Delay prediction system, delay prediction program, and delay prediction method

Abstract

To provide a delay prediction system, a delay prediction program, and a delay prediction method capable of notifying a user of final arrival delay time at a destination when the delay occurs.SOLUTION: A delay prediction system includes a teacher data generation unit for generating teacher data in which information including at least information about the density of a mobile body in a predetermined section is an explanatory variable and an arrival delay time of the mobile body at a destination is an objective variable based on past result data indicating past operation results of a plurality of mobile bodies in the predetermined section, and a learning unit for generating a delay prediction model in which the arrival delay time of the mobile body at the destination is predicted by machine learning with the use of the teacher data.SELECTED DRAWING: Figure 1

Description

[Technical Field]

[0001] The present invention relates to a delay prediction system, a delay prediction program, and a delay prediction method. [Background technology]

[0002] Conventionally, there is known a system that distributes information such as delay time, departure and arrival times of a moving body to a user when a delay occurs in the moving body such as a train or bus. For example, Patent Document 1 discloses a system that, when a train delay occurs, sequentially updates information such as delay time, departure and arrival times of the train based on a planned timetable, which is a train operation plan, and a performance timetable, which indicates the current operation status, and displays the updated information on a user terminal carried by the user. [Prior art documents] [Patent documents]

[0003] [Patent Document 1] Japanese Patent Application Laid-Open No. 2002-274382 Summary of the Invention [Problem to be solved by the invention]

[0004] However, the delay prediction system described in Patent Document 1 uses the difference between the planned schedule and the actual schedule as the delay time, and so the delay time changes from moment to moment depending on the progress of restoration work, changes to the schedule, etc. In other words, the delay time described in Patent Document 1 is merely the delay time at the time of departure, not the delay time at the time of arrival. For this reason, the delay prediction system described in Patent Document 1 may have a large discrepancy between the delay time initially displayed and the final delay time at the time of arrival, which may increase user dissatisfaction.

[0005] The present invention relates to a delay prediction system, a delay prediction program, and a delay prediction method that can notify a user of the final arrival delay time at a destination when a delay occurs. [Means for solving the problem]

[0006] The delay prediction system of the present invention includes a training data generation unit that generates training data based on past performance data showing the past driving performance of multiple mobile bodies in a specified section, using information including at least information related to the density of mobile bodies in a specified section as an explanatory variable and the arrival delay time of the mobile body at the destination as a target variable, and a learning unit that generates a delay prediction model that predicts the arrival delay time of the mobile body at the destination through machine learning using the training data.

[0007] In addition, the delay prediction system of the present invention further includes a prediction unit that predicts the arrival delay time of a mobile body at a destination based on recent actual data indicating the recent driving performance of multiple mobile bodies in a specified section and the delay prediction model.

[0008] Furthermore, in the delay prediction system according to the present invention, the moving body is a train, and the information relating to the density of the moving body includes information relating to the train density from the current location to the destination.

[0009] In addition, in the delay prediction system of the present invention, the information related to the train density from the current location to the destination includes at least one of information regarding the number of stations between the current location and the destination, information regarding the number of trains between the current location and the destination, information regarding the total station stop time of trains between the current location and the destination, and information regarding the planned running time from the current location to the destination.

[0010] Furthermore, in the delay prediction system according to the present invention, the information relating to the density of moving objects includes information relating to the density of trains on the line.

[0011] In addition, in the delay prediction system of the present invention, the information related to the train density on the line section includes at least one of information regarding the number of trains on one track of the line section, information regarding the number of trains on the entire line section, and information regarding the total station stop time of trains on the entire line section.

[0012] Furthermore, in the delay prediction system according to the present invention, the information relating to the density of moving objects includes information relating to the train density at other stations within the line section.

[0013] In addition, in the delay prediction system of the present invention, the information related to train density at other stations within the line section includes at least one of information regarding the train arrival delay time of trains running on one track of the line section subject to delay prediction at a connecting station where multiple line sections are connected, and information regarding the train arrival delay time of trains running on the other track of the line section subject to delay prediction.

[0014] Furthermore, in the delay prediction system according to the present invention, the information relating to the density of moving bodies includes information relating to the train density between the nearest trains.

[0015] In addition, in the delay prediction system according to the present invention, the information relating to the train density between the nearest trains includes information relating to the number of stations between the train itself and the nearest train ahead.

[0016] Furthermore, in the delay prediction system according to the present invention, the information relating to the train density between the nearest trains includes information relating to the number of stations between the train itself and the nearest train behind.

[0017] In the delay prediction system according to the present invention, the explanatory variables include information relating to the occupancy rate of the vehicle.

[0018] Furthermore, in the delay prediction system according to the present invention, the explanatory variables include information regarding the plan, performance, and delay of the mobile object at the current location.

[0019] The delay prediction system according to the present invention further comprises a user terminal that displays the arrival delay time of the mobile body at the destination predicted by the prediction unit.

[0020] The delay prediction program of the present invention is configured to cause a computer to execute a training data generation process for generating training data based on past performance data showing the past driving performance of multiple mobile bodies in a specified section, with information including at least information related to the density of mobile bodies in the specified section as an explanatory variable and the arrival delay time of the mobile body at the destination as a target variable, and a learning process for generating a delay prediction model that predicts the arrival delay time of the mobile body at the destination through machine learning using the training data.

[0021] The delay prediction program of the present invention further includes a prediction step of predicting the arrival delay time of a mobile body at a destination based on the most recent actual data showing the most recent driving performance of multiple mobile bodies in a specified section and the delay prediction model.

[0022] The delay prediction method of the present invention includes a training data generation process for generating training data based on past performance data showing the past driving performance of multiple mobile bodies in a specified section, in which information including at least information related to the density of mobile bodies in a specified section is used as an explanatory variable and the arrival delay time of the mobile body at the destination is used as a target variable, and a learning process for generating a delay prediction model that predicts the arrival delay time of the mobile body at the destination by machine learning using the training data.

[0023] The delay prediction method of the present invention further includes a prediction step of predicting the arrival delay time of a mobile body at a destination based on the most recent actual data showing the most recent driving performance of multiple mobile bodies in a specified section and the delay prediction model. [Effects of the Invention]

[0024] According to the delay prediction system, delay prediction program, and delay prediction method of the present invention, when a delay occurs, the user can be notified of the final arrival delay time at the destination. [Brief explanation of the drawings]

[0025] [Figure 1] 1 is a diagram illustrating an overview of a delay prediction system according to an embodiment of the present invention. [Figure 2] FIG. 2 is a diagram illustrating an example of a database stored in a learning server according to the present embodiment. [Figure 3] FIG. 2 is a diagram showing first information in a database stored in a learning server according to the present embodiment. [Figure 4] FIG. 10 is a diagram showing second information in a database stored in a learning server according to the present embodiment. [Figure 5] FIG. 10 is a diagram showing third information in the database stored in the learning server according to the present embodiment. [Figure 6] FIG. 10 is a diagram showing fourth information in the database stored in the learning server according to the present embodiment. [Figure 7] FIG. 10 is a diagram showing fifth information in the database stored in the learning server according to the present embodiment. [Figure 8] 1 is a flowchart showing the flow of a learning method according to the present embodiment. [Figure 9] 1 is a flowchart showing the flow of a delay prediction method according to the present embodiment. [Figure 10] FIG. 10 is a diagram showing an example of a prediction result displayed on a user terminal according to the present embodiment. DETAILED DESCRIPTION OF THE INVENTION

[0026] Preferred embodiments for carrying out the present invention will be described below with reference to the drawings. Note that the following embodiments do not limit the inventions according to the claims, and not all of the combinations of features described in the embodiments are necessarily essential to the solution of the invention. Also, some components may be omitted in the present embodiments.

[0027] [Configuration of delay prediction system] The delay prediction system 1 of this embodiment is, in outline, a system that uses past performance data (hereinafter simply referred to as past performance data of trains 100) showing the past operating performance of multiple trains 100 in a specified section to generate a delay prediction model that predicts the arrival delay time of train 100 at the destination, and predicts the arrival delay time of train 100 at the destination based on the most recent performance data (hereinafter simply referred to as most recent performance data of trains 100) showing the most recent operating performance of multiple trains 100 in a specified section and the delay prediction model.

[0028] As shown in Figure 1, the delay prediction system 1 according to this embodiment includes a storage server 200 and a learning server 300 managed by a railway company or an affiliated company, and a user terminal 400 carried by a user who uses a train 100. The storage server 200, the learning server 300, and the user terminal 400 are configured to be able to communicate data with each other via a communication network NW. In this embodiment, the storage server 200 and the learning server 300 are described as being independent servers, but this is not limited to this. The storage server 200 and the learning server 300 may be included in a single server, or there may be multiple storage servers 200 and multiple learning servers 300.

[0029] Furthermore, the delay prediction system 1 according to this embodiment is not limited to the train 100, but can also be applied to any moving object that operates according to an operation plan, such as a bus, a ship, or an airplane. In this embodiment, data indicating the operation performance of the train 100 (hereinafter simply referred to as operation performance data of the train 100) is transmitted in real time from the train 100 to the storage server 200 of the delay prediction system 1 via wireless communication. Note that "transmitting in real time" includes delays of several seconds to several minutes, and retransmission when data cannot be transmitted due to radio interference, etc.

[0030] [Terminology] In this embodiment, the "predetermined section" refers to a line section that is a specific section designated by a railway company or the like, and the "line section" refers to a section where the delay prediction system 1 performs learning and a section where arrival delay times can be predicted, and includes the entire route and a specific section of the route. Note that the "predetermined section" is not limited to a line section, and can be arbitrarily determined for each mobile unit, such as a specific section among multiple bus stops if the mobile unit is a bus, a specific section among multiple docks if the mobile unit is a ship, or a specific section among multiple airports if the mobile unit is an airplane.

[0031] In this embodiment, the "train 100's operational performance data" refers to data including the first to fourth information items described below and information related to the objective variables, and includes information related to all trains currently in operation and all stations within the line section, the actual timetable, the planned timetable (operation plan), and the actual timetable. The actual timetable refers to a timetable that is created when the operation plan is changed based on the planned timetable in the event of a delay or other reason for the train 100. The "train 100's past operational performance data" refers to the past operational performance data of the train 100 stored in the storage server 200, and the "train 100's most recent operational performance data" refers to the most recent data among the operational performance data of the train 100 stored in the storage server 200. The "previous station" refers to the station immediately preceding the current location when the current location is used as the base station, and the "succeeding station" refers to the station immediately following the current location when the current location is used as the base station. The "connecting station" refers to a station where multiple different line sections are connected.

[0032] [Storage server configuration] The storage server 200 is a server that accumulates the operation performance data of the train 100 transmitted in real time from the train 100 via wireless communication, and as shown in FIG. 1, includes a memory unit 210 that stores the operation performance data of the train 100, and an extraction unit 220 that extracts the most recent performance data of the train 100 from the operation performance data of the train 100 stored in the memory unit 210. The storage server 200 and the train 100 are configured to be able to communicate data via a communication network NW. The operation performance data of the train 100 (the past performance data of the train 100) stored in the memory unit 210 is transmitted to the learning server 300 as dump data at predetermined time intervals, for example, every 30 seconds.

[0033] The extraction unit 220 is configured to extract the most recent actual data of the train 100 from the actual operation data of the train 100 stored in the storage unit 210. Specifically, the extraction unit 220 is configured to extract the most recent actual data of the train 100 that is the target of delay prediction, based on input information such as line section information, current location and destination station information (e.g., station name), and input time information input by the user operating the user terminal 400. The most recent actual data of the train 100 extracted by the extraction unit 220 is transmitted to the learning server 300. The current location station information input by the user is information on a station selected by the user from among multiple stations within the line section, for example, information on any station within the line section, such as the station where the user boards the train 100 or a station near a location where the user searches for a delay prediction. The destination station information is input is information on a station from among multiple stations within the line section, for which the user wishes to search for the arrival delay time of the train 100, for example, information on any station within the line section, such as the station where the user disembarks the train 100.

[0034] [Learning Server Configuration] The learning server 300 is a server that performs machine learning using past performance data of the train 100 and predicts the arrival delay time of the train 100 at the destination. Specifically, as shown in Figure 1, the learning server 300 includes a control unit 310 that controls the entire learning server 300.

[0035] As shown in Figure 1, the control unit 310 includes a memory unit 311 that stores various data, a teacher data generation unit 312 that generates teacher data necessary for machine learning, a learning unit 313 that performs learning based on the teacher data, and a prediction unit 314 that predicts the arrival delay time of the train 100 at the destination.

[0036] Here, the training data is data that serves as learning material for the learning unit 313 to learn from. In this embodiment, the explanatory variables include information for identifying the date and target (hereinafter referred to as first information), information related to the plan, performance, and delay of the train 100 at the current location (hereinafter referred to as second information), information related to the performance and delay of the train 100 at the previous station (hereinafter referred to as third information), information related to the occupancy rate of the train 100 (hereinafter referred to as fourth information), and information related to the density of the train 100 in a specified section (hereinafter referred to as fifth information). The actual arrival delay time of the train 100 at the destination is included as a response variable. Note that in this embodiment, the explanatory variables have been described as including the first to fifth information, but are not limited thereto and may include at least the fifth information (information related to the density of the train 100 in a specified section).

[0037] The control unit 310 is a CPU, which is a processor, and the storage unit 311 is, for example, a memory such as RAM or ROM, or an HDD or SDD. The learning server 300 according to this embodiment is configured to load the delay prediction program stored in the storage unit 311 into RAM, and have the CPU interpret and execute it to realize the various functions of the teacher data generation unit 312, learning unit 313, and prediction unit 314.

[0038] Here, the delay prediction program is a program configured to cause a computer to execute the following steps: a training data generation process for generating training data based on past actual data of multiple trains 100 in a specified section, with the first information to the fifth information as explanatory variables and the arrival delay time of the train 100 at the destination as the objective variable; a learning process for generating a delay prediction model by machine learning using the training data to predict the arrival delay time of the train 100 at the destination; and a prediction process for predicting the arrival delay time of the train 100 at the destination based on the most recent actual data of multiple trains 100 in the specified section and the delay prediction model.

[0039] The storage unit 311 includes a program storage area for storing a delay prediction program, and a data storage area for storing various data required for executing the program.

[0040] The data storage area stores past performance data of train 100 transmitted from storage server 200, a database into which information regarding training data is input, training data generated by training data generation unit 312, a delay prediction model generated by learning unit 313, and the most recent performance data of train 100 transmitted from storage server 200.

[0041] 2 stored in the storage unit 311 is a database into which information related to the teacher data is input, and the database in a state in which data is input into each item of the explanatory variables and the objective variables by the teacher data generation unit 312 becomes the teacher data input to the learning unit 313. Also, as shown in FIG. 2, in the database stored in the storage unit 311, the items of the explanatory variables are classified into each item of first information to fifth information. Each item of the first information to fifth information will be described later.

[0042] The teacher data generation unit 312 is configured to generate teacher data for learning by the learning unit 313, based on the past performance data of the train 100 stored in the memory unit 311. Specifically, the teacher data generation unit 312 is configured to generate first information to fourth information and information related to the objective variable, based on the past performance data of the train 100 stored in the memory unit 311. The teacher data generation unit 312 is also configured to generate fifth information, based on the first information to fourth information. The teacher data generation unit 312 generates the first information to fifth information and information related to the objective variable for all stations and all trains present in the line section.

[0043] Furthermore, the generation of teacher data by the teacher data generation unit 312 is performed, for example, by daily processing, and the past performance data of the train 100 stored in the storage unit 311 is deleted at the same time as the generation of the teacher data. Furthermore, the first information to the fifth information and information related to the objective variable generated by the teacher data generation unit 312 are stored for each database number shown in FIG. 2 stored in the storage unit 311, and the database in this stored state becomes the teacher data to be input to the learning unit 313.

[0044] Next, the information included in the first information to the fifth information will be described with reference to Figures 3 to 7. For the sake of convenience, the following description will be directed to one train (the train itself 100A) among the trains 100 running within the railway section. The trains other than the train itself 100A will be described as other trains 100B.

[0045] [First information] The first information is information for identifying the date and target. As shown in FIG. 3, the first information is categorized and input into the following fields in the database stored in the storage unit 311: (1) implementation date, (2) weekends and holidays, (3) line section name, (4) direction name, (5) train number, (6) previous station name, (7) current location name, (8) station sequence number, (9) subsequent station name, (10) destination name, (11) terminal station name, (12) starting station name, and (13) departure time from the starting station. The names of multiple stations within the line section are input into the (7) current location name field. That is, the name of a specific station (e.g., Station A) among the multiple stations within the line section is input into one cell, Station B into another cell, Station C into yet another, and so on. The names of each station are input into each cell. The names of stations between the current location and the terminal station are input into the (10) destination name field. That is, if the name of any station (for example, Station A) within the line section is entered in (7) Current Place Name, the names of all stations within the line section after Station A are entered in (10) Destination Name in a brute force manner. For example, in the first column of the database shown in Figure 2, (7) Current Place Name is Station A, and (10) Destination Name is Station B within the line section after Station A, and in the second column of the database, (7) Current Place Name is Station A, and (10) Destination Name is Station C, the station after Station B.

[0046] (1) For the implementation date, the date on which the train 100A ran is entered. (2) For Saturdays, Sundays, and holidays, if the date entered for the implementation date in (1) falls on a Saturday, Sunday, or holiday, the information "1" is entered, and if not, the information "0" is entered. (3) For the line section name, information regarding the name of the line section on which the train 100A runs is entered. (4) For the direction name, information regarding either "up" or "down" is entered. (5) For the train number, the train number of the train 100A is entered. For (6) Previous station name, (7) Current location name, (9) Next station name, (10) Destination name, (11) Terminal station name, and (12) Starting station name, information regarding the names of the stations in question is entered. (8) The station sequence number is input with information regarding the number of stations from the starting station (first station) determined for each line and direction to the current location, and (13) the departure time from the starting station is input with the time when train 100A departs from the starting station.

[0047] [Second information] The second information is information regarding the plan, actual results, and delay of the train 100 at the current location, and as shown in Figure 4, the data is categorized and input into the following items in the database stored in the memory unit 311: (14) planned arrival time, (15) actual arrival time, (16) arrival delay time, (17) planned departure time, (18) actual departure time, (19) departure delay time, and (20) stop time.

[0048] (14) Planned arrival time is entered as the arrival time at the current location in the planned or actual schedule. (15) Actual arrival time is entered as the actual arrival time at the current location. (16) Arrival delay time is entered as the time obtained by subtracting (14) planned arrival time from (15) actual arrival time. If the subtracted value is "-", "0" is entered. (17) Planned departure time is entered as the departure time from the current location in the planned or actual schedule. (18) Actual departure time is entered as the actual departure time. (19) Departure delay time is entered as the time obtained by subtracting (17) planned departure time from (18) actual departure time. (20) Stop time is entered as the time obtained by subtracting (15) actual arrival time from (18) actual departure time.

[0049] [Third information] The third information is information regarding the actual performance and delay of train 100 at the previous station, and as shown in Figure 5, the data is categorized and input into the following items in the database stored in memory unit 311: (21) actual arrival time at previous station, (22) arrival delay time at previous station, (23) actual departure time at previous station, (24) departure delay time at previous station, and (25) elapsed time between previous running stations.

[0050] The actual arrival time at the previous station is input as (21) Actual arrival time at previous station. The time obtained by subtracting the arrival time at the previous station in the planned or actual timetable from (21) Actual arrival time at previous station is input as (22) Arrival delay time at previous station is input as (21) Actual arrival time at previous station. The actual departure time at the previous station is input as (23) Actual departure time at previous station. The time obtained by subtracting the departure time from the previous station in the planned or actual timetable from (24) Actual departure time at previous station is input as (23) Actual departure time at previous station. The time obtained by subtracting the actual departure time from the previous station in the planned or actual timetable from (25) Elapsed time between previous running stations is input as (15) Actual arrival time from the second information.

[0051] [Fourth information] The fourth information is information related to the occupancy rate of the train 100, and as shown in Fig. 6, is inputted by classifying data into the items (26) Number of passengers in ordinary cars and (27) Number of passengers in special cars in the database stored in the memory unit 311. Here, "ordinary cars" refer to cars on the train 100 that passengers can board for the regular fare, and "special cars" refer to cars on the train 100 that passengers can board by paying a fare higher than the regular fare, such as green cars and reserved seats.

[0052] (26) The number of passengers in ordinary cars of the train 100A is input as the number of passengers in ordinary cars of the train 100A. (27) The number of passengers in special cars of the train 100A is input as the number of passengers in special cars of the train 100A. If there are no special cars, "0" is input.

[0053] [5th ​​information] The fifth information is information related to the density of trains 100 in a specified section, and data is calculated based on the first to fourth information. As shown in Fig. 7, the fifth information is input into the database stored in the memory unit 311 after being classified into the following categories: (28) train density from the current location to the destination, (29) train density in the line section, (30) train density at other stations in the line section, and (31) train density between the nearest trains. Here, the "next nearest train" refers to another train 100B running before or after the train 100A among the multiple trains 100 running on the tracks in the line section.

[0054] (28) The train density item from the current location to the destination is classified into the following data items: “Number of stations between the current location and the destination,” “Number of trains between the current location and the destination,” “Total stop time of trains between the current location and the destination,” and “Planned travel time from the current location to the destination.” The fifth information is classified into each data item and input.

[0055] The "number of stations between the current location and the destination" is input with the number of stations between the current location and the destination calculated based on the information (7) current location name and (10) destination name in the first information of the train itself 100A. The "number of trains between the current location and the destination" is input with the number of trains between the current location and the destination calculated based on the information (7) current location name and (10) destination name in the first information of the train itself 100A and the information (7) current location name in the first information of the other train 100B. The "total station stop times of trains between the current location and the destination" is input with the total station stop times of trains between the current location and the destination calculated based on the information (7) current location name and (10) destination name in the first information of the train itself 100A and the information (20) stop time in the second information of the other train 100B. The "planned travel time from current location to destination" field is entered as the time obtained by subtracting the planned departure time from the current location from the planned arrival time at the destination based on the information (7) current location name and (10) destination name in the first information of the train 100A and the planned or actual timetable information.

[0056] (29) The train density items in the line section are classified into the data items of "number of trains on one track of the line section," "number of trains on the entire line section," and "total station stop times of trains on the entire line section," and the fifth information is classified into each item and input. Note that "on one track of the line section" refers to either the track on which the train itself 100A is running or the track on the opposite side of the track on which the train itself 100A is running.

[0057] The "number of trains on one track of the line section" is input with the number of trains on one track of the line section, calculated based on the (4) direction name information in the first information of the own train 100A and the other train 100B and the (18) actual departure time information in the second information. The "number of trains on the entire line section" is input with the number of trains on the entire line section, calculated based on the (18) actual departure time information in the second information of the own train 100A and the other train 100B. The "total station stop times of trains on the entire line section" is input with the total station stop times of trains on the entire line section, calculated based on the (20) stop time information in the second information of the own train 100A and the other train 100B.

[0058] (30) The train density items at other stations within the line section are classified into data items such as "train arrival delay time of trains traveling on one track of the line section subject to delay prediction at a connecting station" and "train arrival delay time of trains traveling on the other track of the line section subject to delay prediction at a connecting station," and the fifth information is input after being classified into each data item. Here, "one track" refers to the track on which the train itself 100A travels, and "the other track" refers to the track located on the opposite side of the track on which the train itself 100A travels. Note that the track located on the opposite side of the track on which the train itself 100A travels may be the "one track," or the track on which the train itself 100A travels may be the "other track." In addition, (30) the information related to train density at other stations within the line section may include at least one of information regarding the "train arrival delay time of trains running on one track of the line section subject to delay prediction at a connecting station" or information regarding the "train arrival delay time of trains running on the other track of the line section subject to delay prediction."

[0059] The "train arrival delay time of the train traveling on one track of the line section subject to delay prediction at the connecting station" and the "train arrival delay time of the train traveling on the other track of the line section subject to delay prediction at the connecting station" are input with the arrival delay time of the later train at the connecting station of the target train of the other trains 100B traveling on the one track and the other track. Specifically, based on the (4) direction name of the first information of the own train 100A and the other train 100B, the other trains 100B traveling on the one track and the other track are identified, respectively, and based on the (7) current location name of the first information of the other train 100B and the (18) actual departure time of the second information of the own train 100A and the other train 100B, the other trains 100B that departed from the connecting stations on the one track and the other track at or immediately before the time the own train 100A departed from the current location are further identified, respectively. Then, the train arrival delay time at the connecting station calculated based on the (16) arrival delay time of the second information of the identified other train 100B is input.

[0060] (31) The train density item between the nearest trains is classified into the following data items: “Number of stations between the train itself and the train immediately ahead” and “Number of stations between the train itself and the train immediately behind.” The fifth information is classified into each data item and input.

[0061] The number of stations between your train and the train immediately ahead, calculated based on (7) the current place name of the first information of your train 100A and the other train 100B running immediately ahead of your train 100A, is input as the "number of stations between your train and the train immediately ahead." The number of stations between your train and the train immediately behind, calculated based on (7) the current place name of the first information of your train 100A and the other train 100B running immediately behind your train 100A, is input as the "number of stations between your train and the train immediately behind."

[0062] The learning unit 313 is configured to generate a delay prediction model that predicts the arrival delay time of the train 100 at the destination by machine learning using a general tool, such as AutoML (Automated Machine Learning), using the training data stored in the storage unit 311. Note that learning by the learning unit 313 may be performed when a certain amount of training data has been generated, may be performed daily, or may be performed simultaneously with the generation of the training data.

[0063] The prediction unit 314 is configured to predict the arrival delay time of the train 100 at the destination based on the latest actual data of the train 100 stored in the memory unit 311 and a delay prediction model. Specifically, the prediction unit 314 is configured to generate first information to fifth information based on the latest actual data of the train 100 stored in the memory unit 311, and input the first information to fifth information into the delay prediction model as explanatory variables, thereby calculating the arrival delay time of the train 100 at the destination as a response variable.

[0064] [User device configuration] 10, the user terminal 400 is a mobile device such as a smartphone or tablet terminal carried by a user who uses the train 100, and is equipped with an operation unit 410 that accepts various input operations from the user, a display unit 420 that displays a search screen for searching for the arrival delay time of the train 100 at the destination and the arrival delay time of the train 100 at the destination predicted by the prediction unit 314 (prediction result), and a control unit 430 within the user terminal 400. Note that the operation unit 410 and the display unit 420 can be configured using functions that are normally provided on mobile devices such as smartphones, and therefore detailed description thereof will be omitted.

[0065] As shown in FIG. 1, the control unit 430 includes a memory unit 431 that stores programs including instructions for operating the operation unit 410 and the display unit 420, programs (applications) for searching arrival delay times, etc., an operation control unit 432 that controls the operation unit 410, and a display control unit 433 that controls the display unit 420.

[0066] The control unit 430 is a CPU, which is a processor, and the storage unit 431 is, for example, a memory such as a RAM or a ROM. In addition, the user terminal 400 according to this embodiment expands a program stored in the storage unit 431 into the RAM, and the CPU interprets and executes the program, thereby converting input operations of the operation unit 410 into electrical signals and outputting the signals, and can display on the display unit 420 a delay time search screen for searching the arrival delay time of the train 100 at the destination, and the arrival delay time of the train 100 at the destination predicted by the prediction unit 314.

[0067] [Delay prediction method] Next, a delay prediction method using the delay prediction system 1 according to this embodiment will be described with reference to FIGS.

[0068] [Learning Method] First, the learning method will be described with reference to FIG. 8. While traveling, the train 100 transmits its operation performance data to the storage server 200 in real time (S1). The storage server 200 receives the operation performance data of the train 100 (S2) and stores it in the memory unit 210. The storage server 200 also transmits the operation performance data of the train 100 (the past performance data of the train 100) to the learning server 300 (S3). The learning server 300 receives the past performance data of the train 100 (S4) and stores it in the memory unit 311. The learning server 300 also generates, via the training data generation unit 312, training data based on the past performance data stored in the memory unit 311, with the first to fifth information as explanatory variables and the arrival delay time of the train 100 at the destination as a response variable (S5). The generated training data is stored in the memory unit 311. Furthermore, the learning server 300 uses the training data stored in the memory unit 311 to perform machine learning using the learning unit 313, and generates a delay prediction model that predicts the arrival delay time of the train 100 (S6). The generated delay prediction model is stored in the memory unit 311.

[0069] [Prediction method] Next, a delay prediction method will be described with reference to FIG. 9. The user operates the operation unit 410 according to a search screen displayed on the display unit 420 of the user terminal 400 to input the line section, current location, and destination. The user terminal 400 transmits input information, such as the input line section information, current location, destination station information, and input time information, to the storage server 200 (S7). The storage server 200 receives the input information, such as the current location and destination, from the user terminal 400 (S8) and stores it in the memory unit 210. Furthermore, the storage server 200, using the extraction unit 220, extracts most recent actual data of the train 100 that is the target of delay prediction from the past actual data of the train 100 stored in the memory unit 210, based on the input line section information, current location, destination station information, and input time information (S9), and transmits the extracted most recent actual data of the train 100 to the learning server 300 (S10). The learning server 300 receives the latest performance data of the train 100 (S11) and stores it in the storage unit 311.

[0070] The learning server 300 uses the prediction unit 314 to predict the arrival delay time of the train 100 at the destination based on the latest actual data of the train 100 stored in the memory unit 311 and the delay prediction model (S12), and transmits the prediction result to the user terminal 400 (S13). The user terminal 400 receives the prediction result (S14) and displays the prediction result on the display unit 420 (S15), as shown in FIG.

[0071] [Advantages of the delay prediction system according to this embodiment] As described above, the delay prediction system 1 according to this embodiment includes a training data generation unit 312 that generates training data based on past performance data showing the past driving performance of multiple mobile bodies in a specified section, with information including at least information related to the density of mobile bodies in the specified section as an explanatory variable and the arrival delay time of the mobile body at the destination as a target variable, and a learning unit 313 that generates a delay prediction model that predicts the arrival delay time of the mobile body at the destination by machine learning using the training data.

[0072] The delay prediction system 1 having such a configuration has the significant advantage that, since the density of mobile objects in a given section, such as the number of mobile objects, has a significant impact on the delay of the mobile objects, by learning information related to the density of mobile objects in a given section, it is possible to generate a delay prediction model that can accurately predict the final delay time when a delay occurs.In addition, since the system is equipped with a teacher data generation unit 312 and a learning unit 313, it has the advantage of improving learning accuracy.

[0073] The delay prediction system 1 according to this embodiment further includes a prediction unit 314 that predicts the arrival delay time of a mobile body at a destination based on the latest performance data indicating the latest driving performance of a plurality of mobile bodies in a predetermined section and the delay prediction model. The delay prediction system 1 having such a configuration has the advantage that, because it includes the prediction unit 314, it can perform everything from generating the delay prediction model to predicting the arrival delay time of a mobile body at a destination in a single system.

[0074] Furthermore, in the delay prediction system 1 according to this embodiment, the moving body is a train 100, and the information relating to the density of moving bodies includes information relating to the train density from the current location to the destination. The delay prediction system 1 having such a configuration has the advantage that the learning unit 313 can learn the degree of train density in the section where the user uses the train 100, and therefore can accurately predict the final delay time at the destination when a delay occurs.

[0075] In addition, in the delay prediction system 1 of this embodiment, the information related to the train density from the current location to the destination includes at least one of the following information: information on the number of stations between the current location and the destination, information on the number of trains between the current location and the destination, information on the total station stop time of trains between the current location and the destination, and information on the planned running time from the current location to the destination.

[0076] According to the delay prediction system 1 having such a configuration, for example, when the information includes information on the number of stations between the current location and the destination and information on the number of trains between the current location and the destination, the learning unit 313 can perform learning taking into account the number of stations between the current location and the destination and the number of trains between the current location and the destination, which are relevant to determining whether the train 100 can run, thereby advantageously being able to accurately predict the final delay time at the destination. Furthermore, when the information includes information on the total stop times of trains between the current location and the destination, the learning unit 313 can perform learning taking into account the stop times of trains, which change depending on the cause of the delay, the progress of restoration work, and other circumstances, thereby advantageously being able to accurately predict the final delay time at the destination. Furthermore, when the information includes information on the planned running time from the current location to the destination, the learning unit 313 can perform learning by comparing the actual running time of the train 100 with the running time in the planned or actual schedule. This advantageously allows for accurate prediction of the final delay time at the destination.

[0077] Furthermore, in the delay prediction system 1 according to this embodiment, the information related to the density of moving objects includes information related to the density of trains on the line. The delay prediction system 1 having such a configuration has the advantage that the learning unit 313 can learn by taking into account the degree of train density in sections other than the section where the user uses the train 100, and therefore can accurately predict the final delay time at the destination.

[0078] Furthermore, in the delay prediction system 1 according to this embodiment, the information related to train density in the line section includes at least one of information related to the number of trains on one track of the line section, information related to the number of trains on the entire line section, and information related to the total stop times of trains on the entire line section. According to the delay prediction system 1 having such a configuration, for example, when information related to the number of trains on one track of the line section or information related to the number of trains on the entire line section is included, the learning unit 313 can learn about the number of trains on the entire line section on which the train 100 runs, as well as the number of trains on the entire line section on which the train 100 runs and the entire line section on which the train 100 runs and the line section on which the train 100 does not run, thereby enabling learning to more appropriately take into account the density of the train 100, thereby providing the advantage of being able to accurately predict the final delay time at the destination. Furthermore, when information related to the total stop times of trains on the entire line section is included, the learning unit 313 can learn by taking into account the total stop times of trains on the entire line section, thereby providing the advantage of being able to accurately predict the final delay time at the destination.

[0079] Furthermore, in the delay prediction system 1 according to this embodiment, the information related to the density of moving objects includes information related to the train density at other stations within the line. The delay prediction system 1 having such a configuration has the advantage that the learning unit 313 can learn by taking into account the train density at stations other than the current location, and therefore can accurately predict the final delay time at the destination.

[0080] Furthermore, in the delay prediction system 1 according to this embodiment, the information related to train density at other stations within the line section includes at least one of information related to the train arrival delay time of trains running on one track of the line section subject to delay prediction at a connecting station where multiple lines are connected, and information related to the train arrival delay time of trains running on the other track of the line section subject to delay prediction. The delay prediction system 1 having this configuration has the advantage that the learning unit 313 can learn by taking into account the delay trends of the train 100 at a connecting station where multiple lines are connected, and can learn by taking into account the effect of the delay time at the connecting station on the arrival delay time at the destination.

[0081] Furthermore, in the delay prediction system 1 according to this embodiment, the information related to the density of moving objects includes information related to the train density between the nearest trains. Furthermore, the information related to the train density between the nearest trains includes information related to the number of stations between the train itself and the train immediately ahead. Furthermore, the information related to the train density between the nearest trains includes information related to the number of stations between the train itself and the train immediately behind. According to the delay prediction system 1 having such a configuration, the learning unit 313 learns about the train density between the nearest trains, thereby being able to determine whether the train itself 100A is in a state where it can run, which has the advantage of being able to accurately predict the final delay time at the destination.

[0082] Furthermore, in the delay prediction system 1 according to this embodiment, the explanatory variables include information related to the occupancy rate of the mobile object. According to the delay prediction system 1 having such a configuration, the learning unit 313 learns about the occupancy rate of the mobile object, and thus the delay time can be predicted taking into consideration the effect of the occupancy rate of the mobile object on the delay time, which has the advantage of being able to accurately predict the final delay time at the destination.

[0083] Furthermore, in the delay prediction system 1 according to this embodiment, the explanatory variables include information about the plan, performance, and delay of the mobile object at the current location. With the delay prediction system 1 having such a configuration, the learning unit 313 learns about the plan, performance, and delay of the mobile object at the current location, and thereby it is possible to predict the delay time based on the planned or actual schedule and the performance schedule, which has the advantage of being able to accurately predict the final delay time at the destination.

[0084] Furthermore, the delay prediction system 1 according to this embodiment further includes a user terminal 400 that displays the arrival delay time of the mobile body at the destination predicted by the prediction unit 314. The delay prediction system 1 having such a configuration can display the arrival delay time predicted by the prediction unit 314 on the user terminal 400 carried by the user, which has the advantage of being able to inform the user of the final delay time at the destination when a delay occurs.

[0085] [Variations] The present invention is not limited to the above-described embodiment, and various modifications can be made within the scope of the technical concept of the present invention.

[0086] For example, in the above-described embodiment, the information related to the density of train 100 was described as including information related to "train density from the current location to the destination," "train density on the line section," "train density at other stations on the line section," and "train density between the nearest trains," but this is not limited to this, and the information may include only one of these pieces of information, or any two or three of these pieces of information, or may include other information such as track circuit information and kilometer distance information in addition to, or regardless of, these pieces of information.

[0087] Furthermore, in the above-described embodiment, the information related to the train density from the current location to the destination was described as including information regarding the "number of stations between the current location and the destination," "number of trains between the current location and the destination," "total station stop times of trains between the current location and the destination," and "planned travel time from the current location to the destination," but this is not limited to this, and it may also be an aspect in which only one of these pieces of information is included, or any two or three of these pieces of information are included.

[0088] Furthermore, in the above-described embodiment, the information related to train density on a line section was described as including information regarding "the number of trains on one track of the line section," "the number of trains on the entire line section," and "the total station stop time of trains on the entire line section," but this is not limited to this, and it may also be an aspect in which only one of these pieces of information is included, or an aspect in which any two of these pieces of information are included.

[0089] Furthermore, in the above-described embodiment, the information related to train density at other stations within the line section was described as including information regarding "train arrival delay times of trains running on one track of the line section subject to delay prediction at the connecting station" and "train arrival delay times of trains running on the other track of the line section subject to delay prediction at the connecting station," but this is not limited to this, and it is also possible for the information to include only one of these pieces of information.

[0090] Furthermore, in the above-described embodiment, the information related to train density between the nearest trains was described as including information regarding "the number of stations between the train itself and the train immediately ahead" and "the number of stations between the train itself and the train immediately behind," but this is not limited to this, and it may also be a case where only one of these pieces of information is included.

[0091] Furthermore, in the above-described embodiment, the explanatory variables were described as including "information for identifying the day or target," "information regarding the plans, performance, and delays of train 100 at the current location," "information regarding the performance and delays of train 100 at the previous station," and "information related to the occupancy rate of train 100," but are not limited to this, and may include only one of these pieces of information, any two, or any three of these pieces of information, or may include other information in addition to or independent of these pieces of information.

[0092] Furthermore, in the above-described embodiment, the delay prediction system 1 is described as being equipped with a user terminal 400 and displaying the arrival delay time of the train 100 at the destination on the display unit 420 of the user terminal 400, but this is not limited to this, and the arrival delay time may also be displayed, for example, on an electronic bulletin board or a website managed by the railway company.

[0093] In addition, in the above-described embodiment, the information regarding "train arrival delay time of trains running on one track of the line section subject to delay prediction at the connecting station" and "train arrival delay time of trains running on the other track of the line section subject to delay prediction at the connecting station" was targeted at one connecting station, but this is not limited to this and multiple connecting stations may be targeted.

[0094] Furthermore, in the above-described embodiment, the information related to train density at other stations within the line section was targeted at connecting stations, but this is not limited to this and may be any station determined by the railway company, etc.

[0095] In the above-described embodiment, the information regarding the "train arrival delay time of a train running on one track in a line section subject to delay prediction at a connecting station" and the "train arrival delay time of a train running on the other track in a line section subject to delay prediction at a connecting station" was described with reference to the train arrival delay time of a train running on a line section subject to delay prediction at a connecting station, but this is not limited to this. For example, instead of or in addition to this, the train arrival delay time of a train running on another line section at a connecting station may also be the target. By taking into account such delay information on another line section that is not the same line section, highly accurate predictions that take into account information about the areas around connecting stations are possible.

[0096] Furthermore, in the above-described embodiment, the explanatory variables are described as including "information regarding the performance and delays of train 100 at the previous station," but are not limited to this and may include, for example, information regarding the performance and delays of train 100 at stations located after the previous station, such as the "station before that" or the "station before that."

[0097] Furthermore, in the above-described embodiment, the information related to the train density from the current location to the destination was described as including information regarding the "planned travel time from the current location to the destination," but it is not limited to this, and may also include information regarding the "distance from the current location to the destination."

[0098] It is clear from the claims that the above-mentioned modifications are included within the scope of the present invention. [Explanation of symbols]

[0099] 1: Delay prediction system 100: Train 200: Storage server 210: Storage section 220:Extraction part 300: Learning server 310: Control unit 311: Storage section 312: Teacher data generation unit 313: Learning Department 314: Prediction Department 400: User terminal 410:Operation unit 420:Display section 430: Control section 431: Storage section 432: Operation control section 433: Display control unit NW: Communication network

Claims

1. a training data generation unit that generates training data based on past performance data indicating past driving performance of a plurality of mobile bodies in a predetermined section, using information including at least information related to the density of mobile bodies in the predetermined section as an explanatory variable and an arrival delay time of the mobile bodies at a destination as a response variable; a learning unit that generates a delay prediction model that predicts the arrival delay time of a mobile object at a destination by machine learning using the training data; A delay prediction system comprising:

2. and a prediction unit that predicts an arrival delay time of the mobile object at the destination based on the latest performance data indicating the latest driving performance of the multiple mobile objects in a predetermined section and the delay prediction model. The delay prediction system according to claim 1 .

3. the moving object is a train, The information relating to the density of moving objects includes information relating to the train density from the current location to the destination.

3. The delay prediction system according to claim 1 or 2.

4. The information relating to the train density from the current location to the destination includes: Information about the number of stations between the current location and the destination and information about the number of trains between the current location and the destination; Information about the total time trains will stop at stations between your current location and your destination, Information about the planned travel time from your current location to your destination, At least one piece of information is included.

4. The delay prediction system according to claim 3.

5. The information related to the density of moving objects includes information related to the train density on the line.

5. The delay prediction system according to claim 3 or 4.

6. The information related to train density in the line section includes: Information about the number of trains on one track of the line; Information about the number of trains on the entire line, Information on the total time trains spend at stations across the entire line, and At least one piece of information is included.

6. The delay prediction system according to claim 5.

7. The information relating to the density of moving objects includes information relating to the train density at other stations within the line.

7. The delay prediction system according to claim 3, wherein the delay prediction system is a system for predicting a delay of a vehicle.

8. The information related to train density at other stations in the line section includes at least one of information related to train arrival delay times of trains running on one track of a line section that is a target for delay prediction at a connecting station where multiple lines are connected, and information related to train arrival delay times of trains running on another track of the line section that is a target for delay prediction.

8. The delay prediction system according to claim 7.

9. The information relating to the density of moving bodies includes information relating to the train density between the nearest trains.

9. The delay prediction system according to claim 3, wherein the delay prediction system is a system for predicting a delay of a vehicle.

10. The information relating to the train density between the nearest trains includes information relating to the number of stations between the train itself and the nearest train ahead. The delay prediction system according to claim 9 .

11. The information relating to the train density between the nearest trains includes information relating to the number of stations between the train and the nearest train behind it.

11. The delay prediction system according to claim 9 or 10.

12. The explanatory variables include information related to the occupancy rate of the vehicle.

12. The delay prediction system according to claim 1.

13. The explanatory variables include information about the plan, performance, and delay of the moving object at the current location.

13. The delay prediction system according to claim 1.

14. The system further includes a user terminal that displays the arrival delay time of the mobile body at the destination predicted by the prediction unit.

3. The delay prediction system according to claim 2.

15. On the computer, a training data generation step of generating training data based on past performance data showing past driving performance of a plurality of mobile bodies in a predetermined section, the training data having information including at least information related to the density of mobile bodies in the predetermined section as an explanatory variable and an arrival delay time of the mobile bodies at the destination as a response variable; a learning process for generating a delay prediction model for predicting the arrival delay time of a mobile object at a destination by machine learning using the training data; A delay prediction program characterized by executing the following.

16. The method further includes a prediction step of predicting an arrival delay time of a mobile object at a destination based on the latest performance data indicating the latest driving performance of a plurality of mobile objects in a predetermined section and the delay prediction model.

16. The delay prediction program according to claim 15.

17. a training data generation step of generating training data based on past performance data showing past driving performance of a plurality of mobile bodies in a predetermined section, the training data having information including at least information related to the density of mobile bodies in the predetermined section as an explanatory variable and an arrival delay time of the mobile bodies at the destination as a response variable; a learning process for generating a delay prediction model for predicting the arrival delay time of a mobile object at a destination by machine learning using the training data; A delay prediction method comprising:

18. The method further includes a prediction step of predicting an arrival delay time of a mobile object at a destination based on the latest performance data indicating the latest driving performance of a plurality of mobile objects in a predetermined section and the delay prediction model.

18. The delay prediction method according to claim 17.

Images