Data generation program, method, and device

Abstract

To generate various training data for building a machine learning model for high accuracy traffic simulation.SOLUTION: A data generation device is configured to: extract, from route information indicating travel status of each of a plurality of mobile bodies in a predetermined geographical range, within a first period, for each of a plurality of points of time, route information of mobile bodies that have started to move within a second period that is included in the first period and shorter than the first period; generate, based on the extracted route information, tally information obtained by aggregating the number of mobile bodies for each combination of a departure point and an arrival point for the traveling mobile bodies included in the route information, and information indicating a degree of traffic congestion in the specific geographical range; and generate training data that uses the tally information as input feature values and the information indicating the degree of traffic congestion as label information, as training data for a machine learning model configured to derive a degree of traffic congestion corresponding to the tally information.SELECTED DRAWING: Figure 13

Description

[Technical field]

[0001] The disclosed technology relates to a data generation program, a data generation method, and a data generation device. [Background technology]

[0002] Conventionally, there is a technology that uses traffic simulation to optimize public transportation schedules. For example, when determining the operation schedule of a special bus during an event, it is necessary to set the departure and arrival times of the special bus according to the traffic conditions to avoid the occurrence and expansion of congestion. In such a case, a traffic simulation is used to search for an optimal schedule that does not cause congestion.

[0003] Building a traffic simulation device is very time-consuming. In addition, the computational load during simulation is high. However, by extracting traffic demand and traffic density from probe data and learning the relationship between the two using machine learning models such as neural networks, it is possible to build a highly accurate alternative simulation model (surrogate model). Because the computational load of the surrogate model is low, it can expand the range of applications of task optimization through simulation. For example, it can be applied to optimizing daily schedules.

[0004] As a technology related to traffic simulation using a surrogate model, a data generation device that efficiently estimates pedestrian and traffic flows has been proposed. The device acquires a first parameter that represents an environment and a second parameter that represents an attribute of the movement of each of a plurality of moving objects in the environment, and classifies the plurality of moving objects into a plurality of groups based on the second parameter. The device also generates a third parameter that indicates the number of moving objects classified into each of the plurality of groups, inputs the first parameter and the third parameter into a machine learning model, and generates estimated information regarding the movement of the plurality of moving objects in the environment. [Prior art documents] [Patent documents]

[0005] [Patent Document 1] JP 2022-131393 A Summary of the Invention [Problem to be solved by the invention]

[0006] In order to build a surrogate model for traffic simulation, training data that covers a wide range of situations is necessary. However, it is difficult to obtain the data used for training because it is expensive or does not exist in the first place (a sufficient amount of actual measurement data for training is not available). Therefore, it is necessary to generate diverse training data for training a highly accurate surrogate model from limited data.

[0007] In one aspect, the disclosed technology aims to generate diverse training data for building highly accurate machine learning models for traffic simulation. [Means for solving the problem]

[0008] In one embodiment, the disclosed technology extracts route information of a moving object that starts moving within a second time period from route information indicating the movement status of each of a plurality of moving objects in a predetermined geographical range within a first time period at a plurality of time points. The second time period is included within the first time period and is shorter than the first time period. The disclosed technology also generates, based on the extracted route information, aggregate information that aggregates the number of moving objects for each combination of starting points and arrival points in the movement of the moving objects included in the route information, and information indicating the traffic congestion in the predetermined geographical range. The disclosed technology then generates training data in which the aggregate information is used as input features and the information indicating the congestion is used as label information, as training data for a machine learning model that derives the traffic congestion according to the aggregate information. Effect of the Invention

[0009] One aspect is that it has the effect of generating diverse training data for building a highly accurate machine learning model for traffic simulation. [Brief description of the drawings]

[0010] [Figure 1] FIG. 13 is a diagram for explaining a traffic simulation. [Diagram 2] FIG. 1 is a schematic diagram showing an example of a road network. [Diagram 3] FIG. 1 is a diagram for explaining the machine learning phase of a surrogate model. [Figure 4] FIG. 13 is a diagram for explaining an optimization phase using a surrogate model. [Diagram 5] FIG. 13 is a diagram for explaining data extension of sequence data. [Figure 6] FIG. 1 is a diagram for explaining a problem that arises when data extension of normal sequence data is applied. [Figure 7] FIG. 2 is a functional block diagram of the data generating device according to the present embodiment. [Figure 8] FIG. 4 is a diagram showing an example of probe data and route information. [Figure 9] FIG. 2 illustrates an example of a time window. [Figure 10] FIG. 11 is a diagram showing an example of a movement start time list. [Figure 11] FIG. 11 is a diagram showing an example of an ID list. [Figure 12] FIG. 11 is a diagram illustrating an example of route information extracted for each time window. [Figure 13] FIG. 13 is a diagram for explaining generation of an OD table and traffic density based on route information extracted for each time window. [Figure 14] FIG. 1 is a block diagram showing a schematic configuration of a computer functioning as a data generating device. [Figure 15] 13 is a flowchart illustrating an example of a data generation process. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

[0011] Hereinafter, an example of an embodiment of the disclosed technology will be described with reference to the drawings.

[0012] <Overview and Issues of Traffic Simulation and Surrogate Model> Before describing the details of the embodiment, an overview and problems of general traffic simulation and surrogate models will be described.

[0013] As shown in Figure 1, traffic simulation takes traffic demand as input, and simulates and outputs traffic conditions according to that traffic demand. For example, as shown in Figure 2, traffic simulation is performed using a road network represented by multiple nodes and links connecting the nodes. In the example in Figure 2, the black circles represent nodes, and the connection lines connecting the nodes represent links. In addition, "ni (i = 1, 2,...)" written next to the nodes is the node identification number, and "ej (j = 1, 2,...)" written next to the links is the link identification number. In the following, the node with identification number ni will be referred to as "node ni", and the link with identification number ej will be referred to as "link ej".

[0014] Traffic demand is represented, for example, by an OD table showing the number of moving objects (people, vehicles, etc.) moving between each point (node). The OD table is represented as a matrix of origin nodes and destination nodes, with the number of moving objects moving from the corresponding origin to the destination stored in each square of the matrix. In addition, a three-dimensional tensor as shown in FIG. 1 may be used as traffic demand using OD by time period. In addition, traffic conditions are, for example, traffic density represented by traffic volume for each link.

[0015] A machine learning model (surrogate model) for traffic simulation is a lightweight alternative model that reproduces the input / output relationship of a traffic simulation as shown in FIG. 1. As shown in FIG. 3, in the machine learning phase of the machine learning model, an information processing device extracts traffic demand and road networks, which are feature quantities, from geographic data and multiple probe data, and also extracts traffic density, which is a correct label. Probe data is data in which location information (latitude, longitude, etc.) of each moving object is recorded together with a timestamp. Probe data is acquired by smartphones, various sensors, etc., and collected via a network. Geographic data is data in which various geographic information, including road network information, is recorded.

[0016] The information processing device then inputs the extracted traffic demand and road network as features into a machine learning model composed of a neural network or the like, and acquires the traffic density inferred by the machine learning model. The information processing device then trains the machine learning model by updating parameters of the machine learning model so as to minimize the error between the acquired traffic density and the traffic density that is the correct label.

[0017] In the optimization phase using the trained machine learning model, as shown in FIG. 4, the information processing device inputs the road network and traffic demand into the machine learning model to obtain the traffic density, which is the objective function. FIG. 4 shows an example of optimizing a bus operation schedule when an event is held. In the example of FIG. 4, the information processing device inputs the event traffic demand for each bus operation schedule proposal as a decision variable into the machine learning model, with the road network and daily traffic demand forecast values ​​as constants. The road network may also be treated as a decision variable. Then, the information processing device compares the traffic density obtained for each proposal, and selects and outputs the proposal with the minimum traffic density (congestion) as the optimal proposal. The optimization may be performed using an existing method such as a genetic algorithm.

[0018] As mentioned above, data used for training machine learning models (surrogate models) for traffic simulations is often difficult to obtain. In order to build a highly accurate machine learning model, diverse training data is required, so it is considered to generate training data using data augmentation methods.

[0019] A typical data extension method for sequence data is to extract data from sequence data using a time window. Using this method, it is possible to generate features (traffic demand) and correct labels (traffic density) based on data extracted from traffic volume data for a first period using a time window for a second period that is shorter than the first period. For example, as shown in Fig. 5, data is extracted using a six-hour time window from traffic volume data for one day (0:00 to 24:00). By setting multiple six-hour time windows within a one-day period, multiple data can be extracted from traffic volume data for one day. In Fig. 5, the traffic volume at each time is represented by different shades of shading.

[0020] When this method is applied to the feature quantity (traffic demand) and the correct answer label (traffic density) used for training the machine learning model, there may be cases where the feature quantity and the correct answer label do not match. For example, as shown in FIG. 6, data within a time window (t=6:00 to 12:00 in the example of FIG. 6) is extracted from the OD table, and data for traffic density is also extracted in the same time window. In this case, the data for the extracted part of the traffic density is affected by the previous time period. In the example of FIG. 6, the traffic density of link e2 is shown as the traffic density. For example, assume that at the stage of t=6:00, 100 vehicles are stuck on link e2 due to congestion, and this stuckness continues until t=7:00. In this case, the data for the part of the time window is still affected by the above-mentioned stuck portion (for example, the shaded part in FIG. 6). On the other hand, since the value of the OD table is the number of vehicles at each time, the data for the extracted part is not affected by the previous time period.

[0021] In this way, when a machine learning model is trained using training data in which the features and the correct labels are not consistent, the inference accuracy of the machine learning model cannot be improved. Therefore, in this embodiment, when extracting data from a time window, data of a moving body that starts moving between the start time and the end time of the time window is extracted. Then, the extracted data is used to generate traffic demand, which is a feature, and traffic density, which is a correct label. As a result, the influence before the start time of the time window is excluded from both the feature side and the correct label side, and the feature and the correct label are consistent.

[0022] <Data Generation Device According to This Embodiment> As shown in Fig. 7, the data generating device 10 according to this embodiment includes a dividing unit 12, an extracting unit 14, a generating unit 16, and a training unit 18. A plurality of probe data are input to the data generating device 10. The data generating device 10 generates training data used for training the machine learning model 30 from the probe data, and trains the machine learning model 30. The machine learning model 30 is configured, for example, by a neural network.

[0023] The division unit 12 acquires a plurality of probe data input to the data generating device 10. The probe data is, for example, data indicating the movement status of each of a plurality of moving objects recorded during one day. That is, in the probe data, the one-day history of a certain moving object is recorded from the time when the recording device for recording the probe data is turned on. Therefore, it is expected that the recording start time of each of the probe data of the plurality of moving objects is concentrated, for example, in the early morning hours. An example of raw data of the probe data is shown in the upper diagram of FIG. 8. The example of FIG. 8 is probe data for a moving object whose identification information (hereinafter, referred to as "moving object ID") is 1, and is expressed in the data format of {moving object ID, time, latitude, longitude}.

[0024] Although details will be described later, in this embodiment, data is extracted in a time window based on the movement start time of the moving object. Therefore, if the method of this embodiment is applied to raw probe data in which the recording start times are concentrated in the early morning hours as described above, almost no data will be extracted from the time window set in the latter half of the first period.

[0025] Therefore, the dividing unit 12 divides the probe data of each mobile object at the boundaries of "activity". Specifically, the dividing unit 12 divides the probe data before and after a portion indicating that the position of the mobile object has remained within a certain range for a certain period of time, based on the time and position information (latitude and longitude) indicated by the probe data. A state in which the position of the mobile object has remained within a certain range for a certain period of time is regarded as a state in which some activity (work, sleep, etc.) is being performed there, and by dividing the probe data before and after the activity, the data indicating movement can be used as route information.

[0026] For example, in the raw probe data in the upper diagram of Fig. 8, the latitude and longitude remain within a range of ±0.0001 for 30 minutes from 9:00 to 9:30. The dividing unit 12 assumes that some activity is taking place in this stagnant portion, and divides the probe data into data before 9:00 and data after 9:30. The time and distance thresholds for determining stagnant periods are preset to values ​​that can exclude temporary stagnant periods such as waiting for traffic lights.

[0027] An example of the probe data after division is shown in the lower diagram of FIG. 8. The division unit 12 regards each piece of divided data as route information and assigns identification information (hereinafter referred to as "route ID") to each piece of route information. In other words, the route information is information in which a route ID is associated with a sequence of position information of a moving object at each of a plurality of time points. In the example of the lower diagram of FIG. 8, the route ID is expressed in a format in which a branch number indicating which divided data it is is assigned to the moving object ID. The division unit 12 collects the route information for a plurality of moving objects and regards it as population data. At this time, the division unit 12 may assign a new route ID such as a serial number to each piece of route information.

[0028] The extraction unit 14 extracts route information of moving objects that are included in the first period and start moving within a second period that is shorter than the first period, from route information that indicates the movement status of each of a plurality of moving objects in a predetermined geographical range within the first period at a plurality of time points. The first period is the period of the population data before being cut out by the time window (e.g., one day). The second period is the period of the time window (e.g., six hours). The predetermined geographical range is an area indicated by a road network.

[0029] Specifically, as shown in FIG. 9, the extraction unit 14 prepares the range (start time-end time) of the time window set for the first period in association with the data name to be given to the data extracted by the time window. Also, as shown in FIG. 10, the extraction unit 14 extracts the movement start time of each piece of route information from the population data, and creates a movement start time list in which the route ID of the route information is associated with the movement start time. The example of FIG. 10 represents population data in which each piece of position information included in each piece of multiple route information is rearranged in chronological order. In this case, the extraction unit 14 extracts the earliest time from among the data to which the same route ID is assigned, as the movement start time to be associated with the route ID. In FIG. 10, the shaded data is the movement start time data of each piece of route information.

[0030] Furthermore, the extraction unit 14 refers to the movement start time list and creates an ID list in which the route ID of the route information including the movement start time in each time window is associated with the data name corresponding to the time window. An example of the ID list is shown in FIG. 11. The extraction unit 14 refers to the ID list and extracts data of the route ID associated with each data name from the data of the population. An example of the extracted data is shown in FIG. 12. Data with the same route ID included in each extracted data is arranged in chronological order to form the route information. In other words, the extracted data is route information extracted for each time window.

[0031] The generation unit 16 generates, for each time window, an OD table that tallies the number of moving objects for each combination of starting points and arrival points in the movement of the moving objects included in the route information, based on the route information extracted by the extraction unit 14, and a traffic density in a specified geographical area.

[0032] Specifically, the generating unit 16 associates the route information with a road network, compares the position information (latitude and longitude) of the mobile object with the position information of each link, and converts the route information, which is a sequence of position information, into route information, which is a sequence of links on which the mobile object exists at each time. The generating unit 16 identifies nodes corresponding to the departure point and arrival point of each route information based on the route information, which is a sequence of links, and extracts ODs, which are combinations of nodes corresponding to the departure point and nodes corresponding to the arrival point, from each route information. The generating unit 16 then counts the number of identical ODs and stores them in the corresponding cells of the OD table, thereby generating an OD table. The generating unit 16 also generates link traffic flow, which is represented by the number of mobile objects existing on each link at each time point, as traffic density based on the route information, which is a sequence of links.

[0033] The generation unit 16 generates training data using the generated OD table as input features and the generated traffic density as correct answer labels as training data for a machine learning model 30 that derives traffic congestion according to traffic demand.

[0034] In this manner, in this embodiment, when data is extracted from a time window, route information whose travel start time is included in the time window is extracted, and traffic demand and traffic density for the time window are generated based on the extracted route information, as shown in Fig. 13. This eliminates the influence of traffic conditions before the start time of the time window from the correct label (traffic density) side as well, and the feature amount and the correct label are consistent.

[0035] Since route information is extracted based on the travel start time, the extracted route information also includes data after the end time of the time window, but this is not a problem for OD tables, since OD is usually tallied at the time of the departure point. Also, for traffic density, the end can be cut off at the end time of the time window.

[0036] The training unit 18 trains the machine learning model 30 using the training data generated by the generation unit 16. Specifically, the training unit 18 inputs the generated OD table as a feature to the machine learning model 30, and acquires the traffic density inferred by the machine learning model 30. Then, the training unit 18 trains the machine learning model 30 by updating the parameters of the machine learning model 30 so as to minimize the error between the acquired traffic density and the traffic density generated as the correct label.

[0037] The data generating device 10 may be realized by, for example, a computer 50 shown in Fig. 14. The computer 50 includes a CPU (Central Processing Unit) 51, a GPU (Graphics Processing Unit) 52, a memory 53 as a temporary storage area, and a non-volatile storage device 54. The computer 50 also includes an input / output device 55 such as an input device and a display device, and an R / W (Read / Write) device 56 that controls reading and writing of data to a storage medium 59. The computer 50 also includes a communication I / F (Interface) 57 that is connected to a network such as the Internet. The CPU 51, the GPU 52, the memory 53, the storage device 54, the input / output device 55, the R / W device 56, and the communication I / F 57 are connected to one another via a bus 58.

[0038] The storage device 54 is, for example, a hard disk drive (HDD), a solid state drive (SSD), a flash memory, etc. The storage device 54 as a storage medium stores a data generation program 60 for causing the computer 50 to function as the data generation device 10. The data generation program 60 has a division process control instruction 62, an extraction process control instruction 64, a generation process control instruction 66, and a training process control instruction 68.

[0039] The CPU 51 reads the data generation program 60 from the storage device 54, expands it in the memory 53, and sequentially executes the control instructions of the data generation program 60. The CPU 51 operates as the division unit 12 shown in FIG. 7 by executing the division process control instruction 62. The CPU 51 also operates as the extraction unit 14 shown in FIG. 7 by executing the extraction process control instruction 64. The CPU 51 also operates as the generation unit 16 shown in FIG. 7 by executing the generation process control instruction 66. The CPU 51 also operates as the training unit 18 shown in FIG. 7 by executing the training process control instruction 68. As a result, the computer 50 that has executed the data generation program 60 functions as the data generation device 10. The CPU 51 that executes the program is hardware. A part of the program may be executed by the GPU 52.

[0040] Furthermore, the functions realized by the data generation program 60 may be realized by, for example, a semiconductor integrated circuit, more specifically, an ASIC (Application Specific Integrated Circuit), an FPGA (Field-Programmable Gate Array), or the like.

[0041] Next, the operation of the data generating device 10 according to the present embodiment will be described. When a plurality of probe data are input to the data generating device 10, the data generating device 10 executes a data generating process shown in FIG. 15. The data generating process is an example of a data generating method of the disclosed technology. The data generating process will be described in detail below.

[0042] In step S10, the division unit 12 acquires the multiple probe data input to the data generating device 10. Next, in step S12, the division unit 12 divides the probe data before and after a portion indicating that the position of the moving object remains within a certain range for a certain period of time, based on the time and position information (latitude and longitude) indicated by the probe data. The division unit 12 assigns a route ID to each piece of data after division to generate route information, and compiles the route information for the multiple moving objects as population data.

[0043] Next, in step S14, the extraction unit 14 prepares the range of the time window (start time-end time) set for the population data and the data name to be given to the data extracted by the time window in association with each other. Next, in step S16, the extraction unit 14 extracts the movement start time of each piece of route information from the population data, and creates a movement start time list in which the route ID of the route information is associated with the movement start time.

[0044] Next, in step S18, the extraction unit 14 refers to the movement start time list and creates an ID list in which the route ID of the route information including the movement start time in each time window is associated with the data name corresponding to the time window. Next, in step S20, the extraction unit 14 refers to the ID list and extracts data of the route ID associated with each data name from the population data. That is, the extraction unit 14 extracts route information for each time window.

[0045] Next, in step S22, the generation unit 16 generates an OD table and traffic density based on the route information extracted for each time window in step S20. Next, in step S24, training data is generated with the OD table as the feature and the traffic density as the correct label, and the data generation process is terminated.

[0046] Then, the training unit 18 trains the machine learning model 30 using the training data generated by the data generation process.

[0047] As described above, the data generating device according to the present embodiment extracts route information of a moving object that starts moving within a second period shorter than the first period from route information indicating the movement status of each of a plurality of moving objects in a predetermined geographical range at each of a plurality of time points within the first period. The data generating device also generates, based on the extracted route information, aggregate information that aggregates the number of moving objects for each combination of a starting point and an ending point in the movement of the moving objects included in the route information, and information indicating the degree of traffic congestion in the predetermined geographical range. The data generating device then generates training data in which the aggregate information is used as input feature and information indicating the degree of congestion is used as label information, as training data for a machine learning model that derives the degree of traffic congestion according to the aggregate information. This makes it possible to generate a variety of training data for constructing a machine learning model for a highly accurate traffic simulation.

[0048] In the above embodiment, the data generating device including the training unit has been described, but the training unit may be configured as a separate computer, and the data generating device may be configured to output the generated training data. In this case, the computer including the training unit trains the machine learning model using the training data output from the data generating device.

[0049] In the above embodiment, the data generation program is stored (installed) in advance in the storage device, but this is not limiting. The program according to the disclosed technology may be provided in a form stored in a storage medium such as a CD-ROM, a DVD-ROM, or a USB memory.

[0050] The following supplementary notes are further disclosed regarding the above embodiment.

[0051] (Appendix 1) extracting route information of a moving object that starts moving within a second time period that is included in the first time period and is shorter than the first time period from route information that indicates the movement status of each of a plurality of moving objects in a predetermined geographical range within the first time period at a plurality of time points; generating, based on the extracted route information, aggregate information that aggregates the number of moving objects for each combination of departure points and arrival points in the movement of the moving objects included in the route information, and information indicating a traffic congestion level in the specified geographical area; generating training data in which the aggregated information is used as input features and the information indicating the congestion level is used as label information, as training data for a machine learning model that derives a traffic congestion level according to the aggregated information; A data generation program for causing a computer to execute a process including the steps of:

[0052] (Appendix 2) A data generation program as described in Appendix 1 for causing the computer to execute a process further including generating the route information by dividing the movement status of each of the multiple moving objects recorded in the first period based on the activity of the moving objects.

[0053] (Appendix 3) The data generation program according to claim 2, wherein a break in the activity of the moving object in the movement status is a part indicating that the position of the moving object remains within a certain range for a certain period of time.

[0054] (Appendix 4) the route information is information in which identification information of the route information is associated with a sequence of position information of the moving object for each of the plurality of time points; creating a list in which the movement start time of the route information is associated with identification information of the route information, and for each of the second time periods having different start times and end times, extracting from the list identification information of the route information whose movement start time falls within the second time period, and extracting the route information associated with the identification information of the extracted route information; The data generating program according to any one of Supplementary Note 1 to Supplementary Note 3.

[0055] (Appendix 5) 5. The data generation program according to any one of claims 1 to 4, for causing the computer to execute a process further including training the machine learning model using the generated training data.

[0056] (Appendix 6) extracting route information of a moving object that starts moving within a second time period that is included in the first time period and is shorter than the first time period from route information that indicates the movement status of each of a plurality of moving objects in a predetermined geographical range within the first time period at a plurality of time points; generating, based on the extracted route information, aggregate information that aggregates the number of moving objects for each combination of departure points and arrival points in the movement of the moving objects included in the route information, and information indicating a traffic congestion level in the specified geographical area; generating training data in which the aggregated information is used as input features and the information indicating the congestion level is used as label information, as training data for a machine learning model that derives a traffic congestion level according to the aggregated information; A data generation method comprising the steps of:

[0057] (Appendix 7) The data generation method described in Appendix 6, wherein the computer executes a process that further includes generating the route information by dividing the movement status of each of the multiple moving objects recorded in the first period based on the activity of the moving objects.

[0058] (Appendix 8) The data generation method according to claim 7, wherein a division of the activity of the moving object in the movement status is a part indicating that the position of the moving object remains within a certain range for a certain period of time.

[0059] (Appendix 9) the route information is information in which identification information of the route information is associated with a sequence of position information of the moving object for each of the plurality of time points; creating a list in which the movement start time of the route information is associated with identification information of the route information, and for each of the second time periods having different start times and end times, extracting from the list identification information of the route information whose movement start time falls within the second time period, and extracting the route information associated with the identification information of the extracted route information; A data generating method according to any one of Supplementary Note 6 to Supplementary Note 8.

[0060] (Appendix 10) 10. The data generation method according to any one of claims 6 to 9, further comprising causing the computer to execute a process that includes training the machine learning model using the generated training data.

[0061] (Appendix 11) an extraction unit that extracts, from route information indicating movement statuses of each of a plurality of moving objects in a predetermined geographical range within a first period at a plurality of time points, route information of a moving object that starts moving within a second period that is included within the first period and is shorter than the first period; a generation unit that generates, based on the extracted route information, aggregated information that aggregates the number of moving objects for each combination of starting points and arrival points in the movement of the moving objects included in the route information, and information indicating the traffic congestion in the specified geographical area, and generates training data in which the aggregated information is used as input features and the information indicating the congestion is used as label information, as training data for a machine learning model that derives the traffic congestion according to the aggregated information; A data generating device comprising:

[0062] (Appendix 12) 12. The data generating device of claim 11, further comprising a division unit that generates the route information by dividing the movement status of each of the plurality of moving objects recorded in the first period based on the activity of the moving objects.

[0063] (Appendix 13) The data generating device according to claim 12, wherein a division of the activity of the moving object in the movement status is a part indicating that the position of the moving object remains within a certain range for a certain period of time.

[0064] (Appendix 14) the route information is information in which identification information of the route information is associated with a sequence of position information of the moving object for each of the plurality of time points; the extraction unit creates a list in which a movement start time of the route information is associated with identification information of the route information, extracts from the list, for each of a plurality of second periods having different start times and end times, identification information of the route information whose movement start time falls within the second period, and extracts the route information associated with the identification information of the extracted route information. The data generating device according to any one of Supplementary notes 11 to 13.

[0065] (Appendix 15) 15. The data generating device according to any one of claims 11 to 14, further comprising a training unit that trains the machine learning model using the generated training data.

[0066] (Appendix 16) extracting route information of a moving object that starts moving within a second time period that is included in the first time period and is shorter than the first time period from route information that indicates the movement status of each of a plurality of moving objects in a predetermined geographical range within the first time period at a plurality of time points; generating, based on the extracted route information, aggregate information that aggregates the number of moving objects for each combination of departure points and arrival points in the movement of the moving objects included in the route information, and information indicating a traffic congestion level in the specified geographical area; generating training data in which the aggregated information is used as input features and the information indicating the congestion level is used as label information, as training data for a machine learning model that derives a traffic congestion level according to the aggregated information; A non-transitory storage medium storing a data generation program for causing a computer to execute a process including the steps of: [Explanation of symbols]

[0067] 10 Data generation device 12 Division 14 Extraction part 16 Generation part 18 Training Department 30 Machine Learning Models 50 Computer 51 CPU 52 GPU 53 Memory 54 Storage device 55 Input / Output Devices 56 R / W device 57 Communication I / F 58 Bus 59 Storage medium 60 Data Generator 62 Split Process Control Instructions 64 Extraction process control instructions 66 Generation process control instructions 68 Training Process Control Instructions

Claims

1. extracting route information of a moving object that starts moving within a second time period that is included in the first time period and is shorter than the first time period from route information that indicates the movement status of each of a plurality of moving objects in a predetermined geographical range within the first time period at a plurality of time points; generating, based on the extracted route information, aggregate information that aggregates the number of moving objects for each combination of departure points and arrival points in the movement of the moving objects included in the route information, and information indicating a traffic congestion level in the specified geographical area; generating training data in which the aggregated information is used as input features and the information indicating the congestion level is used as label information, as training data for a machine learning model that derives a traffic congestion level according to the aggregated information; A data generation program for causing a computer to execute a process including the steps of:

2. The data generation program of claim 1, for causing the computer to execute a process further including generating the route information by dividing the movement status of each of the plurality of moving bodies recorded during the first period based on the activity of the moving bodies.

3. 3. The data generating program according to claim 2, wherein a break in the activity of the mobile object in the movement status is a portion indicating that the position of the mobile object remains within a certain range for a certain period of time.

4. the route information is information in which identification information of the route information is associated with a sequence of position information of the moving object for each of the plurality of time points; creating a list in which the movement start time of the route information is associated with identification information of the route information, extracting from the list, for each of a plurality of second time periods having different start times and end times, identification information of the route information whose movement start time falls within the second time period, and extracting the route information associated with the identification information of the extracted route information; The data generating program according to any one of claims 1 to 3.

5. The data generation program according to any one of claims 1 to 3, for causing the computer to execute a process further including training the machine learning model using the generated training data.

6. extracting route information of a moving object that starts moving within a second time period that is included in the first time period and is shorter than the first time period from route information that indicates the movement status of each of a plurality of moving objects in a predetermined geographical range within the first time period at a plurality of time points; generating, based on the extracted route information, aggregate information that aggregates the number of moving objects for each combination of departure points and arrival points in the movement of the moving objects included in the route information, and information indicating a traffic congestion level in the specified geographical area; generating training data in which the aggregated information is used as input features and the information indicating the congestion level is used as label information, as training data for a machine learning model that derives a traffic congestion level according to the aggregated information; A data generation method comprising the steps of:

7. an extraction unit that extracts, from route information indicating movement statuses of each of a plurality of moving objects in a predetermined geographical range within a first period at a plurality of time points, route information of a moving object that starts moving within a second period that is included within the first period and is shorter than the first period; a generation unit that generates, based on the extracted route information, aggregated information that aggregates the number of moving objects for each combination of starting points and arrival points in the movement of the moving objects included in the route information, and information indicating the traffic congestion in the specified geographical area, and generates training data in which the aggregated information is used as input features and the information indicating the congestion is used as label information, as training data for a machine learning model that derives the traffic congestion according to the aggregated information; A data generating device comprising:

Images