Activity feature extraction device and activity feature extraction method

Abstract

To provide an activity feature extraction device and an activity feature extraction method capable of extracting an activity group having a feature corresponding to an actual state of time and a place where a person actively acts, and allowing a measures taker to appropriately examine a type, execution time, and an execution place of execution measures.SOLUTION: An activity feature extraction device is configured by a server computer including an arithmetic device that executes predetermined arithmetic processing and a storage device that can be accessed by the arithmetic device. The server computer includes: a spatial separation calculation unit configured to calculate spatial separation representing spatial difference between activity path data; a temporal separation calculation unit configured to calculate temporal separation representing temporal difference between the activity path data; a spatiotemporal clustering unit that calculates an activity feature similarity from the spatial separation and the temporal separation and clusters the activity path data using the calculated activity feature similarity; and an output generation unit that outputs information on a cluster generated by the spatiotemporal clustering unit.SELECTED DRAWING: Figure 1

Description

[Technical field]

[0001] The present invention relates to an activity feature extraction device and an activity feature extraction method for analyzing activity route data including movement trajectory information. [Background technology]

[0002] In recent years, with the spread of e-commerce and the introduction of remote work, people's activities are becoming less restricted by physical location and time, and are becoming more diverse. In order for transportation operators and commercial facility operators to implement effective policies in the midst of such social changes, it is important to analyze the actual activities of people and present them in a format that is understandable to policymakers.

[0003] As a related technique of the present technology, there is Patent Document 1 (JP Patent Publication 2016-177452 A). The document describes a travel route data anonymization device including a route data acquisition means for acquiring travel route data including a travel route and a time period during which the travel was made, a route anonymization means for classifying the plurality of travel route data based on the similarity of the travel routes so that the number of travel route data in the same class is equal to or greater than a first threshold value, and unifying the travel routes of the travel route data in the same class into a representative travel route, a time anonymization means for abstracting a time period during which the travel was made in the travel route data so that the travels made in the same time period for the travel route data in the same class are equal to or greater than a second threshold value, and an output means for outputting the travel route data processed by the route anonymization means and the time anonymization means. [Prior art documents] [Patent documents]

[0004] [Patent Document 1] JP 2016-177452 A Summary of the Invention [Problem to be solved by the invention]

[0005] In the aforementioned Patent Document 1 (JP Patent Publication No. 2016-177452), in order to anonymize travel route data, the travel route data, which includes the travel route (location information) and the time period of travel, is classified from the perspective of location, and then abstracted in a format that can be effectively used from the perspective of time period.

[0006] When classifying movement routes from the viewpoints of location and time period, the appropriate number of classification clusters and the threshold value of the time period width in classification from the viewpoint of time period change depending on the tendency of movement route data classified into the same cluster from the viewpoint of location when classifying movement route data according to the actual activity. Similarly, the appropriate number of classification clusters and the threshold value of the location width in classification from the viewpoint of location change depending on the tendency of movement route data classified into the same cluster from the viewpoint of time period. However, when classification of movement routes is divided into two stages, from the viewpoint of location and the viewpoint of time period, the dependency between location and time period is not taken into consideration.

[0007] For this reason, when extracting activity features, it is necessary to take into account the location information and time information of the movement trajectory in an integrated manner in order to classify movement routes in accordance with the actual times and places where people are active. [Means for solving the problem]

[0008] A representative example of the invention disclosed in the present application is as follows: That is, an activity feature extraction device that analyzes activity route data in which human movement trajectory information and time information are associated with each other is configured by a computer having a calculation device that executes a predetermined calculation process and a storage device accessible by the calculation device, and is characterized in that the calculation device includes a spatial separation calculation unit that calculates a spatial separation that represents a spatial difference between the activity route data, a temporal separation calculation unit that calculates a temporal separation that represents a temporal difference between the activity route data, a spatio-temporal clustering unit that calculates an activity feature similarity from the spatial separation and the temporal separation and clusters the activity route data using the calculated activity feature similarity, and an output generation unit that outputs information on the clusters generated by the spatio-temporal clustering unit. Effect of the Invention

[0009] According to one aspect of the present invention, it is possible to extract activity groups that have characteristics that correspond to the actual times and places where people are active, and it is possible for policy examiners to appropriately consider the type, implementation time, and implementation place of the implementation policy. The problems, configurations, and effects other than those described above will be made clear by the explanation of the following examples. [Brief description of the drawings]

[0010] [Figure 1] 1 is a functional block diagram showing a configuration of an activity feature extraction device according to a first embodiment of the present invention. [Diagram 2] 1 is a block diagram showing a hardware configuration of an activity feature extraction device according to a first embodiment of the present invention. [Diagram 3] FIG. 2 is a diagram showing a configuration of an activity route database and an example of data in the first embodiment of the present invention. [Figure 4] 3 is a flowchart of a process executed by an activity feature extraction device according to a first embodiment of the present invention. [Diagram 5] 4 is a flowchart of a spatial separation degree calculation process executed by a spatial separation degree calculation unit according to the first embodiment of the present invention. [Figure 6] 4 is a flowchart of a time separation calculation process executed by a time separation calculation unit according to the first embodiment of the present invention. [Figure 7] 4 is a flowchart of a spatio-temporal clustering process executed by a spatio-temporal clustering unit according to the first exemplary embodiment of the present invention. [Figure 8] 4 is a flowchart of a feature extraction process executed by a feature extraction unit according to the first embodiment of the present invention. [Figure 9] FIG. 2 is a diagram showing an example of a setting input screen output from an output device of the activity feature extraction device according to the first embodiment of the present invention. [Figure 10] FIG. 2 is a diagram showing an example of an activity characteristic display screen output from an output device of the activity characteristic extraction device according to the first embodiment of the present invention. [Figure 11] FIG. 11 is a functional block diagram showing a configuration of an activity feature extraction device according to a second embodiment of the present invention. [Figure 12] 10 is a flowchart of a process executed by an activity feature extraction device according to a second embodiment of the present invention. [Figure 13] FIG. 11 is a diagram showing an example of a person attribute input screen output from an output device of the activity characteristic extraction device according to the second embodiment of the present invention. [Figure 14] FIG. 11 is a diagram showing an example of a spot recommendation screen output from an output device of the activity characteristics extraction device according to the second embodiment of the present invention. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

[0011] (First embodiment) Hereinafter, an embodiment of the present invention will be described with reference to the drawings.

[0012] In this embodiment, the activity feature extraction device 100 clusters the activity route data including the person's movement trajectory information and the time information accompanying the movement trajectory information using the activity feature similarity based on both the spatial difference and the temporal difference between the activity route data, thereby extracting activity groups having characteristics according to the actual time and place where a person is active. Therefore, in this embodiment, it is possible to support the policy examiner in appropriately examining the type, implementation time, and implementation place of the implementation policy.

[0013] FIG. 1 is a functional block diagram showing a configuration of an activity feature extraction device 100 according to a first embodiment of the present invention.

[0014] The activity feature extraction device 100 of this embodiment includes an input device 101 , an output device 102 , and a server computer 103 .

[0015] The input device 101 is an input interface, such as a mouse, a keyboard, or a touch device, for transmitting user operations to the server computer 103. The output device 102 is an output interface configured with a display device such as a liquid crystal display, and is an output interface for displaying data output by the server computer 103 and for interactive operations with the user. For example, the input device 101 and the output device 102 may be integrated by using a so-called touch panel or the like. The output device 102 displays a setting input screen 901 (FIG. 9), an activity characteristic display screen 1001 (FIG. 10), a person characteristic input screen (FIG. 13), and a spot recommendation screen 1401 (FIG. 14), which will be described later. The input device 101 accepts user operations on input fields and buttons in each screen. The output device 102 may be the display device described above, a printer that prints processing results on paper, or a data output interface that transfers processing results to another system. Also, as in the second embodiment, the input device 101 and the output device 102 may be terminals connected to the activity feature extraction device 100 via the network 113. In this case, the activity feature extraction device 100 has a function of a web server, and the terminals access the activity feature extraction device 100 using a predetermined protocol (e.g., http).

[0016] The server computer 103 has, as its functional configuration, a setting input unit 104, an activity route data acquisition unit 105, a spatial separation calculation unit 106, a temporal separation calculation unit 107, a spatio-temporal clustering unit 108, a feature extraction unit 109, an output generation unit 110, and an activity route database 111.

[0017] The setting input unit 104 acquires the target area information, target date and time information, and classification setting information input to the input device 101, sends the target area information and the target date and time information to the activity route data acquisition unit 105, and transmits the classification setting information input to the input device 101 to the spatio-temporal clustering unit 108. Here, the target area information may include at least one of the address of the target area, the distance between the facility of interest and the facility, and a coordinate group representing the polygon shape of the target area range. The target date and time information may include at least one of the date, day of the week, temperature conditions, seasonal conditions, weather conditions, and day of the week conditions. The classification setting information may include at least one of the number of clusters, inter-data similarity threshold, minimum number of data, model structure, and model parameters.

[0018] The activity route data acquisition unit 105 acquires the target area information and the target date and time information from the setting input unit 104, extracts a group of activity route data recorded within the area specified by the target area information and within the date and time specified by the target date and time information from the activity route data stored in the activity route database 114, and transmits them to the spatial separation calculation unit 106 and the temporal separation calculation unit 107. Here, the activity route data includes at least person movement trajectory information and time information associated with the movement trajectory information.

[0019] The spatial separation calculation unit 106 acquires an activity route data group from the activity route data acquisition unit 105, and calculates spatial separation for every combination of two activity route data included in the activity route data group. Here, the spatial separation represents a geographical difference in movement trajectories between the two activity route data.

[0020] The spatial separation is calculated by searching for the closest pairs of observation points that can be determined to be the closest pairs between all observation points of a certain movement trajectory and all observation points of a movement trajectory to be compared while satisfying the constraint of the observation order, for example, as shown in Equation (1), between activity path data A and B, the observation point p included in activity path data A that corresponds to the closest pair k is searched for as follows: A(k) and observation point p included in activity path data B B(k)and the geographical condition coefficient a, which is determined by the characteristics of the location where the two observation points are located. k Using this method, the difference between the movement trajectories themselves is calculated using a scale that can take into account the difference between the movement trajectories of activity route data, taking into account the effects of observation errors that depend on the surrounding environment, such as the size of the facility where each observation point is located, and transportation means, indoor / outdoor status, and radio wave environment.

[0021] Thereafter, the spatial separation calculation unit 106 sends the group of closest pairs to the temporal separation calculation unit 107, and transmits the calculated spatial separation to the spatio-temporal clustering unit .

[0022]

number

[0023] The time separation calculation unit 107 acquires an activity route data group from the activity route data acquisition unit 105, acquires a nearest pair group from the spatial separation calculation unit 106, and calculates a time separation for every combination of two activity route data included in the acquired activity route data group. Here, the time separation represents the time difference between two activity route data.

[0024] For example, the time separation is expressed as follows: between activity path data A and B, the observation point p corresponding to the closest pair k is A(k) The speed of v A(k) , observation point p B(k) The speed of v B(k) , the observation time difference between the two observation points Δt k The difference between the activity route data in at least one of the following aspects is calculated: the time period when the activity was performed, the means of transportation used, and whether the activity was passed through or stayed.

[0025] Thereafter, the time separation calculation unit 107 transmits the calculated time separation to the spatio-temporal clustering unit .

[0026]

number

[0027] The spatio-temporal clustering unit 108 acquires classification setting information from the setting input unit 104, acquires spatial separation from the spatial separation calculation unit 106, acquires temporal separation from the temporal separation calculation unit 107, and clusters the activity route data group using the activity feature similarity based on the spatial separation and temporal separation. The activity feature similarity may be a dissimilarity, and the temporal separation calculation unit 107 converts the temporal separation into a distance dimension by defining it in a format of multiplying the observed time difference by the speed, so that, for example, as in formula (3), the spatial separation D L (A,B) and time separation D T The activity feature similarity can be calculated by directly adding (A, B).

[0028]

number

[0029] The calculation method of the clustering may be changed according to the acquired classification setting information. When the number of clusters is acquired, for example, k-meanS is used to classify the activity route data into the number of clusters based on the activity feature similarity, and an activity route cluster is generated. On the other hand, when the inter-data similarity threshold is acquired, for example, a density-based clustering method such as DenSity-based Spatial clustering of applicationS with noiSe (DBSCAN) is used to cluster the activity route data based on the activity feature similarity, and an activity route cluster is generated so that the activity feature similarity between the activity route data in each cluster satisfies the condition of the inter-data similarity threshold. Then, the activity route cluster is transmitted to the feature extraction unit 109.

[0030] The feature extraction unit 109 acquires the activity route clusters from the time-space clustering unit 108, and extracts cluster feature information from the activity route data in each activity route cluster. Here, the cluster feature information is information on the tendencies of the people who generated the activity route data in each activity route cluster, and preferably includes at least one of age, sex, occupation, work start time, work finish time, and hobby.

[0031] When the activity route data is provided with person attribute information, the cluster feature information may be extracted by aggregating or existing statistical processing the person attribute information of the person who generated the activity route data in the activity route cluster. When the activity route data is not provided with person attribute information, for example, map information including facility information and traffic network information of the planning target area is acquired from the outside via the network 113 and the network interface device 112, and the facilities through which the activity route data passed and stayed and the transportation means estimated to be used are extracted, and the ratio of people who take the activity route data may be extracted by aggregating or existing statistical processing from the user attributes of the facility and / or the transportation means and the time information of the activity route data. After that, the feature extraction unit 109 transmits the activity route cluster and the cluster feature information to the output generation unit 110.

[0032] The output generation unit 110 acquires the activity route cluster and cluster feature information from the feature extraction unit 109, acquires the person feature information and belonging cluster information from the belonging cluster calculation unit 1102, creates an activity feature display screen 1001 (FIG. 10), and outputs it to the output device 102. The configuration of the activity feature display screen 1001 will be described later with reference to FIG.

[0033] The activity route database 111 is a database for storing activity route data. When executing the extraction process, the activity route data acquisition unit 105 accesses the activity route database 111. The configuration of the activity route database 111 will be described later with reference to FIG.

[0034] FIG. 2 is a block diagram showing a hardware configuration of the activity feature extraction device 100 according to the first embodiment of the present invention.

[0035] The server computer 103 is, for example, a general computer having a processor 201 and a storage device 202 connected to each other. The storage device 202 is configured by any type of storage medium. For example, the storage device 202 may include a semiconductor memory and a hard disk drive.

[0036] In the illustrated example, functional units such as the setting input unit 104, the activity path data acquisition unit 105, the spatial separation calculation unit 106, the temporal separation calculation unit 107, the spatio-temporal clustering unit 108, and the feature extraction unit 109 shown in FIG. 1 are realized by a processor 201 executing a processing program 203 stored in a storage device 202.

[0037] In other words, in this embodiment, the processes executed by each of the functional units described above are actually executed by the processor 201 following the instructions written in the processing program 203. Also, the activity route database 111 is configured in the storage device 202. Also, the display by the output device 102 is executed by the processor 201 generating display data for display and outputting it to the output device 102, and the output device 102 displaying according to the display data.

[0038] The server computer 103 further includes a network interface device (NIF) 112 connected to the processor 201. The network interface device 112 is a communication interface that controls communications with other devices in accordance with a predetermined protocol.

[0039] The program executed by the processor 201 is provided to the activity feature extraction device 100 via a removable medium (such as a CD-ROM or a flash memory) or the network 113, and is stored in a non-volatile storage device 202, which is a non-transitory storage medium. For this reason, the server computer 103 may have an interface for reading data from the removable medium.

[0040] The activity feature extraction device 100 is a computer system configured on one physical computer, or on multiple logically or physically configured computers, and may operate on a virtual computer constructed on multiple physical computer resources. For example, each functional unit may operate on a separate physical or logical computer, or multiple functional units may be combined to operate on a single physical or logical computer.

[0041] 3 is a diagram showing a configuration and an example of data of the activity route database 111 according to the first embodiment of the present invention. Note that, although an example in which the activity route database 111 is configured in a table format is shown in the figure, the data format is not limited to the table format and may be other formats.

[0042] The activity route database 111 includes activity route data management information 300 shown in Fig. 3. The table configuration and field configuration of each table shown in Fig. 3 are preferable configurations in the present invention, but tables and fields may be added depending on the application.

[0043] The activity route data management information 300 includes an agent ID field 301 , a data acquisition time field 302 , a latitude field 303 , a longitude field 304 , and a person attribute field 305 .

[0044] The agent ID field 301 holds identification information (hereinafter referred to as a person ID) of the person who performed the action of each piece of trajectory data.

[0045] The data acquisition time field 302 holds information about the time when the trajectory point was acquired.

[0046] The latitude field 303 holds the latitude information of the location where the trajectory point was acquired.

[0047] The longitude field 304 holds the longitude information of the location where the trajectory point was acquired.

[0048] The person attribute field 305 holds the person attribute information of the user corresponding to the person ID described above. Here, the person attribute information includes age, sex, occupation, start and end times of work, hobbies, etc., and available information may be expressed as an array with numerical values ​​or one-hot vectors as elements.

[0049] FIG. 4 is a flowchart of the process executed by the activity feature extraction device 100 according to the first embodiment of the present invention.

[0050] First, when a user inputs target area information, target date and time information, and classification setting information to the input device 101, the input device 101 transmits the input information to the setting input unit 104 (step S401).

[0051] Next, the setting input unit 104 transmits the target area information and the target date and time information to the activity route data acquisition unit 105 (step S402), and transmits the classification setting information to the time-space clustering unit 108 (step S403).

[0052] Next, the activity route data acquisition unit 105 extracts an activity route data group from the activity route database 111 according to the conditions of the target area information and the target date and time information, and transmits the target activity route data group to the spatial separation calculation unit 106 and the temporal separation calculation unit 107 (step S404).

[0053] Next, the spatial separation degree calculation unit 106 extracts a group of closest pairs between the groups of activity path data, calculates the spatial separation, and then transmits the group of closest pairs to the temporal separation degree calculation unit 107 (step S405), and transmits the calculated spatial separation to the spatiotemporal clustering unit 108 (step S406). Details of the spatial separation degree calculation process will be described later with reference to FIG.

[0054] Next, the temporal separation calculation unit 107 calculates the temporal separation using the group of nearest neighbor pairs, and transmits the calculated temporal separation to the spatio-temporal clustering unit 108 (step S407). Details of the temporal separation calculation process will be described later with reference to FIG.

[0055] Next, the spatio-temporal clustering unit 108 calculates an activity feature similarity from the spatial separation degree and the temporal separation degree, generates an activity route cluster according to the activity feature similarity, and then transmits the activity route cluster to the feature extraction unit 109 (step S408). Details of the spatio-temporal clustering process will be described later with reference to FIG. 7.

[0056] Next, the feature extraction unit 109 extracts cluster feature information from the action route cluster, and transmits the action route cluster and the cluster feature information to the output generation unit 110 (step S409). Details of the feature extraction process will be described later with reference to FIG.

[0057] Finally, the output generation unit 110 creates an activity characteristic display screen 1001 (FIG. 10) and transmits it to the output device 102 (step S410). The user can check the processing result by looking at the activity characteristic display screen 1001 displayed on the output device 102.

[0058] FIG. 5 is a flowchart of the spatial separation calculation process executed by the spatial separation calculation unit 106 according to the first embodiment of the present invention.

[0059] First, the spatial separation degree calculation unit 106 acquires an activity route data group from the activity route data acquisition unit 105 (step S501).

[0060] Next, the spatial separation degree calculation unit 106 repeatedly executes the processes of steps S503 to S506 for all combinations of the two activity route data A and B in the acquired activity route data group that do not include overlaps (steps S502, S507).

[0061] In the repetition, first, the spatial separation calculation unit 106 calculates the distance between the observation points using the latitude and longitude information between the observation points of the activity route data A and B (step S503). Next, the spatial separation calculation unit 106 calculates a cumulative distance matrix (step S504) and searches for an optimal route that minimizes the cumulative distance (step S505). For example, dynamic programming can be used to calculate the cumulative distance matrix.

[0062] Thereafter, the spatial separation calculation unit 106 extracts the closest pairs of observation points between the activity route data A and B from the optimal route, and determines the cumulative distance of the optimal route as the spatial separation (S506). Then, the process proceeds to the next activity route data pair (S507).

[0063] Finally, the spatial separation calculation unit 106 outputs all the closest pair groups and their spatial separations (step S508).

[0064] FIG. 6 is a flowchart of the time separation calculation process executed by the time separation calculation unit 107 according to the first embodiment of the present invention.

[0065] First, the time separation degree calculation unit 107 acquires the closest pair group of each activity path data from the spatial separation degree calculation unit 106 (step S601).

[0066] Next, the time separation calculation unit 107 repeatedly executes the processes of steps S603 to S605 for all combinations of the two activity route data A and B in the acquired activity route data group that do not include overlaps (steps S602, S606).

[0067] In the repetition, first, the time separation calculation unit 107 calculates the observed time difference between the nearest pair of activity path data A and B (step S603), and multiplies the observed time difference by the average of two points of the speed at both observation points of the nearest pair to calculate the time separation between the nearest pair (step S604). After that, the time separation calculation unit 107 calculates the time separation between all nearest pairs, and calculates the time separation between the activity path data A and B by adding up the calculated time separations (step S605). Finally, the time separation calculation unit 107 outputs all the calculated time separations (step S607).

[0068] FIG. 7 is a flowchart of the spatio-temporal clustering process executed by the spatio-temporal clustering unit 108 according to the first embodiment of the present invention, and shows the process when the inter-data similarity threshold and the minimum number of data are given.

[0069] First, the spatio-temporal clustering unit 108 obtains the spatial separation from the spatial separation calculation unit 106 and the temporal separation from the temporal separation calculation unit 107 (step S701).

[0070] Next, the spatio-temporal clustering unit 108 adds the spatial separation and the temporal separation for each pair of activity route data to calculate the activity feature similarity between the activity route data (step S702). Since the temporal separation is converted into the dimension of the spatial separation using the speed, the activity feature similarity including the features of both the spatial separation and the temporal separation can be calculated by a simple addition operation. Then, the calculated activity feature similarity enables clustering that takes into account the duration of stay and the means of transportation.

[0071] Thereafter, the spatio-temporal clustering unit 108 randomly selects one activity route data to be focused on as an initial value from the activity route data group (step S703), and repeatedly executes the processes of steps S705 to S710 until all activity route data are assigned to clusters or noise (steps S704, S711).

[0072] In the repetition, first, the spatio-temporal clustering unit 108 extracts all activity route data whose activity feature similarity with the target activity route data is equal to or greater than the inter-data similarity threshold (step S705). At this time, if the number of corresponding activity route data is equal to or greater than the minimum number of data (step S706: Yes), a new cluster is assigned to the target activity route data and the corresponding activity route data (step S707), and the same processes as steps S705 and S707 are performed on all the corresponding activity route data, and the process is repeated until there is no unprocessed corresponding activity route data.

[0073] If the number of corresponding activity path data is smaller than the minimum number of data (step S706: No), the spatio-temporal clustering unit 108 classifies the activity path data of interest into noise that does not constitute a cluster.

[0074] After the process of step S708 or S709 is completed, if there is any activity route data that has not been assigned to a cluster, the spatio-temporal clustering unit 108 randomly selects one activity route data from among them as activity route data of interest.

[0075] Finally, the spatio-temporal clustering unit 108 outputs all the activity route clusters (step S712).

[0076] FIG. 8 is a flowchart of the feature extraction process executed by the feature extraction unit 109 according to the first embodiment of the present invention.

[0077] First, the feature extraction unit 109 acquires activity route clusters from the time-space clustering unit 108 (step S801). At this time, if a person attribute is assigned to the activity route data included in each activity route cluster (step S802: Yes), the feature extraction unit 109 repeatedly executes the process of step S804 for all acquired activity route clusters (steps S803, S805).

[0078] In the repetition, the feature extraction unit 109 extracts cluster features that represent the activity route cluster by tabulating the person attributes in the activity route cluster or by performing statistical processing using an existing statistical method (step S804).

[0079] If the activity route data included in each activity route cluster is not assigned with a person attribute (step S802: No), the feature extraction unit 109 executes the processes of steps S807 to S811 for all the acquired activity route clusters (steps S806, S812).Then, the feature extraction unit 109 executes the processes of steps S808 to S809 for the activity route data included in the activity route cluster (steps S807, S810).

[0080] In the repetition, first, the feature extraction unit 109 acquires map information including facility information and transportation network information of the planning target area from the outside via the network 113 and the network interface device 112, and extracts the facilities passed or stayed in by the activity route data and the transportation means used in the activity route data (step S808). For example, a threshold value of the distance from the facility used for determining whether the user passed or stayed is set, and the movement trajectory information and time information of the activity route data are compared with the set threshold value to extract the facilities passed or stayed in by the activity route data. In addition, the transportation means used in the activity route data is extracted from the movement trajectory information and time information of the activity route data, for example, by using an existing transportation mode estimation method.

[0081] Next, the feature extraction unit 109 acquires user attribute information for the extracted relevant facility and relevant means of transportation at the relevant time (step S809).

[0082] Thereafter, the feature extraction unit 109 extracts user attribute information for all activity route data, and then extracts cluster features that represent the activity route cluster by aggregating the extracted user attribute information or by performing statistical processing using an existing statistical method (step S811).

[0083] Finally, the feature extraction unit 109 outputs all the activity route clusters and the cluster feature information (step S813).

[0084] FIG. 9 is a diagram showing an example of a setting input screen 901 output from the output device 102 of the activity feature extraction device 100 according to the first embodiment of the present invention.

[0085] A setting input screen 901 shown in FIG. 9 includes a target area input area 902 , a target area information input area 903 , a target date and time information input area 904 , a classification setting information input area 905 , and a map display management area 906 .

[0086] In the target area input area 902, a target area marker 902A and a direction display 902B are displayed on a map.

[0087] When the user specifies the area to be analyzed on the map in the target area input area 902, or inputs the target area information in the target area information input area 903, a target area marker 902A is displayed surrounding the periphery of the specified area. After that, the planning target area information is transmitted from the input device 101 to the setting input unit 104.

[0088] In addition, the user can change the scale and display position by specifying on the target area input area 902 using a method different from that used when selecting the target area. At this time, it is preferable that the user can change the angle of the displayed map by specifying on the target area input area 902 using an event to which a key operation, mouse click, or touch method different from that described above is assigned. The direction display 902B indicates the direction of the map displayed in the target area input area 902.

[0089] By displaying the target area input area 902 in this way, characteristic locations such as major facilities and rivers can be displayed together with the analysis target area, making it easy to find and select the analysis target area on the map.

[0090] The target area information input area 903 accepts target area information input by the user. The target area may be specified only by address information input in the target area information input area 903, or may be specified by a facility of interest selected in the target area input area 902 and an area radius (distance from the facility) input in the target area information input area 903.

[0091] The target date and time information input area 904 includes a date and time setting area 904A and a condition setting area 904B.

[0092] The user inputs a period or time in the date and time setting area 904A as the date and time to be analyzed. In the condition setting area 904B, one or more of temperature, season, weather, and day of the week can be selected and input, and target date and time information that satisfies the specified condition within the period specified in the date and time setting area 904A is registered in the setting input unit 104. If no period is input in the date and time setting area 904A, activity route data that satisfies the condition input in the condition setting area 904B among the activity route data for the entire period stored in the activity route database 114 will be analyzed. If no condition is input in the condition setting area 904B, all activity route data within the period specified in the date and time setting area 904A will be analyzed.

[0093] A classification method to be used for activity feature extraction and classification setting information required for the selected classification method are input in a classification setting information input area 905. When a classification method is selected, only setting information required for the selected method is accepted as input.

[0094] After all of the input of the target area information, target date and time information, and classification setting information has been completed, the extraction start button 905A is activated, and the processing of the setting input unit 104 is executed by operating the extraction start button 905A.

[0095] In the map display management area 906, a scale reduction button 906A, a scale enlargement button 906B, a scale 906C, and a home position transition button 906D are displayed.

[0096] In the map display management area 906, the user can operate a scale reduction button 906A with the cursor or touch to reduce the scale of the map displayed in the target area input area 902 and display a wider range. Also, the user can operate a scale increase button 906B to increase the scale of the map displayed in the target area input area 902 and display a narrower range in detail. A scale reduction button 906C displays the distance on the map displayed in the target area input area 902. Also, the user can operate a home position transition button 906D with the cursor or touch to display the position, scale, and direction displayed in the initial screen in the target area input area 902.

[0097] FIG. 10 is a diagram showing an example of an activity feature display screen 1001 output from the output device 102 of the activity feature extraction device 100 according to the first embodiment of the present invention.

[0098] An activity characteristic display screen 1001 shown in FIG. 10 includes a cluster display area 1002, a cluster characteristic display area 1003, and an animation control area 1004.

[0099] In the cluster display area 1002, activity route data 1002A for a target area and a target date and time, and a cluster 1002B to which the activity route data belongs are displayed on a map displayed in the cluster display area 1002.

[0100] Here, the time information accompanying the activity route data may be displayed in a display format appropriate to the activity route data, such as by setting a time axis perpendicular to the map surface, plotting each observation point of the activity route data on a three-dimensional coordinate system of latitude, longitude, and time, and displaying a trajectory connecting the observation points, or by displaying the activity route data corresponding to the time displayed in the animation control area 1004 through animation.

[0101] In addition, the user can change the scale, display position, and viewpoint for displaying the three-dimensional coordinates by operating the cursor or touching the cluster display area 1002. At this time, it is preferable that the user can change the angle of the map displayed by operating on the cluster display area 1002 using an event to which a key operation, mouse click, or touch method different from those described above is assigned.

[0102] Furthermore, cluster 1002B displays ellipses so as to surround the activity route data belonging to each cluster, but the activity route data may be displayed in a distinguishable manner by a display format (different colors, shades of color, etc.) according to the cluster.

[0103] The cluster characteristic display area 1003 is displayed when a cluster is selected by the cursor or touch, and displays the cluster characteristic information extracted by the characteristic extraction unit 109. The cluster characteristic information enables a person considering a policy to easily understand the characteristics of people and appropriately narrow down marketing targets.

[0104] The animation control area 1004 includes a play button 1004A that controls playing and stopping the animation to control the display of the activity path data 1002A and the cluster 1002B to which the activity path data belongs, a seek bar 1004B that indicates the display timing and allows the display timing to be specified, a reverse play button 1004C that allows the animation to be rewound, and a time zone specification field 1004D in which the display time zone can be entered.

[0105] In the animation control area 1004, trajectories with high spatial similarity even if their activity times are different can be displayed as belonging to the same cluster, making it possible to confirm the cluster characteristics (e.g., cluster A). Furthermore, trajectories with high temporal similarity even if they are different can be displayed as belonging to the same cluster, making it possible to confirm the cluster characteristics (e.g., cluster E).

[0106] In the embodiment described above, the time separation is multiplied by the speed to calculate the time separation having the same dimension as the spatial separation, but the spatial separation may be divided by the speed to calculate the spatial separation having the same dimension as the time separation.

[0107] <Functions and Effects of the First Present Example> As described above, the activity feature extraction device 100 of this embodiment can extract activity groups having characteristics that correspond to the actual times and places where people are active, taking into account differences in facilities used and means of transportation. Therefore, policy planners can use the characteristics of the activity groups to appropriately consider the type, implementation time, and implementation location of the policy to be implemented.

[0108] In addition, the activity feature extraction device 100 of this embodiment can provide the person considering a policy with the activity status of the area under consideration, which is the basis for determining the type, implementation time, and implementation location of the policy, from the consideration stage or early stage of the policy when there is no data showing the effect of the implementation of the policy. This reduces the possibility that the effect of the policy after implementation is lower than expected, and reduces the cost of modifying the policy.

[0109] (Second Example) An activity feature extraction device 100 according to a second embodiment will be described. In this embodiment, the same processes as those in the first embodiment are executed up to the extraction of cluster feature information, and further, the user inputs personal attribute information of the user himself or related persons, and the cluster to which the user himself or related persons belong is calculated using the cluster feature information, and facility information and transportation information are recommended taking into account the activity features of the user himself or related persons. Note that in the second embodiment, differences from the first embodiment will be mainly described, and the same configurations and processes as those in the first embodiment will be assigned the same reference numerals, and their description will be omitted.

[0110] The second embodiment is basically the same as the first embodiment, but has a configuration added to the activity feature extraction device 100 of the first embodiment shown in Fig. 1, and further differs in the processing executed by the activity feature extraction device 100 and the display screen created by the output generation unit 110 shown in Fig. 4. Since the other configurations and processing are the same as those of the first embodiment, their explanations will be omitted, and the configuration added to the activity feature extraction device 100 of the first embodiment will be explained with reference to Fig. 11, the flowchart of the processing of the second embodiment will be explained with reference to Fig. 12, and the screen created by the output generation unit 110 with reference to Figs. 13 and 14, respectively.

[0111] FIG. 11 is a functional block diagram showing a configuration of an activity feature extraction device 100 according to the second embodiment of the present invention.

[0112] The server computer 103 has a person attribute input unit 1101, a cluster belonging calculation unit 1102, and a recommended spot extraction unit 1103 in addition to the configuration of the first embodiment.

[0113] The person attribute input unit 1101 acquires person attribute information input by the user from the input device 101, and transmits it to the belonging cluster calculation unit 1102. Here, the person attribute information has at least one of activity route data of the target person and person attribute information. Here, the target person may include not only the user himself / herself but also people related to the user. When the person attribute information of the target person has already been input, for example, the person ID of the target person may be input, and approval for the utilization of the input person attribute information may be obtained from the target person.

[0114] The feature extraction unit 109 obtains the activity route clusters from the time-space clustering unit 108 , extracts cluster feature information from the activity route data in each activity route cluster, and transmits the activity route clusters and the cluster feature information to the output generation unit 110 and the belonging cluster calculation unit 1102 .

[0115] The belonging cluster calculation unit 1102 acquires activity route clusters and cluster feature information from the feature extraction unit 109, acquires person attribute information from the person attribute input unit 1101, calculates the probability that a person having the acquired person attribute information belongs to each activity route cluster, assigns the person to the activity route cluster with the highest probability, and determines the belonging cluster information.

[0116] The probability of belonging to each activity route cluster can be calculated by the following method. For example, when the person attribute information includes activity route data, the activity feature similarity is calculated between the activity route data included in each activity route cluster using the same method as the time-space clustering unit 108, and the belonging probability is calculated to be proportional to the average of the activity feature similarity with the activity route data in each activity route cluster. Also, when the person attribute information includes only person attribute information (no activity route data is included), for example, the cluster feature information of each activity cluster is compared with the person attribute information, and the belonging probability is calculated to be proportional to the attribute similarity between the cluster feature information and the person attribute information. Here, the attribute similarity may be calculated by, for example, converting each item of the person's tendency information included in the cluster feature information and the person attribute information into a one-hot vector and calculating it by the cosine similarity between them. After that, the belonging cluster information is sent to the recommended spot extraction unit 1103.

[0117] The recommended spot extraction unit 1103 acquires the cluster information from the cluster calculation unit 1102, and calculates recommended spots that match the personal attributes of the target person.

[0118] The recommended spots can be calculated by the following method. For example, all facility information visited by the activity route data included in the cluster determined to belong to the target person is weighted by the number of visits to the facility, a matching rate with the target person's preferences is calculated, and recommended spot information is extracted according to the calculated matching rate. At this time, if the target person includes one or more related persons in addition to the user himself / herself, all facility information visited by the activity route data included in the cluster determined to belong to the related person is similarly weighted by the number of visits to the facility to calculate a matching rate with the related person's preferences, and a value obtained by multiplying the matching rate for each facility is set as a final matching rate, and recommended spot information is extracted according to the matching rate. After that, the recommended spot information is transmitted to the output generation unit 110.

[0119] The output generation unit 110 in the second embodiment acquires recommended spot information from the recommended spot extraction unit 1103, creates a spot recommendation screen 1401, and outputs it to the output device 102. The configuration of the spot recommendation screen 1401 will be described later with reference to FIG.

[0120] FIG. 12 is a flowchart of the process executed by the activity feature extraction device 100 according to the second embodiment of the present invention.

[0121] First, when a user inputs personal attribute information of the user himself / herself or a related person into the input device 101 (person attribute input screen 1301 (FIG. 13)), the input device 101 transmits the input personal attribute information to the person attribute input unit 1101 (step S1201).

[0122] Next, the person attribute input unit 1101 transmits the input person attribute information to the belonging cluster calculation unit 1102 (step S1202).

[0123] Furthermore, the feature extraction unit 109 transmits the action path cluster and the cluster feature information to the belonging cluster calculation unit 1102 (step S1203). Here, the steps up to the calculation of the action path cluster and the cluster feature information in the first embodiment (steps S401 to S409) are executed before step S1203.

[0124] Next, the cluster calculation unit 1102 calculates the cluster to which the subject of the person attribute information belongs from the cluster feature information of each activity route cluster and the person attribute information, and transmits the calculated cluster to the recommended spot extraction unit 1103 (step S1204).

[0125] Next, the recommended spot extraction unit 1103 extracts recommended spots and transmits them to the output generation unit 110 (step S1205).

[0126] Finally, the output generation unit 110 creates a spot recommendation screen 1401 (FIG. 14) and transmits it to the output device 102 (step S1206). The user can check the processing result by looking at the spot recommendation screen 1401 displayed on the output device 102.

[0127] FIG. 13 is a diagram showing an example of a person attribute input screen 1301 outputted from the output device 102 of the activity characteristic extraction device 100 according to the second embodiment of the present invention.

[0128] A person attribute input screen 1301 shown in FIG. 13 includes a person attribute information input area 1302 and a related person information input area 1303 .

[0129] The person attribute information input area 1302 accepts input of the user's own personal attributes by the user.

[0130] The related person information input area 1303 accepts input by the user of personal attributes of related persons of the user. Input of related person information is optional. If the personal attribute information of the subject has already been registered in the activity feature extraction device 100, the user may input a person ID as the identification number of the subject. On the other hand, if the personal attribute information of the subject has not already been input in the activity feature extraction device 100, the user inputs the personal attributes of the subject in the related person information input area 1303.

[0131] FIG. 14 is a diagram showing an example of a spot recommendation screen 1401 outputted from the output device 102 of the activity characteristic extraction device 100 according to the second embodiment of the present invention.

[0132] 14 includes a recommended spot display area 1402 and a recommended spot information display area 1403. In the recommended spot display area 1402, a recommended spot pin 1402A is displayed.

[0133] The recommended spot pin 1402A shows the position of the recommended spot acquired from the recommended spot extraction unit 1103 and the recommendation level according to the matching rate on a map, and may be displayed in a display format suitable for displaying location information, such as a pin. The recommendation level may be displayed as a value determined based on the matching rate, and the matching rate itself may be displayed, or the ranking of the matching rates may be displayed.

[0134] The recommended spot information display area 1403 displays recommended spot information. In addition to information about the spot itself, the recommended spot information may display a time period when the activity route data included in the activity cluster to which the user and related parties belong is frequently visited, and the interest level of the related parties.

[0135] <Functions and Effects of the Second Example> As described above, the activity feature extraction device 100 of this embodiment can extract activity groups having characteristics that correspond to the actual times and places where people are active, taking into consideration differences in facilities used and transportation means, and associate the activity groups with users, so that users can learn spot information that matches their own activity times and activity place trends and includes places that they have never visited. In addition, users can learn spot information without providing their own location information.

[0136] The present invention is not limited to the above-described embodiments, and includes various modified examples and equivalent configurations within the spirit of the appended claims. For example, the above-described embodiments have been described in detail to clearly explain the present invention, and the present invention is not necessarily limited to those including all of the configurations described. Furthermore, a part of the configuration of one embodiment may be replaced with the configuration of another embodiment. Furthermore, the configuration of another embodiment may be added to the configuration of one embodiment. Furthermore, a part of the configuration of each embodiment may be added, deleted, or replaced with another configuration.

[0137] Furthermore, each of the aforementioned configurations, functions, processing units, processing means, etc. may be realized in hardware, for example by designing some or all of them as an integrated circuit, or may be realized in software by a processor interpreting and executing a program that realizes each function.

[0138] Information such as programs, tables, and files that realize each function can be stored in a storage device such as a memory, a hard disk, or an SSD (Solid State Drive), or in a recording medium such as an IC card, an SD card, or a DVD.

[0139] In addition, the control lines and information lines shown are those considered necessary for the explanation, and do not necessarily show all the control lines and information lines necessary for implementation. In reality, it can be considered that almost all components are connected to each other. [Explanation of symbols]

[0140] 100 Activity feature extraction device 101 Input Device 102 Output Device 103 Server computer 104 Setting input section 105 Activity Route Data Acquisition Unit 106 Spatial separation calculation section 107 Time resolution calculation section 108 Spatio-Temporal Clustering Department 109 Feature Extraction Unit 110 Output Generation Unit 111 Activity Path Database 112 Network Interface Device 113 Network 114 Activity Path Database 201 Processor 202 Storage device 203 Processing Program 300 Activity route data management information 901 Setting input screen 1001 Activity characteristics display screen 1101 Person attribute input unit 1102 Cluster Computing Department 1103 Recommended Spot Extraction Unit 1301 Person attribute input screen 1401 Spot recommendation screen

Claims

1. An activity feature extraction device that analyzes activity route data in which human movement trajectory information and time information are associated with each other, The computer includes a computing device that executes a predetermined computation process and a storage device that is accessible by the computing device, a spatial separation degree calculation unit configured to calculate a spatial separation degree representing a spatial difference between the activation path data; a time separation calculation unit configured to calculate a time separation representing a time difference between the activity path data; a spatio-temporal clustering unit configured to calculate an activity feature similarity from the spatial separation and the time separation, and cluster the activity route data using the calculated activity feature similarity; The activity feature extraction device, wherein the calculation device further comprises an output generation unit that outputs information on the clusters generated by the spatiotemporal clustering unit.

2. The activity feature extraction device according to claim 1 , The activity feature extraction device, wherein the time separation calculation unit calculates the time separation or the space separation taking into account speed information associated with the movement trajectory information.

3. The activity feature extraction device according to claim 2, The activity feature extraction device, characterized in that the temporal separation calculation unit multiplies the temporal separation by the speed information to calculate a temporal separation having the same dimension as the spatial separation, or divides the spatial separation by the speed information to calculate the spatial separation having the same dimension as the temporal separation.

4. The activity feature extraction device according to claim 2 or 3, the spatial separation calculation unit determines corresponding pairs of observation points included in the movement trajectory information among the activity route data, and calculates the spatial separation by a sum of spatial differences between the corresponding pairs; The activity feature extraction device, characterized in that the time separation calculation unit calculates the time separation based on a difference in the time information of the corresponding pair and a moving speed at the observation points that make up the corresponding pair.

5. The activity feature extraction device according to claim 2 or 3, The activity feature extraction device, characterized in that the spatial separation calculation unit calculates the spatial separation using features of points where people pass through or stay in the movement trajectory information.

6. The activity feature extraction device according to claim 5, The activity feature extraction device is characterized in that the calculation device is equipped with a feature extraction unit that extracts cluster attribute information of the cluster from facilities passed through or stayed in by the activity route data of the cluster and from the means of transportation used in the activity route data of the cluster.

7. The activity feature extraction device according to claim 6, The computing device includes a person attribute input unit for acquiring person attribute information of a user or a related person related to the user; The activity feature extraction device is characterized in that the calculation device includes an affiliation cluster calculation unit that calculates attribute similarity between the person attribute information and the cluster attribute information and calculates affiliation cluster information representing the probability that the user or the related person belongs to the cluster.

8. The activity feature extraction device according to claim 7, The activity characteristic extraction device is characterized in that the calculation device is equipped with a recommended spot extraction unit that extracts recommended spot information that matches the activity tendencies of the user or related parties based on activity route data contained in the cluster to which the user or related parties belong.

9. The activity feature extraction device according to claim 8, The activity feature extraction device is characterized in that the recommended spot extraction unit calculates a matching rate between the user or related party and the recommended spot information based on activity route data contained in the cluster to which the user or related party belongs, and extracts recommended spot information based on the calculated matching rate.

10. The activity feature extraction device according to claim 1 , The activity characteristic extraction device is characterized in that the calculation device is provided with an output device that outputs an activity characteristic display screen that displays the activity route data and the cluster information superimposed on the display of map information and time information on the screen.

11. The activity feature extraction device according to claim 6, An activity feature extraction device characterized in that it displays the activity route data and cluster information superimposed on the display of map information and time information on a screen, and further comprises an output device that outputs an activity feature display screen that displays the cluster attribute information.

12. The activity feature extraction device according to claim 9, An activity feature extraction device characterized by comprising an output device that displays the location of the recommended spot information superimposed on the display of map information on a screen, and further outputs a spot recommendation screen that displays a value determined based on the compatibility rate of the recommended spot information.

13. The activity feature extraction device according to claim 12, the recommended spot extraction unit calculates a mutual matching rate based on both the matching rate of the user of the recommended spot information and the matching rate of the related person; The activity feature extraction device is characterized in that the output device displays the calculated mutual matching rate on the spot recommendation screen.

14. An activity feature extraction method executed by an activity feature extraction device that analyzes activity route data in which human movement trajectory information and time information are associated with each other, comprising: the activity feature extraction device is configured by a computer having a calculation device that executes a predetermined calculation process and a storage device that can be accessed by the calculation device; The activity feature extraction method includes: a spatial separation calculation step in which the computing device calculates a spatial separation representing a spatial difference between the activation path data; a time separation calculation step in which the computing device calculates a time separation representing a time difference between the activity path data; a spatio-temporal clustering step in which the computing device calculates an activity feature similarity from the spatial separation and the temporal separation, and clusters the activity path data using the calculated activity feature similarity; and an output generation step in which the computing device outputs information on the clusters generated in the spatio-temporal clustering step.

15. An activity feature extraction device that analyzes activity route data in which human movement trajectory information and time information are associated with each other, The computer includes a computing device that executes a predetermined computation process and a storage device that is accessible by the computing device, The computing device includes a person attribute input unit for acquiring person attribute information of a user or a related person related to the user; a cluster affiliation calculation unit configured to calculate an attribute similarity between cluster attribute information of a cluster into which the activity route data is clustered and the person attribute information, and to calculate affiliation cluster information representing a probability that the user or the related person belongs to a cluster; The activity characteristic extraction device is characterized in that the calculation device is equipped with a recommended spot extraction unit that extracts recommended spot information that matches the activity tendencies of the user or related parties based on activity route data contained in the cluster to which the user or related parties belong.

Images