JP5980715B2 - Congestion prediction system and method - Google Patents

Description

  The present invention relates to a congestion prediction system and method, and more particularly to a congestion prediction system and method for performing congestion prediction and guidance in a station premises using social media information.

  Emphasis is placed on measures such as transport failures and disasters in public transport in cities. When a large amount of congestion occurs in a public facility such as a station, there is a high possibility that accidents such as a fall of a customer from a home or a shogi fall on an escalator will occur, which is very dangerous. In order to prevent these accidents, it is important to predict the congestion that will occur in the future with high accuracy and take appropriate measures such as guidance of users by station staff and provision of information to users.

  As one technique for predicting passenger congestion at a station, there is known a passenger flow prediction device that predicts passenger flow after the current time of a certain day using past ticket gate entry / exit data (Patent Document 1). , See page 15, figure 19). According to this technology, the distance between the ticket gate entry / exit number vector for the time zone up to the current time on a certain day and the ticket entry / exit number vector of the past date is calculated, and the ticket entry / entry number of the most approximate past date is calculated. Using the data, it is possible to calculate the number of people waiting for the station in the remaining time zone of a day.

  By the way, as one of the technologies for analyzing human behavior, there is a social overview map that uses social media information to analyze what “what”, “where” and “what” events are written, and display them on the map. (For example, Non-Patent Document 1). In this technique, it is possible to display on a map by extracting necessary information from individual writing using a natural language processing technique using machine learning.

JP 2012-196987 A

Nobuhiro Yugami, Nobuyuki Igata, Hirokazu Anai, Hiroya Inakoshi, Analytical Technology Supporting Intelligent Society, Magazine FUJITSU, September 2011 (VOL. 62, No. 5) p. 475-481 (page 477, FIG. 2)

  However, the technique described in Patent Document 1 has a problem that the number of people in the station premises cannot be accurately predicted on a day (for example, when an event is held or an earthquake disaster) whose pattern is significantly different from normal. Further, Patent Document 1 does not describe any suggestion that a passenger flow prediction is performed using social media data.

  The technique described in Non-Patent Document 1 has a problem that events occurring at that time and location can be visualized and the number of people can be calculated, but events after that time cannot be predicted. Furthermore, there is a problem that the number of people cannot be calculated for events other than those who write on social media.

  An object of the present invention is to provide a congestion prediction device and method capable of accurately predicting congestion at a station using social media information.

The congestion prediction system according to the present invention is preferably a congestion prediction system that predicts the congestion of people in a station using a server, and a user who passes through the automatic ticket checker acquired from the automatic ticket checker at the station. The number of people in the station, and / or the number of people in the station using the data of the number of passengers on the train obtained from the video taken by the surveillance camera installed in the station,
Obtaining social media data related to the station from a social media storage server via a network, and predicting the number of people around the station using the social media data;
Based on the number of people in the station premises calculated by the station premises number calculation unit and the number of people in the vicinity of the station calculated by the station number of people calculation unit, the number of people in the station premises of the current time shift It is comprised as a congestion prediction system which has a station premises number prediction part which predicts.

  The congestion prediction method according to the present invention is grasped as a congestion prediction method executed in the above system.

  According to the present invention, congestion of a station can be accurately predicted using social media information.

It is a figure which shows the structural example of the congestion prediction system of the station by one Example. It is a block diagram which shows the function structure of the congestion prediction system in one Example. It is a figure which shows the hardware structural example of the congestion prediction server in one Example. It is a figure which shows the hardware structural example of the guidance server in one Example. It is a flowchart which shows the processing operation by the station premises number calculation part 201, the station periphery number calculation part 202, and the station premises number prediction part 203 in one Example. It is a figure which shows the structural example of the station premises number table in one Example. It is a figure which shows the structural example of the social media data in one Example. It is a figure which shows the structural example of the action prediction determination condition table in one Example. It is a figure which shows the structural example of the event and group information table in one Example. It is a figure which shows the structural example of the station coordinate table in one Example. It is a flowchart which shows the processing operation by the guidance subject extraction part 204 and the guidance guidance part 2050 in one Example. It is a figure which shows the structural example of the allowable number of persons table in one Example. It is a figure which shows the structural example of the movement history data in one Example. It is a figure which shows the structural example of the log | history analysis result table in one Example. It is a figure which shows the structural example of the guidance subject extraction table in one Example. It is a figure which shows the structural example of the access destination address table in one Example. It is a figure which shows the example of a display of the mail reception screen in one Example. It is a figure which shows the example of a display of the screen of the indicator 1041 in one Example.

Hereinafter, embodiments of the present invention will be described in detail with reference to the drawings.
FIG. 1 shows a configuration example of a station congestion prediction system.
In the station congestion prediction system, the social media data storage server 103, the display server 104, the number of attendees counting server 107, the congestion prediction server 108, the guidance server 109, and the entrance / exit history storage server 110 are connected via the network 102. Configured. Usually, the display server 104, the number of attendees counting server 107, the congestion prediction server 108, the guidance server 109, and the entrance / exit history storage server 110 are owned or managed by a railway company, and the social media data storage server 103 is owned by an external company. It is an operating server.

  The social media data storage server 103 receives data such as write contents (text data), write time, current position coordinates, photos, and videos from a plurality of information terminals 101 such as mobile phones and mobile terminals via the network 102. accumulate. Further, the accumulated data can be provided to the information terminal 101 via the network 102 and browsed on the information terminal.

  The automatic ticket gate 105 is provided with a ticket gate with an IC card reader, and data (user-specific card ID, entry / exit station name) recorded on a ride IC card (hereinafter referred to as a transportation IC card) possessed by the user, Read or record data such as gate passage date / time, entrance / exit distinction, usage fee). The visitor count totaling server 107 receives and manages the ticket gate passage data acquired by the plurality of automatic ticket gates 105 via the network 102.

  A plurality of monitoring cameras 106 are installed on the platform of the station, and images near the doors of the arrival train and the departure train are taken. The video data taken by the surveillance camera is transmitted to the attendee count totaling server 107 via the network 102 and managed there. The number of visitors / entry count server 107 counts the number of people in each station at regular intervals based on the user's ticket gate passage data and captured video data. That is, the number of people can be calculated by counting the unique information (user ID) of each user from the ticket passing data acquired by the automatic ticket gate 105, and the image of the passenger is cut out from the captured video data of the monitoring camera 106. Then, the number of passengers and the number of people getting off are calculated.

  The entry / exit history storage server 110 records the data acquired from (or recorded in) the traffic IC card of the user as movement history data (see FIG. 13 for details). The congestion prediction server 108 acquires social media data from the social media data storage server 103 via the network 102, acquires the number of visitors from the number of visitors count server 107, and the station premises after the current time point Predict the number of people. The guidance server 109 acquires the number of people in the station premises after the current time point from the congestion prediction server and the movement history data of a plurality of traffic IC card users from the entry / exit history storage server via the network 102, and To guide guidance. The display server 104 displays guidance information on the signage screens of a plurality of indicators 1041 installed in the station.

Regarding the configuration example realized by this system, the social media data storage server 103, the display server 104, the number of attendees counting server 107, the congestion prediction server 108, the guidance server 109, and the entrance / exit history storage server 110 are personal computers and work The network 102 is composed of Ethernet (registered trademark) (LAN standard devised by Xerox and DEC (currently a division of Hewlett-Packard), 802.3, IEEE 802.3).
Note that the congestion prediction server 108 and the guidance server 109 may be configured as a single server. In addition, an intermediate server that aggregates data of automatic ticket gates and surveillance cameras for each station or route may be prepared.

FIG. 2 is a block diagram showing a functional configuration of the congestion prediction system in the station premises.
The congestion prediction system mainly includes a station premises number calculation unit 201, a station periphery number calculation unit 202, a station premises number prediction unit 203, a guidance target person extraction unit 204, and a guidance guide unit 205. These functions are realized by executing the congestion prediction program 307 and the guidance program 407 on the attendee count totaling server 107, the congestion prediction server 108, and the guidance server 109.

  In the station premises number calculation unit 201, the number of visitors counting server 107 calculates the number of premises in each station at regular time intervals using the ticket gate passage data and photographed video data of the users. The station periphery number calculation unit 202 calculates the number of people around the station using the social media data by the congestion prediction server 108. The station premises number predicting unit 203 is configured such that the congestion prediction server 108 calculates the number of people in the station premises calculated by the station premises number calculating unit 201 and the number of people existing around the station calculated by the station surrounding number calculating unit 202; Based on the number of people moving to the station after the current time estimated from the social media data, the number of people in the station after the current time is predicted. The guidance target person extraction unit 204 uses the information other than the station from the movement history data of the traffic IC card possessed by the user when the number of people in the station premises after the present time exceeds a certain threshold. The target person to be guided to is extracted. The guidance guide unit 205 guides the guidance server 109 to the extracted guidance subject using a communication unit such as an electronic mail.

FIG. 3 is a diagram illustrating a hardware configuration example of the congestion prediction server 108.
The main control unit 301 is a processor (CPU) that performs hardware control and program execution processing. The input unit 302 is an input device such as a keyboard or a mouse through which a system administrator inputs a program execution start instruction, a stop instruction, and the like. The output unit 303 is a display device such as a liquid crystal display that outputs a program execution state and the like. The communication processing unit 304 is an Ethernet or a wireless communication unit that exchanges data with other computers. The storage management unit 305 is a storage unit such as a hard disk or flash memory that reads and records programs and data. A communication unit 306 is a bus that performs data communication between the above-described units. The storage management unit 305 stores a congestion prediction program 307, a station premises number table 308, an action prediction determination condition table 309, an event / group information table 310, and a station coordinate table 311. The table configuration and the processing operation by the program will be described later.

FIG. 4 is a diagram illustrating a hardware configuration example of the guidance server 109.
The main control unit 401 is a processor (CPU) that performs hardware control and program execution processing. The input unit 402 is an input device such as a keyboard or a mouse for inputting a program execution start instruction or a stop instruction from the system administrator. The output unit 403 is a display device such as a liquid crystal display that outputs a program execution state and the like. The communication processing unit 404 is an Ethernet or a wireless communication unit that exchanges data with other computers. The storage management unit 405 is a storage unit such as a hard disk or flash memory that reads and records programs and data. A communication unit 406 is a bus that performs data communication between the above-described units. The storage management unit 405 stores a guide program 407, an allowable number of people table 408, a history analysis result table 409, a guidance target person extraction table 410, and an access destination address table 411. The table configuration and the processing operation by the program will be described later. Note that the input unit and the output unit may include a touch panel.

  Next, with reference to FIG. 5, processing operations by the station premises number calculation unit 201, the station periphery number calculation unit 202, and the station premises number prediction unit 203 will be described. The operation according to this flowchart is executed at regular intervals, for example.

  First, the station premises number calculation unit 201 calculates the number of people in the station premises and stores it in the station premises number table 308 (S501). As shown in FIG. 6, the station premises number table 308 includes a station name 601, a date 602, and a premises number 603. As for the calculation method of the number of people in the station, for example, when finding the number of people in the station A from April 1, 2012 from 10:05 to 10:10, from April 1, 2012 from 10:00 to 10:05 Subtract the number of participants in the ticket gate passage data in time and the number of train passengers in the time calculated by recognizing the person's video from the train getting on and off video data from the number of people in the station A It is obtained by adding the number of visitors in the time in the data and the number of people getting off the train in the time calculated by recognizing the video of the person from the train getting-on / off video data.

  Next, the station periphery number calculation unit 202 acquires social media data written within a certain time (for example, one hour) before the current time point from the social media data storage server 103 (S502). As shown in FIG. 7, social media data includes social media user ID 701, writing date and time 702, writing position coordinates 703, tag name 704 for classifying writing contents, writing contents text 705, and photos and videos attached to writing. It consists of link information 706 to data.

  Next, the station periphery number calculation unit 202 distributes the acquired social media data for each area, and calculates the number of persons existing for each area (S503). Here, the area refers to the station and its surroundings. Further, as a method of allocating social media data for each area, classification can be performed using position coordinates 703 included in the social media data. For example, the station coordinate table 311 is read, and if it is below a certain distance from the position coordinates of a certain station and the position coordinates included in the social media data, it is distributed to the station area.

  As shown in FIG. 10, the station coordinate table 311 is a list including a station name 1001 and position coordinates / area 1002 where the station exists. When a position coordinate / area 1002 where a station is present represents a certain range, for example, a plurality of position coordinates representing rectangular vertices of the range are stored. In the case of a rectangle, it may be distributed to the station area as long as it is a certain distance or less from the center of gravity coordinates of the rectangle and the position coordinates included in the social media data. Or if the position coordinate contained in social media data is located in a rectangle, you may distribute to the station area. Alternatively, if the station name is included as a keyword in the classification 704 or text data 705 included in the social media data, it may be distributed to the station area.

  Next, the station premises number prediction unit 203 checks whether or not there is a writing corresponding to the determination condition registered in the behavior prediction determination condition table 309 in the acquired social media data (S504). The behavior prediction determination condition table 309 includes an event / organization information name 801 and a determination condition 802 as shown in FIG. For example, in the case of an event, the determination condition stores a determination condition for predicting that the event has ended and that the event participant is heading from the event venue to the station. For example, if the social media writing includes a keyword such as “end” or “best”, it is assumed that the event has ended and the event participants have written their impressions of the event. In addition, if a person who has not written for a certain period of time from the writing frequency starts to write more than one time from that point, the event participants will finish the event and write based on the change in the frequency. I guess it is ready. Further, in the case of a group such as a company or a school, a judgment condition for predicting that a member of the group leaves the company or leaves the school and heads for the station is stored. For example, if a keyword such as “return home instruction” is included in social media writing, it is highly likely that a home return instruction will be issued in a group due to an earthquake or the like, and the surrounding stations will be crowded.

  If there is a corresponding event or group in the determination of S504, the station premises number prediction unit 203 reads the event / group information table 310 and the station coordinate table 3110, respectively, and determines the distribution of the number of persons moving to each station (S505). ). As shown in FIG. 9, the event / group information table 310 includes an event / group name 901, position coordinates / area 902 where the event venue or group building exists, the number of participants of the event or the number of members of the group. A list composed of 903. When a position coordinate / area 902 where an event venue or a group building exists is represented, a plurality of position coordinates representing a vertex of a rectangle in the range are stored, for example.

  Here, as a method of determining the number of persons, a record corresponding to the event or group name in S504 is searched from the event / group information table 310, and the number of persons scale 903 is set as the number of persons. Alternatively, if there is no corresponding event or group name in the event / group information table 310, the total number of people may be estimated by multiplying the number of writings and the number of writings by a coefficient. Alternatively, from the link information 706 to the photo and video data included in the social media data, if the photo or video is taken of the event venue scenery, the number of event participants is estimated by analyzing the photo and video data You may do it. As a method for determining the allocation, for example, a station located within a radius of 5 km from the position coordinates of the event venue is extracted and allocated. When there are a plurality of corresponding stations, the distance from the event venue to each station may be obtained from the position coordinates of each station, and it may be calculated so that more people move as the station is closer. Or you may calculate so that a person may be allocated from indices, such as the scale of a station and the number of available routes.

  Next, the station premises number prediction unit 203 calculates the estimated arrival time at each station (S506). For example, the distance is calculated from the position coordinates of the event venue and the station, and the estimated arrival time is calculated according to the distance. Further, the moving speed may be multiplied by a coefficient according to the event type. For example, when the event is for a family or the group is an elementary school, it may be considered that the walking speed of the child is slow.

  When S506 ends or when there is no writing corresponding to the determination condition registered in the behavior prediction determination condition table 309 in S504, the congestion rate prediction for each station is calculated (S507). For example, the predicted number of people A (p1, t1) of the number of people on the premises at time t1 of the station p1 is calculated by Equation 1.

A (p1, t1) = B (p1, t1) + α × C (p1) + β × D (p1, t1) (Formula 1)
Here, B (p1, t1) is the estimated number of people at time t1 of the station p1 calculated from the ticket gate passage data and the captured video data of the surveillance camera. For example, as described in Patent Document 1, the distance between the station premises vector for the time zone up to the current time on a certain day and the station premises vector of the past date is calculated, and the past date of the closest approximation is calculated. The number of people in the station can be calculated for any remaining time period of a day using the number of people in the station.

  C (p1) is the number of people around the station p1 calculated in S503. D (p1, t1) is the predicted number of visitors at time t1 of the station p1 calculated in S505 and 506. For example, if 100 people move to the station p1 in S505, and the estimated arrival time at the station p1 is 12:00 in S506, 11:55 is 10 people according to a certain allocation coefficient for each time before and after that. 12:00 is calculated as 60, 12:05 is calculated as 25, 12:10 is calculated as 5 and so on. α and β are coefficients. For example, since people using social media are a part of the whole, the coefficients may be increased. In addition, α may be a coefficient that is inversely proportional to the difference between the current time and the predicted time (that is, as the prediction for the distant future is less effective, information indicating that the station has been around the station in the past is less effective). . Furthermore, the congestion rate of each station may be calculated in this processing step. Here, for example, the congestion rate is calculated as a ratio of the number of people in the station premises to the allowable number of people in the station. That is, for example, if the number of people in the station is 100 and the allowable number of people in the station is 200, the congestion rate is 50%.

  Finally, the station premises number prediction unit 203 instructs the display server 104 to update the display screen such as the station signage using the calculated values for the station premises number prediction and the station congestion rate prediction, and ends the process. (S508). FIG. 18 shows a display screen example of the station signage on the display 1041 that displays information related to the prediction of the congestion rate of the station. By displaying the congestion rate for each station and route after a certain time from the current point in time, for example, by displaying icons and graphs on the route map, users can easily determine when and how much stations and routes are expected to be crowded. I can grasp.

  Next, with reference to FIG. 11, processing operations of the guidance target person extraction unit 204 and the guidance guide unit 205 will be described. This processing is executed every time the processing operation shown in FIG. 5 ends (that is, every predetermined time).

  First, the guidance target person extraction unit 204 reads the allowable number of people table 408 and checks whether there is a station where the number of people in the station for each station calculated by the processing of FIG. 5 exceeds the allowable number of people in the station (S1101). ). As a result, if there is no station that exceeds the number of people, the process is terminated. As shown in FIG. 12, the allowable number of people table 408 includes a station name 1201 and a accommodated number of people 1202.

  As a result of the determination in S1101, if the corresponding station exists, it is next checked whether or not a predetermined time has elapsed since the previous detour request (S1102). As a result, if the predetermined time or more has not elapsed, the process is terminated. By executing the process of this step, it is possible to prevent the detour request from being transmitted many times to the user within a short time.

  On the other hand, if there is a station that has passed for a certain period of time in the processing of S1102, the guidance target person extraction unit 204 calculates how many people make a detour for each station or route to be an appropriate number of people to avoid congestion ( S1103). For example, the allowable number may be an appropriate number, and for example, 80% of the allowable number may be an appropriate number. The number obtained by subtracting the appropriate number of people from the number of people on the station premises is the number of persons to be guided.

  Next, the guidance target person extraction unit 204 acquires movement history data for a certain period from the entry / exit history storage server 110 (S1104). As shown in FIG. 13, the movement history data is composed of a card ID 1301, a use date 1302, a process name 1303, a terminal name 1304, and a use amount 1305 of a traffic IC card used when the user passes the automatic ticket gate. The

  Next, the guidance target person extraction unit 204 aggregates the behavior patterns for each card ID and stores them in the history analysis result table 409 (S1105). As shown in FIG. 14, the history analysis result table 409 includes a card ID 14010, a home station 1402, a commuting / commuting station 1403, a last use station 1404, presence / absence of a companion 1405 at the time of last use, a use date / time zone 1406, a use route. 1407, frequency 1408, and standard deviation 1409 of entry / exit time. Here, the home station extracts, for example, a station having a high appearance frequency as the first entry station of the day or the last entry station of the day from the movement history data and stores it. As the commuting / commuting station, for example, a station having a high frequency of appearance as a destination station on weekday mornings or a departure station on weekday nights is extracted from the movement history data and stored. The last use station extracts the record of the newest date and time in the movement history data and stores the station of the record. As for the presence / absence of a companion, a record with the latest date and time is extracted from the movement history data, and if a companion is present when the record is moved, the card ID of the companion is stored.

  Here, as to a method for determining whether or not a companion is present from the traffic IC card usage history data, the terminal name 1304 of the same or in the vicinity of the same station (for example, up to one on both sides), the date and time 1302 Are the same time zone (for example, the difference in date and time is within 3 seconds) and the combination of row data with the same process name 1303 (entry or participation) is extracted, the card ID 1301 of the row data is obtained, and the row data If a pair of card IDs appears in a certain amount of movement (for example, from entry to entry excluding transfer), it is determined that a companion is present. The use day / time zone and the use route store the use route for each classification such as weekdays, holidays, and what time. For example, the frequency is stored by counting the movement frequency during one week period for each use day / time zone and use route. The standard deviation of the entry / exit time is calculated and stored, for example, for the use day / time zone and the use route, the standard deviation of the entry / exit time during one week. In other words, the standard deviation becomes smaller as the route is always used on a fixed use date / time zone.

  Next, the guidance target person extraction unit 204 selects a person who requests a detour for each station or route, and stores it in the guidance target person extraction table 410 (S1106). As illustrated in FIG. 15, the guidance target person extraction table includes a card ID 1501 and guidance content 1502. As a method of selecting the person to be guided to be detoured, it is possible to move using the station during times when congestion is expected (for example, moving the station as departure / arrival / transfer / transfer station) Select people who are highly likely and who are likely to bypass. For example, if it is predicted that station B will be crowded at 19:00 on weekdays, from the analysis result of a user's movement history data, the last station used is B station, the home station is A station, and the station B at 19:00 on weekdays ⇒ If the frequency of movement to station A is high, there is a high possibility of being caught in the congestion at station B. In general, when going to school or work, it is necessary to be in time for the start of work, so it is unlikely that you will be detoured. It is extracted as a person to be guided, and is set so as to propose a detour route to a station with less congestion as guidance content. Or, for example, if the card ID user is moving less frequently or is using a route with a large standard deviation (it is highly likely that they are traveling for leisure purposes rather than commuting or attending school) Therefore, it is extracted as a guidance target person, and it is set so as to propose a change of the destination station as guidance content. Or, for example, when it is determined that the user of the card ID is moving with the accompanying person and the accompanying person is a child or a woman, the user is extracted as a guidance target person and is guided to the station with less congestion. Set to suggest alternative routes.

Next, the guidance guide unit 205 sequentially refers to the records in the guidance target person extraction table (S1107), and if there is a next record, creates a detour request message, and according to the access method and access destination address in the access destination address table 16 A message is transmitted (S1108). As shown in FIG. 16, the access destination address table includes a card ID 1601, an access method 1602, and an access destination address 1603. Data such as the access method 1602 and the access destination address 1603 is information registered when the user purchases the transportation IC card. An example of a guide mail reception screen is shown in FIG.
After that, if there is no next record in the guidance target person extraction table in S1107, the process ends.

  In the above, the example which extracts a guidance subject person using traffic IC card use history data was explained. In addition, this invention is not limited to the said Example, It can implement by changing variously.

  For example, the guidance target person may be extracted using social media history data. Alternatively, the traffic IC card use history data and the social media history data may be linked by ID, and the guidance target person may be extracted from the analysis result obtained by combining both the history data. When using social media history data, for example, the latest location information at the time of writing can be used instead of the last use station. When traffic IC card usage history data and social media history data are combined, for example, the home station / last used station can be extracted from the traffic IC card usage history, and the time of arrival at the station can be extracted from the social media history data. Therefore, it is possible to more accurately determine whether or not the user is likely to be involved in the congestion of the station, and perform appropriate guidance.

  Further, according to another example, the display server 104, the number of attendees counting server 107, the congestion prediction server 108, the guidance server 109, the entrance / exit history storage server 110, and the like are not configured as individual servers. It is possible to configure with a few servers. In the case of configuring with one server, it can be realized by having an application program for realizing the functions of each of the above servers in one server.

  Further, in the above-described embodiment, the station premises number calculation unit 201 shown in FIG. Both are used to calculate the number of people in the station. However, according to another example, the station premises number calculating unit 201 can calculate the number of premises using either the ticket gate passage data or the captured video data. When only the ticket gate passage data is used, it is uncertain that the user is at the station platform, but at least the number of people in the station premises can be grasped, and thereby the number of people in the premises can be predicted. It is also possible to install surveillance cameras not only at the platform but also at various locations within the station premises, and calculate the number of people in the premises using these captured video data.

  According to the present embodiment, it becomes possible to realize the congestion prediction with higher accuracy than the station congestion prediction according to the prior art, and in order to ensure passenger safety, the railway company optimizes the arrangement of station staff, It is possible to improve passenger guidance. In addition, it is possible to achieve more effective detour guidance than conventional guidance (stationary signage, signage, and uniform mail delivery using delay information), and leveling of congestion. Can be improved.

DESCRIPTION OF SYMBOLS 101: Information terminal 102: Network 103: Social media data storage server 104: Display server 1041: Display device 105: Automatic ticket gate 106: Surveillance camera 107: Total number of attendees count server 108: Congestion prediction server 109: Guide server 110: Entrance / exit history storage server
201: Number of people in a station premises calculation unit 202: Number of people around a station calculation unit 203: Number of people in a station premises prediction unit 204: Extraction target person extraction unit 205: Guidance guidance unit
301, 401: Main control unit 302, 402: Input unit 303, 403: Output unit 304, 404: Communication processing unit 305, 405: Storage management unit 306, 406: Communication unit 307: Congestion prediction program 308: Number of people in station 309: Action prediction determination condition table 310: Event / group information table 311: Station coordinate table 407: Guide program 408: Allowable number table 409: History analysis result table 410: Target person extraction table 411: Access destination address table 1701: Mail Reception screen 1801: Station signage screen

Claims (11)

A congestion prediction system that uses a server to predict the congestion of people in a station,
Using the data of the number of users passing through the automatic ticket gate acquired from the automatic ticket gate of the station and / or the data of the number of passengers on the train obtained from the video taken by the surveillance camera installed in the station premises A station premises number calculation unit that calculates the number of people in the station premises,
Social that stores social media data including user ID, date and time of writing, position coordinates at the time of writing, tag name for classifying written content, written content text, and link information to image data attached to writing A media storage server;
Via said network from social media storage server obtains the social media data relating to該駅, and around the station number calculation section for predicting the number on the periphery該駅using the social media data,
The number of people existing in the station after the current time based on the number of people existing in the station calculated by the number of people in the station and the number of people around the station calculated by the number of people around the station The number of people prediction section in the station premises,
Congestion prediction system. The station premises number prediction unit determines that an event in which a social media user holding the social media data participates has ended using a keyword included in the social media data , and moves to the station The congestion prediction system according to claim 1, wherein: The station premises prediction unit uses the frequency of writing the social media data to determine that an event for participation of a social media user holding the social media data has ended, and predicts movement to the station The congestion prediction system according to claim 1. 2. The station premises number prediction unit calculates the number of participants in an event in which a social media user holding the social media data participates by using photo or video data associated with the social media data. The congestion prediction system described. The station premises number prediction unit uses the location information in the social media data and the location information of the station to end the event in which the social media user holding the social media data participates. The congestion prediction system according to claim 1, which predicts whether the vehicle will arrive at the station.   Traffic IC card usage history data used when the user passes the automatic ticket gate when the station premises number prediction unit predicts that the number of people existing in the station will be larger than the allowable number of people in the station The congestion prediction system according to claim 1, further comprising: a guidance target person extracting unit that extracts a target person who guides a travel route that does not use the station.   When the number of people in the station premises predicts that the number of people existing in the station will be larger than the allowable number of people in the station, the social media history data is used to identify the target person who guides the travel route that does not use the station. The congestion prediction system according to claim 1, further comprising a guidance target person extraction unit for extraction.   The guidance target person extraction unit estimates a home station for each target person using history data for a certain period, and uses the station when it is determined that the target person moves from the destination to the home. The congestion prediction system according to claim 6, wherein the congestion prediction system is extracted as a target person who guides a moving route that is not to be transmitted.   The guidance target person extraction unit calculates the standard deviation of the entrance time or the departure time of the station for each usage route of the target person using history data for a certain period, and the usage route in which the standard deviation is equal to or greater than a specific threshold value The congestion prediction system according to claim 6, wherein when it is determined that the movement is performed, it is extracted as a target person who guides a movement route that does not use the station.   The said guidance object person extraction part is extracted as an object person who guides the movement route which does not use this station, when it determines with the accompanying person having existed at the time of station use immediately before this object person. Congestion forecasting system. A congestion prediction server comprising a station premises number calculation unit, a station periphery number calculation unit, and a station premises number prediction unit,
A congestion prediction method for predicting congestion of people in a station using the congestion prediction server,
Train obtained from the number of users passing through the automatic ticket gate acquired from the automatic ticket gate at the station by the station number calculation unit and / or video taken by a monitoring camera installed in the station A station premises number calculating step of calculating the number of people in the station premises using the number of passengers in the station,
A user ID related to the station from the social media storage server via the network , a writing date and time, a position coordinate at the time of writing, a tag name for classifying the writing content, and a writing content text , Information on the link to the image data attached to the writing, and acquiring social media data, and predicting the number of people around the station using the social media data,
Wherein the station yard number prediction unit, and the number of people present in該駅premises calculated by該駅premises number calculating step, based on the number of people who are該駅around calculated by該駅near number calculation step, after the current time A station number prediction step for predicting the number of people existing in the station,
A method for predicting congestion.

Images