JP2017059032A - Estimation device - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide an estimation device which estimates the characteristic of a position with higher accuracy regardless of the difference in life pattern of a user based on time.SOLUTION: In a server 10, an information reception section 11 acquires a terminal log being information based on an operation of a terminal 30, a position characteristic estimation section 17 estimates the characteristic of a position using the log of the terminal acquired by the terminal log, and an estimation result output section 18 outputs the characteristic of the position being the estimation result to a residence point cluster extraction storage section 16. In this case, the characteristic of the position where the terminal 30 operates is estimated on the basis of the operation of the terminal 30, thereby enabling more accurate estimation of the characteristic of the position as compared with a case where the characteristic of the position (home or office) is estimated on the basis of day attributes. Namely, the characteristic of the position can be estimated with higher accuracy regardless of the difference in life pattern of a user based on time.SELECTED DRAWING: Figure 1

Description

  The present invention relates to an estimation apparatus that estimates a position characteristic.

  A portable terminal is generally equipped with a positioning function such as GPS (Global Positioning System). The server device communicably connected to the portable terminal acquires information (current position information) indicating the current position specified by the positioning function of the portable terminal from the portable terminal, and what the current position means There is a technique for estimating (see, for example, Patent Document 1). Specifically, the server device acquires current position information from the mobile terminal, extracts only the position where the terminal is estimated to be retained based on the current position information, and extracts a plurality of retained positions. Then, the data is classified into spatial clusters, and a characteristic (for example, home) in a cluster having a high stay rate based on the day attribute (for example, holiday) is estimated. In this way, the home can be estimated from the position of the mobile terminal.

JP 2011-171876 A

  However, the technique described in Patent Document 1 is calculated based on the day attribute, and it is assumed that the life pattern based on time in the user of each terminal is a general one. Therefore, if the user's life pattern based on time is different, the position characteristics may not be estimated appropriately. For example, if a place where a person stays on a regular day on weekdays is estimated as a workplace, the workplace of a night shift user may not be estimated appropriately.

  An object of the present invention is to provide an estimation device that estimates a position characteristic with higher accuracy regardless of a difference in a user's life pattern based on time.

  The estimation apparatus according to the present invention includes a terminal information acquisition unit that acquires operation information that is information based on the operation of the terminal, and a characteristic of a position where the terminal has operated based on the operation information acquired by the terminal information acquisition unit. And an output means for outputting the characteristics of the position estimated by the estimation means.

  According to such an estimation apparatus according to the present invention, since the characteristics of the position where the terminal is operated are estimated based on the operation of the terminal, the position characteristics (home, workplace) are estimated based on the day attribute. The position characteristics can be accurately estimated. That is, the position characteristics can be estimated with higher accuracy regardless of the difference in the user's life pattern based on time.

  In the estimation apparatus according to the present invention, the position acquisition means for acquiring position information indicating the position measured by the terminal, and the plurality of position information acquired by the position acquisition means are referred to, and the position is determined by the terminal within a predetermined period. A position extracting unit that extracts a representative position based on the plurality of positions when the plurality of positions are included in a predetermined range; and the estimating unit includes a representative position extracted by the position extracting unit and a predetermined period. The characteristics of the position where the terminal has operated may be estimated based on the operation information corresponding to.

  According to such an estimation apparatus according to the present invention, since the characteristics at the place where the terminal has been located within a certain range (retained position) for a predetermined period are estimated, it is necessary to estimate the position where the terminal has just passed. It is possible to avoid the estimation with respect to the position where there is no.

  In the estimation apparatus according to the present invention, the terminal information acquisition unit further acquires time information indicating a time at which the terminal has operated, and the estimation unit further determines the terminal information based on the time information acquired by the terminal information acquisition unit. You may estimate the characteristic of the position which acted.

  According to such an estimation apparatus according to the present invention, by using the time information included in the position information, the meaning of the position can be estimated more appropriately in consideration of the usage time zone of the terminal.

  In the estimation apparatus according to the present invention, the estimation unit acquires characteristic information that is information indicating the characteristic of the position of the learning terminal, and learning that is information based on the operation of the learning terminal at the position of the learning terminal. Learning information is generated, and machine learning is performed using learning data with the learning operation information as information corresponding to the explanatory variable and the characteristic information as information corresponding to the objective variable, thereby generating a learning model The position characteristics may be estimated by inputting the operation information acquired by the terminal information acquisition unit to the learning model as an explanatory variable and obtaining the output result of the learning model.

  According to such an estimation apparatus according to the present invention, a learning model is generated based on the characteristics of the actual position in the learning terminal and the operation information based on the operation of the learning terminal at the position, and the learning model is Since the position characteristics are estimated based on the movement information, the position characteristics can be appropriately estimated according to the movement tendency of the learning terminal when the learning terminal is located.

  In the estimation apparatus according to the present invention, the estimation unit stores in advance a condition for estimating the position characteristics based on the operation information acquired by the terminal information acquisition unit, and the operation acquired by the terminal information acquisition unit The position characteristics may be estimated based on whether the information matches the stored condition.

  According to such an estimation apparatus according to the present invention, the conditions for estimating the position characteristics are stored in advance, and the position characteristics are estimated based on the conditions. Characteristics can be estimated.

  In the estimation apparatus according to the present invention, the terminal information acquisition means indicates, as operation information, information indicating the alarm operation of the terminal, information indicating the charging operation of the terminal, and an access point used by the terminal for connection to the network by wireless communication Any information of the information may be acquired.

  According to such an estimation apparatus according to the present invention, since the operation information of the terminal that can estimate the meaning of the position is acquired as the operation information, the position characteristics can be estimated appropriately.

  According to the present invention, it is possible to estimate the position characteristics with higher accuracy regardless of the difference in the user's life pattern based on time.

It is a block diagram of a server concerning an embodiment of the present invention. 2 is a diagram illustrating a hardware configuration of a server 10. FIG. It is a data block diagram which shows a terminal log. It is a data block diagram which shows a position log. It is a figure for demonstrating the process which extracts a stay point cluster. It is a data block diagram which shows staying point information. It is a data block diagram which shows stay point cluster extraction information. It is a figure which shows the example of learning data. It is a flowchart which shows the whole estimation process. It is a flowchart which shows the process which extracts a stay point cluster. It is a flowchart which shows the process which produces | generates a learning model. It is a flowchart which shows the estimation process using a learning model. It is a flowchart which shows the estimation process by a rule base.

  Hereinafter, preferred embodiments of the present invention will be described in detail with reference to the accompanying drawings. In the description of the drawings, the same or equivalent elements will be denoted by the same reference numerals, and redundant description will be omitted.

  As shown in FIG. 1, the server 10 (estimation device) according to the present embodiment is a device for estimating characteristics of a predetermined position. In the present embodiment, the server 10 estimates the position characteristics of the terminal 30. This characteristic is information indicating the meaning of the position for the user of the terminal 30. Specifically, it is a home, a commuting workplace, or the like. As described above, the server 10 can specify the living area of the user of the terminal 30 by estimating the characteristics of the position of the terminal 30. Thereby, for example, advertisement distribution based on a living area can be performed to each user.

  The server 10 and the terminal 30 can transmit / receive information to / from each other via a network. This network includes, for example, a mobile communication network or the Internet.

  The terminal 30 is a portable electronic device such as a mobile phone possessed by a user. The terminal 30 may be other than a mobile phone, and may be, for example, a smartphone, a tablet computer, a personal digital assistant (PDA), or the like.

  The terminal 30 has a positioning function and can measure the position of the terminal 30. This positioning function is realized by a known technique, for example, using a GPS receiver. The terminal 30 performs positioning at a positioning cycle (for example, several minutes) determined in advance by the terminal 30. When the terminal 30 executes positioning, the terminal 30 generates a position log (position information) that is information on the positioning result. Specifically, the terminal 30 has a positioning point ID, which is information for identifying the position log, and information (for example, latitude and longitude) indicating the position of the terminal 30 (position measured by the terminal 30) obtained by this positioning. ) And a positioning time that is a positioning time. The terminal 30 transmits the generated position log to the server 10. Each time the terminal 30 generates a position log, the terminal 30 may transmit the position log to the server 10 or may collectively transmit the position log to the server 10 after a predetermined period has elapsed.

  The terminal 30 includes a rechargeable battery. When the terminal 30 is connected to a charger, the battery can be charged. Further, the terminal 30 can be connected to a network (for example, a wireless local area network (LAN) using WiFi (registered trademark)) by wireless communication. In addition, the terminal 30 has an alarm function that performs voice output, vibration, and the like at a preset time. When the terminal 30 performs operations such as charging, wireless communication, and alarm as described above, a terminal log (operation information) that is a log (information based on the operation of the terminal 30) based on the operation of the terminal 30 is generated. To do. That is, the terminal 30 is information indicating the charging operation of the terminal 30, information used by the terminal 30 to connect to a network by wireless communication (for example, information on an access point to be connected when performing wireless communication), and information on the terminal 30 Generates information indicating alarm action. This terminal log also includes a terminal log generation time (log acquisition time) that is the time when the operation was performed (the time when the terminal log was generated). The terminal 30 transmits a terminal log to the server 10. Each time the terminal 30 generates a terminal log, the terminal log may be transmitted to the server 10, or may be transmitted to the server 10 collectively after a predetermined period of time has elapsed.

  When the terminal 30 is connected to the charger and starts charging the battery of the terminal 30, the terminal 30 generates a terminal log regarding the start of the charging. Specifically, the terminal 30 includes information for identifying a terminal log (terminal log ID), information (user ID) for identifying a user of the terminal 30 stored in advance in the terminal 30, and a type of terminal log. A terminal log including information indicating that charging is performed (charging has been started), information indicating the remaining charge amount, and terminal log generation time (charging start time) is generated as information to be displayed. Each time the terminal 30 generates a terminal log, the terminal 30 generates a terminal log ID that is different from the past terminal log ID. Further, when the connection between the terminal 30 and the charger is released (that is, the charging is finished), the terminal 30 generates a terminal log for the completion of the charging. Specifically, the terminal 30 generates a terminal log ID, a user ID, information indicating that the terminal log has been discharged (charging is terminated), information indicating the remaining charge, and terminal log generation. A terminal log including information indicating time (charging end time) is generated. In this way, the terminal 30 generates a terminal log at the timing when charging of the battery of the terminal itself is started, and generates another terminal log at the timing when charging of the battery is completed. Therefore, the charging time can be calculated based on the terminal log generation time of each terminal log.

  In addition, if the terminal 30 is set to connect to a wireless LAN by a user operation, the wireless LAN of the wireless LAN that is located within a predetermined distance range from the terminal 30 by a known short-range wireless communication or the like can be used. Detect access points. When the terminal 30 detects a wireless LAN access point that can be wirelessly connected, the terminal 30 wirelessly connects to the detected access point and connects to the network via the access point. When the terminal 30 is connected to a network, the terminal 30 generates a terminal log related to the connection to the network. Specifically, the terminal 30 includes a terminal log ID, a user ID, information indicating that the wireless connection is started as information indicating the type of the terminal log (for example, a character string “WiFi”), an SSID (Service Set A terminal log including information indicating an identifier), information indicating a BSSID (Basic Service Set Identifier), and information indicating a terminal log generation time is generated. Here, the SSID is an identification name of an access point to which the terminal 30 is wirelessly connected. The BSSID is a MAC (Media Access Control) address of an access point to which the terminal 30 is wirelessly connected. A plurality of different access points can be provided for one SSID. That is, even if the SSID is the same, the BSSID may be different. When the terminal 30 is wirelessly connected to the access point, the terminal 30 receives the SSID and BSSID, and includes the SSID and BSSID in the terminal log.

  Further, if the terminal 30 deviates from the range in which the terminal 30 can wirelessly communicate with the access point as a result of the movement of the terminal 30 and becomes unable to wirelessly communicate with the access point, the terminal 30 passes the wireless LAN access point Terminate the connection to the network. When the terminal 30 ends the connection to the network, the terminal 30 generates a terminal log related to the end of the connection to the network. Specifically, the terminal 30 has terminal log ID, user ID, information indicating that the wireless connection is terminated as information indicating the type of terminal log (for example, a character string “WiFiOff”), and information indicating the SSID. And a terminal log including information indicating the BSSID and information indicating the terminal log generation time. In this way, the terminal 30 generates a terminal log at the timing of connection to the network, and generates another terminal log at the timing of termination of connection to the network. Therefore, the time connected to the network can be calculated based on the terminal log generation time of each terminal log.

  Further, when the alarm activation time is activated at the preset alarm activation time, the terminal 30 generates a terminal log for the alarm operation. Specifically, the terminal 30 generates a terminal log including a terminal log ID, a user ID, information indicating that the alarm is activated as information indicating the type of the terminal log, and information indicating the terminal log generation time. .

  In addition to the terminal 30 (position characteristics estimation target terminal) that transmits the position log and terminal log as described above, there is a learning terminal that transmits a learning position log and learning terminal log (learning operation information). . The learning position log is a position log for the server 10 to generate a learning model by machine learning described later. The learning terminal log is a terminal log for generating the learning model. This learning terminal can input a characteristic at which the terminal 30 is located by a user input operation. When the characteristic is input by a user operation, the learning terminal generates a position log further including the characteristic as the learning position log. Specifically, the learning terminal is information indicating a positioning point ID and a position obtained by positioning by executing a positioning function according to the input when the characteristic is input by a user input operation. Latitude / longitude, positioning time, information indicating input characteristics (correct data (characteristic information)), and information indicating that it is a learning position log (for example, flag information indicating that it is a learning position log) Generate a location log for learning. The correct answer data is, for example, a numerical value, where “0” indicates a home and “1” indicates a workplace. Further, like the terminal 30, the learning terminal performs positioning at a positioning cycle determined in advance by the learning terminal. When the learning terminal executes positioning, the learning terminal generates a learning position log that is information of the positioning result. When the learning terminal performs positioning in the positioning cycle as described above, the learning terminal generates a learning position log including information indicating that the learning position log is not included in the learning position log. When the learning terminal generates the learning position log and the learning terminal log, the learning terminal transmits the learning position log and the learning terminal log to the server 10.

  Subsequently, the server 10 will be described. The server 10 includes an information receiving unit 11 (terminal information acquisition unit, position acquisition unit), a terminal log storage unit 12, a position log storage unit 13, a stay point extraction unit 14 (position extraction unit), and a stay point storage unit. 15, a stay point cluster extraction storage unit 16, a position characteristic estimation unit 17 (estimation unit), and an estimation result output unit 18 (output unit).

  FIG. 2 is a hardware configuration diagram of the server 10. As shown in FIG. 2, the server 10 includes a CPU 101, a RAM (Random Access Memory) 102 and a ROM (Read Only Memory) 103, which are main storage devices, a communication module 104 for performing communication, and an auxiliary storage device such as a hard disk. The computer is configured to include a computer having hardware such as 105. When these components are operated by a program or the like, the function of each functional element shown in FIG. 1 is exhibited. The server 10 may be configured by a computer system including a plurality of computers. Hereinafter, each functional element shown in FIG. 1 will be described.

  The information receiving unit 11 is a part that receives a terminal log or a location log from the terminal 30. When receiving the terminal log or the position log from the terminal 30, the information receiving unit 11 refers to the received information and determines whether the received information is a terminal log or a position log. When the information receiving unit 11 determines that the received information is a terminal log, the information receiving unit 11 registers the terminal log in the terminal log storage unit 12. Further, when the information receiving unit 11 determines that the received information is a position log, the information receiving unit 11 registers the position log in the position log storage unit 13. For example, the information receiving unit 11 determines the terminal log and the position log based on differences in data size, data format (number of data items), and the like. The information receiving unit 11 receives a learning terminal log and a learning position log, which are information used for machine learning described later, from the learning terminal, and registers the learning terminal log in the terminal log storage unit 12. The position log for learning is registered in the position log storage unit 13. The information receiving unit 11 acquires the terminal log and the position log by receiving the terminal log and the position log from the terminal 30 as described above.

  The terminal log storage unit 12 is a database that stores terminal logs. Here, an example of information stored in the terminal log storage unit 12 is shown in FIG. As illustrated in FIG. 3, the terminal log storage unit 12 includes a “HandsetLogID” that indicates a terminal log ID that identifies a terminal log generated by the terminal 30, a “UserID” that is a user ID that identifies a user of the terminal 30, and “Log_type” that is information indicating the type of the terminal log, “SSID” that is the identification name of the access point to which the terminal 30 is wirelessly connected, the Mac address “BSSID” of the access point to which the terminal 30 is wirelessly connected, and the terminal 30 “Remaining” indicating the remaining charge amount at the time of charging and “Timestamp” indicating the terminal log generation time are stored in association with each other. “HandsetLogID”, “UserID”, “Log_type”, and “Timestamp” are information that is always input regardless of the type of operation to be generated as a terminal log. “Log_type” has different information input depending on the type of operation. “SSID” and “BSSID” are information input when the terminal 30 is connected to the network and terminated. “Remaining” is information that is input when the “Log_type” indicates that it is related to charging.

  The position log storage unit 13 is a database that stores position logs. Here, an example of information stored in the position log storage unit 13 is shown in FIG. As illustrated in FIG. 4, the position log storage unit 13 includes “PosLogID” that is a positioning point ID that is an ID for identifying a position log, “UserID” that is a user ID for identifying a user of the terminal 30, and the terminal 30. Corresponds to “Longitude” indicating latitude, “Latitude” indicating longitude among latitude and longitude indicating the position, and “Timestamp” indicating the time when the terminal 30 has measured the position. Add and remember.

  The stay point extraction unit 14 is a part that refers to the plurality of position logs stored in the position log storage unit 13 and extracts stay points based on the positions indicated by the plurality of position logs. The stay of the user means that the user stays in a narrow area to some extent, that is, the user is continuously located in the predetermined area for a predetermined time or more. The stay point is an area where the user stays or a position representing the area (representative position). This representative position is, for example, the barycentric position of a plurality of positions where the user was located in the predetermined area. Specifically, the stay point extraction unit 14 is configured so that positions indicated by a plurality of position logs (a plurality of positions measured by the terminal 30) are continuously within a predetermined narrow area for a predetermined time (for example, several hours) or more. If it is included, the stay point based on the plurality of positions is extracted.

  In addition, the stay point extraction unit 14 generates a stay point cluster in which a representative position based on a stay point within a certain range is determined when the stay point is equal to or greater than a predetermined threshold value within a certain range. The representative position is the position of the center of gravity of each position of the stay point within the certain range. This representative position is the position of the stay point cluster. The stay point extraction unit 14 starts the stay point extraction process by batch processing at a predetermined time every predetermined period (for example, one week), or requests a position characteristic estimation by the administrator of the server 10. The extraction process of the staying point is started in response to the input indicating.

  Processing until the stay point extraction unit 14 extracts stay points from the position log, clusters the stay points, and generates a stay point cluster will be described with reference to FIG. The stay point extraction unit 14 acquires a position log from the position log storage unit 13 for each user and every day at a predetermined timing. Subsequently, the stay point extraction unit 14 extracts a stay point based on the acquired position log.

  Specifically, the staying point extraction unit 14 determines staying points in the order from the oldest positioning time for the acquired position log. First, the stay point extraction unit 14 sets the position indicated in the position log with the oldest positioning time as the reference position for judging the stay point. The stay point extraction unit 14 compares the position where the positioning time is indicated in the next position log with the reference position. The stay point extraction unit 14 determines whether or not the distance between these positions is smaller than a preset distance threshold for stay point determination. When the stay point extraction unit 14 determines that the distance is smaller than the distance threshold, the stay point extraction unit 14 further compares the position indicated in the next position log with the reference position, and makes a determination based on the same distance threshold as described above. The stay point extraction unit 14 sequentially determines until the distance becomes smaller than the distance threshold (the distance becomes equal to or greater than the distance threshold).

  The staying point extraction unit 14 compares the positioning time related to the previous position log whose distance becomes smaller than the distance threshold with the positioning time related to the position log of the reference position. The stay point extraction unit 14 determines whether or not the time difference between these positioning times is larger than a preset time threshold for stay point determination. When the stay point extraction unit 14 determines that the time difference is larger than the time threshold value, the stay point extraction unit 14 acquires those position logs (position logs from the position log of the reference position to the previous position log whose distance is larger than the distance threshold value). The position log constituting the stay point (position log corresponding to the stay point) is determined. That is, in this case, the stay point extraction unit 14 determines that the stay point can be extracted. If the stay point extraction unit 14 determines that the time difference is not larger than the time threshold value, those position logs (position logs from the position log of the reference position to the previous position log whose distance is larger than the distance threshold value). Is determined not to be a position log constituting a stay point. That is, in this case, the stay point extraction unit 14 determines that the stay point cannot be extracted.

  Further, the stay point extraction unit 14 performs the same determination as described above using the position indicated in the position log whose distance is no smaller than the distance threshold as a reference position for determining a new stay point. Thus, if the user is located in the region from the reference position to the distance threshold for a time exceeding the time threshold, it is determined that there is a staying point. The distance threshold and the time threshold are appropriately set according to the purpose of the stay point extraction. The region for determining the stay point may be a circular region based on the reference position as described above, or a rectangular (mesh) region based on the reference position.

  The stay point extraction unit 14 calculates the position (latitude and longitude) of the stay point from the position log that is supposed to constitute the stay point. The position of the staying point can be, for example, the center of gravity (average position) of the position indicated by the position log that constitutes the staying point. Note that the position of the stay point may be calculated by a method other than this method. The stay point extraction unit 14 acquires a position log from the position log storage unit 13 for each day, and calculates the position of the stay point for each day.

  FIG. 5A is an example in which two stay points are extracted (calculated). The stay point extraction unit 14 extracts the stay point P1 and the stay point P2, and does not extract the positioning points p1 to p9 whose positioning points (latitude and longitude of the position log) are not dense in the surroundings (the stay point). ). Further, the stay point extraction unit 14 extracts stay points for N days (for example, for one week). When the stay point extraction unit 14 extracts the stay point (determines that the stay point is a position log constituting the stay point), the user ID included in the position log corresponding to the stay point and the latitude of the position indicating the stay point are stored. , Longitude, ID of the position log corresponding to the stay point (positioning point ID), the earliest time of the position log, and the last time of the position log are specified as the stay point extraction result.

  When the stay point is extracted, the stay point extracting unit 14 registers information (stay point information) based on the extracted result in the stay point storage unit 15. The data structure stored in the stay point storage unit 15 is shown in FIG. The stay point storage unit 15 includes a “StayPointID” indicating a stay point ID that is unique among the stay points, a “UserID” that is a user ID for identifying the user of the terminal 30, and a latitude / longitude indicating the position of the stay point. , “Longitude” indicating the latitude, “Latitude” indicating the longitude among the latitude and longitude indicating the position of the staying point, and the staying start time (the earliest time in the position log corresponding to the staying point) “From_time”, “to_time” indicating the end time of residence (the last time in the position log corresponding to the stay point), and “PosLogIDs” indicating the ID of the position log corresponding to the stay point are stored in association with each other. To do.

  The stay point extraction unit 14 may extract the stay point by another method. For example, based on the location and time of the location log, well-known technologies (eg, Kyosuke Nishida, Hiroyuki Toda, Ken Kurashima, Kei Uchiyama, stochastic visit POI analysis: user modeling from spatiotemporal behavior trajectory, “multimedia, distributed, The dwell points may be extracted using a spatio-temporal mean shift clustering algorithm, which is a collaboration and mobile (DICOMO2013) symposium, 2C-6, pp. 334-345, 2013.

  As described above, the stay point extraction unit 14 acquires a position log for each day, extracts a stay point from the position log, extracts a stay point for N days, and then generates a stay point cluster based on the stay point. To do. This stay point cluster is a set of stay points located in a range close to each other. For example, as shown in FIG. 5B, the stay point extraction unit 14 generates a stay point cluster by means of a known technology shift when there are stay points P1 to P9. For example, the stay point extraction unit 14 generates the stay point cluster for each user ID.

  The stay point extraction unit 14 generates a stay point cluster based on a preset kernel shift kernel width (width based on distance). For example, the stay point extraction unit 14 clusters all stay points for N days based on the latitude and longitude of the stay point. In this way, the stay point extraction unit 14 clusters stay points that are close to each other, and extracts stay point clusters (locations that have often stayed within a long period (for example, N days)). The stay point extraction unit 14 calculates the representative position (latitude and longitude) of the stay point cluster from the stay points that are supposed to form the stay point cluster. For example, it can be set as the center of gravity (average position) of the positions of the stay points determined to form the stay point cluster. The representative position of the stay point cluster may be calculated by a method other than this method. For example, as shown in FIG. 5B, when there are stay points P1 to P9, the stay points P1, stay points P4, and stay points P5 are dense (the stay points P4 within the kernel width from the stay points P1). Therefore, the stay point extraction unit 14 extracts the stay point cluster C1. Similarly, the stay point extraction unit 14 extracts each stay point cluster (stay point cluster C2, stay point cluster C3, stay point cluster C4).

  When the stay point extraction unit 14 extracts the stay point cluster, the stay point cluster extraction unit 14 registers information (retention point cluster extraction information) based on the extracted result in the stay point cluster extraction storage unit 16. The data structure of the stay point cluster extraction information stored in the stay point cluster extraction storage unit 16 is shown in FIG. The stay point cluster extraction storage unit 16 includes a “StayPointClusterID” indicating a stay point cluster ID that is unique among the stay point clusters, a “UserID” that is a user ID for identifying the user of the terminal 30, and the position of the stay point cluster ( Among the latitude and longitude indicating the representative position), “Longitude” indicating the latitude, “Latitude” indicating the longitude among the latitude and longitude indicating the position of the stay point cluster, and the stay point ID constituting the stay point cluster “StayPointIDs” to be displayed, “Type” to indicate a characteristic result (home or work), and “Score” to indicate the probability of the Type (correct probability) are stored in association with each other. When the stay point cluster extraction information is registered by the stay point extraction unit 14, no information is input in “Type” and “Score”.

  When the stay point extraction unit 14 finishes registering the stay point cluster extraction for all users in the stay point cluster extraction storage unit 16, the registration of the stay point cluster extraction storage unit 16 is completed from the stay point extraction unit 14 to the position characteristic estimation unit 17. Notify that it has been done.

  Further, when generating the learning data to be described later, the staying point extraction unit 14 extracts a staying point for learning using the learning position log for the learning data, or the staying point for learning. A stagnation point cluster (learning point cluster extraction information for learning) is generated using points. In this case, the stay point extraction unit 14 extracts the stay point or the stay point cluster except that the position log to which the information indicating the learning data is attached is acquired from the position log storage unit 13. It is the same as the method of generating. After generating the staying point cluster for learning, the position characteristic estimating unit 17 is notified.

  The position characteristic estimation unit 17 is a part that estimates the characteristic of the position where the terminal 30 has operated based on the terminal log acquired by the information reception unit 11. Further, the position characteristic estimation unit 17 estimates the position characteristic based on the representative position of the stay point cluster extracted by the stay point extraction unit 14 and the terminal log corresponding to the predetermined period to be extracted. Further, the position characteristic estimation unit 17 estimates the position characteristic based further on the log generation time of the terminal log. When the position characteristic estimation unit 17 receives a notification of completion of the stay point cluster extraction storage unit 16 from the stay point extraction unit 14, the position characteristic estimation unit 17 estimates the position characteristics.

  The position characteristic estimation unit 17 includes a machine learning method and a rule base set in advance by the position characteristic estimation unit 17 as a method for estimating the position characteristic.

(Method by machine learning)
The position characteristic estimation unit 17 uses a machine learning method in which learning data is generated in advance, a learning model based on the learning data is generated, and the position characteristic is estimated using the learning model. The learning data includes information indicating the status of the learning terminal (stay status, operation of the learning terminal) based on the terminal log and the stay point cluster information, and the characteristics of the position of the learning terminal in the situation ( Correct data). The learning model is a model that is generated based on the learning data and outputs characteristics according to the input of the operation of the terminal log at the position to be estimated. First, the procedure for generating learning data will be described. As described above, after the stay point cluster for learning is generated by the stay point extraction unit 14, the position characteristic estimation unit 17 acquires the terminal log of the user from the terminal log storage unit 12, and the terminal log and A feature (explanatory variable) is generated based on the stay point cluster. This feature is information indicating the status of the terminal 30 (stay status, operation of the learning terminal) based on the information on the stay point cluster. An example of the features is shown in FIG. The position characteristic estimation unit 17 generates stay / no-stay time, stay time, and stay points for each day of the week / time zone for each stay point cluster of each user, and further calculates stay time, stay points, and stay days for N days. .

  First, the position characteristic estimation unit 17 extracts a staying point cluster (stagnation point cluster extraction information) for a certain user. The position characteristic estimation unit 17 acquires the stay point cluster extraction information from the stay point cluster extraction storage unit 16. The position characteristic estimation unit 17 acquires the stay point information corresponding to the stay point ID included in the “StayPointIDs” of the acquired stay point cluster extraction information from the stay point storage unit 15, and obtains “from_time” of the acquired stay point information. By referring to “to_time” and specifying the staying time zone, the presence / absence of staying in each time zone is generated for each day of the week. Specifically, the position characteristic estimation unit 17 sets the time zone included in “from_time” and “to_time” of the stay point information to “1” and is included in “from_time” and “to_time” of the stay point information. The time period when there is no data is “0”. In addition, the position characteristic estimation unit 17 refers to the “from_time” and “to_time” of the acquired stay point information, and identifies the stay time. Specifically, the position characteristic estimation unit 17 sets the total period from “from_time” to “to_time” of the stay point information for each day of the week as the stay time of each day of the week. The position characteristic estimation unit 17 sets the number of stay point information for each day of the week as the stay point number.

  The position characteristic estimation unit 17 sets the total period of time from all “from_time” to “to_time” of the acquired stay point information as the total stay time. Moreover, the position characteristic estimation part 17 makes the number of acquired stay point information the total stay point number. With reference to “from_time” and “to_time” of the staying point information, it is specified whether or not the user stayed on each day of N days, and the staying days are calculated.

  Further, the location characteristic estimation unit 17, for “WiFi connection”, based on the terminal log, connection presence / absence of each day of the week, connection time, number of connection SSIDs, number of connection BSSIDs, presence / absence of the same SSID connection in other staying point clusters Is calculated. Moreover, the position characteristic estimation part 17 calculates the presence or absence of charge of each day, charge time, and total charge time based on a terminal log about charge. Moreover, the position characteristic estimation part 17 specifies the alarm activation presence or absence of each day of the week based on a terminal log about an alarm. First, the location characteristic estimation unit 17 refers to the “from_time” and “to_time” of the stay point information acquired by the user, and the terminal log corresponding to the time from “from_time” to “to_time” of the stay point information. Is acquired from the terminal log storage unit 12. Specifically, the position characteristic estimation unit 17 acquires a terminal log including “Timestamp” of the terminal log in the time from “from_time” to “to_time” of the acquired stay point information. The location characteristic estimation unit 17 refers to “Log_type” and “Timestamp” of the acquired terminal log, and for the day of the week indicated by “Timestamp” of the terminal log whose “Log_type” is WiFi, the presence / absence of WiFi connection is set to “1”. As for the day of the week when there is no terminal log that is WiFi ON, the presence / absence of WiFi connection is set to “0”.

  The location characteristic estimation unit 17 refers to “Log_type” and “Timestamp” of the acquired terminal log, and determines the period from the time when “Log_type” is “WiFi” to the time when “Log_type” is “WiFiOFF” for each day of the week. And the calculated time is taken as the connection time. The position characteristic estimation unit 17 calculates the number of “SSID” of the acquired terminal log, and calculates the number of connected SSIDs. Further, the position characteristic estimation unit 17 calculates the number of “BSSID” of the acquired terminal log, and calculates the number of connected BSSIDs. The location characteristic estimation unit 17 refers to the “SSID” of the acquired terminal log, and the SSID that is the same as the SSID included in the “SSID” is the same user ID and corresponds to another stay point cluster ID “ Whether or not it is in the “SSID” is specified. As a result, if there is a terminal log “SSID” corresponding to another stay point cluster ID, the position characteristic estimation unit 17 sets the value of the SSID connection to “1” in the other stay point cluster, and sets the other stay point cluster. When there is no “SSID” in the terminal log corresponding to the ID, the value of the SSID connection is set to “0” in the other staying point cluster.

  The location characteristic estimation unit 17 refers to “Log_type” and “Timestamp” of the acquired terminal log, and “Log_type” sets “1” as the presence / absence of charging for the day of the week indicated by “Timestamp” of the terminal log that is charging. For days of the week when there is no terminal log for charging, the presence / absence of charging is set to “0”.

  The location characteristic estimation unit 17 refers to “Log_type” and “Timestamp” of the acquired terminal log, calculates a period from the time when “Log_type” is charging to the time when “Log_type” is discharging for each day of the week, Let the calculated time be the charging time. The position characteristic estimation unit 17 sets the total charging time of each day as the total charging time.

  The position characteristic estimation unit 17 refers to “Log_type” and “Timestamp” of the acquired terminal log, and “1” indicates whether the alarm is activated for the day of the week indicated by “Timestamp” of the terminal log in which “Log_type” is an alarm activation. As for the day of the week for which there is no terminal log for alarm activation, the alarm activation presence / absence is set to “0”.

  The position characteristic estimation unit 17 generates a feature for each user and staying point cluster as described above. And the position characteristic estimation part 17 acquires the correct data contained in the position log corresponding to the stay point corresponding to a stay point cluster, and extracts the correct data used as an objective variable from the acquired correct data. Specifically, the position characteristic estimation unit 17 acquires the correct data as described above, and correct data that is not input in the acquired correct data (correct data of the position log generated when positioning is performed at the positioning cycle). Are extracted from the extraction target, and the largest number of values in the input correct data are associated with the above features as objective variables. Thereby, learning data in which the feature and the objective variable are associated with each other is generated.

  The position characteristic estimation unit 17 generates a learning model that outputs information indicating characteristics by a known technique (for example, SVM, random forest, logistic regression) based on the generated learning data. In addition to the information indicating the characteristic, the position characteristic estimation unit 17 may generate a learning model that outputs the correct answer probability of the characteristic. The correct answer probability is a value indicating the certainty of the characteristic.

  The position characteristic estimation unit 17 generates a feature for each stay point cluster in the same manner as described above for the stay point cluster to be estimated. The position characteristic estimation unit 17 inputs the feature as an explanatory variable to the learning model, and uses the output result of the learning model as a position characteristic. The position characteristic estimation unit 17 outputs the correct answer probability of the position characteristic by a known technique as an output result of the learning model. In this case, the position characteristic estimation unit 17 outputs the stay point cluster ID and the learning model output result (characteristic, correct answer probability) to the estimation result output unit 18.

(Rule-based method)
Subsequently, a rule-based method will be described. The position characteristic estimation unit 17 stores in advance a reference (condition) for estimating the position characteristic based on the terminal log, and whether or not the information of the terminal log corresponding to the stay point cluster matches the reference. Based on this, the position characteristics are estimated. The position characteristic estimation unit 17 stores a home determination criterion and a workplace determination criterion as the reference. The home determination criterion is a criterion that can be estimated to be staying at home, for example. The position characteristic estimation unit 17 stores a plurality of home determination criteria, and determines that the position characteristic is home if it matches a pre-stored threshold value or more. The workplace judgment standard is a standard that can be estimated to be staying at the workplace. The position characteristic estimation unit 17 stores a plurality of workplace determination criteria, and determines that the location characteristic is a workplace if the workplace characteristics match a pre-stored threshold or more. In the rule-based method, the position characteristics are estimated for each stay point cluster as in the case of estimation by machine learning.

  The position characteristic estimation unit 17 uses “WiFi connection always”, “WiFi connection is always the same SSID”, “WiFi connection is always the same BSSID (Mac address)”, “Alarm activation history exists”, “Charge” as home determination criteria. "Holding history", "Longest charging time in one day", "Staying at night", "Staying every day", "Staying for a predetermined time (for example, 5 hours)", and " Ten criteria of “visiting multiple times during N days (multiple staying points)” are stored. In this way, the criteria for wireless communication, alarm activation, and charging are included in the criteria including the staying time.

  The position characteristic estimation unit 17 acquires the stay point cluster extraction information from the stay point cluster extraction storage unit 16, and stores the stay point information corresponding to the stay point ID included in “StayPointIDs” of the acquired stay point cluster extraction information. Obtained from the storage unit 15. The position characteristic estimation unit 17 determines whether each of the ten criteria is met based on the stay point information. In addition, the above-described stay point information is also used to determine whether or not a workplace determination criterion described later is met.

  A method for determining whether or not “Wifi connection exists” will be described. The position characteristic estimation unit 17 acquires a terminal log including “Timestamp” of the terminal log in the time from “from_time” to “to_time” of the acquired stay point information. It meets the criteria of “WiFi connection is always the same SSID”, “WiFi connection is always the same BSSID (Mac address)”, “Alarm activation history”, “Charged history”, “Longest charging time in one day” When determining whether or not, the position characteristic estimation unit 17 determines whether or not each criterion is met using the terminal log.

  The location characteristic estimation unit 17 refers to “Log_type” and “Timestamp” of the acquired terminal log, and if there is at least one terminal log whose “Log_type” is “WiFi”, the criterion “connected to WiFi” If there is no terminal log of “WiFi ON”, it is determined that the criterion “Connected to WiFi” is not satisfied. This standard considers that when the terminal 30 is located at home, for example, an access point such as a wireless LAN router provided at home is used to connect to the wireless LAN for communication.

  Next, a method for determining whether or not “WiFi connection always matches the same SSID” will be described. The location characteristic estimation unit 17 refers to the “SSID” of the acquired terminal log, determines whether or not all are the same SSID, and if all are the same SSID, the criterion of “WiFi connection is always the same SSID” If all SSIDs are not the same, it is determined that the criteria of “WiFi connection is always the same SSID” are not satisfied. The standard is that, when the terminal 30 is located at home, communication is performed by connecting to a wireless LAN using an access point such as a wireless LAN router provided with a specific SSID provided at home. Is taken into account. Next, a method for determining whether or not “WiFi connection always matches the same BSSID (Mac address)” will be described. The location characteristic estimation unit 17 refers to the “BSSID” of the acquired terminal log and determines whether or not all are the same BSSID. If all are the same BSSID, the “WiFi connection always has the same BSSID (Mac address). If all BSSIDs are not the same, it is determined that the “WiFi connection always does not meet the same BSSID (Mac address)” criterion. This standard considers that there is usually one access point such as a wireless LAN router provided at home, and the terminal 30 always uses the same BSSID access point to connect to the wireless LAN for communication. It is a thing.

  Next, a method for determining whether or not “alarm activation history exists” will be described. The position characteristic estimation unit 17 refers to “Log_type” of the acquired terminal log, determines whether there is a terminal log in which “Log_type” is “alarm activation”, and “Log_type” is “alarm activation”. When there is a terminal log that is, it is determined that the criterion of “Alarm activation history is present” is satisfied, and when there is no terminal log whose “Log_type” is “Alarm activation”, the criterion of “Alarm activation history is present” is not satisfied Is determined. This criterion takes into account that an alarm is used at home, for example, for the user to wake up.

  Next, a method for determining whether or not “charged history exists” will be described. The position characteristic estimation unit 17 refers to “Log_type” of the acquired terminal log, determines whether or not there is a terminal log in which “Log_type” is charging, and there is a terminal log in which “Log_type” is charging. In this case, it is determined that the criterion of “having a history of charging” is satisfied, and when “Log_type” does not include a terminal log that is charging, it is determined that the criterion of “having a history of charging” is not satisfied. This standard takes into account that the terminal 30 is normally charged at home.

  Next, a method for determining whether or not “longest charging time in one day” is met will be described. The position characteristic estimation unit 17 refers to “Log_type” and “Timestamp” of the acquired terminal log, and sets the period from the time when “Log_type” is charging to the time when “Log_type” is discharging as the charging time. Then, the position characteristic estimation unit 17 refers to a terminal log different from the terminal log corresponding to the target stay point cluster from the terminal log on the same day as the “Timestamp”, and from the other terminal log as described above. Calculate the charging time. Subsequently, the position characteristic estimation unit 17 compares the charging time calculated from the terminal log corresponding to the staying point cluster with the charging time calculated from the other terminal log, and determines from the terminal log corresponding to the staying point cluster. It is determined whether or not the calculated charging time is “the longest charging time in one day”. When the charging time calculated from the terminal log corresponding to the stay point cluster is the longest charging time in one day, the position characteristic estimation unit 17 determines that the criterion of “longest charging time in one day” is satisfied, and stays If the charging time calculated from the terminal log corresponding to the point cluster is not the longest charging time in one day, it is determined that the criterion of “longest charging time in one day” is not satisfied. This standard takes into account that charging of the terminal 30 at home (for example, charging while the user is sleeping) is usually performed longer than charging at other places.

  Next, a method for determining whether or not “staying at night” is met will be described. The position characteristic estimation unit 17 refers to “from_time” and “to_time” of the acquired stay point information, specifies the staying time zone, and the staying time zone is set in advance by the position characteristic estimation unit 17. It is determined whether it is included in a night time zone (for example, 19:00 to 4 o'clock). When the position characteristic estimation unit 17 determines that the staying time zone is included in the night time zone, the position characteristic estimation unit 17 determines that the criterion “staying at night” is satisfied, and the staying time zone is within the night time zone. If not included, it is determined that the criterion “staying at night” is not satisfied.

  Next, a method for determining whether or not “staying every day” is met will be described. The position characteristic estimation unit 17 refers to “from_time” and “to_time” of the acquired stay point information, and determines whether or not there is stay point information for all the N days. The position characteristic estimation unit 17 determines that the criteria of “staying every day” is satisfied when there is staying point information for all the N days, and there is no staying point information for all the N days. In this case, it is determined that the criterion of “staying every day” is not satisfied.

  Next, a method of determining whether or not “stay for a predetermined time (for example, 5 hours) or more” is met will be described. The position characteristic estimation unit 17 refers to “from_time” and “to_time” of the acquired stay point information, and specifies the stay period. The position characteristic estimation unit 17 determines whether or not the stay period is equal to or longer than a determination period (for example, 5 hours) set in the position characteristic estimation unit 17 in advance. When it is not less than the determination period, it is determined that the criterion “stay for a predetermined time (for example, 5 hours)” is satisfied, and when the stay period is less than the determination period, It is determined that the criterion of “staying (for example, 5 hours) or more” is not satisfied.

  Next, a method for determining whether or not “same multiple visits during N days (multiple stay points)” will be described. The position characteristic estimation unit 17 refers to the “from_time” and “to_time” of the acquired stay point information, and determines whether or not the plurality of stay point information is the same day, whereby the plurality of stay point information is the same day. If it is, it is determined that the criterion “multiple visits during N days” is satisfied, and if the plurality of staying point information is not the same day, it is determined that the criterion “multiple visits during N days” is not satisfied.

  The position characteristic estimation unit 17 determines that the characteristic of the staying point cluster is “home” when the threshold of 6 or more among the above ten criteria is satisfied.

  The location characteristic estimation unit 17 uses, as workplace determination criteria, “Wifi connection is present”, “WiFi connection is always the same SSID”, “one or more BSSIDs of WiFi connection”, “WiFi of the same SSID in another stay point cluster” Connected to "," has a history of charging "," stays in the day "," stays for more than 2 days "," stays for a predetermined time (eg, 3 hours) ", and" N days Nine criteria of “medium / multiple visits (multiple stay points)” are stored. Among them, “There is WiFi connection”, “The WiFi connection is always the same SSID”, “Has a history of charging”, “I have stayed for more than a predetermined time (for example, 3 hours)”, and “N times multiple visits Regarding (multiple stay points), the above-mentioned home determination criteria are “connected to WiFi”, “WiFi connection is always the same SSID”, “has a history of charging”, “stay for a predetermined time (for example, 3 hours) or more ”And“ N-day multiple visits (multiple stay points) ”, and the description is omitted. Note that “staying at noon” in the workplace judgment standard is only different from the “staying at night” in the above-mentioned home judgment standard in the set time zone (for example, 10:00 to 16:00). Since the determination method is the same, detailed description is omitted.

  Among the above workplace determination criteria, “With WiFi connection” means that when the terminal 30 is located at the workplace, the access point provided at the workplace is used to connect to the wireless LAN for communication. It is a thing. “The WiFi connection is always the same SSID”, this standard usually uses an access point such as a wireless LAN router provided at the workplace where the specific and identical SSID is set when the terminal 30 is located at the workplace. Thus, it is considered that communication is performed by connecting to a wireless LAN. “Has a history of charging” usually takes into account that the terminal 30 is charged in the workplace.

  Next, a method of determining whether or not “one or more WiFi connection BSSIDs” are met will be described. The location characteristic estimation unit 17 refers to the “BSSID” of the acquired terminal log, determines whether or not there is one or more BSSIDs, and if there are one or more BSSIDs, “if there is one WiFi connection BSSID. It is determined that the above criterion is satisfied. This criterion is that a workplace may be provided with a plurality of access points at locations corresponding to the same stay point cluster, and the terminal 30 can wirelessly use any one or more access points of BSSID. Considering communication by connecting to a LAN. That is, it is considered that a plurality of access points are used even in the same stay point cluster.

  Next, a method for determining whether or not “stay for two days or more” is met will be described. The position characteristic estimation unit 17 refers to “from_time” and “to_time” of the acquired stay point information, and determines whether or not there is stay point information for two or more days in the N days. The position characteristic estimation unit 17 determines that the staying point information for two days or more satisfies the criterion “stays for two days or more”, and if there is no staying point information for two days or more, “2 days It is determined that it does not satisfy the criteria of “staying above”.

  Next, a method for determining whether or not “connected to WiFi of the same SSID in another staying point cluster” will be described. The position characteristic estimation unit 17 refers to “from_time” and “to_time” of the acquired stay point information, and “Timestamp” of the terminal log at the time from “from_time” to “to_time” of the acquired stay point information. Get terminal logs that contain. The position characteristic estimation unit 17 searches for the SSID included in the terminal log corresponding to the stay point corresponding to the other stay point cluster that is the same as the SSID included in the terminal log. If there is another staying point cluster, the terminal log is used to determine whether or not each of the criteria is met. If there is another staying point cluster, “if another staying point cluster has the same SSID WiFi. If it is determined that the “connection” criterion is satisfied and there is no other stay point cluster, it is determined that the “connection to WiFi of the same SSID at another stay point cluster” is not satisfied. This standard is that, in the workplace, even in different locations (stay point clusters), it may be possible to connect to a wireless LAN with access points of the same SSID (for example, SSID commonly used in the company). Is taken into account.

  The position characteristic estimation unit 17 determines that the characteristic of the staying point cluster is “workplace” when the threshold value of 4 or more is satisfied among the above nine criteria.

  Further, the position characteristic estimation unit 17 determines whether or not the above-described home determination criterion is satisfied, and determines that the target staying point cluster is “home” when it is determined that it is home. , And if the target stay point cluster is not “home”, then it is determined whether the workplace criteria are satisfied, and if it is determined that the subject stay point cluster is “ Determined as “workplace”. Further, when it is determined that the target staying point cluster is “workplace”, and it is determined that “the connection is to another WiFi host having the same SSID with another staying point cluster”, the other staying point cluster is also determined to be “workplace”. (The determination result may be overwritten not only when the other staying point cluster is not yet determined but also when it is determined as “others”). This is because a plurality of access points often have the same SSID at the same workplace. It is determined whether or not the workplace judgment criteria are satisfied, and when it is determined that it is not the workplace, it is determined as “other”.

  The position characteristic estimation unit 17 outputs the stay point cluster ID to be determined and the determination result to the determination result output unit 18. Note that the position characteristic estimation unit 17 may perform estimation using only one of the methods based on machine learning and rule base, or may perform estimation based on a combination of both.

  The determination result output unit 18 is a part that outputs the position characteristics estimated by the position characteristic estimation unit 17. Specifically, the determination result output unit 18 outputs the characteristics by registering the characteristics in the stay point cluster extraction storage unit 16. When only the stay point cluster ID and the characteristic are received, the determination result output unit 18 registers the characteristic in “Type” of the stay point cluster extraction information having the stay point cluster ID in the stay point cluster extraction storage unit 16. In this way, the “home” and “workplace” labels are attached to the stay point cluster. Further, when the determination result output unit 18 receives the correct answer probability, the determination result output unit 18 inputs the correct answer probability to “Score” of the stay point cluster extraction information.

  Thus, since the determination result output unit 18 registers information in which the user ID, the position, and the position characteristic are associated with each other, appropriate advertisement distribution can be performed using the information. For example, an advertisement distribution server that distributes advertisements refers to the above-mentioned information, identifies the user's home position, and distributes advertisements suitable for home around the home (for example, advertisements for daily necessities).

  Subsequently, a processing procedure executed by the server 10 will be described with reference to flowcharts of FIGS. 9 to 13. The flowchart shown in FIG. 9 is a diagram showing an overall process for performing position characteristics. First, the information receiving unit 11 acquires a location log from the terminal 30 and registers the location log in the location log storage unit 13 (step S1). And the information receiving part 11 acquires a terminal log from the terminal 30, and registers the said terminal log in the terminal log memory | storage part 12 (step S2). The information receiving unit 11 may acquire the location log after acquiring the terminal log, or may acquire the location log and the terminal log at the same time.

  The stay point extraction unit 14 acquires a position log from the position log storage unit 13 in response to a predetermined timing or an input of a position characteristic estimation request by an administrator of the server 10. Subsequently, the stay point extraction unit 14 extracts the stay point, and registers the extracted stay point information in the stay point storage unit 15 (step S3). Subsequently, the stay point extraction unit 14 generates a stay point cluster using the stay point information registered in the stay point storage unit 15 and extracts the generated stay point cluster (stay point cluster extraction information). Registration in the storage unit 16 (step S4). The position characteristic estimation unit 17 performs home workplace estimation for each stay point cluster using the information stored in the stay point cluster extraction storage unit 16, the stay point storage unit 15, and the terminal log storage unit 12 (step S5). . The estimation result output unit 18 registers the estimation result in the stay point cluster extraction storage unit 16 (step S6). It is determined whether or not estimation has been performed for all stay point clusters (step S7). If estimation has not been performed for all stay point clusters, step S5 and step S6 are performed for stay point clusters that have not been estimated yet. Perform the process. When the estimation is performed for all the stay point clusters, the process is terminated.

  Next, a detailed processing procedure in the stay point cluster generation process (step S4 in FIG. 9) shown in FIG. 10 will be described. First, the stay point extraction unit 14 obtains a kernel width stored in advance (step S11), obtains stay point information for each user from the stay point storage unit 15, and obtains the latitude and longitude of the stay point ( Step S12). The stay point extraction unit 14 clusters stay points located within the kernel width from the latitude and longitude of the stay point (step S13), calculates the center point of the stay point cluster (step S14), and the stay point including the center point The cluster extraction information is registered in the stay point cluster extraction storage unit 16 (step S15), and the process ends.

  Next, a procedure for generating a learning model to be stored in advance in the method by machine learning will be described using the flowchart shown in FIG. First, the information receiving unit 11 acquires a learning position log (learning user position log) from the learning terminal 30, and registers the position log in the position log storage unit 13 (step S21). Further, the information receiving unit 11 acquires a learning terminal log (learning user terminal log) from the learning terminal 30 and registers the terminal log in the terminal log storage unit 12 (step S22).

  At a predetermined timing, the stay point extraction unit 14 acquires a position log for learning from the position log storage unit 13, extracts a stay point based on the position log, and stores stay point information based on the extraction result. Register in the unit 15 (step S23). Further, the stay point extraction unit 14 generates a stay point cluster for learning using the stay point information for learning registered in the stay point storage unit 15, and stores the stay point cluster extraction information as a generation result as the stay point. Registration in the cluster extraction storage unit 16 (step S24). Note that the processing in step S23 and step S24 is performed for each user. Subsequently, the position characteristic estimation unit 17 generates a feature for each stay point cluster (step S25), and determines whether or not features of all stay point clusters have been generated (step S26). The position characteristic estimation unit 17 repeats the process of step S25 until the features of all the stay point clusters are generated. When the features of all the stay point clusters are generated, the position characteristic estimation unit 17 determines whether or not the stay point clusters for all users are generated (step S27), and the stay point clusters for all users are generated. If not, the process proceeds to step S23, and when the stay point cluster for all users is generated, learning data in which the features of all users are associated with the objective variable is generated (step S28). And the position characteristic estimation part 17 produces | generates the learning model by machine learning (step S29), and complete | finishes a process.

  Next, the processing procedure of the characteristic estimation method by machine learning will be described using the flowchart shown in FIG. The position characteristic estimation unit 17 generates a feature for each stay point cluster to be estimated (step S31), and estimates the home / workplace using the learning model generated in advance using the feature as an input value (step S31). S32).

  Next, a rule-based determination procedure will be described using the flowchart shown in FIG. The position characteristic estimation unit 17 determines whether or not the target stay point cluster satisfies the home determination criterion (satisfies 6 or more of the above ten criteria) (step S41), and satisfies the home determination criterion ( Step S41; YES), it is estimated that the staying point cluster is at home (step S42), and the process ends. If the target stay point cluster does not satisfy the home determination criterion (step S41; NO), the position characteristic estimation unit 17 determines whether the workplace determination criterion is satisfied (step S43). When the workplace characteristic criterion is satisfied (step S43; YES), the position characteristic estimation unit 17 estimates the stay point cluster as a workplace (step S44), and determines whether another stay point cluster is connected with the same SSID. Determination is made (step S45). When the location characteristic estimation unit 17 is connected with the same SSID in another stay point cluster (step S45; YES), the other stay point cluster is also estimated as a workplace (step S46). In step S43, when the workplace determination criterion is not satisfied, the position characteristic estimation unit 17 estimates the stay point cluster as other (step S47), and ends the process.

  In the above-described embodiment, the server 10 extracts the stay point cluster using the position log acquired from the terminal 30, and based on the position of the stay point cluster and the terminal log corresponding to the position, The case of estimating the characteristics of was described. However, instead of acquiring the position log from the terminal 30, the target position log and the terminal log may be acquired from another device, and the position characteristics may be estimated using the terminal log. In addition, since the position characteristic estimation itself does not necessarily require information indicating the position (for example, latitude and longitude), only the terminal log at the position to be estimated is acquired (without acquiring the position log), The characteristics of the position of the estimation target may be estimated based on the terminal log.

  For example, in the other device described above, it is conceivable that the stay point cluster is extracted, the terminal log corresponding to the stay point cluster is stored, and the terminal log is transmitted from the other device to the server 10.

  In the above-described embodiment, the server 10 has estimated the case of estimating the home / workplace. However, the server 10 may estimate characteristics of other positions such as a school destination (elementary school, junior high school, etc.).

  In the above-described embodiment, the case has been described in which the time is also considered, but the time may not be considered. That is, it may be estimated without considering the staying time.

  In the above-described embodiment, the terminal 30 operates based on the information indicating the alarm operation of the terminal 30, the information indicating the charging operation of the terminal 30, and the information indicating the access point used by the terminal 30 to connect to the network by wireless communication. Although the case where the position characteristics are estimated has been described, the position characteristics may be estimated using information indicating other operations of the terminal 30.

(Function and effect)
Then, an effect is demonstrated. In the server 10, the information receiving unit 11 acquires a terminal log that is information based on the operation of the terminal 30, and the position characteristic estimation unit 17 uses the terminal log acquired from the terminal log to determine the position characteristics. presume. As described above, the operation of the terminal 30 can be in accordance with the position characteristics. Therefore, according to the present embodiment, the position characteristic can be estimated more accurately than the case where the position characteristic (home, work) is estimated based on the day attribute. That is, the position characteristics can be estimated with higher accuracy regardless of the difference in the user's life pattern based on time. Therefore, it is possible to appropriately estimate the position characteristics of users such as sales staff who have night shifts, shift systems, and a plurality of workplaces.

  The stay point extraction unit 14 refers to a plurality of position logs, and extracts a stay point based on the plurality of positions when a plurality of positions measured by the terminal 30 are included in a predetermined range within a predetermined period. Then, the position characteristic estimation unit 17 estimates the characteristic of the position where the terminal 30 has operated based on the terminal log corresponding to the period of the stay point. In this case, since the characteristics at the place where the terminal 30 has been located within a certain range for a predetermined period (the position where the terminal 30 stayed) are estimated, the position where the terminal 30 simply passed, etc. is estimated. Can be avoided.

  In addition, the information receiving unit 11 further acquires time information (“Timestamp”) at which the terminal 30 has operated, and the position characteristic estimation unit 17 estimates the characteristic of the position at which the terminal 30 has operated based on the time information. . In this case, the server 10 can more appropriately estimate the meaning of the position in consideration of the usage time zone of the terminal and the like by using the time information of the terminal log.

  The position characteristic estimation unit 17 performs machine learning using the learning data, generates a learning model, inputs features based on the terminal log to the learning model, and obtains an output result of the learning model, thereby acquiring position characteristics. Is estimated. In this case, the server 10 generates a learning model based on the characteristics of the actual position in the learning terminal 30 and the operation information based on the operation of the learning terminal 30 at the position, and uses the learning model. Since the position characteristics are estimated based on the operation information, the position characteristics can be appropriately estimated in accordance with the movement tendency of the learning terminal 30 when the learning terminal 30 is located.

  The position characteristic estimation unit 17 stores a reference (home determination reference, workplace determination reference) in advance, and estimates the position characteristic of the terminal 30 based on whether or not the terminal log matches the reference. Since the server 10 stores a reference for estimating the position characteristic in advance and estimates the position characteristic based on the reference, the position characteristic can be appropriately estimated with a simple configuration.

  The information receiving unit 11 acquires information indicating an alarm operation of the terminal 30, information indicating a charging operation of the terminal 30, and information indicating an access point used by the terminal 30 for connection to a network through wireless communication. In this case, since the server 10 acquires the operation information of the terminal that can estimate the meaning of the position as the operation information of the terminal 30, it is possible to appropriately estimate the characteristics of the position.

  DESCRIPTION OF SYMBOLS 10 ... Server, 11 ... Information receiving part, 12 ... Terminal log storage part, 13 ... Position log storage part, 14 ... Residence point extraction part, 15 ... Residence point storage part, 16 ... Residence point cluster extraction storage part, 17 ... Position Characteristic estimation unit, 18 ... estimation result output unit, 30 ... terminal, 101 ... CPU, 102 ... RAM, 103 ... ROM, 104 ... communication module, 105 ... auxiliary storage device.

Claims (6)

Terminal information acquisition means for acquiring operation information that is information based on the operation of the terminal;
Based on the operation information acquired by the terminal information acquisition unit, an estimation unit that estimates characteristics of a position where the terminal has performed the operation;
Output means for outputting the characteristics of the position estimated by the estimation means;
An estimation apparatus comprising: Position acquisition means for acquiring position information indicating a position measured by the terminal;
A position that refers to a plurality of position information acquired by the position acquisition means, and extracts a representative position based on the plurality of positions when a plurality of positions measured by the terminal within a predetermined period are included in a predetermined range. Extraction means;
Further comprising
2. The estimation according to claim 1, wherein the estimation unit estimates a characteristic of a position where the terminal has performed the operation based on operation information corresponding to the representative position extracted by the position extraction unit and the predetermined period. apparatus. The terminal information acquisition means further acquires time information indicating the time when the terminal performed the operation,
The estimation apparatus according to claim 1, wherein the estimation unit estimates a characteristic of a position where the terminal has performed the operation based further on the time information acquired by the terminal information acquisition unit.   The estimation means acquires characteristic information that is information indicating characteristics of the position of the learning terminal, acquires learning operation information that is information based on the operation of the learning terminal at the position of the learning terminal, A learning model is generated by executing machine learning using the learning operation information using the learning operation information as information corresponding to the explanatory variable and the characteristic information as information corresponding to the objective variable. The position information is estimated by inputting the operation information acquired by the terminal information acquisition unit as an explanatory variable and obtaining the output result of the learning model. Estimating device.   The estimation means stores in advance conditions for estimating the position characteristics based on the motion information acquired by the terminal information acquisition means, and the motion information acquired by the terminal information acquisition means is stored. The estimation apparatus according to claim 1, wherein the position characteristic is estimated based on whether or not a condition is met.   The terminal information acquisition means is one of information indicating an alarm operation of the terminal, information indicating a charging operation of the terminal, and information indicating an access point used by the terminal for connection to a network by wireless communication as the operation information. The estimation apparatus according to claim 1, which acquires information.

Images