JP7779121B2 - Train diagram simulation device, train diagram simulation method, and train diagram simulation program - Google Patents

Description

The present invention relates to a train schedule simulation device, a train schedule simulation method, and a train schedule simulation program for predicting passenger flow when a new train schedule plan is applied.

Railway operators regularly revise train schedules to optimize passenger flow. Once a new train schedule is implemented, it is practically difficult to repeatedly fine-tune the schedule through trial and error while the new schedule is in operation. Therefore, various simulators have been proposed to predict passenger flow when a new train schedule plan is implemented, so that the merits of the created plan can be evaluated before it is implemented (see, for example, Patent Documents 1-3).

Patent Document 1 discloses a device that sets desired arrival times at entry and exit stations for each passenger and estimates which train each passenger will travel on in a new train timetable proposal.
Patent Document 2 discloses a device that predicts the number of passengers and calculates the congestion rate for each section of each train in order to evaluate the suitability of a new train timetable proposal.
In the device disclosed in Patent Document 3, as the simulation time advances moment by moment, virtual passengers appear at each station based on data collected by automatic ticket gates, etc. Each virtual passenger is assigned a behavioral attribute of seat preference or fastest train selection based on a predetermined probability. The movement route of each virtual passenger is estimated according to the behavioral attribute.

Japanese Patent Application Laid-Open No. 2007-237948 Japanese Patent Application Laid-Open No. 2019-209769 Japanese Patent Application Laid-Open No. 2015-229459

However, conventional systems do not fully consider the route that each passenger actually prefers to take when traveling from their entry station to their exit station, so there is room for improvement in the accuracy of passenger flow predictions.
Therefore, the present invention aims to improve the accuracy of passenger flow prediction when estimating passenger routes in a new train timetable.

A train schedule simulation device according to one embodiment of the present invention comprises a data acquisition unit and a new route estimation unit. The data acquisition unit acquires ticket gate passage data and new train schedule data. The ticket gate passage data is collected from automatic ticket gates during operation of the current train schedule. The ticket gate passage data includes passenger entry and exit records, as well as the type of ticket used by the passenger. The new train schedule data indicates a new train schedule. The new route estimation unit references the new train schedule data and estimates a new route to be used by the passenger in the new train schedule based on the passenger entry and exit records and type of ticket.

Here, a "train schedule" refers to a train operation plan. A "current train schedule" refers to a train schedule currently in operation. A "new train schedule" is the subject of passenger flow simulations in the train schedule simulation device of the present invention, and is, for example, one of the proposed train schedules created to be applied in place of the current train schedule in future revisions. "Schedule data" includes, as information indicating the operation plan, information indicating the starting station, terminal station, and intermediate stops for each train in operation, as well as information indicating the departure time from the starting station, the arrival time and departure time at intermediate stops, and the arrival time at the terminal station.

An "entry record" in ticket gate passage data is information that includes, for example, the entrance station and the time the ticket gate was passed through. An "exit record" in ticket gate passage data is information that includes, for example, the exit station and the time the ticket gate was passed through. A set of data consisting of an entrance record and its corresponding exit record (hereinafter, this may be referred to as "trip data") identifies a single trip from entry to exit, including when and where a single passenger departed and when and where they arrived.

Unless otherwise specified, a "passenger" refers to the subject of a trip. A trip corresponds one-to-one with one passenger as its subject.
The "ticket type" in the ticket gate passage data is information that indicates the type of ticket used by a passenger to complete a trip. Examples of ticket types include a commuter pass and a regular ticket.

For passengers with regular tickets, entry times will not change significantly regardless of changes to train schedules, and it is thought that these passengers will not be bothered if their arrival times change slightly. In contrast, passengers with commuter passes have business at their destinations every day, so they are likely to be more strict about arrival times than passengers with regular tickets. Passenger pass users are likely to not want to arrive at their departure station later than their current arrival time, regardless of changes to train schedules, so as not to be late for the start of their business. In this way, it is thought that passenger route preferences differ depending on the type of ticket.

With the above configuration, the new route estimation unit estimates the new route that a passenger will likely use under a new train schedule based on the passenger's entry and exit records (i.e., trip data). This estimation takes into account not only the entry and exit records but also the ticket type. Therefore, the estimated new route can reflect differences in passenger habits depending on the ticket type. This improves the accuracy of passenger flow predictions.

The new route estimation unit may refer to the new timetable data and extract one or more routes whose departure time at the entrance station is near the entrance time included in the entrance record and whose arrival time at the exit station is near the exit time included in the exit record as route candidates for the new route. The new route estimation unit may estimate the new route from the route candidates based on the ticket gate passage data.
According to the above configuration, the new route estimation unit extracts a plurality of route candidates and estimates a new route, which improves the accuracy of passenger flow prediction compared to speculating on one route as the new route.

Among the route candidates, one whose arrival time is earlier than the departure time is considered to be an early arrival candidate, and one whose arrival time is later than the departure time is considered to be a late arrival candidate. If the ticket type is a first ticket type that has a restriction on the arrival time, the new route estimation unit may estimate that there is a high probability that the early arrival candidate will be used as the new route.
According to the above configuration, when the passenger's ticket type is the first ticket type that has a restriction on arrival time, the new route estimation unit estimates that the passenger will more likely use a route candidate whose arrival time is earlier than the departure time. Since the new route is estimated to suit the passenger's propensity, the accuracy of the new route estimation is improved.

The first ticket type may be a commuter pass.
According to the above configuration, the new route is estimated so as to conform to the commuter pass user's way of thinking about arrival time, thereby improving the accuracy of estimating the new route.
Furthermore, it is thought that passengers do not like large changes in route time differences before and after train timetable revisions, and are likely to choose routes with as little time difference as possible. This is particularly true for passengers whose daily routines involve travelling within set distances and at set times, such as commuter pass users.

Therefore, the new route estimation unit may estimate that the smaller the time difference between the entry time and departure time or the time difference between the exit time and arrival time in a route candidate, the higher the probability that the route candidate will be used as a new route.
According to the above configuration, the new route estimation unit estimates a new route that is suited to such passenger tendencies, thereby improving the accuracy of estimating the new route.

The new route estimation unit may calculate an allocation rate for each candidate route based on the ticket gate passage data. The new route estimation unit may allocate passengers to the candidate routes according to the allocation rate.
The "allocation rate" is a quantitative parameter corresponding to the probability that a passenger will select that route candidate, and the new route estimation unit performs an allocation process in which one passenger is divided into a quantity less than one according to the allocation rate. The quantity of passengers can also be expressed as a non-integer. Even if there is one passenger before allocation, it is permissible to allocate a quantity less than one passenger to multiple route candidates.

According to the above configuration, the new route estimation unit calculates the allocation rate based on the ticket gate passage data, so that passengers can be allocated to one or more route candidates in accordance with the passengers' tendencies. As a result, the accuracy of estimating new routes and the accuracy of predicting passenger flow are improved.
The new route estimation unit may calculate the allocation rate based on a boarding difference value associated with the time difference between the entry time and the departure time, and a disembarking difference value associated with the time difference between the exit time and the arrival time.

According to the above configuration, the new route estimation unit can calculate the allocation rate according to the time difference between the routes before and after the train schedule revision. Since the allocation process is performed in accordance with passenger tendencies, the accuracy of the new route estimation is improved.
The new route estimation unit may calculate the allocation rate such that the smaller the boarding differential value or the alighting differential value, the larger the allocation rate.

According to the above configuration, the allocation process is performed in accordance with the passenger's tendencies, thereby improving the accuracy of estimating the new route.
The route time difference is the sum of the boarding difference value and the alighting difference value. The new route estimation unit may calculate the allocation rate such that the reciprocal ratio of the route time differences between the route candidates is the ratio of the allocation rates between the route candidates.

According to the above configuration, the allocation process is performed in accordance with the passenger's tendencies, thereby improving the accuracy of estimating the new route.
When the bill type is the first bill type that has a restriction on arrival time, the new route estimation unit may calculate the allocation rate so that the allocation rate for early arrival candidates is higher than the allocation rate for late arrival candidates.
According to the above configuration, the allocation process is performed in accordance with the commuter pass user's way of thinking about arrival times, thereby improving the accuracy of estimating the new route.

When the allocation rate is calculated based on a boarding difference value associated with the time difference between the entry time and the departure time and a disembarking difference value associated with the time difference between the exit time and the arrival time, the new route estimation unit may correct the disembarking difference value so that the disembarking difference value for late arrival candidates is larger than the disembarking difference value for early arrival candidates.
According to the above configuration, the allocation rate is smaller for route candidates with a large delay in arrival time compared to the departure time. Since the allocation process is performed in accordance with the commuter pass user's perception of arrival time, the estimation accuracy of the new route is improved.

A train schedule simulation method according to one embodiment of the present invention comprises acquiring ticket gate passage data, including passenger entry records, exit records, and ticket types used by the passengers, collected from automatic ticket gates during operation of the current train schedule, and new schedule data indicating a new train schedule; and referencing the new schedule data and estimating a new route that the passenger will use in the new train schedule based on the passenger entry records, exit records, and ticket types.

A train diagram simulation program according to an embodiment of the present invention causes a computer to execute the above method.
The above method and program have the same or corresponding technical features as the above device, and therefore, differences in passenger habits depending on ticket type can be reflected in the passenger flow prediction results.

The present invention makes it possible to improve the accuracy of passenger flow predictions when estimating passenger routes in new train timetables.

1 is a block diagram showing the configuration of a train diagram simulation device according to an embodiment of the present invention and a system including the same. FIG. 10 is a schematic diagram of new timetable data. FIG. 10 is a schematic diagram of ticket gate passage data. 1 is a flowchart showing a train diagram simulation method according to an embodiment of the present invention. 5 is a flowchart showing a part of a new route estimation process in the train diagram simulation method of FIG. 4 . 5 is a flowchart showing a part of the new route estimation process of FIG. 4 . FIG. 6 is an explanatory diagram of the new route estimation process (route candidate extraction process) of FIG. 5 . FIG. 6 is an explanatory diagram of the new route estimation process (route candidate extraction process) of FIG. 5 . FIG. 6 is an explanatory diagram of the new route estimation process (allocation process) of FIG. 5 .

Hereinafter, an embodiment of the present invention will be described with reference to the drawings. Note that the same or corresponding elements throughout the drawings are designated by the same reference numerals, and redundant explanations will be omitted.
(system)
FIG. 1 shows a system 1 including a train diagram simulation device 10 according to an embodiment of the present invention. The train diagram simulation device 10 and the system 1 including the same are suitable for use in railway operators' train diagram revision operations. Railway operators periodically revise train diagrams to optimize passenger flow. Optimizing passenger flow includes reducing congestion on trains or in station premises. In the timetable revision operations, a new train diagram is created to replace the current train diagram.

The train diagram simulation device 10 predicts passenger flow when a new train diagram is applied. This makes it possible to evaluate the merits of the new train diagram before it is actually applied. Therefore, it is possible to modify the new train diagram before the revision is actually implemented, for example, to further optimize passenger flow in specific time periods or sections.
The train diagram simulation device 10 is realized, for example, in a server managed by a railway operator or an information terminal used by an employee of the railway operator who is involved in timetable revision work. The server or information terminal is a computer equipped with a CPU, memory, and an input/output interface.

The train diagram simulation program of the present invention is installed on the computer and stored in its memory. The CPU reads the program from the memory and processes information according to the steps instructed by the program. This allows the computer to execute the train diagram simulation method of the present invention and function as the train diagram simulation device 10.

The system 1 includes a train diagram simulation device 10 as well as an input device 2, a display device 3, and a data server 4 connected to the train diagram simulation device 10.
The input device 2 receives input from an operator of data to be stored in the train diagram simulation device 10 (e.g., new train diagram data D2) and a command to start a simulation. The display device 3 displays the predicted results of passenger flow obtained by the train diagram simulation device 10.

Data server 4 stores ticket gate passage data D3. Ticket gate passage data D3 is generated based on entry and exit records collected from automatic ticket gates 5 installed at each railway station.

(New timetable data)
The new train schedule data D2 is new train schedule data corresponding to a new train schedule, and is the subject of simulation.
The train schedule data is information showing train schedules, i.e., train operation plans. In general rail passenger transport, the daytime is "operation time" when passenger transport occurs, and the period from late at night to early morning is "track maintenance time" when passenger transport is suspended and track maintenance work is performed, with operation time and track maintenance time repeating on a daily cycle. The train schedule data is made up of operation data D4 (see Figure 2) for each train that operates within one operation time period.

Train schedules are created by direction, and train schedules for each direction are further created by day. The new schedule data D2 includes multiple types of data according to direction and day. In this document, the term "new schedule data" may be used to describe multiple types of data without distinguishing between them.
Examples of the types of timetables by direction include an inbound timetable and an outbound timetable. Examples of the types of timetables by day include a weekday timetable that applies to weekday operating hours and a holiday timetable that applies to holiday operating hours. Further examples include a seasonal timetable that applies only on weekends for a limited period, such as in autumn, and in which special trains are operated to transport vacationers, and an event timetable that applies only on days when events that are expected to attract large numbers of people are held, such as soccer games or horse races, and in which special trains are operated to transport event guests.

For ease of explanation, Figure 2 shows a schematic representation of new timetable data D2 for Line P, which is hypothetical. There are eight stations, A to H, on Line P. Four-digit numbers indicate the hour and minute, downward arrows indicate passing stations, and blank spaces indicate non-operating sections. In the illustrated example, weekday operating hours in the current train timetable are set from 5:30 to 24:14.
The operation data D4 includes corresponding train type data, formation data, stop station data, and time data. The type data indicates the type of train (for example, limited express, express, local, etc., and in the case of limited express, whether it is a paid or free train). The formation data includes information related to passenger capacity, such as the number of cars that make up the train. The stop station data includes information indicating the train's starting station, intermediate stops, and terminal station. The time data includes information indicating the departure time from the starting station, the arrival time and departure time at intermediate stops, and the arrival time at the terminal station.

(Ticket gate passage data)
As shown in Fig. 3, the ticket gate passage data D3 is a collection of multiple trip data D5. Each trip data D5 is a set of data including an entry record and a corresponding exit record. The entry record includes the entry station and the time when the train passed through the automatic ticket gate 5 installed at the entry station. The exit record includes the exit station and the time when the train passed through the automatic ticket gate 5 installed at the exit station.

Entry records and exit records are acquired by the automatic ticket gate 5 at different locations and different times. When acquired by the automatic ticket gate 5, the entry records and exit records include ticket ID information assigned to each ticket used to pass through the automatic ticket gate 5 (enter or exit). The collected entry records and exit records are stored on the data server 4 in a state where they are linked to each other via the ticket ID information to form trip data D5.

The trip data D5 indicates what type of ticket a certain passenger used, when he entered which station, and when he exited from which station, and specifies one trip using a railway from entry to exit. One trip corresponds one-to-one with one "passenger" as the subject of the trip.
3 shows an example of trip data showing travel from Station A to Station D in the early morning, and another example of trip data showing travel from Station D to Station A at night. These two trip data contain the same ticket ID information, and therefore are recognized as showing round-trip travel by the same person using a commuter pass. These two trips are defined as different trips by different "passengers" because the direction and time of travel are different.

Each trip data D5 includes ticket type information indicating the type of ticket used by the passenger to complete the journey from the entry station to the exit station. Figure 3 shows that ticket type information is recorded on data server 4 as part of the data group that makes up trip data D5, independently of ticket ID information, but this is just one example. If ticket ID information includes information indicating the ticket type, data indicating the ticket type does not necessarily need to be recorded separately and independently on data server 4. Conversely, ticket ID information may be treated as ticket type information.

(Ticket type and passenger characteristics)
Ticket types include, for example, regular tickets, round-trip tickets, commuter passes (such as commuter passes and student passes), multi-ride tickets (such as regular multi-ride tickets, time-difference multi-ride tickets, and holiday multi-ride tickets), and special tickets. They may be magnetic tickets or IC cards. Furthermore, ticket types may include tickets other than regular tickets that are used to pass through ticket gates (such as admission tickets, free tickets issued for free return trips, and fare-adjusted tickets issued when the missing fare has been settled), as long as they are used by passengers and information about them can be obtained by the automatic ticket gate 5.

In this embodiment, each denomination is classified into a plurality of denomination groups (for example, a first denomination, a second denomination, and a third denomination).
The first ticket type is a group of ticket types that have restrictions on arrival or departure times at the station of departure. For example, commuter passes are classified as the first ticket type. For passengers who use commuter passes, travel within a set section and time period is incorporated into their daily lives. This travel involves business (e.g., work, classes, etc.) at the destination, and the start time of that business is considered to remain unchanged before and after a train schedule change. Therefore, even if the train schedule change occurs, passengers who use commuter passes are considered to not want or be allowed to delay their arrival or departure times at the station of departure.

The second ticket type is a group of ticket types that do not belong to the first ticket type. For example, regular tickets, round-trip tickets, multi-ride tickets, and special tickets are classified as the second ticket type. Passengers using the second ticket type are considered to be more tolerant of delays in arrival and departure times than passengers using the first ticket type.
A common feature of both the first and second ticket types is that passengers are thought to dislike large changes in the time difference between the route they are currently using and the route in the new train schedule that they will use instead. Such an increase in the time difference between routes would entail changes in passenger lifestyles. Passengers using the first ticket type are thought to be particularly susceptible to this tendency. The "time difference between routes" refers to the time difference between entry times, the time difference between exit times, the time difference between departure times at entry stations, the time difference between arrival times at exit stations, the time difference between entry times and departure times, the time difference between exit times and arrival times, or a combination of these.
The third ticket type is a group of special ticket types other than passenger tickets. For example, admission tickets, free tickets, and fare-adjusted tickets are classified as the third ticket type.

(Train diagram simulation device/method)
The train diagram simulation device 10 according to this embodiment estimates new routes D7 (see FIGS. 7 and 8) that will be used by each passenger in the new train diagram, taking into account passenger tendencies that differ depending on the ticket type as described above. This improves the accuracy of passenger flow predictions when the new train diagram is applied.

The new route D7 includes information indicating the entrance station, the exit station, and one or more trains used to travel from the entrance station to the exit station. Since the new route D7 includes information indicating the trains, the new route D7 also includes information indicating the departure time of the train to be boarded at the entrance station (i.e., the boarding time) and the arrival time of the train to be disembarked at the exit station (i.e., the disembarking time). If there is one or more transfers and multiple trains are used, the new route D7 also includes information indicating each transfer station, the arrival time at each transfer station, the departure time at each transfer station, and the waiting time at each transfer station.
As shown in FIG. 1 , the train diagram simulation device 10 includes a storage unit 11 , a data acquisition unit 12 , a new route estimation unit 15 , a passenger number estimation unit 16 , and an output unit 17 .

The memory unit 11 stores various information necessary for predicting passenger flow. In this embodiment, the memory unit 11 is configured to store, as a mere example, new train schedule data D2 input by an operator via the input device 2. However, part of the ticket gate passage data D3 may be pre-stored in the memory unit 11, and part or all of the new train schedule data D2 may be saved in a storage device external to the train schedule simulation device 10.
The operations of the data acquisition unit 12, new route estimation unit 15, passenger number estimation unit 16, and output unit 17 will be described below in accordance with the procedure of the train schedule simulation method according to the present invention shown in Figures 4 to 6. For example, when the operator inputs new schedule data D2 and a simulation start command, execution of the method begins.

<Data acquisition>
First, the data acquisition unit 12 acquires new timetable data D2 and ticket gate passage data D3 (data acquisition process S1). Next, the new route estimation unit 15 references the acquired new timetable data D2 and estimates a new route D7 that will be used by the passenger corresponding to the trip data D5 based on the trip data D5 (entry record, exit record, and ticket type) (new route estimation process S2).

The new route estimation process S2 is performed for each timetable type in the new timetable data D2. Therefore, in the data acquisition process S1, the data acquisition unit 12 acquires one day's worth of ticket gate passage data D3 corresponding to each timetable type. For example, to estimate a new route D7 in weekday timetable data, ticket gate passage data D3 for one weekday is acquired, and to estimate a new route D7 in holiday timetable data, ticket gate passage data D3 for one holiday is acquired.

<New route estimation>
<<Selection of processing target>>
The new route estimation process S2 will be described below with reference to Figures 5 to 9. As shown in Figure 5, the new route estimation unit 15 performs estimation processes for all of the multiple timetable types while sequentially changing the timetable type to be estimated, for example, after completing estimation processes for inbound weekday timetable data, it then starts estimation processes for inbound holiday weekday timetable data. Therefore, as the first process of the new route estimation process S2, the new route estimation unit 15 selects one timetable type to be estimated (S11).
The new route estimation unit 15 refers to one day's worth of ticket gate passage data D3 corresponding to the timetable data selected as the target for estimation processing from the new timetable data D2 (S12), and extracts one piece of trip data D5 from the one day's worth of ticket gate passage data D3 (S13).

<<Route candidate extraction>>
The new route estimation unit 15 extracts route candidates for the new route D7 that can be used by the passenger corresponding to the extracted trip data D5, based on the entry record and exit record of the extracted trip data D5 (S14).

7 and 8 conceptually illustrate the process of extracting route candidates. Fig. 7 shows the process of extracting route candidates in the form of a diagram with the vertical axis representing time, while Fig. 8 shows the extracted multiple route candidates lined up vertically with the horizontal axis representing time.
First, the new route estimation unit 15 sets the entrance and exit stations of the new route D7 to be the same as the entrance and exit stations in the trip data D5. The new route estimation unit 15 searches the new timetable data D2 for routes that satisfy the conditions that the departure time Tdep at the entrance station is near the entrance time Tenter in the trip data D5 and the arrival time Tarr at the exit station is near the exit time Texit in the trip data D5. The new route estimation unit 15 extracts routes that satisfy the conditions as route candidates.

In this embodiment, "departure time Tdep is near entry time Tenter" means that the departure time Tdep is later than entry time Tenter and falls within a range of a predetermined number a, starting from the earliest, or that the departure time Tdep is earlier than entry time Tenter and falls within a range of a predetermined number b, starting from the latest. The same applies to arrival time Tarr and exit time TeTit.

The new route estimation unit 15 extracts candidates for "departing trains" used at the entrance station and candidates for "arriving trains" used at the exit station. In light of the meaning of "nearby" above, the candidates for departing trains are a predetermined number a of "late trains" and a predetermined number b of "early trains." The candidates for arriving trains are a predetermined number a of "early arriving trains" and a predetermined number b of "late arriving trains."
"Late trains" are a predetermined number a of trains that depart the entrance station after the entrance time Tenter, in order from the earliest departure time Tdep (see trains T52 to T55). "Early trains" are a predetermined number b of trains that depart the entrance station before the entrance time Tenter, in order from the latest departure time Tdep (see train T51). More late trains are extracted than early trains, as it is believed that passengers would prefer an earlier departure than a later departure. As a mere example, the predetermined number a is 4 and the predetermined number b is 1 (see Figure 7).

"Early arriving trains" are a predetermined number a of trains that arrive at the exit station before the departure time Texit, in order from the latest arrival time Tarr (see trains T51, T53, T54, and T56). "Late arriving trains" are a predetermined number b of trains that arrive at the exit station after the departure time Texit, in order from the earliest arrival time Tarr (see train T55). More early arriving trains are extracted than late arriving trains. This is because it is believed that passengers prefer delayed arrival to earlier arrival. As a mere example, the predetermined number a is 4 and the predetermined number b is 1 (see Figure 7).

The new route estimation unit 15 extracts a route that satisfies both the above-mentioned conditions for departing trains and conditions for arriving trains from the new timetable data D2.
In addition, for route candidates without transfers, the departing train is the same as the arriving train. Also, if an express train overtakes a preceding train between the entry station and the exit station, it may be a candidate for the departing train but not for the arriving train (see train T52), and vice versa (see train T56).

If such train T52 is used as the departing train, train T52 is not included in the candidate arriving trains, and passengers will be forced to transfer to a candidate arriving train between the entrance and exit stations. The additional waiting time at the transfer station unnecessarily lengthens travel time (see the arrow lengths from the top two nodes at the transfer station in Figure 7 and the dashed-dotted box in Figure 8). Such routes may be excluded from the candidate routes because they are considered unlikely to be selected.

<<Route time difference calculation>>
As shown in Fig. 6, the new route estimation unit 15 determines the bill type of the trip data D5 (S15). If the bill type is the first bill type (S15: Y), the route time difference W for each extracted route candidate is calculated in step S16. Even if the bill type is other than the first bill type (S15: N), the route time difference W for each extracted route candidate is calculated in step S17. In either case, the route time difference W is calculated, but the calculation method is different.

The route time difference W is a parameter that quantitatively indicates the magnitude of the time difference between the route candidate and the entry record and the exit record. As described above, it is considered that passengers, regardless of the ticket type, do not prefer route candidates with a large route time difference W. In light of this passenger tendency, it can be said that the route time difference W is a parameter that is negatively correlated with the probability that the passenger will select the route candidate as new route D7.
As an example, the path time difference W is derived from the following equation (1).

W=│Tdep－Tenter│+│Texit－Tarr│×K...(1)
where K is a weighting factor.
The route time difference W is the sum of the boarding differential value associated with the time difference between the entry time Tenter and the departure time Tdep of each route candidate, and the disembarking differential value associated with the time difference between the exit time Texit and the arrival time Tarr of each route candidate.

The ride difference value corresponds to the first term on the right side of equation (1) and is the absolute value of the time difference between the entry time Tenter and the departure time Tdep. The larger the time difference, the larger the ride difference value, which increases the route time difference W. By using the absolute value, whether the departure time Tdep is earlier or later, the route time difference W increases as this time difference increases.
The drop-off difference value corresponds to the second term on the right side of equation (1) and is the absolute value of the time difference between the exit time Texit and the arrival time Tarr of each route candidate multiplied by a weighting coefficient K. The larger the time difference, the larger the drop-off difference value, which increases the route time difference W. By using the absolute value, whether the arrival time Tarr is earlier or later, the route time difference W increases as this time difference increases.

In this embodiment, the weighting coefficient K is selected from three types: a first early arrival weighting coefficient K1a, a first late arrival weighting coefficient K1b, and a second weighting coefficient K2, depending on the situation. If the value of the weighting coefficient K changes, the alighting difference value and therefore the route time difference W will be different even for the same time difference. Therefore, the weighting coefficient K serves as a parameter that represents the degree of influence that the magnitude of the time difference at alighting has on the route selection probability, or a parameter that represents the sensitivity of the time difference at alighting to the route selection probability. The larger the weighting coefficient K, the larger the alighting time difference, which increases the route time difference W, and the train diagram simulation device 10 estimates a lower route selection probability.

Here, among the route candidates, those whose arrival time Tarr is earlier than the departure time Texit are referred to as "early arrival candidates," and those whose arrival time Tarr is later than the departure time Texit are referred to as "late arrival candidates." In other words, early arrival candidates are route candidates in which an early arriving train is used as the arriving train. Late arrival candidates are route candidates in which a late departing train is used as the arriving train. It does not matter whether the departing train is an early or late train.

The first early arrival weighting coefficient K1a is used when the note type is the first note type and the route candidate is an early arrival candidate (S15: Y → S16, upper part). The first late arrival weighting coefficient K1b is used when the note type is the first note type and the route candidate is a late arrival candidate (S15: Y → S16, lower part). The second weighting coefficient K2 is used when the note type is the first note type (S15: N → S17). If the note type is the first note type, the same coefficient is used regardless of whether the route candidate is an early arrival candidate or a late arrival candidate.

The second weighting coefficient K2 is a value close to 1 (for example, 1). In situations where the ticket type is other than the first ticket type, if the time difference at the time of disembarking and the time difference at the time of boarding are the same value, the disembarking differential value will also be equivalent to the boarding differential value. In other words, when estimating the route used by a passenger based on the route time difference W, the time difference at the time of disembarking is evaluated equally to the time difference at the time of boarding.

The first late arrival weighting coefficient K1b is the largest of the three exemplified values and is greater than 1. In a situation where the ticket type is the first and the arrival time Tarr is later than the exit time Texit, even if the time difference at the time of disembarking and the time difference at the time of boarding are the same, the disembarking differential value will be greater than the boarding differential value. The greater the time difference at the time of disembarking, the greater the route time difference W. In other words, when estimating the route used by a passenger based on the route time difference W, the time difference at the time of disembarking has a stronger influence than the time difference at the time of boarding.

The first early arrival weighting coefficient K1a is a smaller value than the first late arrival weighting coefficient K1b. The first early arrival weighting coefficient K1a may be equal to the second weighting coefficient K2, or may be an intermediate value between the first late arrival weighting coefficient K1b and the second weighting coefficient K2. In a situation where the ticket type is the first and the arrival time Tarr is earlier than the exit time Texit, the time difference at the time of disembarking is evaluated as equal to or slightly larger than the time difference at the time of boarding.

<<Allocation>>
Once the route time difference W for each route candidate is calculated according to the situation in the manner described above, the new route estimation unit 15 determines whether only one route candidate has been extracted, or whether there is a route candidate whose route time difference W is zero (S18).
If at least one of the trip data D5 is true (S18: Y), the new route estimation unit 15 estimates the route candidate as a single new route D7 (S19). The new route estimation unit 15 assigns one passenger corresponding to the trip data D5 as a passenger on the estimated new route (S20).

"Allocating passengers as passengers" means dividing passengers into the number of route candidates and counting the divided passengers as passengers of the trains that make up each route candidate to which they have been allocated. In this example, since there is only one new route D7, one passenger is not divided and is counted as one passenger of the train that makes up the new route D7.
If neither of these applies (S18: N), the new route estimation unit 15 calculates the ratio at which passengers are allocated to each candidate route (hereinafter referred to as the "allocation rate r") based on the calculated route time difference W (S21).

When there are N route candidates, N is a natural number equal to or greater than 2, and i is any natural number from 1 to N, the allocation rate ri of the i-th route candidate is derived from the following equation (11).
ri=(1/Wi)/Σ(1/Wi)...(11)
Here, Wi is the route time difference corresponding to the i-th route candidate, and Σ(1/Wi) is the sum of the reciprocals of the route time differences W1, W2, ... of each route candidate (1/W1 + 1/W2 + ... + 1/WN).

From equation (11), the sum of the allocation rates r1, r2, ... of each route candidate is 1 (i.e., 100%). The larger the route time difference W, the lower the route selection probability, and so the smaller the allocation rate r. In this embodiment, as shown in equation (11), as an example of such a negative correlation between the route time difference W and the allocation rate r (route selection probability), the ratio of the allocation rates ri between route candidates is equal to the reciprocal ratio of the route time difference Wi between the route candidates.

Once the allocation rate r for each route candidate has been calculated in this way, the new route estimation unit 15 allocates one passenger corresponding to the trip data D5 as a passenger to each route candidate according to the allocation rate r (S22). Here, one passenger is allocated to multiple route candidate locations, but the train diagram simulation device 10 allows the number of passengers to be treated as a non-integer number. In the allocation process, passengers in quantities less than one are counted as passengers on trains that make up each route candidate. The number of passengers on that train is estimated by adding up the allocated passenger numbers.

Figure 9 conceptually illustrates the allocation process. When N = 3, W1 = 2, W2 = 5, and W3 = 10, the ratio r1:r2:r3 = 1/2:1/5:1/10 = 5:2:1. The allocation rate r1 for the first route candidate D7(1) is 62.5%, the allocation rate r2 for the second route candidate D7(2) is 25%, and the allocation rate r3 for the third route candidate D7(3) is 12.5%. One passenger is divided into 0.625, 0.25, and 0.125 passengers. 0.625 passengers are allocated to the first route candidate D7(1), 0.25 passengers are allocated to the second route candidate D7(2), and 0.125 passengers are allocated to the third route candidate D7(3).

For route candidates with a zero route time difference W, an allocation rate r of 100% is set (S18: Y → S19), regardless of whether other route candidates exist. This also addresses the arithmetic problem that a zero value cannot be reciprocally calculated. In fact, if a route with no route time difference exists in the new train timetable, passengers are unlikely to have any motivation to choose a different route. By allocating all passengers to route candidates with a zero route time difference W, it is possible to avoid division by zero and reflect such passenger tendencies in the estimation results.

If a large number of route candidates are extracted, passengers will be dispersed across a large number of route candidates during the allocation process. This may result in the number of passengers on each train being averaged out, which could overly blunt the estimated number of passengers. Therefore, if there are no route candidates with a route time difference W of zero, an upper limit on the number of route candidates to which passengers can be allocated may be determined in advance. In this case, route candidates equivalent to the upper limit are re-extracted from the extracted route candidates, in order from the smallest route time difference W. Passengers may be allocated to the re-extracted route candidates.

<<Repeated Processing>>
This completes the process of estimating a new route D7 for one passenger from one trip data. The above process is repeated until processing for all trip data D5 is complete (S23: N → S13). When processing for all trip data D5 is complete (S23: Y), the same process is repeated by selecting each schedule type to be processed one by one until processing for all schedule types is complete (S24: N → S11). When processing for all schedule types is complete (S24: Y), the process returns to the main routine (see FIG. 4).

<Estimated number of passengers>
Next, the passenger number estimation unit 16 calculates the number of passengers for each running section of each train based on the number of passengers allocated by the new route estimation unit 15, for all of the operation data D4 constituting the new timetable data D2 (S3). A "running section" is a section between two adjacent stations among the starting station, intermediate stops, and terminal station of the train. In the example of Figure 2, seven running sections are set for the local train T03, and two running sections are set for the express train T04.

The number of passengers on a certain section of a train is the sum of the number of passengers using the route that includes that section. For example, if new route D7 includes train T01, which departs from station A and terminates at station D, there are seven possible new routes D7: a route that travels between A and B, a route that travels between A and C, a route that travels between A and D, a route that travels between B and C, a route that travels between B and D, and a route that travels between C and D. The number of passengers on section A-B is the sum of the number of passengers on the three new routes D7 between A-B, A-C, and A-D. The number of passengers on section B-C is the sum of the number of passengers on the four new routes D7 between A-C, A-D, B-C, and B-D. The number of passengers on section C-D is the sum of the number of passengers on the three new routes D7 between A-D, B-D, and C-D.

In this way, the number of passengers on each train for each section of travel can be calculated using the results of the new route estimation process S2 described above.
Next, the passenger number estimation unit 16 refers to the operation data D4 of each train and estimates the in-car congestion rate for each running section by dividing the number of passengers for each running section by the capacity of the train (S4).
Next, the output unit 17 outputs the evaluation result of the new train timetable to the display device 3 (S5). By referring to the evaluation result, the operator can easily determine whether the simulation target has achieved a given objective (such as alleviating congestion or improving transportation efficiency).

(Action and effect)
According to the train schedule simulation device 10 of this embodiment described above, the new route estimation unit 15 refers to the new schedule data D2 and estimates the new route D7 that will be used by the passenger in the new train schedule based on the passenger's entry record, exit record, and ticket type.

Ticket type is considered to be a factor related to passenger route tendencies, such as differences in the severity of how passengers perceive the exit time Texit or arrival time Tarr. In this regard, in this embodiment, when estimating new route D7, not only entry and exit records but also ticket type are taken into consideration. This allows the estimation results to reflect differences in passenger tendencies depending on ticket type. This improves the accuracy of passenger flow predictions.

The new route estimation unit 15 refers to the new timetable data D2 and extracts one or more routes in which the departure time Tdep at the entry station is near the entry time Tenter included in the entry record and the arrival time Tarr at the exit station is near the exit time Texit included in the exit record as route candidates for the new route D7. This improves the accuracy of passenger flow prediction compared to simply speculating on one route as the new route.
The new route estimation unit 15 calculates an allocation rate r for each candidate route based on the ticket gate passage data D3, a boarding differential value associated with the time difference between the entry time Tenter and the departure time Tdep, and a disembarking differential value associated with the time difference between the exit time Texit and the arrival time Tarr. In particular, the allocation rate r is calculated based on the route time difference W, which is the sum of the boarding differential value and the disembarking differential value. The new route estimation unit 15 allocates passengers to candidate routes according to the allocation rate r. This allows a passenger to be allocated to one or more candidate routes in accordance with the passenger's tendency that the passenger's route selection probability is correlated with the route time difference W.

The new route estimation unit 15 calculates the allocation rate r so that the smaller the boarding differential value or disembarking differential value, the larger the allocation rate r. In other words, the new route estimation unit 15 estimates that the smaller the time difference between the entry time Tenter and the departure time Tdep or the time difference between the exit time Texit and the arrival time Tarr in a route candidate, the higher the probability that the route candidate will be used as new route D7. This makes it possible to estimate new route D7 in accordance with passenger tendencies, where the passenger route selection probability is negatively correlated with the route time difference W.

The new route estimation unit 15 calculates the allocation rate r so that the reciprocal ratio of the route time difference W between the route candidates is the ratio of the allocation rates r between the route candidates. In this way, when there are multiple route candidates, the allocation rate r (route selection probability) of each route candidate can be easily derived so that it has a negative correlation with the route time difference W.
When the ticket type is a first ticket type (e.g., a commuter pass) that has a restriction on the arrival time Tarr, the new route estimation unit 15 calculates the allocation rate r so that the allocation rate r for the early arrival candidate is higher than the allocation rate r for the late arrival candidate. In other words, when the ticket type is the first ticket type, the new route estimation unit 15 estimates that the early arrival candidate is highly likely to be used as the new route D7. This makes it possible to estimate the new route D7 so as to suit the propensity of passengers using the first ticket type.

In this case, the new route estimation unit 15 corrects the drop-off difference value so that the drop-off difference value for the late arrival candidate is greater than the drop-off difference value for the early arrival candidate. In particular, in this correction process, the weighting coefficient K is multiplied by the absolute value of the time difference between the departure time Texit and the arrival time Tarr. The first late arrival weighting coefficient K1b used for the late arrival candidate is greater than the second early arrival weighting coefficient K1a used for the early arrival candidate. Therefore, even if the absolute values are the same, the drop-off difference value and route time difference W for the late arrival candidate will be greater than those for the early arrival candidate, and the allocation rate for the late arrival candidate will be smaller than that for the early arrival candidate. This makes it possible to estimate a new route D7 that is compatible with the way passengers using the first ticket type perceive arrival times.

(Modification)
The embodiments of the present invention have been described above, but the above configurations can be appropriately modified, added to, and/or deleted within the scope of the present invention.
(A)
Regarding extraction of route candidates, "the departure time Tdep is near the entry time Tenter" may mean that the departure time Tdep falls within a predetermined time range based on the entry time Tenter. The same applies to the arrival time Tarr and the exit time TeTit.

In this case, the new route estimation unit 15 sets a departure time search period and an arrival time search period. The departure time search period is set before or after the entry time Tenter, so that the entry time Tenter falls within the boarding time search period. As just one example, the departure time search period is set to 15 minutes before or after the entry time Tenter, for a total of 30 minutes. The same is true for the arrival time search period. The new route estimation unit 15 searches the new timetable data D2 for routes that depart from the entry station within the departure time search period and arrive at the exit station within the arrival time search period, and extracts the searched routes as route candidates.

(B)
To derive the path time difference W, the following equation (2) may be used instead of the above equation (1).
W=│Tdep−Tenter|+max(0, (Tarr−Texit))×K……(2)
In this calculation, the larger of two values, 0 and the subtraction value obtained by subtracting the exit time Texit from the arrival time Tarr, is selected. The exit difference value is obtained by correcting this larger value with a weighting coefficient K.

The later the arrival time Tarr is relative to the departure time Texit, the larger the drop-off difference value and the larger the route time difference W. If the arrival time Tarr is earlier than the departure time Texit, a time difference itself occurs, but the subtraction value becomes negative and the drop-off difference value becomes 0. Therefore, the route time difference W does not change. If the arrival time Tarr is earlier, the new route estimation unit 15 ignores the time difference between the arrival time Tarr and the departure time Texit when evaluating the magnitude of the route time difference.

It is also considered that passengers are more likely to accept the route candidate (early arrival candidate) if the arrival time is earlier, even if there is a time difference. By employing equation (2), it is possible to estimate new route D7 in light of such passenger tendencies.
Furthermore, in order to derive the path time difference W, the following equation (3) may be employed instead of the above equation (1) or (2).

W=│Tdep−Tenter│ ^m +│Texit−Tarr│ ⁿ ×K……(3)
Here, the boarding differential value and the alighting differential value are power values with the absolute value of the time difference as the base. The exponents m and n may be different, but are equal to or greater than 1. The larger the time difference, the greater the rate of increase in the route time difference W, and the more the allocation rate r decreases.
Either one of the exponents m and n may be 1, and at least one of the boarding differential value and the disembarking differential value may be substantially a power (either one of the boarding differential value and the disembarking differential value may be the absolute value of the time difference itself). Furthermore, formula (3) may be further modified, and the disembarking differential value may be a power value with the largest value in formula (2) as the base.

REFERENCE SIGNS LIST 1 System 2 Input device 3 Display device 4 Data server 5 Automatic ticket gate 10 Train timetable simulation device 11 Memory unit 12 Data acquisition unit 15 New route estimation unit 16 Number of passengers estimation unit 17 Output unit D2 New timetable data D3 Ticket gate passage data D7 New route r Allocation rate Tarr Arrival time Tdep Departure time Tenter Entry time Texit Exit time W Route time difference

Claims (5)

a data acquisition unit that acquires ticket gate passage data including passenger entry records, exit records, and ticket types used by passengers collected from automatic ticket gates during operation of the current train schedule, and new schedule data indicating the new train schedule;
a new route estimation unit that references the new train schedule data and estimates a new route to be used by the passenger in the new train schedule based on the entrance record, the exit record, and the ticket type of the passenger;
Equipped with
The new route estimation unit
With reference to the new timetable data, extract one or more routes whose departure time at the entrance station is near the entrance time included in the entrance record and whose arrival time at the exit station is near the departure time included in the exit record as route candidates for the new route;
estimating the new route from the route candidates based on the ticket gate passage data;
Among the route candidates, a route whose arrival time is earlier than the departure time is defined as an early arrival candidate, and a route whose arrival time is later than the departure time is defined as a late arrival candidate.
the new route estimation unit estimates that there is a high probability that the early arrival candidate will be used as the new route when the bill type is a first bill type that has a restriction on the arrival time;
Train schedule simulation device. The first ticket type includes a commuter pass.
The train diagram simulation device according to claim 1. the new route estimation unit estimates that the smaller the time difference between the entry time and the departure time or the time difference between the exit time and the arrival time in the route candidate, the higher the probability that the route candidate will be used as the new route;
3. The train diagram simulation device according to claim 1 or 2 . a data acquisition step executed by a computer to acquire ticket gate passage data including passenger entry records, exit records and ticket types used by passengers collected from automatic ticket gates during operation of the current train schedule, and new schedule data indicating the new train schedule;
a new route estimation step, executed by a computer, of estimating a new route that the passenger will use in the new train schedule based on the entry record, exit record, and ticket type of the passenger, with reference to the new schedule data;
Equipped with
The new route estimation step includes:
With reference to the new timetable data, extract one or more routes whose departure time at the entrance station is near the entrance time included in the entrance record and whose arrival time at the exit station is near the departure time included in the exit record as route candidates for the new route;
estimating the new route from the route candidates based on the ticket gate passage data;
Among the route candidates, a route whose arrival time is earlier than the departure time is defined as an early arrival candidate, and a route whose arrival time is later than the departure time is defined as a late arrival candidate.
the new route estimation step estimates that, when the bill type is a first bill type that has a restriction on the arrival time, there is a high probability that the early arrival candidate will be used as the new route;
Train schedule simulation method. 5. A train diagram simulation method according to claim 4,
Train schedule simulation program.