Passenger flow predictor, passenger flow prediction method, and passenger flow prediction program
The passenger flow prediction device improves accuracy in predicting railway passenger flows and congestion by using real-time passage data from ticket gates to update OD prediction data, addressing the limitations of conventional methods that rely on past data.
Patent Information
- Application Number
- JP2024147378
- Authority / Receiving Office
- JP · JP
- Patent Type
- Applications
- Current Assignee / Owner
- Filing Date
- 2024-08-29
- Publication Date
- 2024-12-03
- Estimated Expiration
- 2040-10-26
AI Technical Summary
Conventional passenger flow prediction methods in railway transportation heavily rely on past OD data, which may not accurately reflect future passenger flows, especially during the period between entry and exit, leading to inaccuracies in prediction results.
A passenger flow prediction device that includes a passage data acquisition unit, an OD prediction unit, and a congestion degree estimation unit, which uses real-time passage data from ticket gates to update OD prediction data and estimate congestion levels, allowing for improved accuracy and immediacy in predicting passenger flows and congestion.
The solution enables real-time prediction of passenger flows and congestion levels with enhanced accuracy by utilizing current passage data, reducing dependence on external data and hardware, and allowing for precise identification of train congestion based on updated OD prediction data.
Smart Images

Figure 2024167331000001_ABST
Abstract
Description
[Technical field]
[0001] The present invention relates to a passenger flow prediction device, a passenger flow prediction method, and a passenger flow prediction program for predicting passenger flows in rail transport. [Background technology]
[0002] Several methods have been proposed for predicting passenger flow in rail transport. According to the method disclosed in Patent Document 1, passenger flow up to the current time on the day is actually measured based on passage data acquired from an automatic ticket gate. Next, past data that is most similar to the actual measurement result is extracted from past OD (Origin-Destination) data. Next, the extracted past OD data itself, or a finely adjusted version of this data, is set as a predicted value for passenger flow from the current time onwards on the day. [Prior art documents] [Patent documents]
[0003] [Patent Document 1] JP 2010-61321 A Summary of the Invention [Problem to be solved by the invention]
[0004] According to the above method, past OD data is directly or strongly reflected in the predicted value. However, even if the passenger flow up to the present time matches the past data, the passenger flow from the present time onwards does not necessarily match the past data. In addition, although both the transit data and the OD data indicate the passenger flow in the sense that they indicate the location and time of entry and exit, they do not directly indicate the passenger flow during the period between entry and exit.
[0005] Therefore, it is believed that there is room for improvement in the prediction results (e.g., prediction accuracy and prediction content) of conventional methods for predicting passenger flow in rail transport. Therefore, an object of the present invention is to improve the prediction results of passenger flow in rail transport. [Means for solving the problem]
[0006] A passenger flow prediction device according to one embodiment of the present invention predicts passenger flow in rail transport. The passenger flow prediction device includes a passing data acquisition unit, an OD prediction unit, and a congestion degree estimation unit. The passing data acquisition unit sequentially acquires passing data for the day indicating entrance and exit stations and entrance and exit times for each passenger. The OD prediction unit updates OD prediction data indicating the prediction results of the OD data for the day as needed based on the acquired passing data for the day. The congestion degree estimation unit allocates the number of entering passengers indicated by the OD prediction data as the number of train occupants for the day based on the OD prediction data and train operation plan data for the day, and estimates the congestion degree of each train.
[0007] "Passing data" indicates the entrance station, entrance time, exit station and exit time for each passenger. Passing data is collected when each passenger passes through the ticket gate. "OD data" can be generated by aggregating passing data. OD data indicates the number of passengers for each combination of entrance station (origin) and exit station (destination) among multiple stations within the data acquisition range (for example, a railway line on which trains are operated by a railway operator that has introduced the passenger flow prediction device). If the number of stations within the data acquisition range is n, there are n x (n-1) combinations of entrance stations and exit stations.
[0008] According to the above configuration, the OD prediction unit uses the passing data for the day acquired by the passing data acquisition unit to update the OD prediction data, which indicates the prediction result of the OD data, as needed. Since the passing data for the day is used, the OD data can be predicted in real time, improving the prediction accuracy. The congestion degree estimation unit then estimates the congestion degree of the train based on this OD prediction data, which is updated in real time as needed. The congestion degree can also be estimated in real time, improving the estimation accuracy. The congestion degree is an index that shows the flow of passengers during the period between entry and exit, which is difficult to show using passing data or OD data. This makes it possible to expand the prediction content of passenger flow in rail transport. In other words, the prediction results can be improved.
[0009] The passage data acquisition unit may acquire the passage data from a ticket gate installed at each station. According to the above configuration, the passage data is acquired from the ticket gate. The entrance station and the exit station can be easily linked from the ticket used to pass through the ticket gate. Therefore, it is possible to generate highly reliable passage data, and based on this, it is possible to generate highly reliable OD prediction data. Since the congestion degree is estimated based on this OD prediction data, the estimation accuracy is improved.
[0010] In general, ticket gates are railway facilities owned by railway operators. When the passenger flow prediction device is introduced to a railway operator, the railway operator can predict OD data and estimate the congestion level by applying the railway facilities that the railway operator owns. The congestion level can be estimated with reduced or no dependency on data provided from outside. In addition, the congestion level can be estimated even if the train itself is not equipped with hardware (e.g., a weight sensor) for estimating the congestion level. Therefore, the railway operator can obtain the congestion level estimation result at low cost.
[0011] The OD prediction unit may be configured to update a plurality of OD prediction data showing the prediction results of the OD data for the day by time period as needed. The congestion degree estimation unit may be configured to estimate the congestion degree based on the plurality of OD prediction data. According to the above configuration, multiple OD prediction data are generated by time period. For example, the number of passengers from 15:00 to 15:05 is XX people, and the number of passengers from 15:05 to 15:10 is △△ people, and the OD data is predicted by dividing the time period into smaller parts. This makes it possible to instantly and easily determine the number of passengers who have occurred during the target time period of each OD prediction data based on the passing data of the day. Therefore, it is also possible to instantly and easily measure how many passengers who have occurred during the target time period have traveled to which destination, improving the prediction accuracy of the OD data. In addition, since the congestion degree is estimated based on the OD prediction data generated by dividing the time period into smaller parts, it is also easy to identify the trains that passengers who have occurred during each target time period are expected to board, improving the estimation accuracy and immediacy of the congestion degree.
[0012] The OD prediction unit may have an initialization unit that executes an initialization process to set an initial value of the OD prediction data based on past OD data, and an update unit that executes an update process to update the OD prediction data by correcting the initial value based on the passing data of the day. According to the above configuration, first, the initialization unit sets an initial value of the OD prediction data based on past OD data, and then the update unit corrects the initial value and updates the OD prediction data. Compared to a case where the initial value is set randomly, the prediction accuracy of the OD prediction data is improved. Since the OD prediction data is updated based on the passing data of the day, the prediction accuracy is improved.
[0013] The initialization unit may extract data that is estimated to be similar to the OD prediction data from the past OD data, and set the extracted past OD data as an initial value. According to the above configuration, when setting an initial value based on past OD data, the initialization unit extracts past OD data that is estimated to be similar to the OD prediction data. Since the data that is estimated to be similar is set as the initial value, the prediction accuracy of the OD prediction data is improved.
[0014] When the OD prediction unit is configured to periodically update multiple OD prediction data showing the OD data prediction results for the day by time period, the update unit may determine the number of entering passengers at each station during the time period covered by the OD prediction data, based on the passing data acquired at the end of the time period covered by the OD prediction data. According to the above configuration, the OD prediction unit generates OD prediction data by time period, and determines the number of entering passengers at each station during the target time period of the OD prediction data based on the passing data acquired at the end of the target time period. Since the number of entering passengers is determined, the prediction accuracy is improved.
[0015] The update unit may update the OD prediction data by allocating the determined number of entering passengers at each station to the number of passengers leaving other stations in the same ratio as the initial value. According to the above configuration, the initial value generated based on past OD data is updated based on the passing data of the day, real-time updating of OD prediction data can be realized, and prediction accuracy is improved.
[0016] The update unit may update the OD prediction data by partially determining the number of passengers who have entered and exited during the target time period based on passing data acquired after the end of the target time period. According to the above configuration, the number of passengers who have already exited is partially determined based on the passing data acquired after the end of the target time period, thereby improving prediction accuracy.
[0017] The update unit may determine, based on the passing data acquired after the end of the target time period and the longest travel time data that defines the longest travel time expected to be required to travel between two stations for each combination of entry station and exit station, the number of passengers who entered during the target time period of the OD prediction data and whose combination has a longest travel time that is shorter than the period from the start of the target time period to the relevant point in time.
[0018] According to the above configuration, OD data is predicted using the concept of maximum travel time. The maximum travel time is the time expected to be required to travel between two stations, defined for each combination of any two stations, an entry station and an exit station. For example, if the maximum travel time from station A to station B is 10 minutes, it is estimated that all passengers who enter station A and travel to station B during the target time period will leave station B within 10 minutes from the start of the target time period. If the passing data is reacquired 10 minutes or more after the start of the target time period, the number of passengers for the combination of station A as the entry station and station B as the exit station is determined based on this passing data. In this way, the number of passengers who have traveled a relatively short distance is determined as having exited, improving the prediction accuracy.
[0019] The update unit may estimate the number of passengers who have not yet exited by subtracting the confirmed number of passengers traveling between two stations from the confirmed number of entering passengers, and update the OD prediction data by assigning the estimated number of passengers who have not yet exited as the number of exiting passengers traveling between two stations for a combination in which the longest travel time is longer from the starting point to the relevant point in time. According to the above configuration, the number of passengers who have not yet exited is estimated from the confirmed number of entering passengers and the number of passengers who have already exited. This number of passengers who have not yet exited is allocated as the number of passengers traveling long distances. After the number of passengers who have already exited is partially confirmed, it is possible to predict where and how many of the passengers who have not yet exited will exit, improving the prediction accuracy.
[0020] The congestion degree estimation unit may have a train allocation unit that searches for trains that passengers who enter each station within the time period covered by the OD prediction data will use to travel to their exit stations based on specified search conditions based on the operation plan data, and executes a train allocation process that allocates the passengers as passenger numbers to the searched trains. According to the above configuration, the number of passengers on a train is estimated by searching for trains used by passengers based on the predicted value of OD data. The degree of congestion of a train can be estimated based on the operation plan data and the OD prediction data. Therefore, the degree of congestion can be estimated at low cost with less dependency on external data.
[0021] The predetermined search conditions may include two or more conditions. The train allocation unit may allocate passengers who have entered each station in a predetermined ratio according to the number of search conditions, search for trains that the allocated passengers will use to travel to their exit stations based on the corresponding conditions, and allocate the passengers to the searched trains as the number of passengers. When traveling between two stations, passengers may choose routes according to their preferences, such as choosing a route that will take them a shortest time to reach their destination or one that requires fewer transfers. Search conditions of 2 or more correspond to such preferences, and include conditions that give priority to extracting routes that will allow them to reach their destination the fastest, or routes that require the fewest number of transfers.
[0022] According to the above configuration, passengers are allocated in a predetermined ratio according to the number of conditions. Trains likely to be used by passengers allocated for each condition are searched for according to the condition. Then, passengers are allocated as the number of passengers of the searched train. This makes it possible to deal with the possibility that passengers traveling between the same two stations may board different trains depending on their preferences. This improves the accuracy of estimating the degree of congestion on trains.
[0023] A passenger flow prediction method according to one embodiment of the present invention is a passenger flow prediction method for predicting passenger flow in rail transport, and includes a passing data acquisition step, an OD prediction step, and a congestion prediction step. In the passing data acquisition step, passing data for the day indicating entrance and exit stations and entrance and exit times of each passenger are sequentially acquired. In the OD prediction step, OD prediction data indicating the prediction result of the OD data for the day is updated as needed based on the acquired passing data for the day. In the congestion estimation step, the train that the passenger will board is estimated based on the OD prediction data and the train operation plan for the day, the number of passengers entering indicated by the OD prediction data is allocated as the number of passengers on the train to be operated, and the congestion degree of each train is estimated.
[0024] A passenger flow prediction program according to one embodiment of the present invention causes a computer to execute the passenger flow prediction method. The above method and program have technical features corresponding to those of the passenger flow prediction device described above, and provide similar operational effects. Effect of the Invention
[0025] According to the present invention, it is possible to improve prediction results of passenger flow in rail transport. [Brief description of the drawings]
[0026] [Figure 1] 1 is a block diagram showing a passenger flow prediction device according to an embodiment of the present invention. FIG. [Diagram 2] FIG. 1 is a diagram illustrating an application example of an embodiment of the present invention. [Diagram 3] FIG. 11 is a diagram illustrating an example of passage data in a table format. [Figure 4] 1A is a diagram showing an example of OD data in a matrix, and FIG. 1B is a diagram showing an example of time-of-day OD data in a matrix. [Diagram 5] FIG. 2 is a block diagram showing a passenger flow prediction device. [Figure 6] 4 is a time chart of processes executed by an OD prediction unit and a congestion degree estimation unit. [Figure 7] FIG. 11 is a diagram illustrating an initialization process and a first update process. [Figure 8] FIG. 11 is a diagram illustrating a second update process. [Figure 9] FIG. 11 is a diagram illustrating a third update process. [Figure 10] FIG. 11 is a diagram showing a train allocation process. [Figure 11] 1 is a flowchart illustrating a passenger flow prediction method according to an embodiment of the present invention. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS
[0027] Hereinafter, an embodiment of the present invention will be described with reference to the drawings. (Passenger flow prediction device) Fig. 1 shows a passenger flow prediction device 1 according to one embodiment of the present invention. The passenger flow prediction device 1 predicts passenger flows in rail transport. As an example, the passenger flow prediction device 1 is used in passenger transportation operations of railroad operators, and is also suitably applied to information provision services for passengers.
[0028] The passenger flow prediction device 1 includes a central processing unit 11, a memory unit 12, an input unit 13, and an output unit 14, which are interconnected via a communication bus 15. The central processing unit 11 executes the passenger flow prediction method in accordance with a passenger flow prediction program P stored in the memory unit 12. The central processing unit 11 or a passenger flow prediction device 1 equipped with the central processing unit 11 is an example of a computer that executes the passenger flow prediction method.
[0029] The memory unit 12 temporarily or permanently stores data necessary for executing the passenger flow prediction method in addition to the passenger flow prediction program P. In this embodiment, the memory unit 12 includes a main memory unit 12a and an auxiliary memory unit 12b. The main memory unit 12a is realized by a storage device such as a ROM (Read-Only-Memory), a RAM (Random Access Memory), an EEPROM (Electrically Erasable Programmable Read-Only Memory), etc. The main memory unit 12a stores a passenger flow prediction program P. The main memory unit 12a also temporarily stores OD prediction data D20, which will be described later.
[0030] The auxiliary storage unit 12b is a large-scale storage device realized by, for example, a hard disk drive, etc. The auxiliary storage unit 12b may be realized by a cloud server separated from the communication bus 15. The auxiliary storage unit 12b stores various data to be referenced when predicting passenger flow, including past OD data D1, longest travel time data D2, operation data D3, holiday data D4, station data D5, route data D6, and route data D7.
[0031] The input unit 13 collects passage data D10 (see FIG. 3) indicating the movement of each passenger from the passage data collection server 2 via the communication network 9. The central processing unit 11 predicts passenger flow based on the collected passage data and various data stored in the memory unit 12. The output unit 14 outputs information indicating the predicted results of passenger flow to the administrator terminal 3, the signage 4, and the user terminal 5 via the communication network 9.
[0032] The administrator terminal 3 is an information terminal managed by an employee of the railway operator in which the passenger flow prediction device 1 is installed, and includes a display device that displays information output from the output unit 14. The signage 4 is an electronic signboard installed at each station, and includes a display device that displays information output from the output unit 14. The user terminal 5 is an information terminal carried by a passenger, such as a smartphone or tablet. When a dedicated application 5A is installed in the user terminal 5, the passenger can check information showing the passenger flow prediction results on the user terminal 5. The provider of this application 5A is, for example, the railway operator that has installed the passenger flow prediction device 1.
[0033] (passing data, OD data) As shown in FIG. 2, a railway operator owns a set of railway facilities such as tracks and multiple stations, and operates trains on the tracks to transport passengers. The set of railway facilities includes ticket gates 6 installed at each station. In FIG. 2, as a mere example, a railway operator operates trains on two lines: line P, which runs from stations A to F, and line Q, which branches off from line P at station D and runs to stations G and H. A "train" refers to a train formation made up of a single railcar, or two or more coupled railcars.
[0034] Each passenger prepares a ticket necessary to pass through the ticket gate 6, passes through the ticket gate 6 to enter the station of departure (hereinafter also referred to as the "entrance station"), travels on a train operated by a railway operator, and passes through the ticket gate 6 to exit at the station of destination (hereinafter also referred to as the "exit station"). In the following, the series of movements of an individual passenger from entry at the departure station to exit at the destination is referred to as a "trip." "Tickets" include magnetic tickets and contactless IC (integrated circuit) cards, and each ticket has various information recorded on it, such as a ticket ID (identity document) that identifies the ticket. "Tickets" include ordinary tickets, commuter passes, and over-flight tickets.
[0035] The ticket gate 6 reads the information recorded on the ticket. If there is no entry record on the ticket and the passenger passes through the ticket gate 6 moving from outside the station to inside the station, the ticket gate 6 records an entry on the ticket, assuming that the passenger is entering. If there is an entry record on the ticket and the passenger passes through the ticket gate 6 moving from inside the station to outside the station, the ticket gate 6 records an exit on the ticket, assuming that the passenger is exiting (or collects the ticket).
[0036] The ticket gate 6 transmits information regarding the entrance and exit of passengers to the passage data collection server 2, along with information regarding the ticket. When a passenger passes through the ticket gate 6 to enter, the ticket gate 6 transmits at least the entrance station information and the entrance time information. When a passenger passes through the ticket gate 6 to exit, the ticket gate 6 transmits at least the exit station information and the exit time information. In addition, the ticket gate 6 at the exit station transmits the entrance record (entrance station information and entrance time information) read from the ticket, along with the exit station information and the exit time information.
[0037] As shown in FIG. 3, the passage data collection server 2 generates passage data D10 representing one trip, which is a set of entry and exit, based on information transmitted from the ticket gates 6 at multiple stations. The passage data D10 includes entrance station information, entrance time information, exit station information, and exit time information. The entrance station information indicates the entrance station where the ticket gate 6 that read the ticket information is installed. The entrance time information indicates the time when the ticket gate 6 was passed at the entrance station. The exit station information indicates the exit station where the ticket gate 6 that read the ticket information is installed. The exit time information indicates the time when the ticket gate 6 was passed at the exit station.
[0038] While the passenger is moving, there is no exit station information and exit time information (see the passage data D10 in the second row). Therefore, the passage data collection server 2 stores the incomplete passage data D10 until the passenger exits. The passage data collection server 2 can also transmit the incomplete passage data D10 to the passenger flow prediction device 1 in response to a request from the passenger flow prediction device 1. When train operation for the day is completed, all passengers leave the destination station, and no passengers enter the departure station until the first train of the next day. Thus, all trips for the day are confirmed. The passage data collection server 2 generates OD data D1 for this day (see FIG. 4(A)) based on the confirmed group of passage data D10.
[0039] As shown in Fig. 4(A), the OD data D1 indicates the number of passengers (i.e., the number of trips) for each combination of entrance and exit stations among multiple stations within the data measurement range (in the example of Fig. 2, stations A to H on lines P and Q operated by one railway operator). The OD data D1 is represented, for example, in a matrix with entrance stations arranged in rows and exit stations arranged in columns. 2 and 4(A), the symbols "a to h" correspond to stations A to H. Although not shown, the symbols "i, j" used in this document represent any two different stations and can be appropriately replaced with the symbols "a to h". The symbol "Tij" is the total number of passengers who entered at station i and exited at station j. In other words, it is the number of trips with station i as the entry station and station j as the exit station. In other words, it is the number of passing data D10 whose entry station information indicates station i and whose exit station information indicates station j among the aggregated passing data D10 group. The symbol "Gi" indicates the total number of entering passengers at station i, and the symbol "Ai" indicates the total number of exiting passengers at station i. Ga is the sum of Tab, Tac, ..., Tag, and Tah, and Aa is the sum of Tba, Tca, ..., Tga, and Tha. The symbol "Σ" indicates the total number of trips (the number of collected passage data D10), which is the sum of Ga, Gb, . . . , Gg, and Gh, and is also the sum of Aa, Ab, .
[0040] The collection period of the passing data D10 for generating the OD data D1 is not limited to one day, but may be a predetermined time unit shorter than one day (for example, one hour, ten minutes, etc.). In this case, as shown in FIG. 4B, multiple "time period specific OD data D1A" are generated in which the collection period of the passing data D10 differs from each other. Hereinafter, the collection period of the passing data D10 corresponding to each time period specific OD data D1A is referred to as the "target time period T1."
[0041] 4B shows, as a simple example, OD data D1A by time period for the same day when the collection period is set to 5-minute intervals. As examples of multiple pieces of OD data D1A by time period, one set of target time period T1 is 14:50-14:55, another set of target time period T1 is 14:55-15:00, and the next set of target time period T1 is 15:00-15:05. If the collection period is short, such as in 5-minute increments, the trip duration may be longer than the collection period. The OD data by time period D1A covers all trips whose entry time is within the target time period T1, regardless of whether the exit time falls within the target time period T1. In the OD data by time period D1A, the symbol "Σ" indicates the total number of trips whose entry time is within the target time period T1, and the symbol "Gi" indicates the number of entering passengers at station i during the target time period T1.
[0042] The passage data collection server 2 generates OD data D1 after the last train of the day runs and before the first train of the next day runs. The input unit 13 (see FIG. 1) acquires the OD data D1 from the passage data collection server 2 before the first train runs, and the acquired OD data D1 is accumulated and stored in the memory unit 12 (see FIG. 1) as past OD data D1. The passenger flow prediction device 1 may store time-of-day OD data D1A as past OD data D1. The time-of-day OD data D1A may be generated by the passage data collection server 2 or by the passenger flow prediction device 1.
[0043] 5 is a block diagram showing the passenger flow prediction device 1. The passenger flow prediction device 1 executes a passenger flow prediction program P and is provided with a passage data acquisition unit 21, an OD prediction unit 22, and a congestion degree estimation unit 23. The passage data acquisition unit 21 sequentially acquires the passage data D10 for the day from the passage data collection server 2 at a predetermined acquisition cycle. The passage data D10 is acquired by the ticket gates 6 installed at each station. The passage data D10 indicates the entry and exit stations and entry and exit times for each passenger (see FIG. 1(B)). However, the acquired passage data D10 also includes incomplete data that indicates only the entry station and entry time because the passenger has not yet exited. The acquisition cycle of the passage data D10 is not particularly limited. For example, the acquisition cycle is set to the same as the collection period of the passage data D10 at the passage data collection server 2 (for example, 5 minutes).
[0044] (OD prediction) The OD prediction unit 22 updates the OD prediction data D20 indicating the prediction result of the OD data of the day based on the acquired passing data D10 of the day as needed. In particular, the OD prediction unit 22 generates a plurality of time-slot-specific OD prediction data D21. The length of the target time slot T1 of each time-slot-specific OD prediction data D21 is not particularly limited. For example, the length of the target time slot T1 is set to the same as the acquisition cycle of the passing data D10 in the passing data acquisition unit 21 (for example, 5 minutes).
[0045] The OD prediction unit 22 includes an initialization unit 31 and an update unit 32. The initialization unit 31 executes an initialization process for setting an initial value of the OD prediction data D20 based on the past OD data D1 stored in the storage unit 12. The update unit 32 executes an update process in which the initial value of the OD prediction data D20 is corrected based on the passing data D10 of the day that is acquired at any time, and thereby the OD prediction data D20 is updated at any time. That is, the update unit 32 updates the OD prediction data D20 multiple times based on the latest version of the passing data D10 that is acquired intermittently over time.
[0046] In this embodiment, the update is performed three times. The update process includes a first update process, a second update process, and a third update process, and the update unit 32 includes a first update unit 36, a second update unit 37, and a third update unit 38. The first update unit 36 executes the first update process on the OD prediction data after the initialization process. The second update unit 37 executes the second update process on the OD prediction data after the first update process. The third update unit 38 executes the third update process on the OD prediction data after the second update process.
[0047] 6, the first update unit 36 executes the first update process at the end t1 of the target time slot T1 of the OD prediction data D20. The second update unit 37 and the third update unit 38 execute the second update process and the third update process, respectively, at a second update time t2 and a third update time t3, which are later than the end time (first update time) t1 of the target time slot T1. The second update time t2 and the third update time t3 may be set in any manner. In this embodiment, as a mere example, the second update time t2 is set to a time point that is the length of the target time period T1 after the end t1 of the target time period T1 (for example, 10 minutes after the start t0 and 5 minutes after the end t1). The third update time t3 is set to a time point that is the same length of time after the second update time t2 (15 minutes after the start t0, 10 minutes after the end t1, and 5 minutes after the second update time t2).
[0048] The target time slots T1 of the multiple time slot-specific OD prediction data D21 are set to be continuous without interruption. The end time t1 of the target time slot T1 of a certain time slot-specific OD prediction data D21 is the same as the start time t0 of the target time slot T1 of the next time slot-specific OD prediction data D21. The second update time t2 of a certain time slot-specific OD prediction data D21 is the same as the start time t0 of the target time slot T1 of the second succeeding time slot-specific OD prediction data D21. Thus, multiple time slot-specific OD prediction data D21 are generated simultaneously.
[0049] (initialization) FIG. 7 is a diagram showing the initialization process and the first update process. In FIG. 7 and the following figures, for the sake of simplicity, only four stations, A, D, F, and H, are extracted as entry and exit stations, but the number of passengers moving between other two stations is predicted in a similar manner. A and F are terminal stations of line P. D is an intermediate station of line P, a terminal station of line Q, and a transfer station between lines P and Q. H is the terminal station on the opposite side of line Q.
[0050] In the initialization process, the initialization unit 31 extracts data that is estimated to be similar to the currently generated time-slot-specific OD forecast data D21 from the past OD data D1. The initialization unit 31 sets the extracted data as an initial value. This generates initialized time-slot-specific OD forecast data (initial data D21(t0)). There are no particular limitations on the extraction method, the number of past OD data D1 to be extracted, or whether the extracted data needs to be processed. In the illustrated example, the numerical value of the past time-slot OD data D1 for the same day of the week and the same time slot one week ago is set as the initial value as it is. For a line with high demand for commuting to work or school, it is easier to extract similar data because the data is collected on the same day of the week. The initialization unit 31 may extract past time-slot OD data for the same time slot yesterday. On a day when an event (e.g., a sports match, concert, festival, etc.) that attracts many customers is held along the line, the initialization unit 31 extracts past time-slot OD data for the same time slot on the day when the similar event was held. The initialization unit 31 may set the numerical value of a single past time-slot OD data as it is as the initial value, or may set the average value of multiple time-slot OD data as the initial value. The initialization process may be executed at any timing as long as it is completed by the end time t1 of the target time period T1 For example, the initialization process is executed at the start time t0 of the target time period T1.
[0051] (1st update) The first update process is executed at the end t1 of the target time period T1. The first update unit 36 aggregates the passage data D10 whose entry time falls within the target time period T1 from the passage data D10(t1) acquired at the end t1 of the target time period T1. In this way, the first update unit 36 determines the number Gi of entering passengers at each station in the target time period T1. Note that although some passengers may have already exited during the target time period T1, the first update unit 36 does not need the exit station information and exit time information from the passage data D10(t1) in the first update process.
[0052] The first update unit 36 allocates the determined number of entering passengers Gi at each station as the number of exiting passengers at other stations in the same ratio as the initial value. This corrects the initial value, and generates the time-series OD forecast data (first update data D21(t1)) that has been subjected to the first update process. If the number of passengers entering station i in the initial data D21(t0) is Gi(t0) and the number of passengers entering station i and leaving station j in the initial data D21(t0) is Tij(t0), the number of passengers entering station i and leaving station j, Tij(t1), is predicted based on the formula: Tij(t1)=Gi×(Tij(t0) / Gi(t0)). Then, this predicted value Tij(t1) is replaced with the initial value Tij(t0).
[0053] Specifically, the number of passengers Tad(t1) entering at station A and leaving at station D in the first update data D21(t1) is predicted to be 10. This is derived by multiplying the confirmed number of entering passengers at station A Ga (80 people) by the ratio (11 / (11+33+44)) of the number of passengers Tad(t0) entering at station A and leaving at station D in the initial data D21(t0) to the number of entering passengers at station A Ga(t0).
[0054] In other words, in the initial data D21(t0), 88 passengers who entered at Station A exited at other stations (Station D, Station F, Station H) at a ratio of 1:3:4. When generating the first update data D21(t1), the confirmed number of entering passengers at Station A Ga (80 people) is allocated as the number of exiting passengers (Station D, Station F, Station H) (10 people, 30 people, 40 people) at the same ratio of 1:3:4 as the initial value.
[0055] (2nd update) 8 is a diagram showing the second update process. The second update process is executed at a second update time t2 after the end time t1 of the target time slot T1. When the second update time t2 arrives, some of the passengers who entered during the target time slot T1 start to exit. In the second update process, the exit station information and exit time information from the passing data D10 whose entry times are within the target time slot T1 are referenced.
[0056] In the second update process, the second update unit 37 extracts data from the passing data D10(t2) acquired at the second update time t2, in which the number of passengers moving between two stations is greater than the corresponding number of passengers in the first update data D21(t1).Then, the numerical value of the first update data D21(t1) is increased to the extracted number of passengers. Next, the second update unit 37 offsets the increase correction amount by decreasing the number of passengers of other combinations with the same entrance station and different exit stations, thereby maintaining the number of entering passengers at the entrance station at the confirmed number. The share of the decrease correction amount is determined by the ratio of the number of passengers of other combinations in the first update data D21(t1). In this way, the second update processed time-slot-specific OD forecast data (second update data D21(t2)) is generated.
[0057] Specifically, at the second update time t2, the number of passengers Tfd(t2) who entered at station F and left at station D is found to be 12 people from the passing data D10(t2). This is more than the corresponding number of passengers Tfd(t1) of 10 people in the first update data D21(t1). Therefore, the second update unit 37 replaces the number of passengers Tfd(t1) related to the movement between the two stations with the number of passengers Tfd(t2) found from the passing data D10(t2) acquired at the second update time t2. In this manner, an increase correction is performed.
[0058] The amount of increase correction ΔTfd (Tfd(t2)-Tfd(t1)) is 2 people. If the increase corrections were left as they were, the number of entering passengers at Station F, Gf, would increase to 47 people, and the number of passengers that should have been confirmed as 45 would not be maintained. Therefore, the numbers of passengers Tfa(t1), Tfh(t1) of other combinations that use the same Station F as the combination for which the increase correction was made as the entering station, but have different exit stations, are decreased. The subject of the reduction correction is the number of passengers Tfa(t1), Tfh(t1) of all combinations other than the combination for which the increase correction has been applied, assuming that the entrance station is the same. The amount of the reduction correction is equal to the amount of increase correction ΔTfd, and this maintains the number of entering passengers Gf. The share of the reduction correction amount is the ratio of the number of passengers Tfa(t1), Tfh(t1) of the combination that is the subject of the reduction correction, which is 3:2 in this example.
[0059] The number of passengers Tfa(t1) whose departure station is Station A will be subject to a 60% reduction adjustment, decreasing from 21 to 1.2 people. The number of passengers Tfh(t1) whose departure station is Station H will be subject to a 40% reduction adjustment, decreasing from 14 to 0.8 people. The ratio of passenger numbers Tfa and Tfh for combinations subject to reduction adjustment will be maintained at 3:2, as it was before the reduction adjustment, even after the reduction adjustment. Note that while the unit of passenger numbers is people, the predicted value of the number of passengers does not need to be an integer.
[0060] (3rd update) FIG. 9 is a diagram showing the third update process. The third update process is executed at the end t1 of the target time slot T1 and at the third update time t3, which is later than the second update time t2. At the third update time t3, it is considered that all passengers who entered during the target time slot T1 and whose destinations were close to the entry station have already exited. Therefore, the third update unit 38 determines the number of passengers moving between two stations where the entry station and the exit station are close to each other.
[0061] To determine the number of passengers, the third update unit 38 refers to the longest travel time data D2 stored in the memory unit 12. The longest travel time data D2 defines the correspondence between a combination of an entrance station and an exit station and the longest travel time expected to be required to travel from the entrance station to the exit station. The longest travel time for each combination is a value empirically obtained by referring to the passage data D10 accumulated up to that point, and can be determined in advance during the design and manufacturing stages of the passenger flow prediction device 1.
[0062] In the illustrated example, "8" is written in the square where the row of Station A and the column of Station D intersect. That is, in the longest travel time data D2, the combination of Station A as the entrance station and Station D as the exit station corresponds to the longest travel time of 8 minutes. It is possible that the longest time may change if the entrance and exit stations are swapped. This embodiment also handles this case (see "11" in the square where Station D's row and Station A's column intersect). In this embodiment, when the number of stations within the data measurement range is n, n×(n-1) correspondences are defined by the longest travel time data D2.
[0063] The third update unit 38 determines the number of passengers for a combination of two-station movements in which the longest travel time is shorter than the period T3 from the start t0 of the target time slot T1 to the third update time t3, based on the passage data D10(t3) acquired at the third update time t3. The closer the exit station is to the entrance station, the shorter the longest travel time is. By executing this process, the number of passengers for a combination in which the exit station is close to the entrance station is determined.
[0064] Then, for a group of combinations with the same entrance station, the third update unit 38 estimates the number of passengers who have not yet exited by subtracting the number of passengers of the confirmed combination from the number of confirmed entering passengers. The estimated number of passengers who have not yet exited is assigned as the number of passengers of the unconfirmed combination. In this way, the OD forecast data after the third update process (hereinafter referred to as the third update data D21(t3)) is generated.
[0065] Specifically, when entering station A, the longest travel time to exit station D (8 minutes) and the longest travel time to exit station F (10 minutes) are shorter than the above-mentioned period T3 (15 minutes). According to the passing data D10(t3) acquired at the third update time t3, the number of passengers Tad who entered station A and exited station D is known to be 9 people. These 9 people are determined as the number of passengers Tad(t3) for the combination of station A as the entrance station and station D as the exit station. Similarly, 25 people are determined as the number of passengers Taf(t3) for the combination of station A as the entrance station and station F as the exit station.
[0066] On the other hand, the longest travel time (16 minutes) when Station H is the exit station is longer than the above period T3. Therefore, the value (10 people) determined from the passing data D10(t3) is not used to estimate the number of passengers Tah(t3) for the combination with Station A as the entry station and Station H as the exit station. Instead, the sum of the numbers of passengers Tad(t3), Taf(t3) for the combination that has entered and is confirmed to exit Station A is subtracted from the number of entering passengers Ga at Station A. This allows the number of passengers who have entered but not yet exited at Station A to be estimated (80-(9+25)=46).
[0067] In this example, there is only one combination remaining whose longest travel time is longer than the period T3 and whose exit cannot be confirmed. Therefore, the estimated number of passengers who have not yet exited (46 people) is directly allocated as the number of passengers Tah(t3) of the combination. If there are two or more combinations whose longest travel time is longer than the period T3, the estimated number of passengers who have not yet exited may be allocated as the number of passengers of the two or more combinations remaining unconfirmed in the ratio of the number of passengers in the second update data D21(t2). The determination of the number of passengers using the longest travel time may be executed in the second update process at the second update time t2.
[0068] (Estimated congestion level) Returning to FIG. 5 and FIG. 6, the congestion degree estimation unit 23 has a train allocation unit 41 that executes train allocation processing, and a congestion degree derivation unit 42 that executes congestion degree derivation processing.
[0069] In the train allocation process, the train allocation unit 41 allocates the number of entering passengers indicated by the OD prediction data D20 as the number of passengers of trains operated on that day, based on the OD prediction data D20 generated by the OD prediction unit 22 and the train operation plan data for that day. In the congestion degree derivation process, the congestion degree derivation unit 42 estimates the congestion degree of the train, based on the number of passengers of the train estimated by the train allocation unit 41.
[0070] As described above, the OD prediction unit 22 simultaneously generates multiple time-slot OD prediction data D21. The timing (estimated time) at which the congestion degree estimation unit 23 estimates the congestion degree is the same as the start time t0 of a target time slot of a certain time-slot OD prediction data D21, the same as the end time t1 (first update time) of the target time slot of the previous time-slot OD prediction data D21, the same as the second update time t2 of the previous time-slot OD prediction data D21, and the same as the third update time t3 of the previous time-slot OD prediction data D21.
[0071] According to the above-mentioned OD prediction method, at each estimation time point, there are at least initial data D21(t0), first update data D21(t1) and second update data D21(t2) which are being updated, and third update data D21(t3) which has just completed updating. The storage unit 12 may also store one or more pieces of third update data D21(t3) which have completed updating before the estimation time point.
[0072] The congestion degree estimation unit 23 may allocate trains and estimate the congestion degree at each estimation time point based on the initial data D21(t0) that sets the estimation time point as the start time t0 of the target time period. In this case, the congestion degree can be estimated in a situation where there are no passengers entering the target time period T1 at the estimation time point. The congestion degree estimation unit 23 may estimate train allocation and congestion degree at each estimation time point by taking into account one or more update data D21(t1)-(t3) that are being updated or have been updated. In this case, the congestion degree can be estimated with high accuracy based on the OD prediction data that is updated in real time using the passing data D10 of the day.
[0073] (Train allocation) The operation plan data referred to by the train allocation unit 41 is stored in the storage unit 12. The operation plan data includes various information related to train operation schedules and specifications, such as which trains will operate on which days at what times, the operating sections and stopping stations of each train, the stopping times and departure times at each stopping station, the number of cars constituting each train, and the transportation capacity (passenger capacity) of each train.
[0074] The storage unit 12 stores operation data D3 and holiday data D4 as examples of operation plan data. The operation data D3 includes operation pattern information indicating the operation pattern of the train and train information indicating the specifications of the train (e.g., the transportation capacity of the train). The operation pattern information is like a timetable. When trains operate according to different timetables for weekdays, weekends, and holidays, the operation data D3 includes multiple pieces of operation pattern information corresponding to the multiple timetables. When a special train operates on an event day when an event is held along the line, the memory unit 12 includes operation pattern information corresponding to this special train as the operation data D3.
[0075] The holiday data D4 includes calendar information for identifying whether the current day is a weekday, a weekend, or a holiday. The calendar information also includes information indicating whether the current day is an event day. By referring to the holiday data D4, it is possible to determine which operation pattern information should be applied when the operation data D3 includes multiple pieces of operation pattern information. In addition to the operation plan data, the train allocation unit 41 refers to search data necessary for searching for trains in order to estimate which train a passenger who entered each station will board. The storage unit 12 stores station data D5, line data D6, and route data D7 as examples of the search data.
[0076] The station data D5 has information on the time required for transfers (in this example, Station D is a transfer station) and the time required to move from the ticket gate 6 at each station to the platform. The line data D6 has information on a route map within the data measurement range. The route data D7 has information on a route used when moving to an exit station when on the platform of an entrance station at a certain time. A route table is generated by combining this route data D7 with the past passage data D10. A travel route table is generated in advance for each condition, such as the fastest travel route between stations and for each time period, and the minimum travel required for transfers, and these multiple travel route tables are stored in the device 1.
[0077] FIG. 8 is a diagram showing a process executed by the train allocation unit 41. In the illustrated example, the train allocation unit 41 executes the train allocation process by referring to the first update data D21(t1) in which the current estimated time point is the end time t1 of the target time slot T1. As described above, in the first update data D21(t1), the number of entering passengers at each station in the target time slot T1 is determined using the passing data D10(t1) acquired at the end time t1 (i.e., the estimated time point), and the number of passengers at each combination of the entering station and the exit station is predicted based on this determined number of entering passengers and an initial value. In the first update data D21(t1), the number of exiting passengers is not determined at all.
[0078] The train allocation unit 41 searches for trains that passengers who have entered the entrance station will use to travel to the exit station based on specified search conditions, and allocates the passengers to the searched trains as passenger numbers. The search conditions include two or more conditions. Examples of search conditions include a condition for reaching the exit station as quickly as possible (condition 1) and a condition for minimizing the number of transfers (condition 2). The train allocation unit 41 allocates passengers to each condition at a specified ratio according to the number of conditions in the search conditions.
[0079] The train allocation unit 41 further allocates the passengers allocated for each condition according to their entry times. The train allocation unit 41 sets multiple estimated entry times by dividing the target time period T1 by a predetermined time unit (for example, one-minute units). Then, the passengers allocated for each condition are further allocated as passengers who entered at each of the multiple estimated entry times. As an example, the train allocation unit 41 evenly allocates the passengers allocated for each condition as passengers who entered at each estimated entry time.
[0080] The train allocation unit 41 refers to the station data D5 to estimate the time when passengers who entered at each estimated entry time will arrive at the platform. Then, the train allocation unit 41 predicts which train a passenger who arrives at the platform at the estimated time will take to reach the exit station by referring to the operation data D3, the route data D7, or the route table. The above process will be specifically described. For example, in the first update data D21(t1), the number of passengers entering station A and leaving station F is predicted to be 30. The train allocation unit 41 predicts which train these 30 passengers will board to travel to station F.
[0081] The train a passenger chooses is also influenced by the passenger's preferences. These 30 passengers are sorted according to multiple search conditions that correspond to the passenger's preferences. In this example, two conditions are set: condition 1, which is the fastest way to reach the exit station, and condition 2, which is the least number of transfers. The ratio of passengers traveling according to condition 1 and passengers traveling according to condition 2 is set to 6:4, for example. The train allocation unit 41 estimates that, of the 30 passengers, 18 passengers will travel according to condition 1, and 12 passengers will travel according to condition 2. As a result, the 30 passengers are allocated to the two conditions.
[0082] In this example, the target time period T1 is 15:00 to 15:05, and five estimated entry times are set in one-minute units. The train allocation unit 41 evenly allocates the 18 passengers allocated to condition 1 as passengers who entered at each of the five estimated entry times. That is, the train allocation unit 41 estimates that 18 / 5 passengers entered station A by passing through the ticket gate 6 of station A at each estimated entry time. Similarly, the train allocation unit 41 evenly allocates the 12 passengers allocated to condition 2 as passengers who entered at each of the five estimated entry times. Note that the entry time "15:00" is more strictly speaking the time period "15:00:00 to 15:00:59", but here, the time period with seconds omitted is referred to as "time" for convenience. In addition, the estimated number of passengers does not need to be an integer.
[0083] Here, as exemplified by operation data D3, on that day, Saturday, at Station A, a train bound for Station F, which is the terminus of Line P, and a train bound for Station H, which runs directly from Station D to Line Q, depart alternately every five minutes. Also, although detailed illustration is omitted, on Line P, trains travel between Stations D and F, and if you board a train bound for Station H at Station A and get off at Station D, you can smoothly transfer to the round-trip train. However, this operation pattern is merely an example simplified for the sake of convenience.
[0084] As exemplified by station data D5, it is assumed that it takes two minutes to travel from the ticket gate 6 to the platform at Station A. The train allocation unit 41 searches for trains that passengers who have entered station A at each estimated entry time will take to travel to station F, according to the search conditions. A passenger who passes through ticket gate 6 at Station A at 15:00 arrives at the platform at Station A at 15:02. If the passenger who arrives at the platform at 15:02 wants to travel to Station F as quickly as possible, it is estimated that the passenger will board train A bound for Station F departing at 15:04. Thus, train allocation unit 41 estimates that 18 / 5 passengers who entered at 15:00 will board train A departing at 15:04. Similarly, train allocation unit 41 estimates that 18 / 5 passengers who entered at 15:01 and 18 / 5 passengers who entered at 15:02 will board the same train A to travel to Station F as quickly as possible.
[0085] Passengers who entered at 15:03 arrive at the platform at 15:05 after train A departs. If passengers who arrive at the platform at 15:05 want to travel to station F as quickly as possible, they are estimated to board train B bound for station H departing at 15:09 and transfer to a train bound for station F at station D. The train allocation unit 41 estimates that 18 / 5 passengers who entered at 15:03 will board train B departing at 15:09 and a round-trip train between stations D and F. Similarly, the train allocation unit 41 estimates that 18 / 5 passengers who entered at 15:04 will board train B and a round-trip train to travel to station F as quickly as possible.
[0086] On the other hand, if a passenger who enters station A at 15:00 and arrives at the platform at 15:02 wants to travel to station F with the fewest number of transfers, it is estimated that he or she will board train A bound for station F, departing at 15:04. The train allocation unit 41 estimates that 12 / 5 passengers who entered at 15:00 will board train A. The train allocation unit 41 estimates that both the 12 / 5 passengers who entered at 15:01 and the 12 / 5 passengers who entered at 15:02 will board train A in order to travel to station F with the fewest number of transfers.
[0087] If a passenger who entered station A at 15:03 and arrived at the platform at 15:05 wants to travel to station F with the fewest number of transfers, it is estimated that he or she will miss train B bound for station H, which departs at 15:09 and requires a transfer at station D, and board the next train C bound for station F, which departs at 15:14. The train allocation unit 41 estimates that 12 / 5 passengers who entered at 15:03 and travel according to condition 2 will board train C, which departs at 15:14. Similarly, the train allocation unit 41 estimates that 12 / 5 passengers who entered at 15:04 will also board train C in order to travel to station F with the fewest number of transfers.
[0088] The above train allocation process is for the number of passengers Taf(t1) of one combination of entrance and exit stations among all combinations in the first update data D21(t1). The train allocation unit 41 performs the same process as above for all remaining combinations in the first update data D21(t1). Then, the train allocation unit 41 performs the same process as above by referring to the initialization data D21(t0) whose estimated time point is the start time t0 of the target time slot T1.
[0089] Furthermore, the train allocation unit 41 performs the same processing as above by referring to the second update data D21(t2) whose estimated time is the second update time t2, and performs the same processing as above by referring to the third update data D21(t3) whose estimated time is the third update time t3. In the second update data D21(t2) and the third update data D21(t3), unlike the initialization data D21(t0) and the first update data D21(t1), the number of passengers who have exited the station is partially determined. Passengers who have exited the station have already completed their travel by train. Therefore, when using these update data D21(t2) and D21(t3), the train allocation unit 41 estimates how many passengers have boarded which trains for passengers whose exit is not yet determined.
[0090] (Derivation of congestion level) The congestion degree derivation unit 42 executes a congestion degree derivation process after such a train allocation process by the train allocation unit 41. The congestion degree derivation unit 42 derives an estimated value of the congestion degree of each train by dividing the estimated passengers of each train by the transportation capacity (capacity) of the train. As an example, information indicating the transportation capacity of the train may be included as part of the train specifications in the operation data D3 (see the lower part of FIG. 10).
[0091] (output) 1 and 5, the output unit 14 outputs the passenger flow prediction results to the manager terminal 3, the signage 4, and the user terminal 5. Of the prediction results, only the congestion degree of each train may be output to the signage 4 and the user terminal 5. The OD prediction data D20 may also be output to the manager terminal 3 together with the congestion degree as a prediction result.
[0092] (Passenger flow forecast method) 11 is a flowchart showing the passenger flow prediction method according to this embodiment. The operations of the above-mentioned components of the passenger flow prediction device 1 are realized by a passenger flow prediction program P that executes the passenger flow prediction method. Therefore, the passenger flow prediction method will be explained briefly.
[0093] The flow shown in Fig. 11 is executed sequentially at the aforementioned acquisition period (for example, every 5 minutes) of the passage data D10. First, the passage data acquisition unit 21 executes a passage data acquisition step S1 to acquire the passage data D10 from the ticket gate 6 via the communication network 9 and the passage data collection server 2. Next, the OD prediction unit 22 executes an OD prediction step S2. Then, the congestion degree estimation unit 23 executes a congestion degree estimation step S3.
[0094] In the OD prediction step S2, the OD prediction unit 22 executes a process of generating or updating multiple time-slot-specific OD prediction data D21, each of which has a different target time slot T1 at the current processing time. In this embodiment, initialization and three updates are performed for one time-slot-specific OD prediction data D21. Therefore, different processes are simultaneously performed on the four time-slot-specific OD prediction data D21 based on the passing data D10 acquired at the current processing time.
[0095] The OD prediction process S2 includes an initialization process S20, a first update process S21, a second update process S22, and a third update process S23. In the initialization process S20, the initialization unit 31 initializes the time-slot-specific OD prediction data D21, which sets the current processing time as the start time t0 of the target time slot T1. In the first update process S21, the first update unit 36 of the update unit 32 updates the time-slot specific OD forecast data D21(t1), which sets the current processing time as the end t1 of the target time slot, based on the passing data D10 acquired at the current processing time.
[0096] In the second update step S22, the second update unit 37 updates the time-of-day OD forecast data D21(t2), which has the current processing time as the second update time t2, based on the passage data D10 acquired at the current processing time. In a third update step S23, the third update unit 38 updates the time-of-day OD forecast data D21(t3), with the current processing time being the third update time t3, based on the passage data D10 acquired at the current processing time.
[0097] The congestion degree estimation step S3 includes a train allocation step S31 and a congestion degree derivation step S32. In the train allocation step S31, the train allocation unit 41 estimates how many passengers will board which train using the time-slot-specific OD forecast data D21(t1) for which the first update step S21 has been performed at the time of this processing. In the congestion degree derivation step S32, the congestion degree derivation unit 42 estimates the congestion degree of each train according to the estimated number of passengers.
[0098] (Action and effect) According to the passenger flow prediction device 1 of this embodiment, the OD prediction unit 22 updates the OD prediction data D20 indicating the prediction result of OD data as needed, using the passage data D10 of the day acquired by the passage data acquisition unit 21. Since the passage data D10 of the day is used, the OD data can be predicted in real time, improving the prediction accuracy.
[0099] The congestion degree estimation unit 23 estimates the congestion degree of the train based on the OD prediction data that is updated in real time. The congestion degree can also be estimated in real time, improving the estimation accuracy. The congestion degree is an index showing the flow of passengers during the period between entrance and exit, which is difficult to show using the passage data D10 or the OD data D1. It is possible to expand the prediction content of passenger flow in rail transport. In other words, it is possible to improve the prediction results of passenger flow.
[0100] Passage data D10 is acquired from the ticket gate 6 via the passage data collection server 2. The entrance station and the exit station can be easily linked from the ticket used to pass through the ticket gate 6. Therefore, highly reliable passage data D10 can be generated, and highly reliable OD data can be generated based on this. Furthermore, the estimation accuracy of the congestion degree estimation result is improved based on this OD data.
[0101] The ticket gate 6 is a railway facility owned by a railway operator. When the passenger flow prediction device 1 is introduced to a railway operator, the railway operator can predict the OD data D1 and estimate the congestion level by applying the railway facility owned by the railway operator. The congestion level can be estimated with reduced or no dependency on data provided from outside. In addition, the congestion level can be estimated even if the train itself is not equipped with hardware (e.g., a weight sensor) for estimating the congestion level. Therefore, the railway operator can obtain the congestion level estimation result at low cost.
[0102] A plurality of OD prediction data D20 are generated for each time period. By generating OD prediction data by dividing the time into smaller parts, the number of passengers who have occurred during the target time period T1 of the OD prediction data for each time period D21 can be instantly and easily determined based on the passing data D10 for the day. Therefore, it is also possible to instantly and easily measure how many passengers who have occurred during the target time period T1 have traveled to which destination, improving the prediction accuracy of the OD data. In addition, since the congestion level is estimated based on the OD prediction data generated by dividing the time into smaller parts, it is also easy to identify the trains that passengers who have occurred during each target time period are expected to board, improving the estimation accuracy of the congestion level.
[0103] Although the embodiment of the present invention has been described above, the above configurations can be appropriately modified, deleted, and / or added within the scope of the present invention. [Explanation of symbols]
[0104] 1. Passenger flow prediction device 6 Ticket gates 21 Passing data acquisition section 22 OD Prediction Department 23 Congestion degree estimation unit 31 Initialization section 32 Update section 41 Train Allocation Department 42 Congestion degree calculation part D1 Past OD data D2 Longest travel time data D3 Operational Data D20 OD prediction data D21 OD forecast data by time period S1 Passing data acquisition process S2 OD prediction process S3 Congestion estimation process P Passenger flow forecast program
Claims
1. A passenger flow prediction device for predicting passenger flow in rail transport, a transit data acquisition unit that acquires transit data indicating entrance / exit stations and entrance / exit times of each passenger; a congestion degree estimation unit that estimates a congestion degree of the train by allocating the number of entering passengers in a predetermined time period indicated by the passing data based on the acquired passing data and train operation plan data; Equipped with the congestion degree estimation unit has a train allocation unit that executes a train allocation process of searching for trains that passengers who have entered each station during the specified time period will use to travel to their exit stations based on specified search conditions, and allocating the passengers to the searched trains as passenger numbers; The train allocation unit allocates the number of entering passengers according to the predetermined search conditions. Passenger flow prediction device.
2. The train allocation unit further allocates the number of entering passengers allocated for each of the predetermined search conditions as passengers who entered at each of a plurality of entrance times obtained by dividing the predetermined time period by a predetermined time unit. The passenger flow prediction device according to claim 1.
3. the predetermined search criteria include two or more criteria, the congestion degree estimation unit allocates passengers entering each station in a predetermined ratio according to the number of conditions of the search conditions, searches for trains that the allocated passengers will use to travel to their exit stations based on the corresponding conditions, and allocates the passengers to the searched trains as the number of passengers; The passenger flow prediction device according to claim 1 or 2.
4. The predetermined search conditions include a condition for preferentially extracting a route that can reach the destination in the shortest time and a condition for preferentially extracting a route that requires the fewest number of transfers. The passenger flow prediction device according to any one of claims 1 to 3.
5. the train allocation unit evenly distributes the number of entering passengers allocated for each of the predetermined search conditions as passengers who entered at each of the plurality of entry times; The passenger flow prediction device according to any one of claims 1 to 4.
6. The predetermined time unit is one minute. The passenger flow prediction device according to any one of claims 1 to 5.
7. The passage data acquisition unit acquires the passage data from a ticket gate installed at each station. The passenger flow prediction device according to any one of claims 1 to 6.
8. The length of the predetermined time period is set to be the same as the acquisition period of the passing data in the passing data acquisition unit. The passenger flow prediction device according to any one of claims 1 to 7.
9. the congestion degree estimation unit, based on the acquired passing data and train operation plan data, allocates the numbers of entering passengers during a plurality of time periods set to be continuous without interruption as the number of train occupants, and estimates the congestion degree of the train; The passenger flow prediction device according to any one of claims 1 to 8.
10. the train allocation unit estimates the time when passengers who have entered at the plurality of entry times will arrive at the platform based on search data including information on the time required to travel from the ticket gate at each station to the platform in addition to the operation plan data, and searches for trains that passengers who have entered each station during the specified time period will use to travel to their exit stations; The passenger flow prediction device according to claim 2.
11. A passenger flow prediction method for predicting passenger flow in rail transport, comprising: a transit data acquisition step of acquiring transit data indicating entrance / exit stations and entrance / exit times of each passenger; a congestion degree estimation step of estimating a congestion degree of the train by allocating the number of entering passengers in a predetermined time period indicated by the passing data based on the acquired passing data and train operation plan data, as the number of train occupants; Equipped with The congestion degree estimation step includes a train allocation step of searching for trains that passengers who have entered each station during the specified time period will use to travel to their exit stations based on specified search conditions, and executing a train allocation process of allocating the passengers to the searched trains as passenger numbers; In the train allocation step, the number of entering passengers is allocated according to the predetermined search conditions. Passenger flow forecasting methods.
12. A passenger flow forecasting method according to claim 11, Passenger flow prediction program.
Citation Information
Patent Citations
Simulation device
JP2002187551A
Device for estimating number of passengers by train and by between stations, and computer readable storage medium storing program for the same
JP2005284378A
Energy cost calculation device
JP2016015062A
Evaluation system and method for evaluating operation information
JP2016168970A
Passenger flow prediction system
JP2010061321A
Cited By
Road network passenger flow detection method, device, equipment, medium and prediction system
CN119514810A
