (QZSS), and measures the location information of the individual. In this embodiment, the location information from the GNSS 35 is used as the location information of the individual 30 at the start of the competition, and the sensing data of the acceleration sensor 31 is used to measure the movement of the individual 30 during the competition.

The microcomputer 33 records the measurement results obtained by the acceleration sensor 31 and the GNSS 35 as sensing data in the memory 32. Here, the microcomputer 33 records a unique identifier (user ID) for each user and the measurement results as sensing data.

The microcomputer 33 transmits the sensing data and location information recorded in the memory 32 to the server 5 via the communication unit 34. The communication unit 34 transmits the sensing data to the server 5, using a wireless or wired connection, at a timing when communication is possible or at a predetermined timing set by the user or administrator.

Figure 3 is a flowchart showing an example of processing performed by the analysis unit 10 of the server 5.

First, the feature calculation unit 11 acquires sensing data of the target to be processed from the sensor data 22 based on the environmental information 23, and calculates feature information 21 related to the movement of each individual 30 from the target sensor data 22 (S101).

The feature information 21 indicates a feature whose value changes according to the strength of the movement of the individual 30 (or the magnitude of the energy of the movement), such as the amount of body movement, and further indicates the change in the feature in a time series.

In this embodiment, the physical activity intensity is calculated as the characteristic amount, but any characteristic amount related to exercise, such as the amount of exercise, the amount of body movement, the moving speed, the frequency of activity, the pitch, the number of steps, etc., may be calculated as the characteristic amount. In addition, characteristic amounts that change with exercise, such as calorie consumption, heart rate, pulse rate, and amount of sweating, may also be calculated as the characteristic amount.

Next, the network analysis unit 12 uses the calculated feature amount information 21 to perform network analysis between the individuals 30 to generate synchronous network time series data 24 (S102). The network evaluation unit 13 then calculates the central individual of the population and the population activity level from the results of step S102, evaluates the network, and generates network evaluation information 25 (S103). The calculation of the population activity level will be described later.

The clique evaluation unit 15 acquires the synchronous network time series data 24, extracts a preset number of cliques, and classifies the types of the extracted cliques as described below to generate clique evaluation information 40 (S104).

Finally, the display unit 14 outputs the clique evaluation information 40, the synchronous network time series data 24 of the population to be processed, and the network evaluation information 25 (S105). The display unit 14 visualizes and outputs the clique evaluation information 40, the synchronous network time series data 24, and the network evaluation information 25 at a predetermined time interval.

The output of the display unit 14 can be displayed on the display of the input/output device 56 or on an output device of a computer (hereinafter, external computer) connected to the server 5 via the network 4. Note that the information visualized and output by the display unit 14 may include the synchronous network time series data 24 or the network evaluation information 25 in the clique evaluation information 40.

The visualization process performed by the display unit 14 can graph the clique evaluation information 40 and output the synchronous network time series data 24 as a map. The visualization of the clique evaluation information 40 and the synchronous network time series data 24 will be described later.

By executing the process shown in FIG. 3, the analysis unit 10 can use the sensor data 22 and the environmental information 23 to analyze the network between the individuals 30 being analyzed, generate synchronous network time series data 24, and perform clique analysis processing to generate clique evaluation information 40.

The analysis unit 10 can identify combinations of individuals with strong relationships from the generated synchronization network time series data 24 and output them in the form of, for example, a graph. This makes it possible to visualize the network between individuals in a population.

The analysis unit 10 can then extract specific cliques from the synchronous network time series data 24, which is a compilation of the network over time, generate clique evaluation information 40 by classifying the cliques by type, and output the information in a visualized form such as a graph. The clique evaluation information 40 makes it possible to visualize individuals that perform effective movements in relation to the content of the competition in the network between individuals in the population.

Figure 4 shows sensor data 22 for Example 1.

The sensor data 22 includes a user information table 201 that records information about the individual 30 wearing the sensing device 1, and a movement information table 202 that records activity level information for each individual 30.

The user information table 201 includes a user ID 2011, a user name 2012, and a user type 2013. The user ID 2011 records an identifier assigned to each individual 30 in order to recognize the individual 30 wearing the sensing device 1. The user ID stored in the user ID 2011 is stored in the sensing data transmitted from the sensing device 1.

The user name 2012 records the name or nickname of the individual 30 wearing the sensing device 1. The user type 2013 may record age, gender, etc. as user information, and also records various information of the individual 30 that is not limited to these.

For example, if the athletic activity being measured is soccer, the user type 2013 records the team and position, etc., which allows the display unit 14 to evaluate and display each type of individual 30.

One row of the movement information table 202 indicates information on one time (measurement date and time) for one individual 30. The movement information table 202 includes a user ID 2021, a measurement date and time 2022, an acceleration X-axis 2023, an acceleration Y-axis 2024, an acceleration Z-axis 2025, a latitude 2026, and a longitude 2027. The user ID 2021 records an ID for identifying the individual 30, and corresponds to the user ID 2011.

The measurement date and time 2022 stores the date and time when the sensing data was measured. The acceleration x-axis 2023, acceleration y-axis 2024, and acceleration z-axis 2025 store the measurement results of each axis of the three-axis acceleration sensor. The latitude 2026 and longitude 2027 store the measurement values of GNSS35.

As described above, the sensor data 22 stores a user information table 201 for each individual 30 and a time-series movement information table 202.

Figure 5 shows the environmental display input screen 301 displayed by the environmental information input unit 53 in Example 1.

The environmental information input unit 53 provides an environmental display input screen to the display of the input/output device 56 or an external computer, and accepts environmental information input to the input/output device 56 or the server 5 from an operator of the external computer. The environmental information input unit 53 then stores the input environmental information in the environmental information 23 in the auxiliary storage device 55.

By referring to the environmental information 23, the analysis unit 10 is able to analyze data from short-term athletic activities. The analysis unit 10 can also analyze and display networks in athletic activities such as training in which various practice menus are performed in a short period of time while participants are rotated.

The environmental display input screen 301 includes an environmental information input page 302, a participant selection page 303, and a grouping selection page 304.

The environmental information input page 302 is used to input an activity ID 3021 that stores an identifier for the athletic activity, an activity name 3022 that stores the name of the athletic activity, a start time 3023, an end time 3024, an area name 3025 of the area in which the athletic activity was performed, participants in the athletic activity 3026, and a grouping of participants 3027.

The environmental information input page 302 is an example of a screen on which the operator inputs environmental information related to the athletic activity that he/she wishes to analyze. The operator inputs the activity name 3022, start time 3023, end time 3024, etc. into each line of the environmental information input page 302. The operator also uses the tabs to select the area name 3025 of the area where the activity was carried out from candidate locations registered in advance. The operator can also input the area name 3025. The operator then inputs the individuals 30 who will be participating in the athletic activity into participants 3026.

For participants 3026, if all of the wearers (individuals 30) of the sensing device 1 who have been registered in advance are participants, the operator enters a check in the "All" check box. Also, when the operator selects a participant (individual 30), he or she selects "Individual selection" for participants 3026.

When "individual selection" for participant 3026 is selected, the environmental information input unit 53 displays a participant selection page 303. The participant selection page 303 is a screen that displays a list of individuals 30 wearing previously registered sensing devices 1 and allows the operator to select a participant (individual 30). A function for newly registering an individual 30 wearing a sensing device 1 may be added to this screen.

The participant selection page 303 includes selections 3031 as check boxes for selecting a record, and information such as user ID 3032, user name 3033, uniform number 3034, position 3035, and grade 3036 for one record.

Grouping 3027 displays an interface that allows the operator to select how to group the participants. If the operator does not need to group the participants, the operator checks the "None" check box in Grouping 306. If the operator wants to group the participants, the operator selects Advanced Settings.

When the detailed setting of grouping 3027 is selected, the environmental information input unit 53 displays a grouping selection page 304. The grouping selection page 304 displays a list of individuals 30 wearing sensing devices 1 that have been registered in advance. The grouping selection page 304 then displays an interface that allows the operator to select a team for the individuals 30. The page may also have a function that allows the operator to freely set and change team names.

The grouping selection page 304 includes information on user ID 3041, user name 3042, team A 3043, team B 30434, and team C 3045 in one record.

In addition, in this embodiment 1, the individuals 30 are grouped according to the team to which they belong, but this is not limited thereto, and they may be grouped according to the position or grade of the individuals 30, and further, grouping may be performed using various information registered in the user information table 201.

The environmental information received by the environmental information input unit 53 is stored in the environmental information 23 of the database 20. In addition, the environmental information input unit 53 may store the grouping of the individuals 30 input via the grouping selection page 304 in the user type 2013 of the user information table 201.

Figure 6 shows the environment information 23 in Example 1.

The environmental information 23 includes an activity information table 203 that records information about the activity for which the sensor data 22 was measured, an environmental information table 204 that records the environment in which the activity was performed, such as the period and location of the activity, for each activity ID in order to identify the environment in which the activity was performed, and a group information table 205 that records the composition and characteristics of the group in which the activity was performed.

The activity information table 203 includes an activity ID 2031, an activity name 2032, and a group ID 2033. The activity ID 2031 records an ID (identifier) set for each athletic activity in order to recognize the athletic activity for which the sensor data 22 was measured. The activity ID 2031 may be the same as the activity ID entered on the environmental information input page 302, or may be an ID assigned based on that activity ID.

Activity name 2032 records the name of the exercise activity for which sensor data 22 was measured. This allows the display unit 14 to display a screen showing the area name. Activity name 2032 stores the activity name 3022 entered on the environmental information input page 302.

The environmental information table 204 includes an activity ID 2041, a start time 2042, an end time 2043, an area ID 2044, a user ID 2045, and an area name 2046. The activity ID 2041 corresponds to the activity ID 2031 in the activity information table 203. The start time 2042 and the end time 2043 indicate the start time and the end time of the exercise activity indicated by the activity ID 2041. The start time 2042 and the end time 2043 store the start time 3023 and the end time 3024 entered on the environmental information input page 302.

Area ID 2044 indicates the location where the athletic activity indicated by activity ID 2041 was performed. Area ID 2044 stores the ID assigned to area name 3025 entered on environmental information input page 302. User ID 2045 stores the user ID of an individual 30 who is a participant in the athletic activity indicated by activity ID 2041 and who is wearing sensing device 1. Area name 2046 stores the name of the area corresponding to area ID 2044.

The group information table 205 includes a group ID 2051, a group name 2052, a user ID 2053, and a group type 2054. The group ID 2051 corresponds to the group ID 2033. The group name 2052 records the name or nickname of the group. The user ID 2053 stores the user ID of an individual 30 who is part of the group indicated by the group ID 2051 and who is wearing the sensing device 1. The group type 2053 may record the composition and characteristics of the group as group information, and also records various information about the group that is not limited to these.

For example, if the group activity being measured is soccer, the group type 2053 records the formation, etc., and the display unit 14 can display each type, such as by mapping the network on the formation.

Figure 7 is a flowchart showing an overview of step S101 by the feature calculation unit 11 in the first embodiment. The feature calculation unit 11 calculates features related to the movement of the individual 30 from the sensor data 22.

First, in step S1011, the feature calculation unit 11 acquires from the DB 20 the sensor data 22 to be subjected to network analysis and display processing, based on the environmental information 23 input by the environmental information input unit 53. In this embodiment 1, the sensor data 22 is three-axis acceleration data (the acceleration x-axis 2023, the acceleration y-axis 2024, and the acceleration z-axis 2025 of the movement information table 202) measured at a predetermined sampling rate (for example, 20 Hz). The sensor data 22 to be processed can be specified from the input/output device 56 or an external computer. In addition to the three-axis acceleration data, the latitude 2026 and longitude 2027 can be added as location information.

Next, in step S1012, the feature calculation unit 11 applies a predetermined filter to the sensor data 22 acquired in step S1011 to remove noise. In this embodiment 1, a band-pass filter is designed and applied, but the present invention is not limited to a band-pass filter, and any processing such as moving average, low-pass filter, high-pass filter, etc. may be performed depending on the sensing data. Parameters such as the filter order, passband, stopband, etc. may also be set depending on the sensing data. In addition, depending on the sensing data, it is not necessarily necessary to perform step S1012.

In step S1013, the feature calculation unit 11 calculates features from the sensing data to which the above filter has been applied. The features calculated here are information regarding the movement of the individual 30.

The above steps S1011, S1012, and S1013 are applied to the sensor data 22 of all individuals 30 that make up the population.

The feature amount information 21 in Example 1 is METs (Metabolic Equivalents) that indicate exercise intensity. In step S1013, the feature amount information 21 (METs) for each time is calculated.

The feature calculation unit 11 first calculates the scalar quantity S of the acceleration on the three axes, the X-axis, the Y-axis, and the Z-axis, after the filter is applied. Here, the scalar quantity S is calculated by the following formula (1) when the acceleration on each axis is written as Xg, Yg, and Zg.

Scalar quantity S = (Xg^2 + Yg^2 + Zg^2)^(1/2) ... (1)

Next, METs is calculated from the scalar quantity S using the following formula (2).

METs=a×S+1...(2)

where a is a constant. This METs value is at rest, i.e., when the scalar quantity S = 0, then METs = 1, and indicates how many times the exercise intensity is compared to that at rest.

The sensor data 22 described above includes three-axis acceleration, but may include any measurement results that measure the strength of the energy (intensity) of human movement, such as heart rate, breathing rate, and amplitude. The feature calculation unit 11 then calculates feature amounts related to movement, such as breathing rate, heart rate, walking or running patterns, or movement speed, based on these various sensor data 22, enabling the network analysis unit 12 of this embodiment to more effectively analyze the group network.

The feature information 21 may also use values collected at a specific time. For example, features may be calculated from sensor data 22 measured at a sampling frequency of 20 Hz, and the collected values may be used every second, thereby reducing noise associated with sudden changes. Whether or not to use the collected value of the sensor data 22, and to what extent the collected value is used, is determined based on the sensing device 1 used, the environment in which the measurement was performed, and the like.

By the above processing, METs is calculated as an example of feature information 21 from the filtered sensor data 22.

Figure 8 is a diagram showing an example of data stored in the feature information 21 in Example 1. The feature information 21 stores features at each date and time 222 of all individuals 223 for which network analysis was performed. In this example, METs are calculated as features from sensor data 22 measured at a sampling frequency of 20 Hz, and values tallied every second are stored.

Figure 9 is a flowchart showing an overview of step S102 in Figure 3 by the network analysis unit 12 of Example 1. The network analysis unit 12 analyzes the relationships between individuals that make up the population from the features calculated in step S101.

First, in step S1021, the network analysis unit 12 acquires the time interval of the data used to analyze the network between the individuals 30. The time interval of the data may be a preset value as long as it is equal to or greater than the measurement interval (sampling period or a predetermined time interval) of the sensing device 1.

The time interval of the data may be a value already written in the program. Alternatively, an input section (not shown) may be provided in the environmental information input unit 53 to reference the input value. In this embodiment 1, the time interval of the data is set to one minute, and the network can be analyzed and displayed every minute.

In step S1022, the network analysis unit 12 selects two individuals for network analysis. In this embodiment 1, two individuals are selected sequentially to analyze the presence or absence of correlation between the two individuals for all combinations of the individuals that make up the population.

In step S1023, for the individual selected in step S1022 above, the features calculated in step S101 of FIG. 3 are obtained from the feature information 21.

In steps S1024, S1025, and S1026, the relationship between the two individuals is analyzed to determine whether or not a network exists. In this embodiment, the amount of information transmitted between the individuals 30 is calculated using transfer entropy to determine whether or not a network exists.

Transfer entropy is a type of information amount, and is a numerical representation of the flow of information between random variables as the average amount of information. Unlike mutual information, which indicates only the correlation between variables and does not include the flow of information, transfer entropy can identify the flow of information and is an index that measures the relationship between variables. In this embodiment 1, transfer entropy is used, but any index that measures the relationship between multiple variables, such as mutual information, correlation analysis, regression analysis, and multivariate analysis, and any method for calculating the index may be used.

In step S1024, the network analysis unit 12 calculates a probability distribution from the time series data of the variables (features) of each individual 30. A method of approximating the probability distribution using a histogram is used, but a method of estimating by assuming a Gaussian distribution or a mixed Gaussian distribution may also be used, and the method is not limited to these. When approximating using a histogram, the number of divisions into the histogram is determined using the Akaike Information Criterion or Sturges's formula, but is not limited to these.

In step S1025, the network analysis unit 12 calculates the amount of information transmission between the variables of two individuals from the transfer entropy. Here, if the elements of the random variables X and Y at time t are xt and yt, the transfer entropy T(Y→X), which indicates the influence of Y on X, is calculated by the following formula (3).

T(Y → X) takes a value between 0 and 1, with 0 indicating no influence at all, i.e. no relationship at all, and the closer it is to 1, the greater the influence, i.e. the greater the relationship between individuals.

In this Example 1, the influence of element yt of random variable Y on element xt+1 of random variable X was calculated. In other words, when feature information 21 collected every second is used, the influence of the movement of individual Y on individual X one second later is calculated. In addition, by setting the element of random variable X to xt+n, it is also possible to calculate the influence of the movement of individual Y on individual X n seconds later. The value of n can be set arbitrarily.

In step S1026, the network analysis unit 12 uses the value of transfer entropy calculated in step S1025 to determine whether or not a network exists between two individuals. The network analysis unit 12 determines that information is transmitted between individuals 30 whose transfer entropy is equal to or greater than a predetermined threshold, and that a network exists. In this embodiment 1, the threshold is a predetermined constant. The threshold may be a threshold determination method using a x2 distribution, a gamma distribution, or the like, and is set arbitrarily depending on the sensing device 1 used, the environment in which the measurement was performed, and the like.

In step S1027, the network analysis unit 12 stores the presence or absence of the network determined in step S1026 above in the synchronous network time series data 24.

In this embodiment 1, if a network exists between two individuals, the synchronization network time series data 24 is set to 1, and if not, the synchronization network time series data 24 is set to 0. In addition, multiple thresholds may be set and the network strength may be determined and stored in several stages, and the transfer entropy value may also be stored as is.

In step S1028, steps S1022, S1023, S1024, S1025, S1026, and S1027 are repeated until analysis of all combinations of individuals that make up the population is completed.

By the above process, the magnitude of the relationship between the individuals 30 is calculated based on the magnitude of the transfer entropy, the individuals that make up the network are identified, and a network within the population is generated.

Figure 10 is a diagram showing data stored in the synchronous network time series data 24 of Example 1. The synchronous network time series data 24 is a brute force diagram of the network at a certain time, and describes the presence or absence (strength or weakness) of the network at a certain time t for all individuals for which network analysis was performed.

The synchronous network time series data 24 is time series data generated at a predetermined time interval. In the illustrated example, the synchronous network time series data 24 is generated in real time at a predetermined time interval ΔT from time T1, and is analyzed and evaluated by the clique evaluation unit 15.

The first column of the synchronous network time series data 24 is the information and the user ID 242 of the influencing individual, and the first row is the information and the user ID 243 of the affected individual. The value between two individuals shown by the row and column indicates the strength of the network on a five-point scale.

If the relationship between the individuals 30 is strong, "5" is set, and if the relationship between the individuals 30 is weak, "1" is set. By storing the strength of the relationship, it becomes possible to display the relationship between the individuals in more detail on the display unit 14 described below. In this embodiment, the strength of the relationship is described in five stages, but it may be stored in any number of stages. In the above, the strength of the relationship is described in five stages, but it may be stored in any number of stages.

The strength of the relationship can be expressed by degree centrality and closeness centrality, as described below. An example of the relationship between individuals 30 visualized on the display unit 14 is shown in FIG. 19. FIG. 19 is a diagram showing an example in which the degree centrality and closeness centrality of the network evaluation information 25 are added to the synchronous network time series data 24. The example shown in the figure shows the results of a time series network analysis between individuals 30 in a soccer game.

In the illustrated example, the synchronous network time series data 24 at time t1 is displayed in area 81, the network at time t2, a time interval Δt later, is displayed in area 82, and the network at time t3 is displayed in area 83.

Individuals 30 in team A (first group) are displayed as solid circles, and individuals 30 in team B (second group) are displayed as dashed circles. Relationships between individuals 30 are displayed as edges, the magnitude of degree centrality is displayed by the size of the circle, and the magnitude of closeness centrality is displayed by the density of the hatching.

Note that while Figure 19 shows an example in which the relationships between individuals 30 are displayed as a directed graph, the graph may also be an undirected graph (same below).

Figure 11 is a flowchart showing an example of the process performed by the network evaluation unit 13 in step S103. The network evaluation unit evaluates the central individual of the population and the population activity level from the synchronous network time series data 24 calculated in step S102.

In step S1031, the network evaluation unit 13 acquires the synchronous network time series data 24 calculated in step S102 of FIG. 3. In this embodiment, the network shown in FIG. 11 is used.

In step S1032, the synchronized network time series data 24 acquired by the network evaluation unit 13 is used to calculate degree centrality, closeness centrality, betweenness centrality, and the number of edges as evaluation indices of the network. The calculation method of each index will be described later. In this embodiment 1, degree centrality, closeness centrality, betweenness centrality, and the number of edges are used as evaluation indices of the network, but are not limited to these.

In step S1033, the network evaluation unit 13 estimates the individual at the center of the network using the degree centrality, closeness centrality, and betweenness centrality calculated in step S1032. Details will be described later.

In step S1034, the network evaluation unit 13 estimates the group activity of the network in the team using the number of edges calculated in step S1032. Details will be described later.

In step S1035, the network evaluation unit 13 stores the evaluation information calculated in steps S1032, S1033, and S1034 in the network evaluation information 25 (individual evaluation table 250 and group evaluation table 258).

By the above process, network evaluation information 25 is generated using degree centrality, closeness centrality, betweenness centrality, and number of edges for the calculated synchronous network time series data 24.

Figures 12, 13, and 14 are flowcharts showing the details of the processing by the network evaluation unit 13 in Example 1.

Flowchart 60 in Figure 12 shows the process of calculating degree centrality. Degree centrality is an index that highly values vertices (individuals) that have more relationships within a network.

In step S1036, the network evaluation unit 13 counts the in-degree of an individual at a certain time. The in-degree is the number of individuals from which an individual has received information.

In step S1037, the network evaluation unit 13 counts the out-degree of an individual at a certain time. The out-degree is the number of individuals to which an individual transmits information.

In step S1038, the network evaluation unit 13 calculates the in-degree, out-degree, and degree centrality from the in-degree calculated in step S1036 above, the out-degree calculated in step S1037, and the total degree, which is the sum of the in-degree and the out-degree, since the degree centrality is the integrated value of the degrees.

In step S1039, since degree centrality differs depending on the number of vertices included in the network, the network evaluation unit 13 standardizes the degree by dividing it by the theoretical maximum value of the population being processed. In a population with n individuals, the maximum degree is n-1, so the network evaluation unit 13 divides each degree centrality calculated in step S1038 by n-1.

In step S1040, the network evaluation unit 13 repeatedly executes the processes of steps S1036 to S1039 to calculate the in-degree, out-degree, and degree centrality for all individuals (vertices) that have undergone network analysis.

Flowchart 61 in FIG. 13 shows the process of calculating closeness centrality. Closeness centrality is an index that evaluates a vertex (individual) more highly if the distance between the vertex (individual) and other vertices (individuals) is smaller.

In step S1041, the network evaluation unit 13 calculates the sum of the shortest distances from one individual to another individual. The shortest distance is the number of edges that must be traveled to reach from one vertex to another vertex in the shortest time.

In step S1042, since closeness centrality is the inverse of the shortest distance from one individual to another individual, the network evaluation unit 13 calculates the closeness centrality using the shortest distance from one individual 30 to another individual 30 calculated in step S1040 above.

In step S1043, since closeness centrality differs depending on the number of vertices included in the network, the network evaluation unit 13 standardizes it by dividing it by the value that theoretically maximizes closeness centrality in the population being processed. In a population with n individuals, the maximum closeness centrality is 1/(n-1), so the network evaluation unit 13 divides the closeness centrality calculated in step S1040 above by 1/(n-1).

In step S1044, the network evaluation unit 13 repeatedly executes steps S1040 to S1046 to calculate closeness centrality for all individuals that have undergone network analysis.

Flowchart 62 in FIG. 14 shows the process of calculating betweenness centrality in the network evaluation unit 13. Betweenness centrality is an index that evaluates a vertex (individual) more highly on the shortest path between other vertices (individuals). An individual with high betweenness centrality can be considered an important individual for maintaining the connectivity of the network.

In step S1045, the network evaluation unit 13 calculates the sum of the shortest distances from one individual to another individual. The shortest distance is the number of edges that must be traveled to reach from one vertex to another vertex in the shortest time.

In step S1046, the network evaluation unit 13 calculates the number of shortest paths from one individual to another individual. The number of shortest paths is calculated as follows. The number of shortest paths from the starting point to the starting point itself is set to 1. The number of paths to a vertex reached during the search process is equal to the number of paths to the immediately preceding vertex. If there are multiple immediately preceding vertices, it is equal to the sum of the numbers of paths to those immediately preceding points.

In step S1047, the network evaluation unit 13 repeatedly executes steps S1045 and S1046 for all individuals 30 (vertices) to calculate the sum of the shortest distances and the number of shortest paths for all individuals that have undergone network analysis.

In step S1048, using information about the shortest paths examined in steps S1045 and S1046, the network evaluation unit 13 examines the degree to which each vertex is located on the shortest path between other vertices.

Here, vertex w is the vertex immediately following vertex v on the shortest path starting from vertex s, and g _sv and g _sw are the numbers of shortest paths to vertex v and vertex w, respectively. In this case, the extent to which vertex v is on the shortest path to another vertex, b _sv , is calculated by the following formula (4).

The network evaluation unit 13 uses the above formula (4) to calculate whether each vertex included in the graph of each individual 30 is located on the shortest path to other vertices when the starting point is each vertex.

In step S1049, the network evaluation unit 13 calculates the betweenness centrality of each vertex by adding, for each vertex, the degree b _sv of each vertex being located on the shortest path between other vertices calculated in step S1048 above.

In step S1050, since betweenness centrality differs depending on the number of vertices included in the network, the network evaluation unit 13 standardizes it by dividing it by the value that theoretically maximizes betweenness centrality in the group. In a group with n individuals, the maximum betweenness centrality is (n-2) x (n-1), so the network evaluation unit 13 divides the closeness centrality calculated in step S1049 above by (n-2) x (n-1).

Through the above process, the network evaluation unit 13 calculates the betweenness centrality of each vertex (individual).

Flowchart 63 in FIG. 15 shows an example of the process of calculating the central individual of the population in step S1033 shown in FIG. 11. In step S1051, the network evaluation unit 13 calculates the individual with the maximum centrality shown in the following formula (5) using the degree centrality x, closeness centrality y, and betweenness centrality z of each individual 30 at each time calculated in step S1032 in FIG. 11.

ax+by+cz...(5)

In the above formula (5), a, b, and c are coefficients. In this embodiment, the individual for which formula (5) is maximized is calculated as the central individual of the population, but the method is not limited to this.

Flowchart 64 in FIG. 16 shows the process of calculating group activity in step S1034 shown in FIG. 11.

In step S1052, the network evaluation unit 13 counts the degree of the entire group for which the group activity is to be calculated. In addition, when multiple individuals 30 exist on the same field 100, such as in soccer, the relationship between groups can be evaluated by calculating the degree within the group, the outdegree from within the group to another group, the indegree from outside the other group to the group, and the degree outside the group.

In step S1053, the network evaluation unit 13 standardizes the degree by dividing it by the theoretically maximum value, and calculates the population activity. The population activity takes a value between 0 and 1, for example, and the closer to 1 the value is, the more activated the population is. In this embodiment 1, the population activity is calculated using the degree of the entire population, but the method is not limited to this.

Figure 17 is a diagram showing an example of an individual evaluation table 250 stored in the network evaluation information 25 of Example 1. The individual evaluation table 250 of the network evaluation information 25 stores the evaluation results for each date and time 251 of each individual for which the network evaluation was performed in step S103 of Figure 3.

The individual evaluation table 250 includes a date and time 251, indegree centrality 252, outdegree centrality 253, degree centrality 254, closeness centrality 255, betweenness centrality 256, and centrality 257 in one entry.

Indegree centrality 252, outdegree centrality 253, and degree centrality 254 store the indegree, outdegree, and degree centrality calculated in the flowchart 60 shown in FIG. 12, closeness centrality 255 stores the closeness centrality calculated in the flowchart 61 shown in FIG. 13, betweenness centrality 256 stores the betweenness centrality calculated in the flowchart 62 shown in FIG. 14, and centrality 257 stores the centrality calculated in the flowchart 63 shown in FIG. 15.

The individual evaluation table 250 of the network evaluation information 25 is a table that is generated for each date of the individual 30.

Figure 18 is a diagram showing the data of the group evaluation table 258 stored in the network evaluation information 25 of Example 1. The group evaluation table 258 stores the evaluation results for each date for each group for which a network evaluation was performed in step S103 of Figure 3.

The group evaluation table 258 includes, in each entry, a date and time 259, an in-group degree 260, an out-group degree 261, an in-group degree 262, an out-group degree 263, a total degree 264, a liveness level 265, an individual with the highest degree centrality 266, an individual with the highest closeness centrality 267, an individual with the highest betweenness centrality 268, and a central individual 269.

The date and time 259 in the group evaluation table 258 corresponds to the date and time 251 in the individual evaluation table 250.

The degree within a group 260, the out-degree to outside the group 261, the in-degree to the group 262, the out-degree to the group 263, the total degree 264, and the activity 265 store the degree within a group, the out-degree from within a group to another group, the in-degree from outside another group to the group, the degree outside the group, the total degree, and the group activity calculated in the flowchart 64 of FIG. 16.

The individual with the highest degree centrality 266 stores the individual with the highest degree centrality calculated in the flowchart 60 of FIG. 12, the individual with the highest closeness centrality 267 stores the individual with the highest closeness centrality calculated in the flowchart 61 of FIG. 13, the individual with the highest betweenness centrality 268 stores the individual with the highest betweenness centrality calculated in the flowchart 62 of FIG. 14, and the central individual 269 stores the individual with the highest centrality calculated in the flowchart 63 of FIG. 15.

Figure 20 is a diagram showing the relationship between the sensor data 22 and the synchronous network time series data 24. The illustrated example shows a portion of the synchronous network time series data 24 generated from acceleration data obtained from the movement information table 202 of the sensor data 22, displayed as a graph, to show the time-series changes in the relationship between multiple individuals 30.

At time t1, the individual 30 with ID="1" has relationships with two individuals 30 with ID="2" and "4", and the individual 30 with ID="3" has relationships with an individual 30 with ID="6". At time t2, after a predetermined time interval Δt (e.g., one minute) has elapsed, ID="1" changes the individual 30 with which it is related from ID="4" to ID="5", and has relationships with two individuals 30 with ID="2" and "5". At time t3, ID="1" has only one individual 30 with ID="5".

The synchronous network time series data 24 is a network of synchronization between individuals 30, and by comparing the synchronous networks between individuals 30 over time, it is possible to analyze sports involving intense group movements, such as soccer, rugby, or basketball.

Figure 21 shows an example of extracting a clique from the synchronous network time series data 24. As shown in Figure 20, a graph can be generated from the synchronous network time series data 24, and changes in the related destinations of an individual 30 can be extracted in time series.

The clique evaluation unit 15 of the analysis unit 10 extracts a predetermined number of cliques from the synchronous network time series data 24, and aggregates and outputs the cliques at each time in time series. As shown in the figure, the clique evaluation unit 15 of this embodiment extracts two cliques consisting of two individuals 30 and three cliques consisting of three individuals 30 as evaluation targets.

The graph at time t1 in Figure 20 can be decomposed into a relationship between two cliques and three cliques as shown in Figure 21.

In FIG. 21, the clique evaluation unit 15 connects related individuals 30 with one edge, and extracts a graph with the nodes at both ends of the edge as 2 cliques. In other words, a combination of nodes with degree 1 forms a 2 clique. For example, in the graph at time t1, IDs "1" and "2" form 2 cliques, and furthermore, IDs "1" and "4", IDs "2" and "4", and IDs "3" and "6" are extracted as 2 cliques. Note that in this embodiment, the direction of the edges is not evaluated, and the graph is treated as an undirected graph.

When the destination node of two cliques is related to a third node, and the source node of the relationship is also related to the same third node, the clique evaluation unit 15 extracts these three nodes as a three-clique.

Each node in the 3-clique has degree 2, and the three nodes are related to each other. In the graph at time t1, in the 2-clique with IDs "1" and "2", node "1" is the source of the relationship, and node "2" is the destination of the relationship with "1". The destination node "2" has a relationship with another (third) node "4", and node "1" of the source of the relationship has a relationship with the same other node "4". In this way, a configuration in which the edges connecting the three nodes are closed is called a 3-clique.

Figure 22 is a flowchart showing an example of the processing performed by the clique evaluation unit 15. This processing is performed in step S104 of Figure 3.

The clique evaluation unit 15 first acquires data to be analyzed from the synchronous network time series data 24 (S1141). The clique evaluation unit 15 extracts two cliques for each date and time (or time) of the synchronous network time series data 24 (S1142).

The clique evaluation unit 15 classifies the data of the two extracted cliques into three types as shown in FIG. 23 (S1143). If the individual 30 is competing with two teams, Team A and Team B, the types of the two cliques will be the following three types.

A combination in which the two individuals 30 are only from team A is type A2, a combination in which the two individuals 30 are only from team B is type B2, and a combination in which one individual 30 each from team A and team B is type A1B1. Types A2 and B2 can be considered to be a state in which allies are working together. Also, clique type A1B1 can be considered to be a state in which allies and enemies are facing each other and marking each other.

The clique evaluation unit 15 can detect the type of relationship between a group of individuals 30 by classifying the type of clique according to the group to which the individuals belong and the number of individuals 30 for the extracted clique.

The clique evaluation unit 15 aggregates the data of the two cliques classified by type by time and stores it in the two clique information 410 that constitutes the clique evaluation information 40 (S1143). FIG. 24 is a diagram showing an example of the two clique information 410. The two clique information 410 includes, in one record, a date and time 4101 that stores the date and time of the synchronous network time series data 24, a first node 4102 that stores the identifier of the node (individual 30) that is the source of the relationship, a second node 4103 that stores the identifier of the individual 30 of the node that is the destination of the relationship, and the classified clique type 4104.

Next, the clique evaluation unit 15 chronologically aggregates the data of the two cliques classified by type, and generates and outputs time series data for each type (S1144). FIG. 25 is a graph showing an example of time series data 420 for each type of two cliques. The time series data 420 for each type of two cliques is also data that constitutes the clique evaluation information 40. The time series data 420 for each type of two cliques is displayed on the display of the input/output device 56 or the display of an external computer.

The time series data 420 for two cliques by type is a time series graph of the two clique information 410 by type, with the horizontal axis representing time and the vertical axis representing the number of cliques. The time series data 420 for two cliques by type can represent the frequency of occurrence of a state in which each individual 30 cooperates with a teammate (type A2 or B2) in the entire group, and a state in which each individual marks each other (type A1B1). In particular, in the case of soccer, types A2 and B2 can be considered as states in which the ball is being kept.

Next, the clique evaluation unit 15 chronologically aggregates the two-clique information 410 classified by type for each identifier of the individuals 30 to generate and output time-series data for each individual (S1145). FIG. 26 is a graph showing an example of individual-specific time-series data 430 for two cliques. The individual-specific time-series data 430 for the two cliques is displayed on the display of the input/output device 56 or the display of an external computer.

The individual time series data 430 of the two cliques is a time series graph with the horizontal axis representing time and the vertical axis representing the number of cliques, for each identifier of the individual 30 (first node 4102 of the two clique information 410). The illustrated example shows an example of displaying time series data by type for the individual 30 with ID = "1".

The individual time series data 430 of the two cliques can represent, in time series, the frequency of occurrence of each individual 30 in a state where it is cooperating with allies (type A2) and in an attacking or defensive state (type A1B1). By displaying the individual time series data 430 of the two cliques on the display of the input/output device 56 or the display of an external computer, the user of the server 5 can observe the changes in the manner in which the individuals 30 cooperate with the group in time series and evaluate each individual 30.

By processing steps S1142 to S1145 above, the clique evaluation unit 15 generates the two-clique information 410, the species-specific time series data 420, and the individual-specific time series data 430 that constitute the clique evaluation information 40.

In steps S1142 to S1145, the three cliques are processed in the same manner as in steps S1142 to S1145. The clique evaluation unit 15 can execute the processes of steps S1142 to S1145 and steps S1142 to S1145 in parallel. However, the clique evaluation unit 15 may execute the processes of steps S1142 to S1145 and steps S1142 to S1145 sequentially.

The clique evaluation unit 15 extracts three cliques for each date and time (or) of the synchronous network time series data 24 (S1146).

The clique evaluation unit 15 classifies the extracted data of the three cliques into four types as shown in FIG. 27 (S1147). If the individual 30 is competing with two teams, Team A and Team B, the types of the three cliques will be the following four types.

A combination in which the three individuals 30 are only from team A is type A3, a combination in which the three individuals 30 are only from team B is type B3, a combination in which the three individuals 30 are only from team B is type A2B1, and a combination in which the three individuals 30 are only from team A and team B is type A1B2.

When the clique type is A3 or B3, it can be understood as a state where teammates are working together. When it is type A2B1, it can be understood as a state where team A has a numerical advantage. When it is type A1B2, it can be understood as a state where team B has a numerical advantage.

The clique evaluation unit 15 collects the data of the three cliques classified by type by time and stores it in the three clique information 440 constituting the clique evaluation information 40. FIG. 28 is a diagram showing an example of the three clique information 440. The three clique information 440 includes, in one record, a date and time 4401 that stores the date and time of the synchronous network time series data 24, a first node 4402 that stores the identifier of the individual 30 of the node that is the source of the relationship, a second node 4403 that stores the identifier of the individual 30 of the node that is the destination of the relationship, a third node 4404 that stores the identifier of the individual 30 of another node that is the destination of the relationship, and the type of the classified clique 4405.

Next, the clique evaluation unit 15 chronologically aggregates the data of the three cliques classified by type, and generates and outputs time series data for each type (S1148). Figures 29 and 30 are graphs showing an example of time series data 450 for each type of three cliques. The time series data 450 for each type of three cliques is also data that constitutes the clique evaluation information 40. The time series data 450 for each type of three cliques is displayed on the display of the input/output device 56 or the display of an external computer.

Note that Figure 29 shows type-specific time series data 450 for types A3 and B3, and Figure 30 shows type-specific time series data 450 for types A2B1 and A1B2. The graphs are divided into two for ease of viewing, but they may be one graph.

The three clique type time series data 450 is a time series graph of the three clique information 440 type, with the horizontal axis representing time and the vertical axis representing the number of cliques. The three clique type time series data 450 can represent the frequency of occurrence of a state in which the individuals 30 cooperate with their teammates in the entire group (type A3 or B3) and a state of numerical superiority (type A2B1, A1B2) in a time series. In particular, in the case of soccer, types A3 and B3 can be considered to be states in which ideal cooperative play is possible, avoiding the opponent's mark.

Next, the clique evaluation unit 15 chronologically aggregates the 3-clique information 440 classified by type for each identifier of the individual 30 to generate and output time-series data for each individual (S1149). FIG. 31 is a graph showing an example of individual time-series data 460 for the 3 cliques. The individual time-series data 460 for the 3 cliques is displayed on the display of the input/output device 56 or the display of an external computer.

The individual time series data 460 of the three cliques is a time series graph with the horizontal axis representing time and the vertical axis representing the number of cliques, for each identifier of the individual 30 (first node 4402 of the three clique information 440). The illustrated example shows an example of displaying time series data by type for the individual 30 with ID = "1".

The individual time series data 460 for the three cliques can show the frequency of occurrence of a state in which individual 30 (ID = "1") cooperates with allies (type A3) and a state in which it has a numerical advantage (types A2B1, A1B2) in time series.

By processing steps S1146 to S1149 above, the clique evaluation unit 15 generates the three clique information 440 constituting the clique evaluation information 40, the time series data for each type of the three cliques 450, and the time series data for each individual of the three cliques 460.

By performing the above process, the relationships between the individuals 30 are extracted from the sensing data collected by the sensing device 1, and clique evaluation information 40 is generated and output in real time, making it possible to grasp the relationships between the individuals 30 during a competition in real time, and to send accurate instructions to each individual 30.

A manager or coach of a group can plan a strategy for a competition by referring to each piece of information in the clique evaluation information 40. For example, in the case of soccer, the two most frequently used cliques are considered to dominate the rhythm of the entire network. In addition, the diversity of each type of the three cliques can be used to understand the organizational play of the team and the rhythm of the entire team.

Therefore, if the two most frequently occurring cliques among the individuals 30 of the opposing team are suppressed, it is expected that the diversity of the opposing team's three cliques will be destroyed. Therefore, by sending an instruction to the team to mark or press the two most frequently occurring cliques, it is possible to improve the content of the game.

In addition, by acquiring the movements of each individual 30 in real time from the various clique evaluation information 40 of 2-clique and 3-clique, it is possible to extract individuals 30 that perform effective movements regardless of whether the ball is present or not.

For example, an individual 30 that frequently appears in type A3 based on information from three cliques can be estimated to be a central player. Also, an individual 30 that appears in types A1B2 or A2B1 based on information from three cliques can be evaluated as making effective moves such as marking an opponent or attracting an opponent, regardless of whether or not the ball is in hand.

As described above, the network analysis system of this embodiment provides clique evaluation information 40 in real time from sensing data from the sensing device 1 attached to the individual 30, making it possible to visualize the individual 30 that is making effective moves in relation to the content of the competition, and making it possible to plan strategies and evaluate the individual 30 in real time.

In addition, the two-clique or three-clique clique evaluation information 40 makes it possible to evaluate an individual 30 regardless of whether or not there is a ball or the centrality of the game, making it possible to accurately evaluate players in games with high levels of movement.

The network analysis system of this embodiment is versatile and can be applied to a variety of sports. As mentioned above, the analysis subject of this embodiment is not limited to soccer, but can also be applied to rugby, volleyball, etc.

In addition, since the network analysis system of this embodiment uses only the number of cliques as an explanatory variable, there is no need to refer to a large number of parameters as in the conventional technology, and the behavior of the individual 30 can be easily understood and quickly evaluated.

In addition, in the above Example 1, examples of two and three cliques are described as being extracted from the synchronous network time series data 24, but the number of cliques to be extracted is not limited to the above and can be changed as appropriate depending on the type of sport to be analyzed, etc.

In addition, in the above Example 1, an example was shown in which the clique types of individuals 30 competing in two teams, Team A (first group) and Team B (second group), were classified, but the clique types can be set according to the number of teams competing simultaneously, and are not limited to the above two teams.

Figure 32 shows a second embodiment of the present invention, and is a block diagram showing an example of the configuration of a network analysis system. In the first embodiment, an example was shown in which the server 5 generates clique evaluation information 40 in real time based on acceleration and position information acquired by the sensing device 1 attached to the individual 30, but there are cases in which it is not possible to attach the sensing device 1 to all of the individuals 30 participating in the competition.

In this embodiment, a camera is installed on a stadium or the like to collect video data of the competition without using the sensing device 1, and clique evaluation information 40 is generated. In this embodiment, one or more cameras 70 are installed on the field 100 to obtain video data of the individuals 30 competing, and the video data is stored in the video data 27 of the server 5.

In the server 5, the image tracking unit 16 reads the video data 27, identifies the individual 30, extracts the location information, calculates the moving speed, etc. The other configurations are the same as in the first embodiment, and the same components are given the same reference numerals and redundant explanations are omitted.

Figure 33 is a flowchart showing an example of processing performed by the server 5. First, the image tracking unit 16 in the server 5 acquires the video data 27 stored in the auxiliary storage device 55 (S211).

The image tracking unit 16 identifies the individual 30 from the loaded video data 27 and calculates its position on the field 100 (S212). Note that the individual 30 can be identified by applying well-known techniques such as face recognition. Also, the identified individual 30 may be associated with the user ID 2021 (movement information table 202) by an administrator who operates the input/output device 56 or an external computer.

The image tracking unit 16 calculates the acceleration (or movement speed) from the position information of the individual 30 detected for each frame of the video data 27, and sets it as the accelerations 2023 to 2025 in the movement information table 202 (S213).

When the analysis unit 10 of the server 5 has completed the calculation of the acceleration for each individual, the feature calculation unit 11 calculates the feature information 21 using the movement information table 202 as an input, in the same manner as in Example 1. Then, the network analysis unit 12 generates the synchronous network time series data 24 based on the feature information 21, in the same manner as in Example 1 (S214).

Next, the clique evaluation unit 15 acquires the synchronous network time series data 24 (S215), and extracts two-cliques and three-cliques in the same manner as in Example 1 to generate clique evaluation information 40 (S216).

The display unit 14 outputs the generated clique evaluation information 40 to the display of the input/output device 56 or an external computer, and presents a two-clique graph or a three-clique graph to the administrator of the group, etc.

As described above, in this embodiment, even in a situation where the sensing device 1 cannot be attached to the individual 30, it is possible to evaluate the individual 30 based on the clique evaluation information 40, as in the first embodiment.

<Conclusion>
As described above, the network analysis systems of the first and second embodiments can be configured as follows.

(1) A network analysis method in which a computer (server 5) including a processor (CPU 51) and a memory (memory 54) analyzes a network between individuals (30) in a group, the method comprising: a sensing data collection step in which the computer (5) collects time-series information (sensing data) regarding the movements of individuals exercising in a group as sensor data (22); a feature calculation step (feature calculation unit 11) in which the computer (5) calculates features (feature information 21) regarding the movements of each individual from the sensor data (22); The network analysis method includes a network analysis step (network analysis unit 12) in which the relationship between the individuals (30) for combinations of individuals in the population is analyzed based on the feature amount, and a network between the individuals is generated in time series from the analysis result of the relationship between the individuals, and a clique evaluation step (clique evaluation unit 15) in which the computer (5) extracts a group consisting of a plurality of individuals (30) from among the individuals (30) having a relationship in the network as a clique, and analyzes the clique in time series.

The above configuration makes it possible to analyze the network between individuals 30 in a population and visualize individuals 30 that make effective moves from multiple cliques.

(2) The network analysis method according to (1) above, wherein the individual (30) belongs to either a first group (Team A) or a second group (Team B), and the clique evaluation step (15) classifies the clique into a clique type that is preset according to a combination of the individual (30) and the group to which the individual (30) belongs.

With the above configuration, the clique evaluation unit 15 can detect the type of relationship between a group of individuals by classifying the type of clique according to the group to which the individuals 30 belong and the number of individuals 30 for the extracted clique.

(3) The network analysis method according to (2) above, wherein the clique evaluation step (15) includes aggregating the classified clique types for each individual (30) in a time series manner and outputting the results as individual time series data (430, 460).

With the above configuration, the time series data 430 for each individual is displayed on the display of the input/output device 56 or on the display of an external computer, allowing the user of the server 5 to observe the changes in the manner in which the individual 30 interacts with the group over time and evaluate each individual 30.

(4) The network analysis method according to (2) above, wherein the clique evaluation step (15) includes aggregating the classified clique types in a time series and outputting the types as time series data (420, 450).

With the above configuration, the time series data 420 by type is displayed on the display of the input/output device 56 or on the display of an external computer, allowing users of the server 5 to observe the changes in the individuals 30 that are cooperating within the population over time.

The present invention is not limited to the above-described embodiments, but includes various modified examples. For example, the above-described embodiments are described in detail to clearly explain the present invention, and are not necessarily limited to including all of the configurations described. It is also possible to replace part of the configuration of one embodiment with the configuration of another embodiment, and it is also possible to add the configuration of another embodiment to the configuration of one embodiment. In addition, the addition, deletion, or replacement of part of the configuration of each embodiment with other configurations can be applied alone or in combination.

The above configurations, functions, processing units, and processing means may be realized in part or in whole in hardware, for example by designing them as integrated circuits. The above configurations and functions may be realized in software by a processor interpreting and executing a program that realizes each function. Information on the programs, tables, files, etc. that realize each function may be stored in a memory, a recording device such as a hard disk or SSD (Solid State Drive), or a recording medium such as an IC card, SD card, or DVD.

In addition, the control lines and information lines shown are those considered necessary for the explanation, and not all control lines and information lines on the product are necessarily shown. In reality, it can be assumed that almost all components are interconnected.

REFERENCE SIGNS LIST 1 Sensing device 5 Server 10 Analysis unit 11 Feature amount calculation unit 12 Network analysis unit 13 Network evaluation unit 14 Display unit 15 Clique evaluation unit 21 Feature amount information 22 Sensor data 24 Synchronous network time series data 25 Network evaluation information 40 Clique evaluation information 51 CPU
52 Memory 55 Auxiliary storage device

Claims (12)

A network analysis method in which a computer including a processor and a memory analyzes a network between individuals in a population, comprising:
a sensing data collection step in which the computer collects, as sensor data, time-series information on the movements of individuals exercising in a group;
a feature calculation step in which the computer calculates a feature related to the movement of each of the individuals from the sensor data;
a network analysis step in which the computer analyzes relationships between individuals for combinations of individuals in the population based on the feature amount, and generates a network between individuals in time series from the analysis result of the relationships between the individuals;
a clique evaluation step in which the computer extracts a group of a plurality of individuals having mutual relationships among the individuals having relationships in the network as a clique, and analyzes the clique in a time series;
A network analysis method comprising: 2. The network analysis method of claim 1, comprising:
The individual is
Belonging to either the first group or the second group,
The clique evaluation step includes:
A network analysis method comprising: classifying the cliques into predetermined clique types according to a combination of the individuals and a group to which the individuals belong. 3. A network analysis method according to claim 2, comprising:
The clique evaluation step includes:
A network analysis method comprising: tabulating the classified types of cliques for each individual in time series and outputting the results as individual time series data. 3. A network analysis method according to claim 2, comprising:
The clique evaluation step includes:
A network analysis method comprising: tabulating the classified types of cliques in time series and outputting the types as time series data. a server including a processor and a memory;
A network analysis system including a sensing device that acquires time-series information regarding the movement of an individual,
The sensing device transmits time-series information regarding the movements of individuals exercising in a group to the server;
The server,
an information collecting unit that collects, as sensor data, time-series information regarding the movement of the individual received from the sensing device;
a feature extraction unit that calculates a feature related to the movement of each individual from the sensor data;
a network analysis unit that analyzes relationships between individuals for combinations of individuals in the population based on the feature amount, and generates a network between individuals in time series from a result of the analysis of the relationships between the individuals;
a clique evaluation unit that extracts a group of a plurality of individuals having mutual relationships among the individuals having relationships in the network as a clique and analyzes the clique in a time series;
A network analysis system comprising: 6. The network analysis system of claim 5,
The individual is
Belonging to either the first group or the second group,
The clique evaluation unit includes:
A network analysis system comprising: a network analysis unit that classifies the cliques into predetermined clique types according to a combination of the individuals and a group to which the individuals belong. 7. The network analysis system of claim 6,
The clique evaluation unit includes:
A network analysis system characterized in that the classified types of cliques are compiled in time series for each individual and output as individual time series data. 7. The network analysis system of claim 6,
The clique evaluation unit includes:
A network analysis system characterized in that the classified types of cliques are compiled in time series and output as time series data for each type. A server including a processor and a memory,
an information collecting unit that receives time-series information on the movements of individuals exercising in a group and collects the information as sensor data;
a feature extraction unit that calculates a feature related to the movement of each individual from the sensor data;
a network analysis unit that analyzes relationships between individuals for combinations of individuals in the population based on the feature amount, and generates a network between individuals in time series from a result of the analysis of the relationships between the individuals;
a clique evaluation unit that extracts a group of a plurality of individuals having mutual relationships among the individuals having relationships in the network as a clique and analyzes the clique in a time series;
A server comprising: 10. The server of claim 9,
The individual is
Belonging to either the first group or the second group,
The clique evaluation unit includes:
A server which classifies the cliques into predetermined clique types according to a combination of the individuals and a group to which the individuals belong. 11. The server of claim 10,
The clique evaluation unit includes:
A server that counts the classified types of cliques for each individual in time series and outputs the counted types of cliques as individual time series data. 11. The server of claim 10,
The clique evaluation unit includes:
A server that aggregates the classified clique types in time series and outputs the aggregated types as time series data.

Images