JP7490599B2 - Information processing device, information processing method, and information processing program - Google Patents

Description

Embodiments of the present invention relate to an information processing device, an information processing method, and an information processing program.

Technologies are known that predict the routes that moving bodies such as vehicles will travel. For example, one known route prediction technology is one that predicts congestion on a route. For example, systems that provide a graph of trends obtained by simply aggregating past measurement data using moving averages or the like, systems that provide a single trend graph for a specific period such as the New Year holidays, and systems that incrementally predict near-future speed changes have been disclosed. Also disclosed is a technology that associates congestion speeds and average speeds with each section on a map in advance, stores them, and outputs congestion speeds for each section.

JP 2017-97890 A JP 2004-272408 A

However, using conventional technology, it was difficult to predict the route prediction index with high accuracy.

An information processing device of an embodiment includes a control unit that clusters a group of speed data for each time of a plurality of sections of a moving body moving in a section obtained by dividing a route into the plurality of sections into a number of clusters with an optimal information criterion , identifies, for each of the plurality of sections, a group of speed data belonging to a cluster including at least one speed data below a target speed upper limit corresponding to the number of clusters as a group of target speed data that is a target speed, predicts a speed distribution for each time based on the identified group of target speed data, and predicts a prediction index of the route for a prediction period after a set period based on the speed distribution.

1 is a schematic diagram of an information processing apparatus according to a first embodiment. FIG. 4 is an explanatory diagram of a traffic counter. FIG. An explanatory diagram of the congestion forecast index prediction. An explanatory diagram of the congestion forecast index prediction. An explanatory diagram of the congestion forecast index prediction. Schematic diagram of the congestion prediction index. 1 is a flowchart of the flow of information processing. FIG. 13 is a schematic diagram of an information processing apparatus according to a second embodiment. FIG. 1 is an explanatory diagram of prediction of upper traffic speed limits. FIG. 1 is an explanatory diagram of prediction of upper traffic speed limits. FIG. 1 is an explanatory diagram of prediction of upper traffic speed limits. FIG. 1 is an explanatory diagram of prediction of upper traffic speed limits. FIG. 1 is an explanatory diagram of prediction of upper traffic speed limits. FIG. 1 is an explanatory diagram of prediction of upper traffic speed limits. 1 is a flowchart of the flow of information processing. 13 is a flowchart of a traffic congestion upper speed limit prediction process. Clustering flow chart. FIG. 13 is a schematic diagram of an information processing apparatus according to a third embodiment. FIG. 1 is a diagram illustrating predictor learning. FIG. Schematic diagram of the search result display screen. 1 is a flowchart of the flow of information processing. FIG. 13 is a schematic diagram of an information processing apparatus according to a fourth embodiment. Schematic diagram of an interval similarity map. 1 is a flowchart of the flow of information processing. Hardware configuration diagram.

The information processing device, information processing method, and information processing program of this embodiment will be described in detail below with reference to the attached drawings. Note that in this embodiment, parts that have common functions, etc. are given the same reference numerals and detailed descriptions are omitted.

First Embodiment
FIG. 1 is a schematic diagram showing an example of an information processing apparatus 10A according to the present embodiment.

The information processing device 10A is an example of the information processing device 10. The information processing device 10 is a device that predicts traffic congestion for mobile objects moving along a route.

A mobile object is an object that can move. Examples of mobile objects include vehicles (motorcycles, automobiles, bicycles), carts, robots, ships, flying objects (airplanes, unmanned aerial vehicles (UAVs), drones, etc.). In this embodiment, the mobile object is described assuming that it is a vehicle.

A route is a path along which a moving object travels. A route may be any path along which a moving object can travel. In this embodiment, an example in which the route is a road is described. In this embodiment, the movement of a moving object along a route may be described as a vehicle traveling on a road.

The information processing device 10A includes a memory unit 20, a communication unit 22, an input unit 23, a control unit 24, and a display unit 29. The memory unit 20, the communication unit 22, the input unit 23, the control unit 24, and the display unit 29 are connected so as to be able to communicate with each other.

The storage unit 20 stores various types of data. The storage unit 20 is, for example, a semiconductor memory element such as a RAM (Random Access Memory), a flash memory, a hard disk, an optical disk, or the like. The storage unit 20 may be a storage device provided outside the information processing device 10A. The storage unit 20 may also be a storage medium. Specifically, the storage medium may be a storage medium in which programs and various types of information are downloaded and stored or temporarily stored via a LAN (Local Area Network) or the Internet. The storage unit 20 may also be composed of multiple storage media.

In this embodiment, the memory unit 20 stores traffic counter data. Traffic counter data is data obtained from traffic counters installed on roads.

Figure 2 is an explanatory diagram of a traffic counter 28 installed on road R. In Figure 2, an expressway is shown as an example of road R. Traffic counters 28 are installed at predetermined intervals on road R. The predetermined intervals are, for example, several hundred meters, but are not limited to this. The traffic counters 28 measure at least two of the following: the traffic volume, speed, and road occupancy rate of vehicles C traveling on road R. The remaining value is calculated by a formula using the distance between the traffic counters 28 and the two measured values.

Therefore, the traffic counter 28 obtains measurement data for each section P of the road R. A section P refers to each section obtained by dividing the road on which the vehicle C travels into predetermined distances. For example, adjustments can be made so that one section P includes at least one traffic counter 28.

Figure 3 is an explanatory diagram of an example of measurement data. In Figure 3, "P-001-001-F", "P-001-001-B", and "P-001-002-F" are examples of identification information for section P. In Figure 3, Q represents the traffic volume of vehicle C, V represents the speed of vehicle C, and Qcc represents the road occupancy rate of vehicle C. More specifically, the speed of vehicle C means the average speed of vehicles C traveling in the same direction in section P. Note that if there are multiple lanes in one driving direction on road R, the traffic counter 28 may measure the measurement data for each driving direction, treating multiple lanes as one.

The traffic counter 28 obtains measurement data for each of the multiple sections P at each date and time. That is, the traffic counter 28 obtains measurement data including speed data that is the average speed of the vehicle C at each date and time for each of the multiple sections P. When multiple vehicles C are present at the same date and time for the same section P, the moving average speed of these multiple vehicles C in the section P is obtained as the speed data for that date and time for that section P. Note that FIG. 3 shows an example in which the measurement interval is one minute. However, it is sufficient that measurement data is measured at each of multiple timings in a day, and the measurement interval is not limited to one minute.

Returning to FIG. 1, the explanation will be continued. The communication unit 22 communicates with an external information processing device, a traffic counter 28, and the like, via a known communication network such as a network. In this embodiment, the communication unit 22 sequentially receives measurement data measured by each of the multiple traffic counters 28 at each time from each of these multiple traffic counters 28, and stores it in the memory unit 20. Therefore, the memory unit 20 sequentially stores the latest traffic counter data for each date and time for each of the multiple sections P. That is, the memory unit 20 stores measurement data including speed data, traffic volume data, and road occupancy data for each time of each of the multiple days for each of the multiple sections P as traffic counter data.

The input unit 23 accepts various operations by the user. The input unit 23 is, for example, a keyboard, a mouse, a pointing device, a microphone, etc.

The display unit 29 displays various types of information. The display unit 29 is, for example, a display or a projection device. The display unit 29 and the input unit 23 may be integrated together to form a touch panel.

In addition, at least one of the memory unit 20 and the control unit 24 may be configured to be mounted on an external information processing device, such as a server device, connected via a network and the communication unit 22. In addition, at least one of the functional units included in the control unit 24, which will be described later, may be mounted on an external information processing device, such as a server device, connected to the control unit 24 via a network and the communication unit 22.

Next, the control unit 24 will be described in detail. The control unit 24 executes various information processes in the information processing device 10A.

The control unit 24 includes a collection unit 24A, a speed distribution prediction unit 24B, a predicted index prediction unit 24C, and a display control unit 24D.

At least one of the collection unit 24A, the speed distribution prediction unit 24B, the prediction index prediction unit 24C, and the display control unit 24D is realized, for example, by one or more processors. For example, each of the above units may be realized by having a processor such as a CPU (Central Processing Unit) execute a program, i.e., by software. Each of the above units may be realized by a processor such as a dedicated IC (Integrated Circuit), i.e., by hardware. Each of the above units may be realized by using a combination of software and hardware. When multiple processors are used, each processor may realize one of the units, or two or more of the units.

The collection unit 24A collects speed data of a vehicle C, which is an example of a moving body traveling in a section P. In this embodiment, the collection unit 24A collects speed data for each time of day for a vehicle C traveling on a road R in each of a plurality of sections P. The collection unit 24A collects speed data for each date and time for each of a plurality of sections P by reading traffic counter data stored in the memory unit 20.

The collection unit 24A may collect speed data for each time of one or more days in the set period for each section P. The set period may be a period consisting of one or more days that are consecutive in chronological order. For example, the set period may be one week, two weeks, etc., but is not limited to these. Below, an example will be described in which the set period is composed of multiple days that are consecutive in chronological order.

The speed distribution prediction unit 24B predicts the speed distribution based on the speed data collected by the collection unit 24A. In detail, in this embodiment, the speed distribution prediction unit 24B predicts the speed distribution of the speed data for each time for each of the multiple sections P based on the speed data collected by the collection unit 24A.

The prediction index prediction unit 24C predicts a prediction index of a road R, which is an example of a route, based on the speed distribution for each of the multiple sections P. The prediction index of the road R represents a prediction index related to the road R. In this embodiment, a form in which the prediction index of the road R is a prediction index of congestion of a vehicle C moving on the road R is described. Hereinafter, the prediction index of congestion is described as a congestion prediction index. That is, in this embodiment, the prediction index prediction unit 24C predicts a congestion prediction index. Also, in this embodiment, a prediction index of a prediction period after the set period is predicted based on the speed distribution. The prediction period may be a period after the set period, and is a period including one or more days. In this embodiment, a form in which the prediction index prediction unit 24C predicts a prediction index of a prediction date after the set date is described as an example. In this embodiment, the prediction index prediction unit 24C calculates a congestion probability for each time based on the speed distribution for each of the multiple sections P, and predicts the calculated congestion probability as a congestion prediction index, which is a prediction index for the prediction date.

The predicted date is a date that follows a certain set date. The set date may be any date on which measurement data is obtained. The set date may be a specific date in the past, or may be the most recent date on which measurement data is obtained (for example, today). The predicted date may be any date that follows one or several days after the set date. The set date and predicted date may be changeable as appropriate by the user's operation instruction via the input unit 23, etc.

Detailed explanation using diagrams.

Figures 4A to 4C are explanatory diagrams showing the prediction of the congestion forecast index.

Figure 4A is a diagram showing an example of speed data collection results 30 for a certain section P, for each of multiple days in a set period. The collection unit 24A collects speed data for each time of day for each of the multiple days, for each of the multiple sections P. Therefore, for example, the progress of speed data with respect to time on each of the multiple days in a certain section P is as shown in Figure 4A. Each of the multiple waveforms in Figure 4A represents the progress of speed data on each of the different days.

Figure 4B is a schematic diagram showing an example of a speed distribution 32 at a specific time obtained from the speed data of a certain section P shown in Figure 4A. Figure 4B shows 8 p.m. as an example of a specific time. In other words, the speed distribution 32 at 8 p.m. in a certain section P is as shown in Figure 4B.

The speed distribution prediction unit 24B predicts the speed distribution 32 for each time of day for each of the multiple sections P from the speed data collection results 30 for each section P. For this reason, the speed distribution prediction unit 24B predicts the speed distribution 32 as shown in FIG. 4B for each time of day for each of the multiple sections P. The speed distribution prediction unit 24B may predict the speed distribution 32 using a multi-peak distribution such as a Gaussian mixture model. In other words, the speed distribution 32 is the prediction result of what speeds are most frequent at what times for each section P.

The prediction index prediction unit 24C calculates the congestion probability 34 for each time based on the speed distribution 32 for each time for each of the multiple sections P. The congestion probability 34 is the probability that congestion will occur.

For example, the prediction index prediction unit 24C pre-sets a traffic jam speed that is considered to be a traffic jam. The traffic jam speed is, for example, 20 km/h, but is not limited to this. The traffic jam speed is an example of a speed that indicates a state that should be detected. The prediction index prediction unit 24C predicts a traffic jam probability 34, which is the probability that the speed distribution 32 will be below the traffic jam speed.

For example, assume that the speed distribution 32 shown in FIG. 4B is predicted as the speed distribution 32 at 20:00 for a certain section P. In this case, the forecast index prediction unit 24C calculates the proportion of the speed distribution 32 that is less than 20 km/h, which is the congestion speed, as the congestion probability 34 at 20:00 for that section P on the predicted day. For example, the forecast index prediction unit 24C calculates the congestion probability 34 by calculating the proportion of the integral value of the entire integral value of the speed distribution 32 that is less than the congestion speed of 20 km/h.

In the example shown in FIG. 4C, the predicted index prediction unit 24C calculates the percentage of days during a set period in which speed data below the congestion speed was collected at 20:00 for a certain section P as the congestion probability 34 for that section P at 20:00.

The forecast index prediction unit 24C calculates the congestion probability 34 for each of the other times in the same manner. The forecast index prediction unit 24C also calculates the congestion probability 34 for each of the other sections P in the same manner. In this manner, the forecast index prediction unit 24C calculates the congestion probability 34 for each of the multiple sections P at each time.

Then, in the forecast index prediction unit 24C of this embodiment, the traffic congestion probability 34 for each time calculated for each of the multiple sections P is treated as being common across future days. In other words, the forecast index prediction unit 24C treats the traffic congestion probability 34 for each of the multiple sections P for each time as being common even on forecast dates after the set date.

Therefore, the prediction index prediction unit 24C predicts the congestion probability 34 for each time of day calculated based on the speed distribution 32 for each of the multiple sections P as a congestion prediction index 36 for each of the multiple sections P for each time of day on the predicted date.

Figure 5 is a schematic diagram of an example of a traffic congestion prediction index 36 for each time of day on a predicted date for a certain section P. By performing the above process, the prediction index prediction unit 24C predicts the traffic congestion prediction index 36 for each time of day on a predicted date for each of multiple sections P.

Returning to FIG. 1, the explanation will be continued. The display control unit 24D displays the congestion prediction index 36, which is the prediction result by the prediction index prediction unit 24C, on the display unit 29. For example, the display control unit 24D displays a display screen 50 including the congestion prediction index 36 shown in FIG. 5 on the display unit 29. FIG. 5 shows an example in which the congestion prediction index 36 corresponding to each time is visualized as an index value of 11 levels from 0 to 100%, with the last digit rounded off. Note that the way in which the congestion prediction index 36 is expressed is not limited to an 11-level value.

For this reason, the display unit 29 displays the congestion prediction index 36 for each of the multiple sections P at each time on the predicted date.

Note that FIG. 5 shows an example of a display screen 50 including the congestion prediction index 36 for the predicted day for one section P. However, the display control unit 24D may display, on the display unit 29, a display screen 50 including the congestion prediction index 36 for the predicted day for each of a plurality of sections P. The display control unit 24D may also display, on the display unit 29, the congestion prediction index 36 for the section P input by the user through an operation instruction of the input unit 23.

Next, an example of the flow of information processing executed by the control unit 24 of the information processing device 10A of this embodiment will be described.

Figure 6 is a flowchart showing an example of the flow of information processing executed by the information processing device 10A of this embodiment.

The collection unit 24A collects speed data for each of the multiple sections P at each time of each of the multiple days in the set period (step S100).

The speed distribution prediction unit 24B predicts the speed distribution 32 for each time for each of the multiple sections P based on the speed data collected in step S100 (step S102).

The forecast index prediction unit 24C calculates the congestion probability 34 for each hour based on the speed distribution 32 for each hour of the multiple sections P predicted in step S102. The forecast index prediction unit 24C then predicts the calculated congestion probability 34 as a congestion forecast index 3 for each hour of the predicted day (step S104).

The display control unit 24D displays the traffic congestion forecast index 36 predicted in step S104 on the display unit 29. Then, this routine ends.

As described above, the control unit 24 of the information processing device 10A of this embodiment includes a collection unit 24A, a speed distribution prediction unit 24B, and a prediction index prediction unit 24C. The collection unit 24A collects speed data of a moving body such as a vehicle C moving in a section P, which is obtained by dividing a route such as a road R into a plurality of sections P. The speed distribution prediction unit 24B predicts a speed distribution 32 based on the collected speed data. The prediction index prediction unit 24C predicts a prediction index of the route based on the speed distribution 32.

In this way, the information processing device 10A of this embodiment predicts the speed distribution 32 for the section P where the speed data is measured. Then, the information processing device 10A predicts the predicted index of the route from the speed distribution 32, assuming that the predicted result of the speed distribution 32 is common across future days.

Therefore, the information processing device 10A of this embodiment can predict the predicted index of the route with high accuracy.

The information processing device 10A of this embodiment also predicts the congestion probability 34, which is predicted based on the speed distribution 32, as a congestion prediction index 36, which is a prediction index for the predicted day. Then, the information processing device 10A of this embodiment displays the congestion prediction index 36 on the display unit 29. Therefore, in addition to the above effects, the information processing device 10A of this embodiment can provide time periods when congestion is likely to occur in an easily checkable manner.

In addition, the information processing device 10A of this embodiment can predict the congestion prediction index 36 for a predicted date one day or multiple days after the set date. Therefore, by using the information processing device 10A of this embodiment, the user can easily and effectively consider travel plans for a specific predicted date, use of road R to avoid busy times, etc.

Second Embodiment
In this embodiment, a form will be described in which the congestion prediction index 36 is predicted by clustering the speed data.

Figure 7 is a schematic diagram showing an example of information processing device 10B of this embodiment. Information processing device 10B is an example of information processing device 10.

The information processing device 10B includes a memory unit 20, a communication unit 22, an input unit 23, a control unit 25, and a display unit 29. The memory unit 20, the communication unit 22, the input unit 23, the control unit 25, and the display unit 29 are connected so as to be able to communicate with each other. The information processing device 10B of this embodiment has the same configuration as the information processing device 10A of the above embodiment, except that it includes a control unit 25 instead of the control unit 24.

The control unit 25 executes various information processing operations in the information processing device 10B.

The control unit 25 includes a collection unit 25A, a speed distribution prediction unit 25B, a predicted index prediction unit 25C, a display control unit 25D, an upper limit prediction unit 25E, a determination unit 25F, and a clustering unit 25G.

At least one of the collection unit 25A, the speed distribution prediction unit 25B, the predicted index prediction unit 25C, the display control unit 25D, the upper limit prediction unit 25E, the identification unit 25F, and the clustering unit 25G is realized, for example, by one or more processors. For example, each of the above units may be realized by having a processor such as a CPU execute a program, i.e., by software. Each of the above units may be realized by a processor such as a dedicated IC, i.e., by hardware. Each of the above units may be realized by using a combination of software and hardware. When multiple processors are used, each processor may realize one of the units, or two or more of the units.

The collection unit 25A collects speed data for each time of day for each of the multiple sections P. Similar to the collection unit 25A, the collection unit 25A reads the traffic counter data stored in the memory unit 20 to collect speed data for each time of day on which measurements have been performed for each of the multiple sections P. Note that in this embodiment, the collection unit 25A collects speed data for each time of day on each of the multiple days during the set period for each section P. The definition of the set period has been described above, so description will be omitted.

The upper limit prediction unit 25E clusters a group of speed data for each time period collected for each of the multiple sections P on each of the multiple days of the set period into a number of clusters that optimizes the information criterion. The upper limit prediction unit 25E then predicts the upper traffic congestion speed limit for each section P according to the number of clusters.

In detail, the upper limit prediction unit 25E treats the speed data for each time of each of the multiple days in the set period collected by the collection unit 25A as one group regardless of time, and classifies the data into a specified number of clusters. The upper limit prediction unit 25E then calculates an information criterion when the data is classified into the specified number of clusters using a known method. The upper limit prediction unit 25E changes the specified number of clusters, and each time the specified number of clusters is changed, it repeatedly executes a series of processes of clustering into the specified number of clusters and calculating the information criterion.

Then, the upper limit prediction unit 25E determines the specified number of clusters for which the information amount criterion is optimal as the number of clusters. Furthermore, the upper limit prediction unit 25E clusters the group of speed data into the determined number of clusters. Furthermore, the upper limit prediction unit 25E predicts the upper limit of traffic congestion speed according to the determined number of clusters.

The upper limit prediction unit 25E executes this series of processes for each of the multiple sections P. For this reason, the upper limit prediction unit 25E predicts the number of clusters for which the information criterion is optimal and the upper congestion speed limit for each of the multiple sections P. Furthermore, the upper limit prediction unit 25E clusters the groups of speed data for each of the multiple days in the set period for each of the multiple sections P into the determined optimal number of clusters and assigns cluster labels.

The following will be explained in detail using figures. Figures 8A to 8C are explanatory diagrams for predicting upper traffic congestion speed limits. Figures 8A to 8C show an example in which the number of clusters for which the information criterion is optimal is 1.

Figure 8A is a diagram showing an example of speed data collection results 30 for a certain section P, for each of multiple days in a set period. The collection unit 25A collects speed data for each time of day for each of multiple days, for each of multiple sections P. Therefore, for example, the progress of speed data with respect to time for each of multiple days in a certain section P is, for example, as shown in Figure 8A.

Figure 8B is a schematic diagram showing an example of a speed distribution 32 at a specific time obtained from the speed data of a certain section P shown in Figure 8A. Figure 8B shows an example in which the specific time is 8 p.m. Therefore, the speed distribution 32 at 8 p.m. obtained from the speed data collection results 30 in a certain section P is as shown in Figure 8B.

The upper limit prediction unit 25E first sets an initial congestion speed upper limit as an initial condition. The initial congestion speed upper limit may be set to, for example, a speed slightly lower than the speed limit of road R in the corresponding section P. For example, assume that the speed limit of road R, which is a highway, is 50 km/h. In this case, the upper limit prediction unit 25E sets, for example, 48 km/h, which is slightly lower than 50 km/h, as the initial congestion speed upper limit. For example, the upper limit prediction unit 25E may set a speed that is 0.9 times the speed limit of road R in the corresponding section P as the initial congestion speed upper limit for the section P.

The upper limit prediction unit 25E may set an initial congestion speed upper limit for each section P according to the speed limit of the section P, or may set an initial congestion speed upper limit common to multiple sections P. When setting an initial congestion speed upper limit common to multiple sections P, a speed slightly lower than the lowest speed limit on the road R consisting of multiple sections P may be set as the initial congestion speed upper limit for each of these multiple sections P.

The upper limit prediction unit 25E treats the speed data collected by the collection unit 25A as one group for each section P, regardless of time. The upper limit prediction unit 25E then clusters the groups of speed data for each section P into an optimal number of clusters using a clustering method that can automatically determine the number of clusters, and determines the groups that belong to each cluster.

To automatically determine the number of clusters, for example, a Gaussian mixture model may be used. For example, the upper limit prediction unit 25E uses a Gaussian mixture model to determine the distance between the representative vector of each cluster and the group to which it belongs as the likelihood, and to determine the number of parameters of each distribution and the group as variables. The upper limit prediction unit 25E may then determine the number of clusters (k) using the Bayesian information criterion or the like using these likelihoods and variables. k is a number representing the number of clusters and is an integer equal to or greater than 1.

That is, the upper limit prediction unit 25E predetermines an upper limit for the number of clusters, changes the designated number of clusters by one at a time from 1 to the upper limit, and repeatedly performs classification of the designated number of clusters into clusters and calculation of the information amount criterion for each designated number of clusters. Then, the upper limit prediction unit 25E determines the designated number of clusters for which the information amount criterion is optimal as the number of clusters. Furthermore, the upper limit prediction unit 25E predicts the upper limit of traffic congestion speed according to the determined number of clusters.

As shown in FIG. 8B, when there is one peak in the speed distribution 32, the number of clusters determined is "1". That is, k=1. When the group of collected speed data is all equal to or greater than the initial congestion speed upper limit, the upper limit prediction unit 25E determines that no congestion is occurring in that section P, and predicts the initial congestion speed upper limit as the congestion speed upper limit.

Figure 8C is an explanatory diagram of prediction of the upper traffic congestion speed limit. For example, in Figure 8B, assume that the initial upper traffic congestion speed limit is 48.0 km/h. Also assume that all of the collected speed data groups are 48.0 km/h or higher, as shown in Figures 8B and 8C. In this case, the upper limit prediction unit 25E assumes that no congestion is occurring in that section P, and predicts that the initial upper traffic congestion speed limit of 48.0 km/h is the upper traffic congestion speed limit.

In addition, there are cases where the number of clusters determined by the upper limit prediction unit 25E is "1" and all of the groups of collected speed data are below the initial congestion speed upper limit. In this case, the upper limit prediction unit 25E predicts an infinite speed as the congestion speed upper limit, assuming that there is no transition period at a stable non-congestion speed.

The upper limit prediction unit 25E may predict an infinite speed as the upper limit of the traffic congestion speed when 1% or more of the collected speed data is below the initial traffic congestion speed upper limit. Also, when less than 1% of the collected speed data is below the initial traffic congestion speed upper limit, the upper limit prediction unit 25E may predict the initial traffic congestion speed upper limit as the upper limit of the traffic congestion speed. The value "1%" may be changeable as appropriate in response to a user's operation instruction on the input unit 23, etc.

Figures 9A to 9C are explanatory diagrams for predicting upper traffic congestion speed limits. Figures 9A to 9C show an example in which the number of clusters for which the information criterion is optimal is two or more.

Figure 9A is a diagram showing an example of speed data collection results 30 for a certain section P, for each of multiple days in a set period. The collection unit 25A collects speed data for each time of day for each of multiple days, for each of multiple sections P. Therefore, for example, the progress of speed data with respect to time for each of multiple days in a certain section P is, for example, as shown in Figure 9A.

Figure 9B is a schematic diagram showing an example of a speed distribution 32 at a specific time obtained from the speed data of a certain section P shown in Figure 9A. Figure 9B shows an example in which the specific time is 8 p.m. For example, the speed distribution 32 at 8 p.m. obtained from the speed data collection results 30 in a certain section P is as shown in Figure 9B.

As described above, the upper limit prediction unit 25E sets an initial congestion speed upper limit as an initial condition. For example, the upper limit prediction unit 25E will be described assuming that the initial congestion speed upper limit is set to 48 km/h, which is slightly lower than the speed limit of road R.

The upper limit prediction unit 25E treats the speed data collected by the collection unit 25A as one group for each section P, regardless of time. The upper limit prediction unit 25E then clusters the groups of speed data for each section P to an optimal number of clusters using a clustering method that can automatically determine the number of clusters, and determines the groups of speed data that belong to each cluster. The automatic determination of the number of clusters is the same as described above.

As shown in FIG. 9B, when there are multiple peaks in the speed distribution 32, the number of clusters to be determined is also multiple. That is, k>1. In this case, the upper limit prediction unit 25E calculates the slope Ai of the speed of the group of speed data belonging to each of the multiple clusters clustered to that number of clusters. The slope Ai is expressed by the following formula (1).

Ai = (Maximum speed - Minimum speed) / Number of speed data belonging to a cluster Equation (1)

In formula (1), Ai represents the velocity gradient, and i represents the number of clusters k. i is an integer equal to or greater than 1.

Then, the upper limit prediction unit 25E predicts the maximum speed among the speed data belonging to the cluster with the largest slope Ai as the upper traffic congestion speed limit.

For example, as shown in FIG. 9C, let us assume that the maximum speed in the speed data belonging to cluster a2, which has the maximum speed gradient Ai of SM among the three clusters a1, a2, and a3, is 63.0 km/h. In this case, the upper limit prediction unit 25E predicts 63.0 km/h as the upper traffic congestion speed limit.

The upper limit prediction unit 25E then executes these information processes for each of the multiple sections P. For this reason, the upper limit prediction unit 25E predicts the number of clusters for which the information amount criterion is optimal and the upper congestion speed limit for each of the multiple sections P. Furthermore, the upper limit prediction unit 25E clusters the group of speed data for each of the multiple sections P into clusters with the predicted number of clusters. Furthermore, the upper limit prediction unit 25E assigns a cluster label to each cluster.

Returning to FIG. 7 for further explanation, the identification unit 25F identifies, for each of the multiple sections P, a group of speed data belonging to a cluster that includes at least one piece of speed data that is less than the upper congestion speed limit, as a group of congestion speed data that is the congestion speed.

The identification unit 25F identifies a cluster that includes at least one speed data that is less than the upper traffic congestion speed limit predicted by the upper limit prediction unit 25E, from among the multiple clusters clustered by the upper limit prediction unit 25E. The identification unit 25F then identifies the speed data that belongs to the identified cluster as a group of traffic congestion speed data.

That is, the identification unit 25F collects, for each section P, speed data that is below the upper traffic congestion speed limit predicted by the upper limit prediction unit 25E from the speed data for each time of each of the multiple days in the set period as a traffic congestion speed data group.

The identification unit 25F also excludes speed data that is equal to or exceeds the upper congestion speed limit predicted by the upper limit prediction unit 25E from subsequent processing, treating it as a group of speed data that is not congested, i.e., that has progressed at a free flow. The identification unit 25F then stores the group of speed data that has progressed at a free flow in the storage unit 20 for each cluster clustered by the upper limit prediction unit 25E for each section P, in association with the cluster label of the cluster.

Then, the identification unit 25F outputs the group of traffic jam speed data identified for each of the multiple sections P to the clustering unit 25G. That is, the identification unit 25F selects a group of speed data that is less than the upper traffic jam speed limit from the collected speed data, and outputs the group to the clustering unit 25G.

The clustering unit 25G clusters the multiple speed data constituting the group of traffic congestion speed data for each of the multiple sections P into multiple traffic congestion clusters with an optimal number of clusters based on the information amount criterion. Then, the clustering unit 25G assigns traffic congestion cluster labels to the traffic congestion clusters.

In detail, first, the clustering unit 25G receives a group of traffic congestion speed data for each of the multiple sections P from the identification unit 25F. Then, the clustering unit 25G clusters the group of traffic congestion speed data received from the identification unit 25F into a number of clusters with an optimal information criterion. That is, the clustering unit 25G treats the group of traffic congestion speed data for each time of each of the multiple days of the set period as one group for each of the multiple sections P, regardless of the date and time. Then, like the upper limit prediction unit 25E, the clustering unit 25G clusters the group of traffic congestion speed data into multiple traffic congestion clusters with a number of clusters with an optimal information criterion, using a clustering method that can automatically determine the number of clusters. Then, the clustering unit 25G assigns a traffic congestion cluster label to the traffic congestion cluster as a cluster label.

That is, by the processing of the upper limit prediction unit 25E, the identification unit 25F, and the clustering unit 25G, the speed data collected for each of the multiple sections P on each of the multiple days of the set period is clustered into one or more non-congestion clusters and one or more congestion clusters. In addition, by assigning cluster labels to these clusters, a cluster label is assigned for each of the multiple sections P for the measurement date on which the speed data was measured.

In this embodiment, a cluster to which only non-congestion speed data belongs may be referred to as a non-congestion cluster. As described above, speed data belonging to a congestion cluster is a group of speed data that includes at least one congestion speed data. In this embodiment, when describing congestion clusters and non-congestion clusters collectively, they may be referred to simply as clusters. When describing cluster labels of non-congestion clusters and congestion cluster labels of congestion clusters collectively, they may be referred to simply as cluster labels.

The clustering unit 25G outputs the congestion clusters and congestion cluster labels to the speed distribution prediction unit 25B.

The speed distribution prediction unit 25B predicts the speed distribution 32 of the speed data for each time for each of the multiple sections P based on the speed data belonging to the congestion cluster received from the clustering unit 25G. That is, in this embodiment, the speed distribution prediction unit 25B predicts the speed distribution 32 for each congestion cluster for each of the multiple sections P.

The speed distribution prediction unit 25B may predict the speed distribution 32 for each cluster based on the speed data belonging to each of the non-congestion cluster and the congestion cluster. In this embodiment, an example will be described in which the speed distribution prediction unit 25B excludes the non-congestion cluster from the processing and predicts the speed distribution 32 for each cluster based on the congestion cluster.

In this embodiment, the speed distribution prediction unit 25B receives non-congestion clusters from the upper limit prediction unit 25E. The speed distribution prediction unit 25B uses the speed data belonging to the received non-congestion clusters to predict the speed distribution 32 of the speed data for each cluster.

The speed distribution prediction unit 25B uses the congestion clusters received from the clustering unit 25G to predict the speed distribution 32 for each cluster for each of the multiple sections P. The speed distribution prediction unit 25B may predict the speed distribution 32 for each cluster using a method similar to that used by the speed distribution prediction unit 24B.

For this reason, the speed distribution prediction unit 25B can predict the speed distribution 32 of the speed data corresponding to each time for each cluster for each of the multiple sections P, in the same manner as the speed distribution prediction unit 24B. In detail, the speed distribution prediction unit 25B can predict the speed distribution 32 for each time of the speed data belonging to each cluster for each of the non-congestion clusters and the congestion clusters.

The predicted index prediction unit 25C predicts the congestion predicted index 36 for the predicted day for each of the multiple sections P for each cluster in the same manner as the predicted index prediction unit 24C in the above embodiment.

In detail, the forecast index prediction unit 25C receives the speed distribution 32 predicted for each cluster of the multiple sections P from the speed distribution prediction unit 25B. The forecast index prediction unit 25C then calculates the congestion probability 34 for each time of day for each cluster of the multiple sections P in the same manner as the forecast index prediction unit 24C. The forecast index prediction unit 25C then predicts the calculated congestion probability 34 for each time of day as the congestion forecast index 36 for each cluster on the prediction date.

The display control unit 25D displays the congestion prediction index 36, which is the prediction result by the prediction index prediction unit 25C, on the display unit 29. Therefore, the display unit 29 displays the congestion prediction index 36 for each time of day for a certain section P on the prediction date for each cluster.

In other words, according to the information processing device 10B of this embodiment, it is possible to calculate the traffic congestion prediction index 36 for each time by cluster. In addition, by displaying the traffic congestion prediction index 36 for each time by cluster on the display unit 29, it is possible to visualize and provide the traffic congestion prediction index 36 for each time for each cluster.

For each cluster belonging to each section P, the center vector of the multiple speed data included in the cluster is assigned as a representative vector. For this reason, the display control unit 25D may display, for example, the representative vector with the lowest daily average speed among the representative vectors of the clusters belonging to each section P on the display unit 29 as the "least efficient pattern in that section P."

The forecast index prediction unit 25C may also calculate the proportion of the cluster with the lowest daily speed average as the frequency of traffic congestion in each section P. The display control unit 25D may further display the calculation results of the frequency of traffic congestion for each of the multiple sections P on the display unit 29.

Next, an example of the flow of information processing executed by the control unit 25 of the information processing device 10B of this embodiment will be described.

Figure 10 is a flowchart showing an example of the flow of information processing executed by the information processing device 10B of this embodiment.

The collection unit 25A collects speed data for each time of day for each of the set periods for each of the multiple sections P (step S200).

The upper limit prediction unit 25E clusters the speed data for each time of day for each of the multiple days in the set period collected in step S200 into a number of clusters that optimizes the information criterion, and predicts the upper limit of traffic congestion speed according to the number of clusters (step S202).

Figure 11 is a flowchart of the traffic jam speed limit prediction process in step S202.

The upper limit prediction unit 25E classifies the speed data for each time of day for the set period collected for each of the multiple sections P in step S200 into a specified number of clusters (step S300).

Next, the upper limit prediction unit 25E calculates the information criterion when classified into the specified number of clusters using a known method (step S302).

Then, the upper limit prediction unit 25E changes the designated number of clusters and executes the series of processes in steps S300 and S302 each time the designated number of clusters is changed.

Then, the upper limit prediction unit 25E determines the number of clusters specified by the optimal information amount criterion among the information amount criteria calculated by repeating the series of processes from step S300 to step S302 as the number of clusters (step S304).

Next, the upper limit prediction unit 25E predicts the upper traffic congestion speed limit according to the number of clusters determined in step S304 (step S306).

By the processing of steps S304 and S306, the number of clusters and the upper traffic congestion speed limit corresponding to the number of clusters are predicted for each of the multiple sections P and for each of the multiple days in the set period.

Then, the upper limit prediction unit 25E clusters the groups of speed data for each time period collected for each of the multiple sections P in step S200 into clusters with the number of clusters determined in step S304 (step S308). At this time, the upper limit prediction unit 25E also assigns a cluster label to each cluster. Then, this routine ends.

Returning to FIG. 10, the explanation will continue. By executing step S202, i.e., the process shown in FIG. 11, the upper limit prediction unit 25E predicts the upper traffic congestion speed limit for each of the multiple sections P and for each of the multiple days in the set period. In addition, the upper limit prediction unit 25E clusters the speed data collected in step S200 into the number of clusters determined in step S202 for each of the multiple sections P and for one or more of the set periods, and assigns cluster labels.

Next, the identification unit 25F identifies traffic congestion speed data from the speed data collected in step S200 (step S204). The identification unit 25F identifies a cluster that includes at least one piece of speed data that is below the upper traffic congestion speed limit predicted in step S02 from among the multiple clusters clustered in step S202. The identification unit 25F then identifies the speed data that belongs to the identified cluster as a group of traffic congestion speed data.

The clustering unit 25G clusters the group of traffic congestion speed data identified in step S204 into traffic congestion clusters (step S206).

Figure 12 is a flowchart of the process of clustering into congestion clusters in step S206.

The clustering unit 25G classifies the groups of traffic congestion speed data for each time period identified in step S204 for each of the multiple days in the set period into a specified number of clusters (step S400).

Next, the clustering unit 25G calculates the information criterion when classified into the specified number of clusters using a known method (step S402).

Then, the clustering unit 25G changes the designated number of clusters and executes the series of processes in steps S400 and S402 each time the designated number of clusters is changed.

Then, the clustering unit 25G determines the number of clusters to be the number of clusters specified by the optimal information criterion among the information criterions calculated by repeating the series of processes from step S400 to step S402 (step S404).

Then, the clustering unit 25G clusters the group of traffic congestion speed data identified in step S204 into traffic congestion clusters with the number of clusters determined in step S404 (step S406). At this time, the clustering unit 25G also assigns a traffic congestion cluster label to each of the traffic congestion clusters. Then, this routine ends.

Returning to FIG. 10, the explanation will continue. Through the processing of steps S202 to S206, the speed data collected for each of the multiple sections P on each of the multiple days of the set period is clustered into one or more non-congestion clusters and one or more congestion clusters. In addition, by assigning cluster labels to these clusters, each of the multiple sections P is assigned a cluster label for the date on which the speed data was measured.

Next, the speed distribution prediction unit 25B predicts the speed distribution 32 of the speed data for each time for each cluster for each of the multiple sections P based on the speed data belonging to the congestion clusters clustered in step S206 (step S208).

The predicted index prediction unit 25C uses the speed distribution 32 predicted in step S208 to predict the congestion prediction index 36 for the predicted day for each cluster of the multiple sections P (step S210).

The display control unit 25D displays the congestion prediction index 36 predicted in step S210 on the display unit 29 (step S212). Therefore, the congestion prediction index 36 for each time of day for a certain section P on the predicted date is displayed for each cluster on the display unit 29. Then, this routine ends.

As described above, in this embodiment, the upper limit prediction unit 25E clusters a group of speed data for each time period collected for each of the multiple sections P on each of the multiple days of the set period into a number of clusters that is optimal for the information criterion, and predicts the upper congestion speed limit according to the number of clusters. The identification unit 25F identifies, for each of the multiple sections P, a group of speed data belonging to a cluster that includes at least one piece of speed data that is less than the predicted upper congestion speed limit, as a group of congestion speed data that is the congestion speed. The clustering unit 25G clusters the multiple speed data constituting the group of congestion speed data for each of the multiple sections P into multiple congestion clusters with the optimal number of clusters based on the information criterion. The speed distribution prediction unit 25B predicts the speed distribution 32 of the speed data for each time period for each cluster, based on the speed data belonging to the congestion cluster.

For this reason, the prediction index prediction unit 25C predicts the congestion prediction index 36 for each cluster of the multiple sections P at each time of the prediction date.

The multiple sections P for which congestion is predicted may include sections P that repeat a specific pattern, such as a speed reduction that continues for a certain period of time and then resolves. The multiple sections P may also include sections P where no speed reduction occurs at all. In this way, the multiple sections P may include sections P that show a specific congestion tendency.

In the first embodiment described above, an example is shown in which the speed distribution 32 of the speed data for each time is calculated by mixing time periods with congestion and time periods without congestion, without distinguishing between the congestion trends of each of the multiple sections P. However, when managing congestion on road R, congestion analysis may be performed from a perspective according to congestion trends. For example, congestion analysis may be performed such as "What kind of speed change will occur in a certain section P when congestion is severe?".

In the information processing device 10B of this embodiment, for each of the multiple sections P, a group of speed data corresponding to each time is clustered into a number of clusters with an optimal information criterion based on the speed data for each time for each of the multiple days of the set period collected. Therefore, in the information processing device 10B of this embodiment, it is assumed that there is a congestion tendency pattern for congestion that occurs with a certain frequency in each of the multiple sections P, and the speed data can be clustered into multiple clusters according to the congestion tendency. In other words, in the information processing device 10B of this embodiment, the speed data for each of the multiple sections P is clustered into clusters for each group that represents a uniformly similar speed transition for each section P. Then, in the information processing device 10B, the congestion prediction index 36 is predicted for each speed data belonging to the cluster. Therefore, in the information processing device 10B of this embodiment, the congestion prediction index 36 can be predicted for each congestion tendency pattern, such as speed transition, for each of the multiple sections P.

Therefore, in addition to the effects of the above embodiments, the information processing device 10B of this embodiment can perform traffic congestion predictions with even higher accuracy.

Third Embodiment
In this embodiment, a predictor that outputs a congestion cluster label is trained and used to predict the congestion prediction index 36.

Figure 13 is a schematic diagram showing an example of information processing device 10C of this embodiment. Information processing device 10C is an example of information processing device 10.

The information processing device 10C includes a memory unit 20, a communication unit 22, an input unit 23, a control unit 26, and a display unit 29. The memory unit 20, the communication unit 22, the input unit 23, the control unit 26, and the display unit 29 are connected so as to be able to communicate with each other. The information processing device 10C of this embodiment has the same configuration as the information processing device 10A of the above embodiment, except that it includes a control unit 26 instead of the control unit 24.

The control unit 26 executes various information processing operations in the information processing device 10C.

The control unit 26 includes a collection unit 25A, a speed distribution prediction unit 25B, a predicted index prediction unit 25C, a display control unit 26D, an upper limit prediction unit 25E, an identification unit 25F, a clustering unit 25G, a learning unit 26H, a label output unit 26I, a cluster-specific predicted index prediction unit 26J, a reception unit 26K, and a search unit 26L.

At least one of the collection unit 25A, the speed distribution prediction unit 25B, the predicted index prediction unit 25C, the display control unit 25D, the upper limit prediction unit 25E, the identification unit 25F, the clustering unit 25G, the learning unit 26H, the label output unit 26I, the cluster-specific predicted index prediction unit 26J, the reception unit 26K, and the search unit 26L is realized, for example, by one or more processors. For example, each of the above units may be realized by having a processor such as a CPU execute a program, i.e., by software. Each of the above units may be realized by a processor such as a dedicated IC, i.e., by hardware. Each of the above units may be realized by using a combination of software and hardware. When multiple processors are used, each processor may realize one of the units, or two or more of the units.

The collection unit 25A, the speed distribution prediction unit 25B, the predicted index prediction unit 25C, the upper limit prediction unit 25E, the identification unit 25F, and the clustering unit 25G are the same as those in the second embodiment described above.

The learning unit 26H learns the predictor. The predictor is a learning model that receives speed data for each time of day on multiple days in the third measurement period before the set date as input, and outputs a congestion cluster label that belongs to a predicted date that is a day after the set date. The third measurement period will be defined later.

Figure 14 is an explanatory diagram of an example of predictor learning.

The learning unit 26H uses, for each of the multiple sections P, a group of traffic congestion speed data belonging to each of the multiple traffic congestion clusters derived by the clustering unit 25G for each of the multiple days included in the first measurement period before the specific day, as an explanatory variable. That is, the learning unit 26H uses the group of traffic congestion speed data derived by the clustering unit 25G using the speed data for the first measurement period as input data when learning the predictor.

The specific date may be any of the dates on which measurement data is available. FIG. 14 shows an example in which measurement data is available from April 1, 2015 to April 7, 2017. FIG. 14 also shows an example in which the specific dates are from April 8, 2015 to March 30, 2017.

The first measurement period is the period before the specific date, during which measurement data is obtained. In FIG. 14, the one-week period from the day before the specific date to the day one week before is shown as the first measurement period. Specifically, an example is shown in which the first measurement period for the specific date of April 8, 2015 is the seven days from April 1, 2015 to April 7, 2015. Similarly, for the other specific dates, the one-week period from the day before the specific date to the day one week before is shown as the first measurement period.

14 also shows a group of congestion speed data for the first measurement period for each of a plurality of sections P. In FIG. 14, P-N _P -N _L -B is an example of identification information for each section P. N _P and N _L are each an integer between 001 and 999. The learning unit 26H may acquire the speed data for the first measurement period for each section P from the collection unit 25A.

The learning unit 26H also uses the congestion cluster label of the second measurement date, which is after the specific date, as the objective variable during learning. The second measurement date is N days (N is an integer equal to or greater than 1) after the specific date. That is, the learning unit 26H uses the congestion cluster label of each of the one or more congestion clusters derived by the clustering unit 25G using the speed data of the second measurement date as output data of the predictor during learning. In other words, the learning unit 26H uses the congestion cluster label of the second measurement date as the objective variable.

Then, the learning unit 26H uses these explanatory variables and objective variables to learn a predictor for the second measurement date for each interval P. In other words, the learning unit 26H learns a predictor that corresponds to each day N days after the specific date for each of the multiple intervals P.

To train a predictor, an algorithm that can learn categorical data, such as random forest, as explanatory and objective variables can be used.

Figure 14 shows April 9, 2015 to March 31, 2017 as second measurement dates N days after the specific dates, which correspond to the specific dates April 8, 2015 to March 30, 2017.

Figure 14 also shows an example in which the traffic congestion cluster label one day after the specific date for section P with identification information "P-003-010-B" is used as the objective variable. Figure 14 also shows an example in which the traffic congestion speed data for the first measurement period one week prior to the specific date for each section P with identification information other than "P-003-010-B" is used as the explanatory variable.

In this way, the learning unit 26H learns a predictor for each of a plurality of specific dates that differ from each other for each of a plurality of sections P. Therefore, the learning unit 26H can learn a plurality of predictors for outputting each congestion cluster label N days after a specific date for each of a plurality of sections P.

For example, assume that April 8, 2015 to March 30, 2017 are set as specific dates. In this case, the learning unit 26H can use two years' worth of speed data from April 1, 2015 to March 31, 2017 to learn multiple predictors that output congestion cluster labels for the days following each of the specific dates (April 9, 2015 to March 31, 2017). That is, in this case, the learning unit 26H can learn, for each of the multiple sections P, a predictor for outputting congestion cluster labels belonging to each predicted date from one day to two years after the set date.

Returning to FIG. 13 for further explanation, the label output unit 26I uses the predictor trained by the learning unit 26H to output the congestion cluster label belonging to the predicted date, which is each of the multiple days after the set date.

First, the label output unit 26I identifies a corresponding predictor for each of the multiple intervals P. That is, the label output unit 26I identifies a predictor corresponding to each of the multiple intervals P, N days after the specific date, as a predictor corresponding to each of the multiple days after the set date, which are prediction dates.

Then, the label output unit 26I inputs to each of the identified predictors the speed data for each time of the multiple days included in the third measurement period before the set date in the corresponding section P. As described above, the set date is the previous day predicted, such as a specific day in the past or the most recent day for which measurement data is available (for example, today). The third measurement period is a day before the set date that is part of the period for which measurement data is available. For example, the third measurement period is one week before the set date, but is not limited to this.

For example, assume that the set date is April 8, 2017. In this case, the label output unit 26I inputs the speed data for each time of day included in the third measurement period, which is from April 1, 2017 to April 7, 2017, to the predictor. The label output unit 26I can obtain the speed data from the collection unit 25A.

For each of the multiple sections P, speed data for each time of the third measurement period is input to a predictor corresponding to each of the N days after the set date, and each predictor outputs a congestion cluster label belonging to each of the predicted dates, which are each of the days N days after the set date.

For example, assume that the set date is April 8, 2017. Also assume that predictors for 1 day to 31 days later have been trained for each section P as predictors that output congestion cluster labels N days after a specific date. In this case, the label output unit 26I inputs the speed data for each time of day included in the third measurement period, April 1, 2017 to April 7, 2017, to the predictors corresponding to each of the days 1 day later to 31 days later. Through this process, the label output unit 26I can output a congestion cluster label for each predicted day, with each day in the month from April 9, 2017 to May 8, 2017 being the predicted day.

In this way, the label output unit 26I uses a predictor to output, for each of the multiple sections P, a congestion cluster label that belongs to each of the predicted dates, which are N days after the set date.

The cluster-specific predicted index prediction unit 26J predicts the traffic congestion predicted index 36 belonging to the traffic congestion cluster identified by the traffic congestion cluster label output from the label output unit 26I as the traffic congestion predicted index 36 for each traffic congestion cluster on each predicted date.

That is, the cluster-specific predicted index prediction unit 26J predicts the congestion prediction index 36 for the predicted day for each cluster using the congestion cluster label output from the label output unit 26I and the congestion prediction index 36 for each cluster received from the predicted index prediction unit 25C.

In detail, the cluster-specific predicted index prediction unit 26J identifies a cluster identified by the congestion cluster label output from the label output unit 26I for each predicted date N days later from the clusters received from the predicted index prediction unit 25C. Then, the cluster-specific predicted index prediction unit 26J predicts the congestion prediction index 36 belonging to the identified cluster as the congestion prediction index 36 for each congestion cluster for the corresponding predicted date N days later.

The reception unit 26K receives search conditions including at least one of the section P, the predicted date, the search time period, and the range of the congestion prediction index 36. For example, the user operates the input unit 23 to input at least one of the section P for which a congestion prediction is desired, the predicted date which is the search target date for which a congestion prediction is desired, the search time period which is the time period for which a congestion prediction is desired, and the range of the congestion prediction index 36. Through these operations of the input unit 23 by the user, the reception unit 26K receives search conditions including at least one of the section P, the predicted date, the search time period, and the range of the congestion prediction index 36.

The search unit 26L searches for at least one of the section P, the congestion cluster, the time period to which the congestion cluster belongs, and the congestion prediction index 36 for the time period that matches the received search criteria, based on the congestion prediction index 36 for each congestion cluster on the predicted date predicted by the cluster-specific prediction index prediction unit 26J.

The display control unit 26D displays, on the display unit 29, at least one of the congestion prediction index 36 for each congestion cluster of the multiple sections P predicted by the cluster-specific prediction index prediction unit 26J and the search results of the search unit 26L.

Figure 15A is an explanatory diagram of an example of search criteria 40. Figure 15B is a schematic diagram of an example of a search result display screen 42.

For example, assume that the reception unit 26K receives, as search conditions 40, a section P from "Kuuko Chuo" to "Oi Minami", which is an entrance/exit point of the expressway (road R), forecast dates from January 11, 2021 to January 15, 2021, and a congestion forecast index of 30% or less.

In this case, for example, the display control unit 26D displays the search result display screen 42 shown in FIG. 15B on the display unit 29. The display screen 42 displays, for example, that the congestion prediction index 36 for time period 42A on each of the predicted dates of January 11, 2021 to January 15, 2021 in section P from "Airport Center" to "Oi Minami" is 30% or less. In addition, for time period 42B, it displays that the congestion prediction index 36 exceeds 30%.

Next, an example of the flow of information processing executed by the control unit 26 of the information processing device 10C of this embodiment will be described.

Figure 16 is a flowchart showing an example of the flow of information processing executed by the information processing device 10C of this embodiment.

In the information processing device 10C, the processing of steps S500 to S506 is executed in the same manner as steps S200 to S206 (see FIG. 10) in the information processing device 10B of the second embodiment.

In detail, the collection unit 25A collects speed data for each time of day for each of the multiple sections P for each of the multiple days in the set period (step S500). The upper limit prediction unit 25E clusters the speed data for each time of day for each of the multiple days in the set period collected in step S500 into a number of clusters with an optimal information criterion, and predicts the upper traffic speed limit according to the number of clusters (step S502).

Next, the identification unit 25F identifies a group of traffic congestion speed data from the speed data collected in step S500 (step S504). The clustering unit 25G clusters the group of traffic congestion speed data identified in step S504 into a traffic congestion cluster (step S506).

By the processing of steps S502 to S506, the speed data collected for each of the multiple sections P on each of the multiple days of the set period is clustered into one or more non-congestion clusters and one or more congestion clusters. In addition, by assigning cluster labels to these clusters, a cluster label is assigned for each of the multiple sections P for the date on which the speed data was measured.

Next, the learning unit 26H learns the predictor (step S508). The learning unit 26H learns the predictor using the congestion cluster labels of the congestion clusters clustered in step S506.

The label output unit 26I inputs the speed data for each time of the multiple days included in the third measurement period before the set date to the predictor trained in step S508. Then, the label output unit 26I outputs the congestion cluster label belonging to each predicted date for each section P as the output from the predictor corresponding to each predicted date, which is each day N days after the set date (step S510).

On the other hand, the speed distribution prediction unit 25B predicts the speed distribution 32 of the speed data for each time for each cluster for each of the multiple sections P based on the speed data belonging to the congestion clusters clustered in step S506 (step S512).

The predicted index prediction unit 25C uses the speed distribution 32 predicted in step S512 to predict the congestion prediction index 36 for the predicted day for each cluster of the multiple sections P (step S514).

The cluster-specific predicted index prediction unit 26J identifies the congestion prediction index 36 belonging to the congestion cluster identified by the congestion cluster label output in step S510 from the congestion prediction index 36 for each cluster predicted in step S514. Then, the cluster-specific predicted index prediction unit 26J predicts the identified congestion prediction index 36 as the congestion prediction index 36 for each congestion cluster for each prediction date (step S516).

Next, the reception unit 26K determines whether or not search conditions have been received from the input unit 23 (step S518). If the determination in step S518 is negative (step S518: No), the process proceeds to step S520.

In step S520, the display control unit 26D displays on the display unit 29 the congestion prediction index 36 for each congestion cluster of each of the multiple sections P predicted in step S516 (step S520). Then, this routine ends.

On the other hand, if the answer is YES in step S518 (step S518: Yes), the process proceeds to step S522. In step S522, the search unit 26L searches for at least one of the section P, the congestion cluster, the time period to which the congestion cluster belongs, and the congestion prediction index 36 for the time period that matches the search criteria received in step S522 (step S522). The display control unit 26D displays the search result of step S522 on the display unit 29 (step S524). Then, this routine ends.

As described above, in the information processing device 10C of this embodiment, the learning unit 26H learns a predictor by inputting a group of traffic congestion speed data belonging to each of the multiple congestion clusters derived for the first measurement period before the specific date for each of the multiple sections P, and outputting the traffic congestion cluster label for the second measurement date after the specific date. The label output unit 26I inputs the speed data for each time of the third measurement period before the set date for each of the multiple sections P to the predictor, and outputs the traffic congestion cluster label belonging to the prediction date, which is a day after the set date, as the output of the predictor. The cluster-specific prediction index prediction unit 26J predicts the traffic congestion prediction index belonging to the traffic congestion cluster identified by the output traffic congestion cluster label as the traffic congestion prediction index for each traffic congestion cluster on the prediction date.

In this manner, the information processing device 10C of this embodiment uses a trained predictor to predict the congestion cluster label on the predicted date for each of the multiple sections P. Therefore, the information processing device 10C of this embodiment can predict the congestion cluster label that belongs to each predicted date N days after the set date, for example, for one month, such as 1 day to 31 days after the set date.

Therefore, the information processing device 10C of this embodiment can easily predict the congestion prediction index 36 for each of multiple sections P on multiple prediction dates after the set date.

Therefore, in addition to the effects of the above embodiments, the information processing device 10C of this embodiment can perform traffic congestion predictions with even higher accuracy.

In addition, in the information processing device 10C of this embodiment, the search unit 26L searches for at least one of the section P, the congestion cluster, the time period to which the congestion cluster belongs, and the congestion prediction index 36 for the time period that matches the received search criteria. Then, the display control unit 26D displays the search results on the display unit 29.

The speed distribution 32 at each time is composed of the speed data belonging to each of the congestion clusters to which a congestion cluster label has been assigned. Therefore, the search unit 26L can use the speed distribution 32 to search for at least one of the section P, the congestion cluster, the time period to which the congestion cluster belongs, and the congestion prediction index 36 for that time period that matches the search criteria. The display control unit 26D then displays the search results on the display unit 29, making it possible to easily visualize the congestion prediction index 36 for the predicted day and provide it to the user.

That is, according to the information processing device 10C of this embodiment, the learning unit 26H predicts the congestion cluster label of each of the multiple sections P on a predicted date for a certain future period. Then, the information processing device 10C can predict the congestion prediction index 36 for each time on a future predicted date. Then, the information processing device 10C can set the section P to be searched, the search date, the search range of the congestion prediction index 36 (e.g., less than 10%, 80% or more), etc. as search conditions, search for a time period of the section P that matches the search conditions, and display it on the display unit 29. Therefore, as described above, the display unit 29 displays, for example, the search result display screen 42 shown in FIG. 15B.

Therefore, in addition to the effects of the above-described embodiment, the information processing device 10C of this embodiment can also provide traffic congestion prediction results according to the search conditions desired by the user.

Fourth Embodiment
In this embodiment, a form will be described in which the clustering unit 25G calculates the similarity for each section P based on the similarity between groups of congestion cluster labels.

Figure 17 is a schematic diagram showing an example of an information processing device 10D of this embodiment. The information processing device 10D is an example of the information processing device 10.

The information processing device 10D includes a memory unit 20, a communication unit 22, an input unit 23, a control unit 27, and a display unit 29. The memory unit 20, the communication unit 22, the input unit 23, the control unit 27, and the display unit 29 are connected so as to be able to communicate with each other. The information processing device 10D of this embodiment has the same configuration as the information processing device 10A of the above embodiment, except that it includes a control unit 27 instead of the control unit 24.

The control unit 27 executes various information processing operations in the information processing device 10D.

The control unit 27 includes an upper limit prediction unit 25E, a specification unit 25F, a clustering unit 25G, a section similarity calculation unit 27M, a generation unit 27N, and a display control unit 27D.

At least one of the collection unit 25A, upper limit prediction unit 25E, identification unit 25F, clustering unit 25G, section similarity calculation unit 27M, and display control unit 27D is realized, for example, by one or more processors. For example, each of the above units may be realized by having a processor such as a CPU execute a program, i.e., by software. Each of the above units may be realized by a processor such as a dedicated IC, i.e., by hardware. Each of the above units may be realized by using a combination of software and hardware. When multiple processors are used, each processor may realize one of the units, or two or more of the units.

The collection unit 25A, upper limit prediction unit 25E, identification unit 25F, and clustering unit 25G are the same as those in the second embodiment. Although not shown, the information processing device 10D may further include a speed distribution prediction unit 25B, a predicted index prediction unit 25C, and a display control unit 25D, as in the second embodiment.

The section similarity calculation unit 27M identifies a group of congestion cluster labels of the congestion cluster for each of the multiple sections P. Then, the section similarity calculation unit 27M calculates the section similarity between the sections P for each of the multiple days in the set period using the similarity between the identified groups of congestion cluster labels.

For example, the section similarity calculation unit 27M identifies, for each of the multiple sections P, a feature represented by a group of congestion cluster labels generated by the clustering unit 25G. The feature is represented, for example, by a sequence of congestion cluster labels.

Then, the section similarity calculation unit 27M calculates the similarity between sections P for each of the multiple set periods using the similarity between the features represented by the identified group of congestion cluster labels.

In detail, first, the section similarity calculation unit 27M calculates the dissimilarity in a brute-force manner for the series of congestion cluster labels for each of the multiple sections P to be compared. Note that the congestion cluster labels assigned by clustering are categorical, and there is no relationship of magnitude between the label values. For this reason, the section similarity calculation unit 27M calculates the similarity between the sections P from the series of congestion cluster labels belonging to each of the multiple sections P to be compared.

To calculate the similarity, it is sufficient to apply the Adjusted Rand Index or Normalized Mutual Information, which can define the similarity based only on the matching rate of the congestion cluster labels of the mutual sections P.

The generating unit 27N generates a section similarity map that shows the sections P in different display forms for each group of sections P whose similarity is equal to or greater than a threshold value.

First, the generation unit 27N uses the Adjusted Rand Index and Normalized Mutual Information used to calculate the similarity to convert the similarity between sections P into a dissimilarity vector whose value becomes 0 (zero) the closer the patterns of the series of congestion cluster labels are.

Then, the generating unit 27N converts the dissimilarity vector calculated for each section P into a distance between sections P, etc. For the distance, for example, cosine similarity or the like may be used.

Then, the generation unit 27N applies non-metric MDS to the transformed distance matrix as input, transforming it into a space in which the higher the similarity, the closer the labels are to each other, according to the similarity of the congestion cluster labels of each section P in the low-dimensional space.

Then, in a space constructed using MDS (Multi-Dimensional Scaling) until the number of groups reaches a preset number, the generation unit 27N constructs similarity groups by disconnecting paths in order from the longest path until the partial network reaches the number of groups, starting from a state in which the paths of the section network are fully connected.

Through these processes, the generation unit 27N classifies multiple sections P into groups of sections P whose similarity is equal to or greater than a threshold value.

Then, for each group of sections P whose similarity is equal to or greater than a threshold, the generating unit 27N generates a section similarity map in which each section P is displayed in a different display form.

The display control unit 27D displays the section similarity map generated by the generation unit 27N on the display unit 29.

Figure 18 is a schematic diagram showing an example of a section similarity map 44. In this manner, the section similarity map 44 is displayed on the display unit 29, in which the sections P are displayed in a different color or other display form for each similar section P.

Next, an example of the flow of information processing executed by the control unit 27 of the information processing device 10D of this embodiment will be described.

Figure 19 is a flowchart showing an example of the flow of information processing executed by the information processing device 10D of this embodiment.

In the information processing device 10D, the processing of steps S600 to S606 is executed in the same manner as steps S200 to S206 (see FIG. 10) in the information processing device 10B of the second embodiment.

In detail, the collection unit 25A collects speed data for each time of day for each of the multiple sections P for each of the multiple days in the set period (step S600). The upper limit prediction unit 25E clusters the speed data for each time of day collected in step S600 into a number of clusters with an optimal information criterion for each of the multiple days in the set period, and predicts the upper limit of traffic congestion speed according to the number of clusters (step S602).

Next, the identification unit 25F identifies a group of traffic congestion speed data from the speed data collected in step S600 (step S604). The clustering unit 25G clusters the group of traffic congestion speed data identified in step S604 into a traffic congestion cluster (step S606).

By the processing of steps S602 to S606, the speed data collected for each of the multiple sections P on each of the multiple days of the set period is clustered into one or more non-congestion clusters and one or more congestion clusters. In addition, by assigning cluster labels to these clusters, a cluster label is assigned for each of the multiple sections P for the date on which the speed data was measured.

Next, the section similarity calculation unit 27M calculates the similarity between sections P (step S608). Next, the generation unit 27N generates a section similarity map 44 using the similarity calculated in step S608 (step S610).

The display control unit 27D displays the section similarity map 44 generated in step S610 on the display unit 29 (step S612). Then, this routine ends.

As described above, in the information processing device 10D of this embodiment, a section similarity map 44 is generated using the similarity between groups of congestion cluster labels belonging to each of multiple sections P, and is displayed on the display unit 29.

Therefore, in addition to the effects of the above-described embodiments, the information processing device 10D of this embodiment can visualize and provide groups of sections P with similar congestion times.

Next, an example of the hardware configuration of the information processing device 10 of the above embodiment will be described.

Figure 20 is an example of a hardware configuration diagram of an information processing device 10, which is each of the information processing devices 10A to 10D of the above embodiment.

The information processing device 10 of the above embodiment includes a control device such as a CPU 86, storage devices such as a ROM (Read Only Memory) 88, a RAM (Random Access Memory) 90, and a HDD (Hard Disk Drive) 92, an I/F unit 82 that interfaces with various devices, an output unit 80 that outputs various information such as output information, an input unit 94 that accepts operations by the user, and a bus 96 that connects each unit, and has a hardware configuration that utilizes a normal computer.

In the information processing device 10 of the above embodiment, the CPU 86 reads a program from the ROM 88 onto the RAM 90 and executes it, thereby realizing each of the above-mentioned parts on the computer.

The programs for executing the above processes executed by the information processing device 10 of the above embodiment may be stored in the HDD 92. Also, the programs for executing the above processes executed by the information processing device 10 of the above embodiment may be provided in advance in the ROM 88.

The program for executing the above-mentioned processes executed by the information processing device 10 of the above-mentioned embodiment may be stored in an installable or executable file format on a computer-readable storage medium such as a CD-ROM, CD-R, memory card, DVD (Digital Versatile Disk), or flexible disk (FD) and provided as a computer program product. The program for executing the above-mentioned processes executed by the information processing device 10 of the above-mentioned embodiment may be stored on a computer connected to a network such as the Internet and provided by downloading the program via the network. The program for executing the above-mentioned processes executed by the information processing device 10 of the above-mentioned embodiment may be provided or distributed via a network such as the Internet.

Although an embodiment of the present invention has been described above, the above embodiment is presented as an example and is not intended to limit the scope of the invention. This new embodiment can be implemented in various other forms, and various omissions, substitutions, and modifications can be made without departing from the gist of the invention. This embodiment and its modifications are included in the scope and gist of the invention, and are included in the scope of the invention and its equivalents described in the claims.

10, 10A, 10B, 10C, 10D Information processing device 24A, 25A Collection unit 24B Speed distribution prediction unit 24C, 25C Prediction index prediction unit 24D, 25D, 26D, 27D Display control unit 25E Upper limit prediction unit 25F Identification unit 25G Clustering unit 26H Learning unit 26I Label output unit 26J Cluster-specific prediction index prediction unit 26K Reception unit 26L Search unit 27M Section similarity calculation unit 27N Generation unit 29 Display unit

Claims (15)

A route is divided into a plurality of sections, and a group of speed data for each of the plurality of sections for a moving object moving in the plurality of sections is clustered into a number of clusters that is optimal for an information criterion ;
For each of the plurality of sections, a group of speed data belonging to a cluster including at least one speed data item that is less than a target speed upper limit according to the number of clusters is identified as a group of target speed data items that are a target speed;
Predicting a speed distribution for each time based on the identified group of target speed data;
predicting a prediction index of the route for a prediction period after a set period based on the speed distribution;
An information processing device comprising a control unit. The control unit is
a collection unit that collects speed data of the moving object moving through the section;
a speed distribution prediction unit that predicts the speed distribution based on collected speed data;
a predicted index prediction unit that predicts the predicted index based on the speed distribution;
Equipped with
The information processing device according to claim 1 . The collecting unit includes:
Collecting speed data for each hour of the day for each of a number of days;
The speed distribution prediction unit
predicting the velocity distribution of the velocity data over time;
The information processing device according to claim 2 . The predicted index prediction unit
predicting the prediction index for a prediction date after a set date based on the speed distribution;
The information processing device according to claim 2 . The target speed is a traffic jam speed,
The predicted index prediction unit
Calculating a congestion probability for each hour of the day based on the speed distribution of the speed data predicted for each hour of the day for each of a plurality of the sections, and predicting the congestion probability as the prediction index for the prediction day.
The information processing device according to claim 4. The target speed is a traffic jam speed,
The control unit is
an upper limit prediction unit that clusters a group of speed data for each time period collected for each of the one or more days of the set period for each of the multiple sections into the number of clusters that is optimal for an information criterion, and predicts a traffic jam upper limit speed, which is the target upper limit speed, according to the number of clusters;
A determination unit that determines, for each of the plurality of sections, a group of speed data belonging to a cluster including at least one speed data item that is less than the upper traffic congestion speed limit as a group of traffic congestion speed data items that are the target speed data;
A clustering unit that clusters a plurality of speed data constituting the group of congestion speed data for each of the plurality of sections into a plurality of congestion clusters having an optimal cluster number based on an information amount criterion;
Equipped with
The speed distribution prediction unit
predicting the speed distribution of the speed data for each cluster at each time based on the speed data belonging to the congestion cluster;
The information processing device according to claim 4. a learning unit that learns a predictor using, as an input, a group of the congestion speed data belonging to each of the plurality of congestion clusters, derived for a first measurement period before a specific date, for each of the plurality of sections, and using, as an output, a congestion cluster label of the congestion cluster for a second measurement date after the specific date;
a label output unit that inputs speed data for each time of a third measurement period before the set date to the corresponding predictor for each of the multiple sections, and outputs, as an output of the predictor, the congestion cluster label that belongs to the predicted date that is a date after the set date;
A cluster-specific predicted index prediction unit that predicts the predicted index belonging to the congestion cluster identified by the output congestion cluster label as the predicted index for each congestion cluster on the prediction date;
Equipped with
The information processing device according to claim 6. a reception unit that receives search conditions including at least one of the section, the prediction date, the search target time period, and the range of the prediction index;
a search unit that searches for at least one of the section, the congestion cluster, the time period to which the congestion cluster belongs, and the predicted index for the time period, which match the received search criteria, based on the predicted index for each congestion cluster on the predicted date predicted by the cluster-specific predicted index prediction unit;
A display control unit that displays the search results on a display unit;
The information processing device according to claim 7 . a section similarity calculation unit that specifies a group of the congestion cluster labels of the congestion cluster for each of the plurality of sections, and calculates a similarity between the sections using a similarity between the specified groups of the congestion cluster labels;
a generation unit that generates a section similarity map for each group of the sections whose similarity is equal to or greater than a threshold, the section similarity map being displayed in a different display form from each other;
a display control unit that displays the segment similarity map on a display unit;
The information processing device according to claim 7 . The information processing device according to any one of claims 1 to 9, wherein the route is a road. The information processing device according to any one of claims 1 to 10, wherein the moving body is a vehicle. The information processing device according to any one of claims 2 to 9, wherein the prediction index prediction unit predicts a congestion prediction index for the route as the prediction index. a control unit which divides a route into a plurality of sections, clusters a group of speed data for each of the sections at each time point into a number of clusters with an optimal information criterion , identifies, for each of the plurality of sections, a group of speed data belonging to a cluster including at least one speed data below a target speed upper limit according to the number of clusters as a group of target speed data that is a target speed, predicts a speed distribution for each time point based on the identified group of target speed data, and predicts a prediction index of the route for a prediction period after a set period based on the speed distribution;
A display control unit that displays the predicted index on a display unit;
An information processing device comprising: 1. A computer-implemented information processing method, comprising:
A route is divided into a plurality of sections, and a group of speed data for each of the plurality of sections for a moving object moving in the plurality of sections is clustered into a number of clusters that is optimal for an information criterion ;
For each of the plurality of sections, a group of speed data belonging to a cluster including at least one speed data item that is less than a target speed upper limit according to the number of clusters is identified as a group of target speed data items that are a target speed;
Predicting a speed distribution for each time based on the identified group of target speed data;
predicting a prediction index of the route for a prediction period after a set period based on the speed distribution;
Information processing methods. A step of dividing a route into a plurality of sections, and clustering a group of speed data for each of the plurality of sections at each time point into a number of clusters that is optimal for an information criterion ;
identifying, for each of the plurality of sections, a group of speed data belonging to a cluster including at least one speed data item that is less than a target speed upper limit according to the number of clusters, as a group of target speed data items that are a target speed;
A step of predicting a speed distribution for each time based on the identified group of target speed data;
predicting a prediction index of the route for a prediction period after a set period based on the speed distribution;
An information processing program for causing a computer to execute the above.

Images