JP2012196987A - Traveler flow prediction device - Google Patents

Abstract

PROBLEM TO BE SOLVED: To forecast traveler's flow including the traveler with uncertain arrival station in normal circumstances, and especially at the incident situation in a specific route by high accuracy.SOLUTION: A past date that has the pattern approximated to the pattern of the flow of the traveler actually obtained about the time period on the partway of a certain day is obtained, and the pattern of each time period of the remainder at the date of a bygone concerned during a day is obtained. The probability of each route from which the route is selected by the traveler and calculated by a prescribed formula when a plurality of routes from the departure station toward the arrival station exist, and the number of travelers is distributed to the plurality of pathways. The plurality of trains that can get on at the departure station in each time period of the remainder during each route and a day are extracted, and the number of travelers is distributed to the train. When the number of travelers is distributed to the plurality of pathways, the learning processing that compares the number of travelers who distribute each train with the actual measurement value is done by using the utility function, and the parameter of the utility function is optimized.

Description

  The present invention relates to a passenger flow prediction apparatus.

The train is operated according to a predetermined schedule. If the normal schedule cannot be maintained due to vehicle failure, bad weather, excessive congestion, etc., it is necessary to organize the operation to review the operation plan as needed. In order to perform such driving arrangement safely and reliably, it is possible to predict with high accuracy how many passengers are present between which stations in which time zone of the day. When a route is delayed, it is necessary to simulate how a passenger flows to another detour route.
For example, the technique of Patent Document 1 uses a virtual passenger model in which a departure station, an arrival station, and a movement route are set in advance, and when the train operation is interrupted, the virtual passenger waits until the operation resumes, Predict whether to bypass.
The technology of Patent Document 2 is based on the train congestion degree collected in real time from the vehicle side and the station congestion degree based on the ticket gate data collected in real time on the station side, and the subsequent diagram of the train that has started to be delayed I expect.

JP 2009-61984 (paragraph 0004) JP 2008-189180 A (paragraph 0006)

However, since the technology of Patent Document 1 is based on the premise that the arrival station of the virtual passenger is specified, the arrival station is found for the first time when the passenger enters the ticket gate, as in the case where the passenger uses an unsigned prepaid card. It is impossible to predict the movement of passengers.
In the technique of Patent Document 2, the arrival station of the passenger is not taken into consideration at all, and the prediction accuracy is low because there is a time lag between the ticket gate data and the actual train operation.

  Therefore, an object of the present invention is to predict with high accuracy the flow of passengers including passengers whose arrival station is unknown, even in normal times, particularly when a failure occurs on a specific route.

The passenger flow prediction device of the present invention stores history information including history of passenger direction, number of passengers, date and time zone at the ticket gates of the departure station and arrival station, and the station where the train stops for each train. A storage unit storing diagram information storing arrival time and departure time every time, and extracting a temporal pattern characterized by history information about a time zone up to the middle of any one day from the storage unit, A past date having a pattern that is recognized to approximate the extracted pattern is acquired from the storage unit, and history information on the acquired past date is acquired from the storage unit at least for any remaining time period of one day, When there are multiple routes from the station to the arrival station, use the utility function that outputs the utility of the route with a predetermined constant as an input value. Set for each route, based on the set utility, calculate the selection probability that the route is selected by the passenger for each route, and select the number of passengers in the history information acquired for any remaining time of the day Allocating to a plurality of routes based on the probability, and for each of the allocated routes and every remaining time zone of one day, extracts the schedule information of a plurality of trains that a passenger can board at the departure station from the storage unit, The number of passengers in the history information acquired for each remaining time of the day is distributed to the extracted trains as the number of passengers who get on at the departure station and get off at the arrival station. And a control unit that optimizes the input value by comparing the allocated number of passengers with the actual measurement value.
Other features will be described in the detailed description.

  According to the present invention, it is possible to predict the flow of passengers including passengers whose arrival station is unknown, with high accuracy even during normal times and when a failure occurs on a specific route.

(A) And (b) is a figure explaining the route network which concerns on this embodiment. It is a lineblock diagram of a passenger flow prediction device concerning this embodiment. (A) is a figure showing an example of route data concerning this embodiment. (B) is a figure showing an example of station data concerning this embodiment. (C) is a figure showing an example of transfer time data concerning this embodiment. It is a figure which shows an example of the diamond information which concerns on this embodiment. (A) is a figure showing an example of train passenger number data concerning this embodiment. (B) is a figure showing an example of camera measurement data concerning this embodiment. (A) is a figure which shows an example of the ticket gate entrance / exit information which concerns on this embodiment. (B) is a figure showing an example of OD data concerning this embodiment. (A) is a figure showing an example of a route distribution constant table concerning this embodiment. (B) is a figure which shows an example of the optimal path | route distribution constant table which concerns on this embodiment. (C) is a figure which shows an example of the failure information which concerns on this embodiment. It is a figure which shows an example of the passenger bundle data which concerns on this embodiment. It is a figure showing an example of passenger flow data concerning this embodiment. It is a figure which shows an example of the route data concerning this embodiment. (A), (b) and (c) is a figure explaining the fundamental view of the OD data determination processing procedure which concerns on this embodiment. It is a figure which shows an example of the data for displaying the passenger bundle data which concerns on this embodiment. (A) And (b) is a figure which shows the example of a display of passenger bundle data which concerns on this embodiment. (A) is a figure showing an example of a failure monitor concerning this embodiment. (B) is a figure which shows an example of the detour display which concerns on this embodiment. (A) And (b) is a figure which shows the example of a display of the passenger flow data which concerns on this embodiment. (A) And (b) is a figure which shows the example of a display of the passenger flow data which concerns on this embodiment. (A) And (b) is a figure which shows the example of a display of the passenger flow data which concerns on this embodiment. (A) And (b) is a figure explaining the shortest route search model which concerns on this embodiment. It is a flowchart of the OD data determination processing procedure which concerns on this embodiment. It is a flowchart of the route determination processing procedure which concerns on this embodiment. It is a detailed flowchart of step S503 of the route determination processing procedure according to the present embodiment. It is a detailed flowchart of step S506 of the route determination processing procedure according to the present embodiment. It is a flowchart of the train allocation process procedure which concerns on this embodiment. It is a flowchart of the route distribution constant learning process procedure which concerns on this embodiment.

  Hereinafter, a mode for carrying out the present invention (referred to as “the present embodiment”) will be described in detail with reference to the drawings.

(Explanation of terms, etc.)
The route network will be described with reference to FIG. The route network consists of a plurality of stations and routes. In FIG. 1A, there are six stations, station a, station b, station c, station d, station e, and station f. A route is defined as a line between one station and another station. For example, in FIG. 1A, a route A exists between a station a and a station c. The route A has a station b between the station a and the station c. Similarly, a route B exists between the station c and the station d, a route C exists between the station c and the station e, a route D exists between the station e and the station f, and the station d. There is a route E between the station e and the station e. There is no station in the middle of these lines. Two stations at both ends defining a route are also referred to as “characteristic stations” of the route. In FIG. 1A, all stations except for the station b are characteristic stations.

  The route network usually has a mesh shape or a star shape connected to one. That is, a passenger can reach all other stations by using a train from one station. When two stations are assumed, there are often a plurality of paths from one of the stations to another station. For example, in FIG. 1A, there are two routes from the station a to the station f: “route A → route C → route D” and “route A → route B → route E → route D”. A route is conceivable.

  Of these two stations assumed from the viewpoint of passengers, the starting point is called the “departure station”, and the end station is called the “arrival station”. Such route candidates are called “routes”. A route may include a plurality of routes, and may be completed within one route. Of the routes when there is no failure, the route that takes the shortest time from the departure station to the arrival station is called a “normal route”, and one or more other routes are “detoured” This is called “route”. For one combination of departure station and arrival station, one normal route and usually one or more detour routes are defined. However, there may be no detour route. When there are a plurality of detour routes, they may be referred to as “detour route 1”, “detour route 2”,... In the shortest time from the departure from the departure station to the arrival station. In addition, it is assumed that the train operation has a delay time from the planning time. Even if the route is interrupted due to a “signal failure” or the like, it is delayed from the train operation at the time of planning and is included in the delay.

  In FIG. 1A, it is assumed that the normal route among the routes from the station “a” as the departure station to the station “f” as the arrival station is a route “route A → route C → route D”. In the present embodiment, unless there is a situation such as a route failure, the flow of passengers is assumed to pass exclusively through a normal route. In FIG. 1B, a failure has occurred in the route C included in the normal route, and a delay has occurred in the route. In this case, a part or all of the flow of the passenger who has passed the normal route “Route A → Route C → Route D” is changed to “Route A → Route B → Route E → Route D” which is a detour route. The number of passengers to be switched to the detour route is determined in consideration of the required time difference between the normal route and the detour route, the number of transfers, the fare, etc. in the event of a failure.

Although details will be described later, “passenger bundle” and “passenger flow” exist as concepts for explaining the flow of passengers. “Passenger bundle” is a concept specified by a departure station, an arrival station, a route, and the number of passengers moving from the departure station to the arrival station. “Passenger flow” is a concept specified by a departure station, an arrival station, a route, the number of passengers moving from the departure station to the arrival station, and the train. In other words, passenger flow is a subordinate concept of passenger bundle, and is divided into trains for each train.
Roughly speaking, the processing of this embodiment is as follows: (1) Predict the passenger bundle and passenger flow from the current point of time on the basis of given data, and (2) display them in a form that is easy to see. (3) When a failure occurs on a certain route, the passenger bundle and passenger flow on the normal route are allocated to the passenger bundle and passenger flow on the normal route and the detour route, respectively, and (4) the distribution is performed accurately. Is to learn the optimal route allocation constant (detailed later).

(Passenger flow prediction device)
The passenger flow prediction apparatus 1 will be described with reference to FIG. The passenger flow prediction device 1 is a general computer. The passenger flow prediction device 1 includes a central control device 11, a main storage device 12, an auxiliary storage device 13, an input device 14, and an output device 15. These are connected to each other via a bus. The auxiliary storage device 13 includes route data 31, station data 32, transfer time data 33, diagram information 34, train passenger data 35, camera measurement data 36, ticket gate entrance / exit information 37, OD (origin destination) data 38, route allocation. The constant table 39, the optimum route distribution constant table 40, and the failure information 41 are stored (details will be described later).

The main storage device 12 stores an OD prediction unit 21, a route distribution unit 22, a train allocation unit 23, and a route distribution constant prior evaluation unit 24. These are programs. In the following description, when the subject is described as “XX part is”, the central control device 11 reads each program from the auxiliary storage device 13 and loads it into the main storage device 12, and then the function of each program (details will be described later). ). Each program may be stored in the auxiliary storage device 13 in advance, or may be taken into the passenger flow prediction device 1 when necessary via another storage medium or communication medium.
The main storage device 12 also stores passenger bundle data 42, passenger flow data 43 and route data 44. These data are data temporarily created and stored in the main storage device 12 in the course of processing by each program (details will be described later). These data may be permanently stored in the auxiliary storage device 13.

One passenger flow prediction device 1 is usually installed in the vicinity of a user who manages the operation of a train. However, for example, the configuration including each program may be made independent and necessary data may be transmitted and received between each other via a network.
Furthermore, one or a plurality of user terminal devices (not shown) which are independent general computers having a central control device, a main storage device, an auxiliary storage device, an input device and an output device, and a passenger flow prediction device 1 However, it is good also as a structure by which communication is ensured via the network (The wireless network between a station and a train, the wired or wireless network in a train is included). In this way, the station staff at the station, the conductor carrying the user terminal device in the train, and the screen (detailed below) created by the passenger flow prediction device 1 are displayed on the output device of the user terminal device. It is possible to organize the operation after visually confirming and cooperating with each other.

(Route data)
The route data 31 will be described with reference to FIG. In the route data 31, in association with the route name stored in the route name column 101, the feature station name is in the feature station column 102, the travel time is in the travel time column 103, and the connectable route is in the connectable route column 104. A name and a failure flag are stored in the failure flag column 105.
The route name in the route name column 101 is the name of the route.
The feature station name in the feature station column 102 is the name of the feature station on the route. Since there are two characteristic stations for one route, the characteristic station name is a combination of two station names such as “station a, station b”.

The travel time in the travel time column 103 is a time (unit: second) required for the train to travel on the route from one characteristic station to the other characteristic station. The travel time is not limited to a train that stops at each station, but is also the shortest time that takes into account a train that does not stop at a specific station. The travel time is obtained by adding the time required for traveling between all stations on the route and the average stop time at each station. The actual travel time of the route is longer than the travel time described in the example. Here, a small value such as “900 seconds” is adopted for the sake of simplicity and ease of explanation.
The connectable route name in the connectable route column 104 is a name of a route that can be changed at any station in the route.
The failure flag in the failure flag column 105 is “ON” indicating that a failure has occurred in the route and the route is delayed, or the route does not have a failure and the route operates as usual. It is either “OFF” indicating that it is being performed. Note that the failure flag of the record of the route data 31 having the route name of the record of the failure information 41 (FIG. 7C) to be described later is “ON”.
There are as many records (rows) of route data as there are routes.

(Station data)
The station data 32 will be described with reference to FIG. In the station data 32, the characteristic station name is stored in the characteristic station 1 column 112, the required time is stored in the required time 1 column 113, and the characteristic station name is stored in the characteristic station 2 column 114 in association with the station name stored in the station name column 111. The required time 2 column 115 stores the required time, and the route column 116 stores the route name.

The station name in the station name column 111 is the name of the station.
The characteristic station name in the characteristic station 1 column 112 is the name of one characteristic station among the two characteristic stations on the route to which the station belongs.
The required time in the required time 1 column 113 is the time required for the train to travel from the station to the characteristic station in the characteristic station 1 column 112.
The feature station name in the feature station 2 column 114 is the name of the other feature station among the two feature stations on the route to which the station belongs.
The required time in the required time 2 column 115 is the time required for the train to travel from the station to the characteristic station in the characteristic station 2 column 114.
The route name in the route column 116 is the name of the route to which the station belongs. When the station belongs to a plurality of routes, it is a set of names of the plurality of routes.
When the station is a characteristic station, the characteristic station 1 column 112, the required time 1 column 113, the characteristic station 2 column 114, and the required time 2 column 115 are blank (for example, the record in the first row corresponds). There are as many records of the station data 32 as the number of stations.

(Transfer time data)
The transfer time data 33 will be described with reference to FIG. In the transfer time data 33, in association with the route name stored in the route name column 121, the transfer route 1 column 122 has a route name, the transfer time 1 column 123 has a transfer time, and the transfer route 2 column 124 has a route name. The route name is stored in the transfer time 2 column 125.

The route name in the route name column 121 is the same as the route name in FIG.
The route name in the transfer route 1 column 122 is one of the routes that can be transferred from the route (the route in the column 121).
The transfer time in the transfer time 1 column 123 is the time required for the passenger to transfer from the route to the route in the transfer route 1 column 122.
The route name in the transfer route 2 column 124 is another name of routes that can be transferred from the route.
The transfer time in the transfer time 2 column 125 is the time required for the passenger to transfer from the route to the route in the transfer route 2 column 124.
There are as many transfer route columns and transfer time columns as there are transferable routes.

(Diamond information)
The diamond information 34 will be described with reference to FIG. In the schedule information 34, in association with the train number stored in the train number column 131, the route name in the route column 132, the capacity in the capacity column 133, the station name in the station name column 134, and the station name in the arrival time column 135 The arrival time is stored in the departure time column 136.
The train number in the train number column 131 is an identifier for uniquely identifying a train traveling on all routes, and in this embodiment, is a combination of the number and the common letter “let” indicating the train.
The route name in the route column 132 is the name of the route on which the train travels, and when the station described in the station name column 134 is a station other than the terminal station, the train immediately after the train leaves the station. The train shows which route is actually running. When the station described in the station name column 134 is an end station, it indicates which route the train is currently traveling immediately before the train arrives at the end station. Incidentally, when the station d becomes the starting station, the value in the route column 132 is either the route B or the route E.
The capacity of the capacity column 133 is the number of passengers who can get on the train.
The station name in the station name column 134 is the name of the station where the train handles the passenger.
The arrival time in the arrival time column 135 is the time when the train arrives at the station. The actual arrival time is expressed in hours, minutes and seconds. However, here, for simplicity and ease of explanation, the “second” portion is omitted (the same applies to the departure time described below).
The departure time in the departure time column 136 is a time when the train leaves the station.

  There are as many records of the schedule information 34 as the number of combinations of train numbers and station names. When the station name is the first station, the arrival time column 135 is blank. When the station name is the terminal station, the departure time column 136 is blank. Incidentally, the records in the first to fourth lines in FIG. 4 indicate the following. That is, the train with the train number “1” has a capacity of 800 people, departs station a at “8:30”, arrives at station b at “8:37”, Depart at 8:37, arrive at station c at 8:45, leave station c at 8:45, arrive at station d at 8:58, and to station c Means traveling on route A and traveling on route B to station d. Different schedule information 34 may be prepared, for example, for each route and / or for each day, but in this embodiment, there is only one diagram information 34 for all trains traveling on all routes. It shall be.

(Train passenger data)
The train passenger number data 35 will be described with reference to FIG. In the train passenger number data 35, the departure station name is stored in the departure station column 142, the departure time is stored in the departure time column 143, and the number of passengers is stored in the passenger number column 144 in association with the train number stored in the train number column 141. Is remembered.
The train number in the train number column 141 is the same as the train number in FIG.
The departure station name in the departure station column 142 is the name of the station where the train actually started.
The departure time in the departure time column 143 is the time when the train actually leaves the station.
The number of passengers in the number of passengers column 144 is the number of passengers who are considered to actually get on the train at the departure time. An error in the number of passengers here is allowed.

  For example, a variable load device (not shown) mounted on the floor of the train keeps track of the number of passengers in real time, and associates it with the train number, departure station name, and departure time at each station (such as when the door is closed). It is assumed that the number of passengers is transmitted to the passenger flow prediction device 1 via a wired and / or wireless network. The train passenger number data 35 is created for each day, and is given the date of the day and stored in the auxiliary storage device 13. There are as many records as the number of combinations of train numbers and departure station names in the train passenger number data 35 of each date.

(Camera measurement data)
The camera measurement data 36 will be described with reference to FIG. In the camera measurement data 36, in association with the time zone stored in the time zone column 151, an observation location is stored in the observation location column 152, and the waiting number is stored in the number of people column 153.
The time zone in the time zone column 151 is a range obtained by dividing one day into equal intervals, and in this embodiment, the interval is 15 minutes.
The observation place in the observation place column 152 is a place where the trains arrive and depart. In this embodiment, it is assumed that there are two platforms for each station: a home (home a) where an up-bound train arrives and departs and a home (home b) where a down-bound train arrives and departs.
The number of people waiting in the number of people column 153 is an average value of the number of passengers waiting for the train at the observation site.

  For example, a camera device (not shown) installed at the home acquires the number of passengers waiting for the train at the home at predetermined intervals. When the length of the time zone is 15 minutes and the predetermined interval is 1 minute, the camera device averages the acquired 15 values as “waiting number of people”. It is assumed that the camera device transmits the waiting number in association with the time zone and the observation place to the passenger flow prediction device 1 via a wired and / or wireless network. The camera measurement data 36 is created for each day, and the date of the day is attached and stored in the auxiliary storage device 13. There are as many records in the camera measurement data 36 for each date as there are combinations of time zones and observation locations.

(Ticket entrance / exit information)
The ticket gate entry / exit information 37 will be described with reference to FIG. In the ticket gate entry / exit information 37, the entry / exit column 162 stores the entry / exit direction and the number of people column 163 stores the number of people in association with the time zone stored in the time zone column 161.
The time zone in the time zone column 161 is the same as the time zone in FIG.
The direction of entry / exit in the entry / exit column 162 is the direction in which the passenger passes through the ticket gate of the station, and is either “entrance” entering the ticket gate or “entrance” exiting from the ticket gate.
The number of people in the number of people column 163 is the number of passengers who have passed through the ticket gate.

For example, an automatic ticket gate (not shown) installed at each station acquires the number of passengers who passed the automatic ticket gate (or the number of tickets and commuter passes read by the automatic ticket gate) during a certain period of time. Calculate the total number of passengers and tickets. It is assumed that the automatic ticket gate transmits the number of people to the passenger flow prediction device 1 through a wired and / or wireless network in association with the time zone and the direction of entry / exit. The ticket gate entry / exit information 37 is created every day and every station. Each time a record is created, the created record is sent to the passenger flow prediction device 1 with the date and station name of the day, and stored in the auxiliary storage device 13. That is, at the time when a certain time zone has passed, the passenger flow prediction device 1 stores a record relating to the time zone and a record relating to a time zone before that for each station.
In the ticket gate entrance / exit information 37 for a specific station on a specific date, there are as many records as the number of combinations of time zones and entrance directions.

(OD data)
The OD data 38 indicating the number of people moving will be described with reference to FIG. In the OD data 38, the departure station name is stored in the departure station column 172, the arrival station name is stored in the arrival station column 173, and the number of people is moved in the movement number column 174 in association with the time zone stored in the time zone column 171. Has been.
The time zone in the time zone column 171 is the same as the time zone in FIG. The time included in the time zone is the time when an automatic ticket gate (not shown) installed at each station (departure station) reads a boarding ticket or a commuter pass that was introduced by the passenger.
The departure station name in the departure station column 172 is the name of the departure station.
The arrival station name in the arrival station column 173 is the name of the arrival station.
The number of people in the number of people column 174 is the number of passengers heading from the departure station to the arrival station.
For example, an automatic ticket gate installed at each station (departure station) reads the “arrival station” of the boarding ticket and commuter pass that the passenger has entered. The number of times read is accumulated for each “arrival station”. It is assumed that the automatic ticket gate transmits the number of passengers to the passenger flow prediction apparatus 1 via a wired and / or wireless network in association with the time zone, the departure station name, and the arrival station name.

  In some cases, the arrival station name cannot be acquired at the departure station, as in the case where the passenger inserts an unsigned prepaid card. However, the automatic ticket gate at the departure station acquires the identification number and the name of the departure station at the departure station, and transmits these information to the passenger flow prediction device 1. The automatic ticket gate at the arrival station After acquiring the identification number and the arrival station name, the information is transmitted to the passenger flow prediction device 1, and then the passenger flow prediction device 1 collates the same identification numbers, so that the passenger can check the ticket gate of the arrival station. At the time of exiting, the data necessary for creating the record of the OD data 38 is prepared. The OD data 38 is created every day and every station, and the date and station name of the day are attached and stored in the auxiliary storage device 13. In the OD data 38 of a specific station on a specific date, there are records corresponding to the number of combinations of time zone, departure station name, and arrival station name. The OD data 38 is similar to the passenger bundle data 42 (detailed later), but differs from the passenger bundle data 42 in that the route is not considered.

(Route allocation constant)
When there are a plurality of routes from a certain departure station to a certain arrival station, it is generally performed to calculate the “utility” of each route.

U _m = β ₁ × X _1m + β ₂ × X _2m + α (Formula 1)

Formula 1 is an example of a “utility function” that outputs utility U _m when X _1m and X _2m are input as variables. Here, m indicates an individual route (m = 1, 2,..., N), “n” is the total number of routes from a specific departure station to a specific arrival station, and U _m is This is the utility of the route “m”. The variable X _1m is the time required for the route “m” (details will be described later). The variable X _2m is “0” if the route “m” is a normal route, and “1” if the route “m” is a detour route. The coefficients β ₁ and β ₂ and the constant term α of the variables X _1m and X _2m , respectively, are parameters that can be set with various values. A set of these parameters (β ₁ , β ₂ , α) is collectively called a route allocation constant.

  When there are a plurality of routes for which utility is calculated, the probability that each route is selected by the passenger can be calculated.

P _m = exp (U _m ) / (exp (U ₁ ) + exp (U ₂ ) +... + Exp (U _n ))
(Formula 2)

In Expression 2, exp (U _m ) is an exponential function of U _m with e (base of natural logarithm) as the base. P _m is the probability that the route “m” is selected from the n route candidates. Then, the passenger bundle of the normal route is distributed to a plurality of detour routes in a form proportional to the probability. Furthermore, the route allocation constant can be optimized by learning by changing the route allocation constant in various ways (details will be described later).

(Route allocation constant table)
The route distribution constant table 39 will be described with reference to FIG. In the route allocation constant table 39, the value of the parameter β ₁ is stored in the β ₁ column 182, the value of the parameter β ₂ is stored in the β ₂ column 183, and the value of the parameter β ₂ is stored in the α column 184 in association with the number stored in the number column 181. Stores the value of the parameter α.
The number in the number column 181 is an identifier that uniquely identifies a set of parameters.
The value of the parameter β _{1 in} the β ₁ column 182, the value of the parameter β _{2 in} the β ₂ column 183 and the value of the parameter α in the α column 184 are arbitrary set values. In the present embodiment, on the condition that there are no redundant route allocation constants in which all three parameters have the same value, regular (for example, only a certain parameter value is changed little by little). ), Or randomly, the result of generating a set of parameters is stored.
The number of records in the route allocation constant table 39 is arbitrary.

(Optimum route distribution constant table)
The optimum route distribution constant table 40 will be described with reference to FIG. In the optimum route allocation constant table 40, the value of the parameter β ₁ is stored in the β ₁ column 192, the value of the parameter β ₂ is stored in the β ₂ column 193, and the α column 194 is associated with the date stored in the date column 191. Stores the value of the parameter α, and the prediction error column 195 stores the prediction error.
The date in the date column 191 is a past date.
The value of the parameter β _{1 in} the β ₁ column 192, the value of the parameter β _{2 in} the β ₂ column 193, and the value of the parameter α in the α column 194 are the values of the parameter β _{1 in} the β ₁ column 182 of FIG. Value, the same as the value of the parameter β _{2 in} the β ₂ column 183 and the value of the parameter α in the α column 184, and these constitute a path allocation constant. However, the parameter values stored in FIG. 7B are optimized for each date as a result of so-called “supervised learning” (detailed later).

The prediction error in the prediction error column 195 includes the passenger flow data 43 of the date predicted using the route allocation constant, the train passenger number data 35 actually measured (FIG. 5A), and / or camera measurement. This is a difference from the data 36 (FIG. 5B) (details will be described later).
By the way, the record on the first line uses a plurality of candidates for route allocation constants, and based on Equation 1 and Equation 2, how the passenger bundle and the passenger flow are detoured for the past year, month, and month X As a result, the path allocation constant (β ₁ , β ₂ , α) = (− 0.004, 2.2, 10) is predicted to be closest to the actually measured data. It shows that the value has been calculated. There are as many records in the optimum route allocation constant table 40 as the number of past dates. The past date may be any date in the past, but is preferably the date on which a failure has occurred on a certain route.

(Fault information)
The failure information 41 will be described with reference to FIG. In the failure information 41, in association with the route name stored in the route name column 201, the type of failure is stored in the failure type column 202, and the delay expected time is stored in the delay expected time column 203.
The route name in the route name column 201 is the same as the route name in FIG.
The type of failure in the type of failure column 202 is the cause of the delay of the route, such as “signal failure”, “personal accident”, “strong wind”, and the like.
The estimated delay time in the estimated delay time column 203 is a time that will take more than the normal time to move and arrive on the route. If the route is temporarily disconnected due to a “signal failure” or the like, the estimated recovery time from a state where the signal is disconnected such as “signal failure” is assumed. When a train is operating on a route, the average time that the train operation on the route at the present time is delayed from the planning time is assumed to be the delay expected time. In actual routes, the expected delay time is often much larger than the numerical value described as an example. However, a small value such as “5 minutes” is adopted here for the sake of simplicity and ease of explanation.
It is assumed that the record information of the failure information 41 is input by the user via the input device 14, for example. There are as many records of the fault information 41 as the number of routes delayed at the present time.

(Passenger bundle data)
The passenger bundle data 42 will be described with reference to FIG. In the passenger bundle data 42, in association with the time zone stored in the time zone column 211, the departure station column 212 has a departure station name, the arrival station column 213 has an arrival station name, and the route column 214 has a route. The number of people column 215 stores the number of people moving, and the bypass flag column 216 stores a bypass flag.
The time zone in the time zone column 211 is the same as the time zone in FIG.
The departure station name in the departure station column 212 is the same as the departure station name in FIG.
The arrival station name in the arrival station column 213 is the same as the arrival station name in FIG.
The route in the route column 214 is a route from the departure station to the arrival station, and is represented by the name of one or more routes. When there are a plurality of routes, they are arranged from the left in the order in which the train travels. The route of the record with the detour flag “OFF” is a normal route, and the route of the record with the detour flag “ON” is a detour route (details will be described later).
The number of people in the number of people column 215 is the number of passengers heading from the departure station to the arrival station through the route.
The detour flag in the detour flag column 216 is either “ON” indicating that the route in the route column 214 is a detour route, or “OFF” indicating that the route in the route column 204 is a normal route.

If there are no delayed routes, the number of records of the passenger bundle data 42 is equal to the number of combinations of time zone, departure station name, and arrival station name. When there are delayed routes, there are as many combinations as the time zone, departure station name, arrival station name, and route (normal route and detour route).
Incidentally, paying attention to the record in the fourth row in FIG. 8, the route “route A, route C” is a normal route. If the route C is delayed, a record with the route “route A, route B, route E” and the detour flag “ON” is created corresponding to the record (line 5 in FIG. 8). ). In this case, the only routes that can replace the delayed route C are “route B, route E”. However, the actual route network is much more complicated than the example of the present embodiment, and the number of routes is large, so that a plurality of detour routes can correspond to one combination of the departure station name and the arrival station name. In the example of FIG. 8, for example, two or more records having different paths can be created corresponding to the record in the fourth row. The number of people moved “10” (the sum of “3” in the fourth row and “7” in the fifth row) is allocated to those three or more routes (details will be described later).

(Passenger flow data)
The passenger flow data 43 will be described with reference to FIG. In the passenger flow data 43, in association with the train number stored in the train number column 221, the route column 222 has a route name, the capacity column 223 has a capacity, the station name column 224 has a station name, and the arrival time column 225. Is the arrival time, the departure time column 226 is the departure time, the boarding / exiting number column 227 is the number of boarding / exiting people, the boarding number of people column 228 is the number of boarding people, and the detouring boarding / exiting number column 229 is the number of boarding / exiting people. The number of people on the detour boarding 220 stores the number of people on the detour boarding.
The description of the train number column 221 to the departure time column 226 is the same as the description of the train number column 131 to the departure time column 136 of the diagram information 34 (FIG. 4).

The number of passengers in the boarding / exiting number column 227 is the number of passengers who get on the train at the station (number of passengers), the number of passengers who get off the train at the station (number of people getting off), and the number obtained by deducting the number of getting off from the number of passengers Three values of (number of people getting on and off the net) are arranged in that order, and each value is a predicted value.
The number of passengers in the number of passengers column 228 is a predicted value of the number of passengers on the train immediately after leaving the station. When the station is the terminal station, the number of passengers is always “0”.

The number of people getting on and off in the bypass number of passengers column 229 is the number of passengers who get on the train at the station when the route becomes a part of the detour route as a result of a delay of another route. (Detour passengers), the number of passengers getting off the train at the station (the number of people getting around by detour) and the number of people getting off the detour from the detour passengers (number of people getting around the Internet) Values are arranged in that order, and each value is a predicted value.
The number of people on the bypass in the number of people on the bypass 230 column is a predicted value of the number of passengers who are on the train immediately after leaving the station, resulting from the detour. When the station is the terminal station, the number of detour passengers is always “0”.
That is, the number of people who get on and off and the number of people who get on and around the detour are the numbers of people who get on and off and people who get on and off, respectively. The detour passenger number column 229 and the detour boarding number column 230 are blank when there is no delayed route.

  There are as many records of the passenger flow data 43 as combinations of train numbers and station names. When the station name is the first station, the arrival time column 225 is blank. When the station name is the terminal station, the departure time column 226 is blank. Incidentally, a description will be given by taking the records in the first to fourth lines in FIG. These records are records for the train whose train number described in FIG. 4 is “1”. Referring to the boarding / exiting number field 227 and the number-on-boarding number field 228, as a result of 400 people getting on and 0 people getting off at the station a, the number of boarding people immediately after departure is 400. At station b, 100 people get on and 100 people get off. As a result, the number of people on board immediately after departure is 400. At station c, 200 people get on and 100 people get off. As a result, the number of people on board immediately after departure is 500. At station d, 0 people get on and 500 people get off, so the number of people on board is 0. The same applies to the detour passenger number column 229 and the detour boarding number column 230, except that the value is an internal number.

(Route data)
The route data 44 will be described with reference to FIG. In the route data 44, in association with the departure station name stored in the departure station column 231, the arrival station column 232 has an arrival station name, the normal route column 233 has a normal route, and the required time column 234 has a required time. The detour route 1 column 235 has a detour route, the required time 1 column 236 has a required time, the influence range 1 column 237 has an influence range, and the right side has a detour route 2 (3,...) Column. 238, 241, ..., required time 2 (3, ...) columns 239, 242, ..., influence range 2 (3, ...) columns 240, 243, ... in that order. It is memorized repeatedly. That is, for a combination of the departure station and the arrival station, when there are a plurality of detour routes, the second and subsequent combinations of the detour route, the required time, and the influence range are stored on the right side of the column 237.

The departure station name in the departure station column 231 is the same as the departure station name in FIG.
The arrival station name in the arrival station column 232 is the same as the arrival station name in FIG.
The normal route in the normal route column 233 is a normal route from the departure station to the arrival station, and is represented by the name of one or more routes.
The required time in the required time column 234 is the time required for the train to travel from the departure station to the arrival station through the normal route. The required time in the required time field 234 of the record when the detour path 1 field 235 (the detour path 2 field 238, the detour path 3 field 241,...) Of a record is not blank is the detour path 1 of the record. When the column 235 and others are blank, that is, when the failure has not occurred, the value is larger by the delay expected time (FIG. 7C) than the required time in the required time column 234 of the record.

The detour route in the detour route 1 column 235 is a detour route from the departure station to the arrival station, and is represented by the name of one or more routes. The detour route differs depending on which route of the normal route is delayed.
The required time in the required time 1 column 236 is the time required to travel from the departure station to the arrival station through the detour route in the detour route 1 column 235.
The influence range in the influence range 1 column 237 indicates a route name that is a difference between the normal route and the detour route in the detour route 1 column 235. For example, “Route C → Route B, Route E” means that “Route C” in the normal route is replaced with “Route B, Route E” in the alternative route in the alternative route 1 column 235, that is, “Route C”. As a result of the delay, it is shown that the passenger is flowing on “Route B, Route E”.
The columns 238, 239, 240, 241, 242, 243,... Are the same as those described in the columns 235 to 237. There are as many records of the route data 44 as there are combinations of departure station names and arrival station names.

The route data 44 in FIG. 10 is an example of the route data 44 created when a failure occurs in the route C and the route C is delayed (FIG. 1B).
For the records in the first to third lines, the detour path 1 column 235 and the right column 236,... Are blank. This indicates that the normal routes “Route A” and “Route A, Route B” are not affected by the delay of the route C.
For the records on the fourth and fifth lines, information is stored in the detour path 1 column 235 and the columns 236 and 277 on the right side thereof. This is because the normal routes “Route A, Route C” and “Route A, Route C, Route D” are affected by the delay due to the delay of the route C, and are substituted for those routes. A, route B, route E "and detour route" route A, route B, route E, route D "are present.

(Processing procedure)
The processing procedure includes (1) OD data determination processing procedure, (2) route determination processing procedure, (3) train allocation processing procedure, and (4) route allocation constant learning processing procedure. In order to perform the process (2), it is assumed that the processes (1) and (4) have been completed. In order to perform the process (3), it is assumed that the process (2) has been completed. The processing procedures of (1), (2) and (3) are executed while the train is operating, whereas the processing procedure of (4) is performed at an appropriate timing (usually at night batch processing). ) Is executed.

(OD data decision processing procedure)
The basic concept of the OD data determination processing procedure will be described with reference to FIGS. 11 (a), 11 (b), and 11 (c). FIG. 11 (b) is a representation of the ticket gate entrance / exit information 37 (FIG. 6 (a)) at a certain station of the past year, year, month, and month X in a histogram. (Numbers of odd-numbered bar graphs counted from the left) and the number of participants (numbers of even-numbered bar graphs counted from the left) are described. FIG. 11 (c) is a representation of the ticket gate entrance / exit information 37 at the station on the past year, month, month, and month Y, and shows the number of people involved in the entry and the number of people involved in the participation for all time zones. Has been.

  FIG. 11 (a) is a representation of the ticket gate entrance / exit information 37 at the corresponding station on the current day in a histogram, but the present time is, for example, a certain time in the morning, and of course all the time zones (until midnight) The number of people involved and the number of people involved are not acquired. Needless to say, the ticket entry / exit information 37 on a specific date tells the flow of passengers on that day. For example, when the weather is rainy, the number of people involved in the entry and the number of people involved in the entry tend to be lower and stable compared to the time when the weather is sunny. In addition, the situation of congestion and accidents is also reflected. In an extreme example, if both the number of people involved in admission and the number of people involved in participation are 0, it may be determined that the time zone is out of service.

Looking at Fig. 11 from this point of view, it is highly likely that it was rainy for a while from the morning on year Y, month Y, and it was possible that year Y, month X and "the day" were clear from the morning. High nature.
By comparing the portion of the histogram that has already been acquired on the day with the corresponding portion of the histogram in the past (the portion surrounded by the broken line), the portion in FIG. 11A becomes the portion in FIG. It is possible to determine which of the portions in FIG. That is, the comparison is based on past date ticket gates having a pattern that is recognized as approximating a temporal pattern characterized by the direction of passengers and the number of passengers actually acquired for a time period up to the middle of a day. A comparison to determine entry / exit information. In the case of this example, the part in FIG. 11A is more approximate to the part in FIG. And it is statistically rational to predict that the time zone after the current time of the day will be the result approximated to FIG. In this way, in the OD data determination processing procedure, the passenger flow prediction device 1 determines the past ticket entrance / exit information 37 that approximates a part of the ticket entrance / exit information 37 on the current day, and the OD data of the past date. Can be identified.

For convenience of explanation, the description of FIGS. 12 to 18 will be postponed, and the OD data determination processing procedure will be described with reference to FIG.
In step S401, the OD prediction unit 21 that predicts the OD data of the day acquires the ticket gate entry / exit information of the day. Specifically, the OD predicting unit 21 stores all the records of the ticket gate entrance / exit information 36 (FIG. 6A) having the date of the day stored in the auxiliary storage device 13 up to the current time point in all the records. Get about the station. For example, when the current time is “9:05” on May 10, 2010, as described above, the passenger flow prediction device 1 is attached with “May 10, 2010: station a”. Ticket gate entry / exit information 36 having records up to the time zone “8: 45-9: 0” is already stored. The same applies to other stations.

  In step S402, the OD prediction unit 21 acquires the number of people on the day. Specifically, the OD prediction unit 21 acquires the number of all records acquired in step S401. For example, when the time zone starts from “8: 30-8: 45” and the current time is “9:05”, for “station a”, “station a, 8: 30-8: 45, “10” for “admission”, “20” for “station a, 8: 30-8: 45, entry”, “15” for “station a, 8: 45-9: 0, admission”, “station a, 8 : 4-9: 00, “25” is acquired for “participation” (FIG. 6A). Further, although not shown in FIG. 6A, regarding “station b”, “station b, 8: 30-8: 45, entrance” is “12”, “station b, 8: 30-8” ： 45, “Enter” “22”, “Station b, 8: 45-9: 0, Admission” “18”, “Station b, 8: 45-9: 0, Enter” “28” It is assumed that the same applies to the other stations c, d,.

  In step S403, the OD prediction unit 21 creates an entrance / exit number vector for the day. Specifically, the OD prediction unit 21 determines the number of people acquired in step S402 in the order of station a, station b,... (10, 20, 15, 25, 12, 22, 18, 28,. ) And arrange them as the entrance / exit number vector for the day. In the place of “...”, The number of stations c, station d,.

  In step S404, the OD prediction unit 21 creates a past entry / exit number vector. Specifically, the OD prediction unit 21 acquires the ticket gate entrance / exit information 37 having a certain date from the auxiliary storage device 13. Then, the number of persons is acquired by the method described in steps S402 and S403 for all the time zones prior to the time zone including the current time point of the record of the obtained ticket entrance / exit information 37, and for all the stations. Create entrance / exit information vector. This process is repeated for all past dates. When the repetition is completed, the past entrance / exit number vector is created for the number of past dates.

  In step S405, the OD prediction unit 21 calculates the distance between the entry / exit vector on the current day and the past entry / exit vector. Specifically, the OD prediction unit 21 calculates the Mahalanobis distance between the entrance / exit number vector of the day created in step S403 and the past entrance / exit number vector created in step S404. This process is repeated for all past attendance vector. When the repetition is completed, the Mahalanobis distance is calculated by the number of past dates. Since the calculation of the Mahalanobis distance is a well-known technique, description thereof is omitted.

  In step S406, the OD prediction unit 21 determines past ticket gate entry / exit information that most closely approximates that day. Specifically, the OD prediction unit 21 determines a past date corresponding to the minimum Mahalanobis distance among the Mahalanobis distances calculated in step S405.

In step S407, the OD prediction unit 21 acquires past OD data that is most approximated on the current day. Specifically, the OD prediction unit 21 searches the auxiliary storage device 13 using the past date determined in step S406 as a search key, and acquires the corresponding OD data 38 (FIG. 6B). The acquired OD data 38 is referred to as “OD data of the day”. The OD data 38 of the day is acquired for each station by the number of stations.
The ticket gate entrance / exit information 37 on that day is compared with the ticket entrance / exit information 37 having the past date acquired in step S406 for all time zones before the time zone including the current time point. The OD data on the current day may be corrected by adding the average value of the difference in the number of persons to the number of people on the OD data on the current day.

  The OD prediction unit 21 may calculate a distance defined by another method instead of calculating the Mahalanobis distance between vectors. Furthermore, instead of the distance between the vectors, the inner product of the vectors and the angle formed by the vectors may be calculated. In other words, any value may be calculated as long as the degree of approximation between the current day's entrance / exit vector and the past entrance / exit vector can be quantified.

(Route determination procedure)
The route determination processing procedure will be described with reference to FIG.
In step S501, the path distribution unit 22 acquires a record of OD data. Specifically, the route allocation unit 22 records the earliest time zone (referred to as “processing target time zone”) among unprocessed records having a time zone after the time zone including the current time from the OD data 38 of the current day. To get.

In step S502, the route distribution unit 22 creates a route data record. Specifically, the route distribution unit 22 first creates as many new records as the number of combinations of departure station names and arrival station names in which each column of the route data 43 (FIG. 10) is blank.
Secondly, the combination of the departure station name and the arrival station name is stored in the departure station column 231 and the arrival station column 232 of the newly created record.

  In step S503, the route distribution unit 22 completes route data. Details of this step will be described later. When the step is completed, the route data 44 (FIG. 10) is stored in the main storage device 12 in a completed state. However, at this stage, the influence range 1 (2, 3,...) Columns 237, 240, 243,... Remain blank.

  In step S504, the route distribution unit 22 determines whether there is a delayed route. Specifically, the route distribution unit 22 refers to the failure information 41 (FIG. 7C), and if a record exists (step S504 “YES”), the route name of the record (“delayed route name”) The type of failure and the expected delay time are held, and the process proceeds to step S505. In other cases (step S504 “NO”), nothing is held and the process proceeds to step S507.

In step S505, the route distribution unit 22 reviews the normal route and the detour route. Specifically, the route distribution unit 22 firstly uses the delay route name as a search key, the normal route column 233, the detour route 1 column 235, the detour route 2 column 238 of the route data 43 (FIG. 10),.・ Search. Then, among the route names included in the normal route (or detour route), “#” is added to the same route name as the delay route name, for example, “route A #”.
Second, for a record having a route including a route to which “#” is added (referred to as a “review record”), the detour route having the delay route and the corresponding required time are deleted (the column is left blank). Furthermore, the detour route that is left without being deleted and its required time are moved so as to fill in the blanks in order from the left.
Third, if a route with “#” added is included in the normal route in the normal route column 233, the estimated delay time held in step S504 is added to the required time in the required time column 234 of the record. Add and store.
Fourthly, with respect to the review record, the normal route and the detour route are compared, and the route that is the difference is stored in the influence range 1 column 237 and others as “route C → route B, route E”. The left side of “→” is a route to which “#” is added, and the right side of “→” is a route that substitutes for that route.
Fifth, “#” added in “first” is deleted, and for example, “route A #” is restored to “route A”.

  In step S506, the route distribution unit 22 distributes the number of moving people to each route. Details of this step will be described later. At the end of the step, in association with all combinations of departure station name and arrival station name, normal route, number of people on normal route, detour route 1, number of people on detour route 1, detour route 2, movement of detour route 2 The number of persons,... Is stored in the main storage device 12.

In step S507, the route distribution unit 22 creates passenger bundle data. Specifically, the route distribution unit 22 first creates a new record of the passenger bundle data 42 (FIG. 8) in which each column is blank.
Secondly, the departure station name, arrival station name, normal route, and normal stored in the main storage device 12 in the departure station column 212, arrival station column 213, route column 214, and moving number column 215 of the newly created record. Memorize the number of people on the route. When one or a plurality of detour routes are associated with one combination of the departure station name and the arrival station name, a new record is created separately, and the departure station column 212, the arrival station column 213, the route column 214, and the moving number column 215 are created. In addition, the departure station name, the arrival station name, the detour route, and the number of people traveling on the detour route are stored (for example, records on lines 5 and 7 in FIG. 8).
Third, the processing target time zone is stored in the time zone column 211 of the newly created record.
Fourthly, “OFF” is stored in the bypass flag column 216 when the route of the newly created record is a normal route, and “OFF” is stored in the bypass flag column 216 when the route of the newly created record is a bypass route. "ON" is memorized.

  The route distribution unit 22 repeats the processing of steps S501 to S507 for every processing target time zone.

  In step S508, the route distribution unit 22 creates data for displaying passenger bundle data. Specifically, the route distribution unit 22 groups the records of the passenger bundle data 42 (FIG. 8) with records having the same time zone. Then, for all groups, the number of people in each record is distributed between stations and totaled between stations.

For example, for the time zone “8: 30-8: 45”, the route allocation unit 22 displays data for displaying passenger bundle data from the records in the first to seventh rows of the passenger bundle data 42 (FIG. 8) (FIG. The case where the records in the first to seventh lines of 12) are created will be described.
(1) The departure station name and arrival station name of the record on the first line in FIG. 8 are stored in the departure station column 251 and arrival station column 252 on the first line in FIG. Then, a station interval which is included in the route of the route column 214 of the record in the first row in FIG. 8 and is between the departure station and the arrival station is specified. In this case, there is only "station a-station b" between the stations. Next, the number of people in the record in the first row in FIG. 8 is stored in the column between the identified stations in FIG. In this example, “10” is stored in the station a-station b column 253.
(2) The departure station name and arrival station name of the record on the second line in FIG. 8 are stored in the departure station column 251 and arrival station column 252 on the second line in FIG. Then, it identifies between stations that are included in the route of the route column 214 of the record in the second row in FIG. 8 and are between the departure station and the arrival station. In this case, there are two stations, “station a-station b” and “station b-station c”. Next, the number of people in the record in the second row in FIG. 8 is stored in the column between all the specified stations in FIG. In this example, “20” is stored in the station a-station b column 253, and “20” is also stored in the station b-station c column 254. Since “20” is the number of persons moving from the station a to the station c (second line in FIG. 8), the same “20” is applied to both the station a-station b column 243 and the station b-station c column 244. Is stored. (Of the 20 people, no passengers get off at station b.)
(3) Similarly, records on lines 3-7 in FIG. 12 are created based on the records on lines 3-7 in FIG. However, when the detour flag “ON” is stored in the record of the passenger bundle data 42, “*” is added to the number of persons moving in the record and stored in FIG. 12 (lines 5 and 7 in FIG. 12). Eye).

In FIG. 12, if there is no delay route, the number of people with “*” should be added to the number of people without “*” in records that have the same departure station and arrival station. It was the number of people who moved.
For example, first, delete the record in the fifth line where “*” exists, and instead add “7 ^* ” to “3” between all stations in the record in the fourth line to obtain “10”. Overwrite. Similarly, delete the record on the 7th line where “*” exists, and instead add “9 ^* ” to “3” between all stations in the record on the 6th line and overwrite it as “12”. To do. The total row value is then calculated again. By doing in this way, the number of people traveling between stations when “Route C” is not delayed can also be calculated. This calculation is called “normalization simulation”.

In this way, the route distribution unit 22 totals the number of persons moving between stations after the transfer to the data (FIG. 12) for displaying the passenger bundle data for all the records of the passenger bundle data 42. In FIG. 12, the value of the total row becomes smaller toward the right. However, since “station b”, “station c”,... Are arranged in the departure station column 212 of the records in the eighth and subsequent lines in FIG. It becomes.
The route distribution unit 22 repeats the process of step S508 for all time zones.

In step S509, the route distribution unit 22 displays passenger bundle data. Specifically, the route distribution unit 22 displays a total value of data for displaying passenger bundle data on the output device 15 for each time zone and between stations.
FIGS. 13A and 13B are display examples. The height of the trapezoid 261 displayed in “station c-station d” between the stations in FIG. 13A corresponds to the value “31” in the “station c-station d” column 255 of the total row in FIG. . The height of the trapezoid 262 displayed in “station c-station d” between stations in FIG. 13B is “station c-station d” in the total row when the normalization simulation is performed in FIG. This corresponds to the value “15” in the column 255. The route distribution unit 22 executes a normalization simulation in response to an instruction from the user, and displays the result.
Lines “transportation amounts” 263 and 264 in FIGS. 13A and 13B are numbers obtained by extracting trains passing between the stations in the time zone from the diagram information 34 (FIG. 4) and summing up the capacity. It is. In the figure, although it is displayed as a horizontal line between the stations, the height of the line is calculated for each station.

In step S510, the route distribution unit 22 compares and displays the number of people traveling on the route. Specifically, the route distribution unit 22 firstly bypasses a record having the detour flag “ON” among the records of the passenger bundle data 42 (FIG. 8), the detour station name, and the arrival station name. Records whose flag is “OFF” are acquired as a set. In the example of FIG. 8, the records in the fourth and fifth rows correspond to one of the sets acquired in this way.
Secondly, the difference between the routes of the paired records and the number of moving people are acquired. In the case of the example, the difference “route C” and the moving number “3” are acquired from the record on the fourth row, and the difference “route B, route E” and the moving number “7” are acquired from the record on the fifth row. Is done. This means that the route B and the route E are used in place of the delayed route C.
3rdly, a route and a characteristic station are displayed on the output device 15 as a detour display for every time slot | zone (refer FIG.14 (b)). At this time, the thickness of the line representing the route is assumed to have a thickness proportional to the moving number of people so that the number of moving people on the route can be easily compared, or to have a color for each stage of the moving number of people.
Fourth, an arrow heading from the normal route to the detour route is displayed with the number of people moving along the detour route. In the case of FIG. 14B, the arrow is an arrow from “Route C” to “Route B, Route E”, and the number of people moving is “7”.

In step S511, the route distribution unit 22 displays a failure monitor (FIG. 14A). Specifically, the route distribution unit 22 first acquires the route name and the type of failure from the record of the failure information 41 (FIG. 7C). The route data 31 (FIG. 3 (a)) is searched using the acquired route name as a search key, and a characteristic station is acquired. Then, the type of the acquired failure and the two characteristic stations are displayed in the “failure” column 271 of the failure monitor (FIG. 14A) via the output device 15 and “signal failure between station c and station e”. Is displayed. When there are a plurality of records of the failure information 41, the acquired failure type and a plurality of combinations of two characteristic stations are displayed.
Second, when the user accepts selection of any one of information such as “signal failure between station c and station e” displayed in the “failure” column 271 (selection) Then, the route distribution unit 22 acquires the route name, the type of failure, and the expected delay time of the corresponding failure information 41 record. Then, the influence range 1 column 237, the influence range 2 column 240, the influence range 3 column 243,... Of the route data 44 (FIG. 10) are searched by using the acquired route name as a search key to acquire the influence range.
Third, the route data 31 is searched based on the influence range acquired in the “second” (for example, “Route C → Route B, Route E”, etc.), and the stations between the characteristic stations corresponding to the influence range are acquired. . It is assumed that the influence range is “Route C → Route B, Route E”. The characteristic stations of line C are station c and station e. The characteristic stations of route B are station c and station d. The characteristic station of the route E is the station d and the station e. Eventually, the movement between the station c and the station e affects between the station c and the station d and between the station d and the station e. Therefore, the route distribution unit 22 displays a message “The movement between the station c and the station e has an influence between the station c and the station d and between the station d and the station e” in the failure monitor column 272. . Then, along with the type of failure acquired in “second” and the expected delay time, the route in the designed influence range is displayed in the “failure status” column 273 of the failure monitor.
Thereafter, the route determination processing procedure is terminated.

(Details of Step S503 of Route Determination Processing Procedure)
Step S503 is processing in which the route distribution unit 22 determines a normal route and one or a plurality of detour routes for each given combination of departure station and arrival station. In this process, the route distribution unit 22 uses the “shortest route search model” (FIG. 18A).

(Shortest route search model)
FIG. 18A shows a route network. The route network is a route network exclusively for explaining the shortest route search model, and is different from the route network described in FIG.
There are four characteristic stations (characteristic station p, characteristic station q, characteristic station r, characteristic station s), and 11 stations that are not characteristic stations (station 1 to station 11). Stations with branching routes and dead-end stations are always characteristic stations. Within this route network, "assumed passengers" move freely according to the following rules.
(Rule 1) An assumed passenger is first placed at a certain departure station.
(Rule 2) When there are a plurality of directions to proceed at the departure station, the assumed passenger is divided into the number of directions. And each alternation advances in each direction.
(Rule 3) When an assumed passenger reaches a characteristic station that branches into a plurality of routes, the passenger is divided into the number of routes to branch. And each alternation follows each route.
(Rule 4) The assumed passenger turns back the original route after reaching the dead end station.
(Rule 5) The assumed passenger stores the passed station names in the order of passage.
At first, after placing one assumed passenger at a certain station (departure station), when a predetermined time elapses, a plurality of alternations exist in the route network. And the number of alternations increases over time. For example, in order to acquire a route from a certain departure station to a certain arrival station together with the required time, the route distribution unit 22 performs the following processing.

The details of step S503 of the route determination processing procedure will be described with reference to FIG.
In step S601, the route distribution unit 22 places the assumed passenger at the departure station. Thereafter, the assumed passenger and its alternation (hereinafter simply referred to as “assumed passenger”) are allowed to proceed freely until a predetermined time elapses. As an example, it is assumed that the departure station is “Station 2” in FIG.

  In step S602, the route distribution unit 22 orders all assumed passengers to stop. The timing for ordering the stop may be any timing.

  In step S603, the route distribution unit 22 acquires the station names (referred to as “station order list”) stored in the order in which the assumed passengers have passed from each assumed passenger.

  In step S604, the route distribution unit 22 acquires a portion to the destination station from the station order list. As an example, it is assumed that the arrival station is “station 10”. The route distribution unit 22 discards the station order list that does not include “station 10” in the station order list acquired in step S603. And the part after reaching "station 10" for the first time is deleted from the remaining station order list. There are a plurality of station order lists obtained in this way. For example, “station 2 → station 3 → feature station q → station 6 → station 7 → station 8 → feature station r → station 9 → station 10” (FIG. 18B).

  In step S605, the route distribution unit 22 creates a combination of departure station → characteristic station → arrival station. Specifically, the route distribution unit 22 determines that a station sandwiched between two characteristic stations, a station sandwiched between the departure station and the characteristic station, and an arrival station from all the station order lists acquired in step S604. Delete a station between features. The station order list in the above example changes to “station 2 → characteristic station q → characteristic station r → station 10”.

  In step S606, the route distribution unit 22 acquires a route between the characteristic stations and a travel time. Specifically, the route distribution unit 22 searches the route data 31 (FIG. 3A) using a combination of two feature stations appearing adjacent to each other included in the station order list as a search key, and the route of the corresponding record. Get name and travel time. In this example, it is assumed that the route name “route F” and the travel time “650 seconds” are acquired using the combination of feature stations “feature station q, feature station r” as a search key. If there are other combinations of two characteristic stations that appear adjacent to each other in the station order list, the next route name and travel time are acquired in the same manner as the search keys.

In step S607, the route distribution unit 22 acquires the route to the characteristic station and the required time. Specifically, the route distribution unit 22 first searches the station data 32 (FIG. 3 (b)) using the station name of the departure station and the station name of the characteristic station included in the station order list as search keys. The time required for the record (the time required immediately to the right of the column corresponding to the station name of the feature station used as the search key) and the route name are acquired. In this example, it is assumed that “200 seconds” and “route G” are acquired using “station 2” and “characteristic station q” as search keys.
Secondly, the station data 32 (FIG. 3B) is searched using the station name of the arrival station and the station name of the feature station included in the station order list as search keys, and the time required for the corresponding record (the feature station using the search key). (The required time just to the right of the column corresponding to the station name) and the route name. In this example, it is assumed that “150 seconds” and “route H” are acquired using “station 10” and “characteristic station r” as search keys.

In step S608, the route distribution unit 22 acquires the route and the required time of the route. Specifically, the route distribution unit 22 firstly creates a route by combining the route name (which may be plural) acquired in step S606 and the route name acquired in step S607. In this example, “route G, route F, route H” are created.
Second, the transfer time between routes is acquired from the created route. The transfer time data 33 (FIG. 3c) is searched using two consecutive routes in the route as a key, and the transfer time between the routes is acquired. In this example, it is assumed that the transfer time “300 seconds” is acquired using “Route G” and “Route F” as search keys, and the transfer time “0 seconds” is acquired using “Route F” and “Route H” as search keys.
Third, add the required time acquired in step S607 and the transfer time acquired in “second” to the travel time acquired in step S606 (the sum of the travel times if multiple travel times are acquired). To get the time required for the route. In this example, 200 + 650 + 150 + 300 + 0 = 1300 seconds is acquired.
Fourth, the required time of the route acquired in “third” is temporarily stored in the main storage device 12 in association with the route created in “first”.

  The route distribution unit 22 repeats the processes in steps S605 to S608 for all the station order lists acquired in step S604. When the repetition is completed, there are a plurality of combinations of a route and a required time of a route such as “Route G, Route F, Route H”: “1300 seconds” for a combination of a predetermined departure station and a predetermined arrival station. It will be stored.

In step S609, the route distribution unit 22 ranks the routes. Specifically, first, the route distribution unit 22 temporarily arranges the combinations of the routes and the required time of the routes acquired in the repetitive processing of steps S605 to S608 in the main storage device 12 in order from the shortest required time of the route. Remember me.
Secondly, the route having the shortest route required time is defined as “normal route”, and the other routes are “detour route 1”, “detour route 2”, in order of increasing route time.・ Let's say.

  Furthermore, the route distribution unit 22 repeats the processing of steps S601 to S609 for all combinations of departure station names and arrival station names stored in “second” of step S502. When the repetitive processing (outer loop) is completed, “normal route, normal route required time, alternative route 1, alternative route 1, required time, alternative route are associated with the combination of“ departure station, arrival station ”. 2 ”, the required time of the detour route 2,... The number of detour paths and their required time may be zero or more.

In step S610, the route distribution unit 22 completes route data. Specifically, the route distribution unit 22 adds the normal route, the required time for the normal route, the detour route 1, the required time for the detour route 1, the detour route 2, and the detour in the record of the route data 44 newly created in step S502. The time required for the route 2 is stored in association with the combination of the departure station name and arrival station name already stored in the new record.
Thereafter, the process returns to step S504.

(Details of Step S506 of Route Determination Processing Procedure)
Details of step S506 of the route determination processing procedure will be described with reference to FIG.
In step S701, the route distribution unit 22 acquires an optimum route distribution constant. Specifically, the route distribution unit 22 searches the optimum route distribution constant table 40 (FIG. 7B) using the date of the OD data 38 acquired in step S407 of the OD data determination processing procedure as a search key, relevant record parameters beta ₁ value, to obtain the value of the parameter beta ₂ values and parameter alpha. It is assumed that the parameter values acquired at this time are (β ₁ , β ₂ , α) = (− 0.004, 2.2, 10).

  In step S702, the route distribution unit 22 determines a route whose utility should be compared. Specifically, the route distribution unit 22 includes a normal route, a required time, a detour route (there may be a plurality of routes) and an arbitrary record (referred to as “moving number distribution target record”) of the route data 44 (FIG. 10). Get the required time (there may be more than one) of the detour route. It is assumed that the record acquired at this time is the record on the fifth line in FIG.

In step S703, the route distribution unit 22 calculates the utility of each route. Specifically, the path distribution unit 22 firstly sets the value of the parameter β ₁ , the value of the parameter β ₂ , and the value of the parameter α acquired in step S 701 to β ₁ , β _2, and α in Equation 1, respectively. Is assigned. Formula 1 becomes “U _m = −0.004 × X _1m + 2.2 × X _2m +10”.
Second, the utility of the normal path is calculated by substituting the required time acquired in step S702 and “0” for X _1m and X _2m in Equation 1, respectively. In the case of the above example, the utility of the normal route “Route A, Route C, Route D” is −0.004 × 2200 + 2.2 × 0 + 10 = 1.2.

Thirdly, the utility of the detour path is calculated by substituting the time required for the detour path acquired in step S702 and “1” into X _1m and X _2m of Equation 1, respectively. The route distribution unit 22 executes “third” processing for each bypass route when there are a plurality of bypass routes, and does not execute “third” processing when there is no bypass route. In the case of the above example, the utility of the detour route 1 “Route A, Route B, Route E, Route D” is −0.004 × 2550 + 2.2 × 1 + 10 = 2.0.

In step S704, the route distribution unit 22 calculates the selection probability of each route. Specifically, the route distribution unit 22 calculates the normal route selection probability and the selection probability (P) of each detour route using Equation 2. When there is no detour route, the selection probability of the normal route is “1”.
In the case of the above example, the selection probability of the normal route “Route A, Route C, Route D” is exp (1.2) / (exp (1.2) + exp (2.0)) = 0.31. The selection probability of the alternative route 1 “route A, route B, route E, route D” is exp (2.0) / (exp (1.2) + exp (2.0)) = 0.69.

In step S705, the route distribution unit 22 distributes the number of moving people to each route. Specifically, the route allocation unit 22 records records having the same departure station name and arrival station name as the departure station name and arrival station name of the movement number allocation target record of the route data 44 among the records in the processing target time zone of the OD data 38. Get the number of people moving. In the example, “10” is acquired.
Second, the number of moving people acquired in “first” is multiplied by the selection probability of the normal route and the selection probability of each detour route acquired in step S704, and the number of people moving on the normal route and the movement of each detour route are calculated. Calculate the number of people.
In the case of the example, the number of people traveling on the normal route is 10 × 0.31 = 3 (rounded off the decimal point), and the number of people traveling on the detour route is 10 × 0.69 = 7 (rounded off the decimal point). is there.

The route distribution unit 22 repeats the processing of steps S702 to S705 for all the records of the route data 44 (FIG. 10). At the end of the repetition, in association with all combinations of departure station name and arrival station name, normal route, number of people on normal route, detour route 1, number of people on detour route 1, detour route 2, movement of detour route 2 The number of persons is stored in the main storage device 12. When there is no detour route, there is no moving number of people on the detour route.
Thereafter, the process returns to step S507.
In the above description, the utility of each path is calculated using a predetermined utility function such as Equation 1. However, the route distribution unit 22 may accept that the user inputs a setting value indicating the utility of each route via the input device 14.

(Train allocation procedure)
A train allocation process procedure is demonstrated along FIG.
In step S801, the train assignment unit 23 acquires a record of passenger bundle data. Specifically, the train assignment unit 23 acquires the time zone, departure station name, arrival station name, route, and number of people in the first record of the passenger bundle data 42 (FIG. 8). Here, for example, it is assumed that “8: 30-8: 45”, “station a”, “station b”, “route A”, and “10” of the record in the first row are acquired.

In step S802, the train assignment unit 23 acquires a train. Specifically, the train allocation unit 23 specifies all trains that simultaneously satisfy all the following conditions from the diagram information 34 (FIG. 4), and acquires all the records of the specified trains.
(Condition 1) The route column 132 has a route name included in the route acquired in step S801.
(Condition 2) The station name column 134 has the departure station name acquired in step S801.
(Condition 3) The departure time column 136 has the departure time included in the time zone acquired in step S801.
In the case of the example, it is assumed that the trains “1” and “3” in FIG. 4 are specified and the records on the 1st to 8th rows in FIG. 4 are acquired.

In step S803, the train allocation unit 23 allocates moving personnel to the train. Specifically, the train assigning unit 23 firstly determines the interval between the departure time at which the train specified in step S802 leaves the departure station and the departure time at which the train immediately before the train leaves the departure station (“ "Departure interval"). In the case of the example, although not shown in FIG. 4, it is assumed that there is a train that departs station a at “8:18” immediately before “1”. And the train that departs from the station a immediately before “3rd” is “1th”, and the departure time is “8:30”. At this time, a departure interval of 12 minutes is acquired for “1” and a departure interval of 5 minutes is acquired for “3”.
Secondly, the number of moving persons acquired in step S801 is distributed to the identified train based on the departure interval. In the example, 7 people (= 10 × 12 / (12 + 5)) are allocated to “1”, and 3 people (= 10 × 5 / (12 + 5)) are allocated to “3”. Is allocated.

  In step S804, the train assignment unit 23 determines the number of passengers. Specifically, the train assignment unit 23 first determines the number of people getting on and off for each train and for each station. In the case of the example, it is predicted that the moving person number “7” gets on “1” at “station a” and gets off from “1” at “station b”. At this time, the combination “7, 0, 7” of the number of passengers, the number of people getting off and the number of people getting on and off the net is stored in the main storage device 12 in association with “1” and “station a”. In association with “1” and “station b”, a combination “0, 7, -7” of the number of passengers, the number of people getting off and the number of people getting on and off the net is also stored. Similarly, the same combination “0, 0, 0” is stored in association with “1” and “station c”, and the same combination “0, 0” is stored in association with “1” and “station d”. 0, 0 "is stored.

  Similarly, it is predicted that the moving number “3” gets on “3” at “station a” and gets off from “3” at “station b”. At this time, the combination “3, 0, 3” of the number of passengers, the number of people getting off and the number of people getting on and off the net is stored in the main storage device 12 in association with “3” and “station a”. Then, the combination “0, 3, −3” of the number of passengers, the number of people getting off and the number of people getting on and off the net is also stored in association with “3” and “station b”. Similarly, a similar combination “0, 0, 0” is stored in association with “3” and “station c”, and a similar combination “0, 0” is associated with “3” and “station d”. 0, 0 "is stored.

  The train assignment unit 23 repeats the processing of steps S801 to S804 for all the records of the passenger bundle data 42 (FIG. 8). When the repetition is completed, a plurality of (number of passengers, number of people getting off, number of people getting on and off the net) is stored in association with the combination of the train number and the station name of each train.

  In step S805, the train assignment unit 23 creates passenger flow data. Specifically, the train allocating unit 23 adds a passenger number column 227, a passenger number column 228, a detour passenger number column 229, and a detour passenger number column 230 to the right of the diagram information 34 (FIG. 4). Let it be flow data 43 (FIG. 9).

In step S806, the train assignment unit 23 calculates the number of people on board. Specifically, the train assigning unit 23 first sets a plurality of sets of “passenger number, number of people getting off, number of people getting on and off the Internet” stored in the main storage device 12 in step S804 to the same train number and station name. The associated items are summed up, and the summation result is stored in the passenger number column 227 of the record having the train number and the station name in the passenger flow data 43 (FIG. 9).
Secondly, the number of passengers column 228 is filled for each train. In the record of the first departure station, the number of people getting on and off the net stored in the number of passengers column 227 is stored as the number of people on board. In subsequent station records, the number obtained by adding the number of people on the net of the record to the number of people in the previous record is stored as the number of people in the record.

In step S807, the train assignment unit 23 calculates the number of people on the detour. Specifically, first, the train assignment unit 23 sets the detour flag to “ON” among a plurality of sets of “number of passengers, number of people getting off, number of people getting on and off the net” stored in the main storage device 12 in step S804. Only those processed with respect to the record of the passenger bundle data 42, which are associated with the same train number and station name, are totaled, and the detour passenger number column of the record having the train number and station name of the passenger flow data 43 In 229, the summed result is stored.
Secondly, the number-of-passengers boarding field 230 is filled for each train. In the record of the first departure station, the number of passengers on the net stored in the bypass passenger number column 229 is stored as the number of passengers on the bypass. In subsequent station records, the number obtained by adding the number of people on the net of the record to the number of people on the detour in the previous record is stored as the number of people on the detour.

  Before explaining the next step, a display example of the passenger flow data 43 will be described along FIGS. 15 (a), 15 (b), 16 (a) and 16 (b). In these figures, the horizontal axis is time, and the vertical axis is the distance from the terminal station (station d). Horizontal lines 281 to 284 are drawn at a level corresponding to the distance from the terminal station of each station. These horizontal lines 281 to 284 represent each station. When a train is represented as a movement of a point having a time and a position at that time (distance from the terminal station) as coordinate values, a train heading from station a to station d is represented as a straight line pointing downwards from station d to station d. The train going to a is represented as a straight line descending to the right. In actual train operation, the speed of each train differs from station to station, the stopping time at each station varies, and the departure interval between trains also varies. However, in FIG. 15A and others, for simplification of explanation, the distance between all stations and the departure interval between all trains are equal, and all trains always run at the same speed. The stop time is discarded. That is, each train is represented as a straight line having the same inclination between the station a and the station d. A thick broken line 275 drawn vertically represents the current time.

  In FIG. 15A and FIG. 15B, a straight line representing a train (for example, reference numeral 286) is divided into a plurality of line segments by horizontal lines 281 to 284 representing stations. Each line segment has a thickness (reference numeral 287a, etc., indicated by a dot between a thick solid line and a thin broken line) representing the number of passengers between the stations. In addition, when the departure interval with the immediately preceding train increases, as a result of the passenger concentration on the train, the line segment of the train becomes thicker (reference numeral 287b in FIG. 15B). In the figure, the straight line indicating a certain train is displayed in the same way between all stations, but the thickness is defined for each station (for each line segment).

  In FIG. 16A and FIG. 16B, the horizontal lines 281 to 284 representing a station are divided into a plurality of line segments by a straight line representing a train (for example, reference numeral 286). A right triangle (for example, reference numeral 288a) whose hypotenuse rises (or falls) as time progresses is displayed above or below the line segment. The height of the right triangle in the direction of the vertical axis represents the number of passengers waiting at each station (referred to as “station waiting number”) in order to get on the next train. The height of the right triangle below the horizontal lines 281 to 284 representing the station represents the number of people waiting on the station for the train from the station a to the station d, and the height of the upper right triangle is from the station d to the station a. This represents the number of people waiting for the station on the train to go. In addition, when the space | interval with the last train becomes large, as a result of a passenger concentrating on the said train, the right triangle of the said train becomes higher (code | symbol 288b of FIG.16 (b)).

  In step S808, the train assignment unit 23 displays the number of people on board. Specifically, when receiving an instruction from a user who sets the “train passenger display” button 292 to “ON”, the train allocating unit 23 firstly selects a train number from the passenger flow data 43 (FIG. 9). A plurality of records are acquired, and the station name, departure time, and number of people on board of these records are acquired. And on each horizontal line which shows each station name, the point showing the departure time corresponding to the said station name is plotted, and points are connected by a line segment. Next, a first thickness proportional to the number of passengers in the corresponding board is added to each line segment (for example, reference numeral 293a in FIG. 17A).

Second, a plurality of records having the train number are acquired from the passenger flow data 43 (FIG. 9), and the station name, departure time, and number of people on the detour boarding of these records are acquired. And on each horizontal line which shows each station name, the point showing the departure time corresponding to the said station name is plotted, and points are connected by a line segment. Next, a second thickness proportional to the number of people on the corresponding detour boarding is added to each line segment (for example, reference numeral 293b in FIG. 17A).
In FIG. 17A, a first thickness and a second thickness are added to each line segment. The second thickness 293b indicates the number of passengers who will get on the train as a result of detouring, and is the number of the first thickness 293a.

In step S809, the train assignment unit 23 displays the station waiting number. When the train assignment unit 23 receives an instruction from the user who sets the “station waiting number display” button 291 to “ON”, first, a plurality of records having a certain train number from the passenger flow data 43 (FIG. 9) are recorded. Acquire the station name, departure time, and number of passengers of these records. Then, starting from a point representing the departure time on each horizontal line 281 to 284 indicating each station name, a line segment representing “the number of passengers” out of the number of passengers corresponding to the station name is raised directly above (or directly below) Fall). Furthermore, a line segment is described from the end point of the started (or lowered) line segment toward the point representing the departure time of the train immediately before on the horizontal lines 281 to 284 representing the station, and the first right angle A triangle is completed (for example, a right triangle having a height of 294a in FIG. 17B).
When the “Train Passenger Number Display” button 292 is set to “ON (OFF)”, the “Station Waiting Number Display” button 291 is automatically switched to “OFF (ON)”.

Secondly, a plurality of records having the train number are acquired from the passenger flow data 43 (FIG. 9), and the station name, departure time and detour passenger number of these records are acquired. Then, starting from a point representing the departure time on each horizontal line 281 to 284 indicating each station name, the number of detour passengers corresponding to the station name from the number of passengers corresponding to the station name (obtained in the “first”) The line segment representing the value obtained by subtracting the “number of detoured passengers” is raised directly above (or just below). Furthermore, a line segment is described from the end point of the started (or lowered) line segment toward the point representing the departure time of the train immediately before on the horizontal lines 281 to 284 representing the station. A triangle is completed (for example, a right triangle having a height of 294b in FIG. 17B).
In FIG. 17B, the first right triangle and the second right triangle are displayed in an overlapping manner. The difference between the height 294a of the first right triangle and the height 294b of the second right triangle indicates the number of people waiting on the station who will get on the train as a result of detouring, and the height of the first right triangle It is an inner number of 294a.
Thereafter, the train allocation processing procedure is terminated.

(Route allocation constant learning processing procedure)
The route distribution constant learning process procedure will be described with reference to FIG.
In step S901, the route allocation constant prior evaluation unit 24 acquires all OD data having a certain date. A certain date is referred to as a “learning reference date”.

In step S902, the route allocation constant prior evaluation unit 24 acquires a set of route allocation constants. Specifically, the path allocation constant prior evaluation unit 24 determines the value of the parameter β ₁ and the parameter β _{2 of the} record with the smallest number among the unprocessed records from the path allocation constant table 39 (FIG. 7A). And the combination of the value of the parameter α.

In step S903, the route distribution unit 22 executes steps S501 to S508 of the route determination processing procedure. The processing of each step is as described above.
However, regarding the processing in step S501, “in step S501, the route distribution unit 22 acquires a record of OD data. Specifically, the route distribution unit 22 determines from all the OD data acquired in step S901. Get the record of the earliest time zone (referred to as “processing time zone”). "
Further, regarding the process of step S701, which is the details of step S506, “in step S701, the route distribution unit 22 acquires candidates for the optimum route distribution constant. Specifically, the route distribution unit 22 determines in step S902. The combination of the acquired parameter β ₁ value, parameter β ₂ value, and parameter α value is used as a candidate for the optimum route allocation constant as it is, and the acquired parameter values are (β ₁ , β ₂ , “α) = (− 0.004, 2.2, 10)”.

In step S904, the train assignment unit 23 executes the processes of steps S801 to S807 of the train assignment processing procedure.
At this stage, the passenger flow data 43 for the learning reference date is completed.

In step S905, the route allocation constant prior evaluation unit 24 determines evaluation points. The evaluation points are, for example, the value of passenger flow data 43 (predicted value) and train passenger data 35 (FIG. 5A) or camera measurement data, such as the number of passengers immediately after leaving a certain station. Any item can be used as long as it can be compared with the value (actual value) of 36 (FIG. 5B). The route allocation constant pre-evaluation unit 24 may determine the evaluation points by accepting the user's designation. However, the larger the number of evaluation points, the longer the processing time is required. Reliability increases.
Here, it is assumed that the number of passengers at the starting station of all trains is determined as an evaluation point, and the following processing will be described on the assumption.

In step S906, the route allocation constant prior evaluation unit 24 compares the predicted value with the actually measured value. Specifically, the route allocation constant prior evaluation unit 24 first acquires the combination of the train number, the station name, and the number of passengers in the record of the passenger flow data 43 in which the arrival time column 225 is blank. . The combinations are acquired by the number of train numbers. Taking FIG. 9 as an example, “1 train, station a, 400” and “3 train, station a, 200” are acquired.
Secondly, train passenger number data 35 (FIG. 5A) having a learning reference date is acquired. Then, using the train number and station name of the combination acquired in “first” as a search key, the acquired train boarding data 35 is searched, and the number of passengers in the corresponding record is acquired. In the example, “974” is acquired for “1 station, station a, 400”, and “300” is acquired for “3 station, station a, 200”.
Third, the sum of squares of the difference between the predicted value and the actually measured value is calculated. In the case of the example, (400−974) ² + (200−300) ² is calculated, and the value is set as a prediction error.
Fourth, the prediction error is stored in the main storage device 12 in association with the path distribution constant acquired in step S902.

  The route allocation constant prior evaluation unit 24 repeats the processing of steps S902 to S906 for all the records in the route allocation constant table 39 (FIG. 7A).

In step S907, the route allocation constant prior evaluation unit 24 determines an optimum route allocation constant. Specifically, the route allocation constant prior evaluation unit 24 determines the route allocation constant associated with the minimum prediction error as the optimum route allocation constant. Then, a new record of the optimum route distribution constant table 40 (FIG. 7B) is created, the learning reference date is stored in the date column 191, and the optimum route is stored in the β ₁ column 192, the β ₂ column 193, and the α column 194. Each value of the allocation constant is stored, and the minimum prediction error is stored in the prediction error column 195.

The route allocation constant prior evaluation unit 24 repeats all the processes in steps S901 to S907 for the past date.
Thereafter, the route distribution constant learning process procedure is terminated.

(Modification 1)
In addition to the example described above, there are examples of evaluation points. For example, the route allocation constant prior evaluation unit 24 may determine the number of station standby persons at a certain time at all stations as an evaluation point in step S905.
In this case, the processing of step S906 is as follows: “In step S906, the route allocation constant prior evaluation unit 24 firstly records the passenger flow data 43 before the time at which the departure time is, For each station, a combination of the station name, departure time, and the number of passengers in all the records that are close is obtained. Taking FIG. 9 as an example, if a certain time is “8:40”, “station a, 8:35, 200” and “station b, 8:37, 100” are acquired.
Second, camera measurement data 36 (FIG. 5B) having a learning reference date is acquired. Then, the camera measurement data 36 is searched using the observation location including the station name acquired in “first” and the time zone including the departure time acquired in “first” as a search key, and the number of the corresponding records is determined. get. In the example, “150” is acquired for “station a, 8:35, 200”, and “170” is acquired for “station b, 8:37, 100”.
Third, the sum of squares of the difference between the predicted value and the actually measured value is calculated. In the case of the example, (200−150) ² + (100−170) ² is calculated, and the value is set as a prediction error.
Fourth, the prediction error is stored in the main storage device 12 in association with the path distribution constant acquired in step S902. To read.

(Modification 2)
Further, a search condition 1 field 196, a search condition 2 field 197,... May be provided on the right side of the prediction error field 195 of the optimum route distribution constant table 40 (FIG. 7B).
For example, the search condition 1 column 196 may store the weather of the date as “rain” or “snow”, and the search condition 2 column 197 may store the delay line name on the date.
In step S701, the route distribution unit 22 accepts that the user inputs the weather of the day and / or the delay route name of the day, and uses the received weather and / or the delay route name of the day as a search key as an optimum route distribution constant. The table 40 may be searched to obtain the optimum route distribution constant of the corresponding record. When there are a plurality of corresponding records, the average value of each parameter value may be set as the optimum route distribution constant.

  Further, the train passenger number data 35 (FIG. 5A) or the camera measurement data 36 (FIG. 5B) is stored in the auxiliary storage device 13 in a usable state for a part of the day (for example, only in the morning). May be. In step S701, the route distribution unit 22 uses the method described in the OD data determination processing procedure for the train passenger number data 35 or camera measurement data 36 on the day and the past train passenger number data 35 or camera measurement data 36. The date of the past train passenger data 35 or camera measurement data 36 that is most similar to the train passenger data 35 or camera measurement data 36 on the current day may be acquired. Then, the optimal route distribution constant table 40 may be searched using the acquired date as a search key, and the optimal route distribution constant of the corresponding record may be acquired.

  The present invention is not limited to the above-described embodiment, and can be modified without departing from the gist of the present invention.

1 Passenger Flow Prediction Device 11 Central Controller 12 Main Storage Device 13 Auxiliary Storage Device 14 Input Device 15 Output Device 21 OD Prediction Unit 22 Route Allocation Unit 23 Train Allocation Unit 24 Route Allocation Constant Pre-Evaluation Unit 31 Route Data 32 Station Data 33 Transfer Time data 34 Timetable information 35 Train passenger data 36 Camera measurement data 37 Ticket gate entrance / exit information 38 OD data 39 Route allocation constant table 40 Optimal route allocation constant table 41 Fault information

Claims (7)

A passenger flow prediction device that predicts passenger movement from a departure station to an arrival station using a train,
Stores history information including history regarding the direction of passengers, number of passengers, date and time zone at the ticket gates of the departure station and the arrival station,
A storage unit storing diagram information storing the arrival time and departure time for each train and for each station where the train stops,
A time-dependent pattern characterized by the history information for a time period up to the middle of an arbitrary day is extracted from the storage unit, and a past date having a pattern recognized to be approximate to the extracted pattern is stored in the storage unit And get from
The history information on the acquired past date is acquired from the storage unit at least for each remaining time zone of the arbitrary day,
In the case where there are multiple routes from the departure station to the arrival station,
A selection probability that the utility of the route is set for each route using a utility function that outputs the utility of the route with a predetermined constant as an input value, and the route is selected by a passenger based on the set utility For each route,
Distributing the number of passengers of the history information acquired for each remaining time zone of any one day to the plurality of routes based on the selection probability;
For each of the allocated routes and for each remaining time zone of any one day, the schedule information of a plurality of trains that passengers can board at the departure station is extracted from the storage unit,
The number of passengers in the history information acquired for each remaining time zone of the day, the number of passengers boarding at the departure station and getting off at the arrival station, is distributed to the extracted trains,
A controller that optimizes the input value by comparing the number of passengers allocated to each of the plurality of trains with an actual measurement value;
Having
Passenger flow prediction device characterized by. The controller is
The number of passengers allocated to the plurality of routes,
Displaying via the output device for each remaining time zone of the given day and for each route;
The passenger flow prediction device according to claim 1. The storage unit
For each combination of the departure station and the arrival station, a normal route with the shortest travel time and a detour route with the shortest travel time are stored,
The controller is
Displaying the number of passengers allocated to the detour route among the number of passengers on the normal route in association with the detour route via the output device;
The passenger flow prediction device according to claim 2, wherein: The controller is
Displaying the number of passengers allocated to the plurality of trains as the number of passengers boarding the train via an output device for each train and between stations;
The passenger flow prediction device according to claim 1. The storage unit
For each combination of the departure station and the arrival station, a normal route with the shortest travel time and a detour route with the shortest travel time are stored,
The controller is
Via the output device, displaying the number of passengers boarding the train divided into the number of passengers on the normal route and the number of passengers on the detour route;
The passenger flow prediction device according to claim 4. The controller is
Displaying the number of passengers distributed to the plurality of trains as the number of station standbys waiting for the train via an output device for each train and for each station;
The passenger flow prediction device according to claim 1. The storage unit
For each combination of the departure station and the arrival station, a normal route with the shortest travel time and a detour route with the shortest travel time are stored,
The controller is
Via the output device, the station waiting number is divided into the number of passengers on the normal route and the number of passengers on the detour route, and output,
The passenger flow prediction device according to claim 6.

Images