JP2020135231A - Traffic demand prediction device and traffic demand prediction system - Google Patents

Abstract

To enable prediction of a traffic demand even in a region where traffic data containing movement results necessary for traffic demand prediction is hard to obtain.SOLUTION: A traffic demand prediction device in accordance with an embodiment of the present invention includes: a model memory unit that stores a basic model for use in predicting a traffic demand for each section; a prediction condition acquisition unit that acquires prediction conditions including at least a prediction object section and a moving route in the prediction object section; a prediction model selection unit that has a feature of selecting a basic model, which meets the prediction object conditions, from the model memory unit, and that, when no basic model meeting the prediction conditions is obtained in the prediction object section, selects the basic model in a section different from the prediction object section as a prediction model for the prediction object section; and a traffic demand calculation unit that calculates a traffic demand, which complies with the prediction conditions using the selected prediction model.SELECTED DRAWING: Figure 1

Description

The present invention relates to a traffic demand forecasting device and a traffic demand forecasting system for predicting traffic demand in a target section or a target area.

There are various means of transportation for road transportation by car and urban transportation such as public transportation such as buses and railroads. Proper operation of transportation means promoting smooth movement of people and reducing the economic burden and environmental burden of society as a whole. In order to properly operate the means of transportation, it is necessary to control congestion and improve the moving environment such as roads with respect to the generated / concentrated traffic volume (the index "OD traffic volume" that indicates where the traffic originates and where it concentrates). , And measures such as setting an appropriate transportation fee are necessary. Therefore, the importance of the traffic demand forecasting technology for predicting the demand for each means of transportation, the route of use of the means of transportation, and the system for the OD traffic volume is increasing.

As a traffic demand forecasting technique, a method such as a daily aggregation model that targets individual traffic behavior and an aggregation model that handles collective behavior for zones is generally used. For example, in the aggregate model, a four-step estimation method is used, which is divided into stages such as prediction of traffic generation / concentration, prediction of distributed traffic, distribution of transportation means, and surplus of distributed traffic. As human behavior data used in the aggregate model, person trip survey data in the survey target area is generally used.

The person trip survey data is the daily traffic data (also called trip data) of each person such as the starting point (O) of the traffic behavior, the ending point (D), the purpose of travel, the means of transportation, and the travel time. Is. The traffic data is the record of traffic behavior to be surveyed (hereinafter referred to as "movement record"). Since the person trip survey data can be obtained as trips (behavioral information) that have been traveled using a plurality of means of transportation, it is possible to examine the division of means of transportation.

However, since the person trip survey data is a questionnaire survey, it is currently conducted as needed or as often as once every 10 years. Therefore, it is difficult to immediately reflect changes in human behavior due to new construction of urban structures such as facilities and road railroads and changes in their operating hours in the model, and to grasp changes in traffic demand. In particular, in emerging countries, population growth and traffic development in urban areas are remarkable, and human behavior changes in a short period of time, so it is difficult to keep track of the actual situation through questionnaire surveys.

In recent years, the use of traffic data obtained from automatic toll collection systems (Electronic Toll Collection: ETC), traffic IC cards, probe car systems, probe person systems, etc. is expected. The probe car system is a system that uses a vehicle as a sensor, acquires driving data such as traveling speed and position from a traveling vehicle, and generates road traffic information. By using the probe car, it is possible to acquire position measurement data (hereinafter referred to as "probe information") such as the position, time, speed, and route actually passed by the vehicle.

In addition, the probe person system is not limited to roads by acquiring movement data such as the movement speed and position of people from mobile terminals such as mobile phones and smartphones equipped with satellite positioning systems such as GPS (Global Positioning System). , It is a system that generates information on the movement route of people such as in railways and facilities. Here, the movement data (position, time, etc.) of a person obtained by the probe car system or the probe person system is collectively referred to as probe data. The probe data is a movement locus measured at a predetermined cycle, and the movement route can be estimated.

A method of predicting traffic demand by modeling traffic behavior between ODs based on the movement routes of a large number of users and the movement record of transportation means using such traffic data is being studied. Since traffic demand fluctuates over time depending on the weather, travel environment such as travel time and travel charges, and user conditions such as travel purpose, age, and gender, a forecast model that responds to such fluctuations is handled. ..

However, a considerable computational load is required to learn the dependency between the traffic day and the moving environment from a huge amount of traffic data to generate a prediction model, and to predict the traffic demand by the prediction model. Therefore, Patent Document 1 discloses a technique for efficiently predicting traffic demand by using multitask learning.

Japanese Unexamined Patent Publication No. 2015-125775

The technique described in Patent Document 1 is a method of predicting traffic demand by learning movement data of each segment (moving section, corresponding to OD of the embodiment of the present invention) to generate a prediction model. Measure the similarity of changes in traffic volume (time-series data) with respect to the travel time of each segment, cluster similar time-series, and perform multitask learning on each cluster to associate it with the travel environment. Generate a prediction model with. Then, for the segment to be predicted, a mobile environment is given to the prediction model of the corresponding cluster, and the traffic demand is predicted. Clustering is performed among segments with similar time series of traffic demand, and a forecast model is generated for each cluster to predict traffic demand. Therefore, forecast of traffic demand in a huge number of segments reflecting changes in the mobile environment can be predicted in real time. It is effective to execute.

However, in the technique of Patent Document 1, in the segment where the traffic data cannot be observed or the segment where the traffic data which can be an important factor for forecasting the traffic demand is not observed, clustering due to the similarity of the traffic data (time series data) can be performed. Multitasking learning can be difficult and it is difficult to predict traffic demand.

The present invention has been made in view of the above-mentioned situation, and an object of the present invention is to make it possible to predict traffic demand even in a section where it is difficult to obtain traffic data including travel records necessary for forecasting traffic demand. To do.

In order to solve the above problems, the traffic demand prediction device of one aspect of the present invention includes a model storage unit that stores a basic model used for forecasting traffic demand for each section, and at least a prediction target section and the prediction target section. It has a function to select a basic model corresponding to the prediction condition from the prediction condition acquisition unit and the model storage unit to acquire the prediction condition including the movement route in the inside, and the basic model corresponding to the prediction target section could not be obtained. In this case, the prediction model selection unit that selects the basic model of the section different from the prediction target section as the prediction model of the prediction target section, and the traffic demand calculation that calculates the traffic demand according to the prediction conditions using the selected prediction model. It has a department.

According to at least one aspect of the present invention, a basic model for predicting traffic demand is prepared in advance for each section, and for a prediction target section for which a basic model does not exist, the basic model for other sections is substituted for traffic demand. Predict. As a result, it is possible to forecast traffic demand for more areas, including sections where it is difficult to obtain traffic data.
Issues, configurations and effects other than those described above will be clarified by the description of the following embodiments.

It is a figure which shows the traffic demand forecasting system to which the traffic demand forecasting apparatus which concerns on 1st Embodiment of this invention is applied. It is a figure which shows the hardware configuration example of the traffic demand forecasting apparatus which concerns on 1st Embodiment of this invention. It is a flowchart which shows the procedure example of the traffic demand calculation processing which is executed by the traffic demand forecasting apparatus in response to the request from the terminal apparatus which concerns on 1st Embodiment of this invention. It is a flowchart which shows the procedure example of the process of selecting the basic model of another area by the prediction model selection part in the traffic demand calculation process (step S15) of FIG. It is a flowchart which shows the procedure example of the process which calculates and modifies the traffic demand prediction model which concerns on 1st Embodiment of this invention. It is a flowchart which shows the procedure example of the model correction processing by the model correction part in the calculation and correction processing (step S35) of the prediction model of FIG. It is a flowchart explaining the procedure example of the process of calculating a new basic model from a plurality of basic models which concerns on 1st Embodiment of this invention. It is a figure for demonstrating the feature of the basic model newly created from the plurality of basic models which concerns on 1st Embodiment of this invention. It is a flowchart for demonstrating the procedure example of the process which calculates the general-purpose model which reflected the similarity of the traffic environment of a plurality of regions as the prediction model of the traffic demand which concerns on 1st Embodiment of this invention. It is a figure for demonstrating the basic model of the area group which concerns on 1st Embodiment of this invention. It is a figure which shows the traffic demand forecasting system to which the traffic demand forecasting apparatus which concerns on 2nd Embodiment of this invention is applied.

Hereinafter, examples of embodiments for carrying out the present invention (hereinafter, referred to as “embodiments”) will be described with reference to the accompanying drawings. In the present specification and the accompanying drawings, components having substantially the same function or configuration are designated by the same reference numerals, and duplicate description will be omitted.

<1. First Embodiment>
[Overall configuration of traffic demand forecasting system]
First, the first embodiment of the present invention will be described. FIG. 1 shows a traffic demand forecasting system to which the traffic demand forecasting device according to the first embodiment is applied.

The traffic flow evaluation system 10 shown in FIG. 1 includes a traffic demand forecasting device 1, an operation information providing device 2, a moving information providing device 3, and a terminal device 4. The devices are connected to each other so as to be able to communicate with each other via a communication network 5 such as the Internet.

The traffic demand forecasting device 1 receives a forecasting condition for predicting traffic demand from an external device (for example, a terminal device 4), and determines the traffic demand for each transportation means or traveling route corresponding to the forecasting condition. Calculate continuously at time intervals. For example, as forecast conditions, designation of intercity or travel section (OD) to execute traffic demand forecast and OD traffic volume, travel environment such as weather, forecast date and time, travel fee, user information such as travel purpose, age, gender, etc. including. In the present embodiment, it is assumed that the prediction condition includes at least information about a moving section and a moving route within the moving section. The traffic demand forecasting device 1 mutually exchanges necessary information with the operation information providing device 2, the moving information providing device 3, and the terminal device 4 via the communication network 5.

The operation information providing device 2 provides information related to an operation plan such as public transportation. More specifically, the operation information providing device 2 has a transport capacity such as an operation route (operation route) of public transportation such as a bus or a railroad, an operation schedule, and the number of seats and a passenger capacity of a moving body via a communication network 5. Information is provided to the traffic demand forecasting device 1.

The movement information providing device 3 is obtained from the OD traffic volume generated for each OD, the transportation means used for the movement of each section, the movement route, the ETC, the traffic IC card, the probe car system, the probe person system, and the like. Traffic data such as movement results are provided to the traffic demand forecasting device 1.

Here, the OD traffic volume is the traffic volume from the traffic generation zone (O zone) to the concentrated zone (D zone), which is aggregated or estimated at predetermined time intervals. In addition, the movement route information has a starting point at an arbitrary point in the O zone and an arbitrary point in the D zone as a destination, and the road section from the starting point to the destination point (a road link connecting nodes in the map data, or It is the data that enumerates the nodes).

The terminal device 4 is equipped with a traffic flow evaluation application for acquiring forecast data of traffic demand between cities or sections to be predicted from the traffic demand forecasting device 1 and predicting and evaluating future traffic flow. .. The traffic flow evaluation application is composed of, for example, a Web application. The terminal device 4 specifies forecast conditions to the traffic demand forecast device 1 and acquires a traffic demand forecast result by using a traffic flow evaluation application. The traffic flow evaluation application accepts the designation of the environmental model of the city (section) targeted for traffic demand forecast, and determines the position and travel time of moving objects such as automobiles, buses, trains, and pedestrians in a predetermined time zone. The traffic flow of the designated city (section) is predicted by continuously calculating at the time interval ti of. Hereinafter, in the present specification, "intercity or section" may be abbreviated as "section". Unless otherwise specified, the term "section" means "intercity or section".

Not limited to the above, a device that provides information effective for forecasting traffic demand, such as a traffic information providing device (not shown) that provides traffic information such as traffic congestion and regulations between cities or sections, is connected to the communication network 5. It may be provided to the traffic demand forecasting device 1.

As shown in FIG. 1, the traffic demand forecasting device 1 includes a communication interface 101, a reception processing unit 102, a prediction model selection unit 103, a traffic demand calculation unit 104, a transmission processing unit 105, and a prediction model calculation unit 106. , A model diagnosis unit 107, a model correction unit 108, a basic model storage unit 110, a traffic data storage unit 111, and a regional data storage unit 112.

In the traffic demand forecasting device 1, at least one of the reception processing unit 102, the prediction model selection unit 103, the traffic demand calculation unit 104, the transmission processing unit 105, the prediction model calculation unit 106, the model diagnosis unit 107, and the model correction unit 108 is , A program (for example, a program stored in a program memory such as ROM 21b) may be realized.

Further, the basic model storage unit 110, the traffic data storage unit 111, and the regional data storage unit 112 are provided with a plurality of storage devices (the types of storage devices such as semiconductor memory and hard disk may be different). Each data may be stored in the corresponding storage device. Further, the storage area may be divided in one or a plurality of storage devices, and each data may be stored in each storage area. In FIG. 1, the flow of data in the traffic demand forecasting device 1 is shown by a solid line, and the reference of information (providing information) is shown by a broken line.

In the present embodiment, in the traffic demand forecasting device 1, the OD traffic volume of the automobile using the road is acquired from the movement information providing device 3 via the communication network 5, and the road demand for the use of the automobile is targeted. The embodiment for predicting traffic demand will be described as an example, but the embodiment is not limited thereto. Regarding public transportation, the usage demand for the entire public transportation and the usage demand for each system or flight may be predicted based on the OD traffic volume acquired from the transportation IC card collection system (not shown).

Similarly, indoors, based on the OD traffic volume (number of migrants) at two predetermined points of moving objects such as humans indoors, the demand for using elevators and carts moving indoors, and by route and facility. You may try to predict the usage demand of. In each case, the OD traffic volume and the movement information are acquired from the movement information providing device 3 and the operation information providing device 2.

In FIG. 1, the basic model storage unit 110 stores a basic model used for forecasting traffic demand for each section. The traffic data storage unit 111 stores traffic data indicating a movement record of an arbitrary section. The traffic data includes travel sections, routes, dates of use, departure times, travel times, travel charges, operation data (operation status) of transportation means, traffic conditions, user information such as purpose of travel, age, and gender. In the present embodiment, it is assumed that the traffic data includes at least information about a moving section and a moving route within the moving section.

The regional data storage unit 112 stores regional data showing regional characteristics such as transportation network and household information at least for each region. For example, regional characteristics are determined by features such as the number of nodes and link lengths that make up the transportation network and movement routes of each region, the angle of intersection between links, the width of roads, the range of population distribution, and the population density.

The communication interface 101 controls communication between the traffic demand forecasting device 1 and the communication network 5, and transmits / receives data to / from the operation information providing device 2, the mobile information providing device 3, and the terminal device 4 via the communication network 5. Do.

The reception processing unit 102 (an example of the prediction condition acquisition unit) interprets the traffic demand forecast request and the prediction condition received from the terminal device 4 via the communication interface 101, and outputs the interpreted result to the prediction model selection unit 103. ..

The prediction model selection unit 103 searches the basic model storage unit 110 based on the prediction conditions input from the reception processing unit 102, and selects a basic model for predicting the traffic demand between cities or sections corresponding to the prediction conditions. However, it has a function of setting it as a prediction model in the traffic demand calculation unit 104. Prediction conditions include at least between cities or sections to be predicted and their routes. Further, when the prediction model selection unit 103 cannot obtain the prediction model of the intercity or section corresponding to the prediction condition from the basic model storage unit 110, the basic model storage unit 110 is the basic between other sections or other cities. Select a model. In addition to the basic model prepared in advance for other sections and between other cities, the basic model is calculated based on one or more basic models between other sections and other cities and stored in the basic model storage unit 110. There is also.

The traffic demand calculation unit 104 reads the prediction model selected by the prediction model selection unit 103 and the prediction conditions interpreted by the reception processing unit 102. Then, the traffic demand calculation unit 104 sets the prediction conditions read into the selected prediction model, and calculates the traffic demand for each transportation means or route that can be used to move between cities or sections to be predicted (). Prediction), and the prediction result is output to the transmission processing unit 105.

For example, the traffic demand is the OD traffic volume (traffic demand) generated for each OD from the O zone to the D zone of traffic, the distribution value for each moving body between the ODs, the distribution value for each vehicle type, and the movement route. Information on the distribution value and its movement route. For example, the distribution value for each moving body is the ratio of the transportation volume of automobiles and trains (the number of users of automobiles and trains). Further, for example, the distribution value by vehicle type is the ratio of the number of vehicles by vehicle type such as ordinary cars and buses. Further, for example, the distribution value for each movement route is the ratio of the transportation amount of the moving body moving on the plurality of movement routes connecting the ODs.

The transmission processing unit 105 determines the number of users by means of transportation, the number of users by route or system, the number of vehicles, etc. for each section as the traffic demand forecast result of the section or city output by the traffic demand calculation unit 104. , The process of transmitting to the terminal device 4 via the communication interface 101 is executed. The section here is assumed to be between ODs, but the number of users is not limited to the number of users associated with each route or system between ODs. For example, it may be the number of users or the number of vehicles associated with the road section (road link) constituting each route, or the number of users associated with the stop section constituting each system.

The prediction model calculation unit 106 reads traffic data from the traffic data storage unit 111 for the ODs to be calculated by the prediction model. Then, the prediction model calculation unit 106 learns on what basis the traffic user moving between the ODs selects the transportation means and the route based on the read traffic data, and obtains the traffic demand from the traffic behavior. Generate a prediction model to predict. The variables that make up the prediction model are selected from the usage date and time obtained as traffic data, the travel environment such as travel time and travel charges, and user attributes. Generally, as a model, since the variable items that are candidates are multi-items, a multi-item logit regression model, a multiple regression model, or the like is used. The calculated prediction model is output to the model diagnosis unit 107. As will be described later, the basic model of the section or the city can be calculated based on the basic model of the other section or the city.

The model diagnosis unit 107 compares the characteristics of the prediction model between predetermined cities or sections calculated by the prediction model calculation unit 106 (for example, prediction variables and weight parameters) with the characteristics of the basic model stored in the basic model storage unit 110. , Diagnose whether a valid model has been generated.

Further, when the model diagnosis unit 107 diagnoses as a valid model, the model diagnosis unit 107 outputs the calculated prediction model to the basic model storage unit 110 as a basic model between the cities or sections. On the other hand, when the model diagnosis unit 107 diagnoses the model as having no validity (the validity is lower than a predetermined reference), the model diagnosis unit 107 outputs the calculated prediction model to the model correction unit 108.

The model correction unit 108 reads the basic model for predicting the traffic demand from the basic model storage unit 110, and the model diagnosis unit 107 reads the prediction model determined to be invalid (low) based on the read basic model. Fix it. Then, the model modification unit 108 outputs the modified prediction model to the basic model storage unit 110 as a basic model between the corresponding cities or sections.

[Hardware configuration of traffic demand forecaster]
FIG. 2 is a block diagram showing a hardware configuration example of the traffic demand forecasting device 1.
As shown in FIG. 2, the traffic demand forecasting device 1 can be configured by a computer 20 (information processing device), and has a CPU 21a, a ROM 21b, a RAM 21c, and a display unit (display device) connected to the system bus 26, respectively. A 22, an operation unit 24, a storage unit 25, and a communication interface 101 are provided. Each part in the computer 20 is connected so as to be able to send and receive data to and from each other via the system bus 26.

The control unit 21 is composed of the CPU 21a, the ROM 21b, and the RAM 21c. The control unit 21 is used as an example of a computer in a narrow sense that controls the operation of each unit in the traffic demand forecasting device 1. The CPU 21a is an arithmetic processing unit that reads a program code of software that realizes each function according to the present embodiment from the ROM 21b and executes it to control each part and perform various operations. Instead of the CPU 21a, another arithmetic processing unit such as an MPU (Micro Processing Unit) may be used.

The ROM 21b is used as an example of a non-volatile memory (recording medium), and stores programs, data, and the like necessary for the CPU 21a to operate. The RAM 21c is used as an example of a volatile memory, and variables, parameters, and the like generated during the arithmetic processing by the CPU 21a are temporarily written.

For example, a liquid crystal display monitor is used in the display unit 22, and the menu screen, the result of processing performed by the computer 20, and the like are displayed. For example, a keyboard, a mouse, a touch panel, or the like is used as the operation unit 24, and the user can perform predetermined operation inputs and instructions using the operation unit 24.

The storage unit 25 is a non-volatile large-capacity storage device (an example of a recording medium), and stores programs, programs such as an OS (Operating System), data, and the like. For example, in the case of the traffic demand forecasting device 1, the basic model storage unit 110, the traffic data storage unit 111, and the regional data storage unit 112 are composed of the storage unit 25. As the storage unit 25, for example, an HDD, an SSD (Solid State Drive), an optical disk, a magneto-optical disk, a CD-ROM, a CD-R, a magnetic tape, a memory card, or the like is used.

For example, a NIC (Network Interface Card), a modem, or the like is used as the communication interface 101, and various data are transmitted and received to and from each device via a communication network 5 such as a LAN to which terminals are connected or a dedicated line. To do.

The operation information providing device 2, the moving information providing device 3, and the terminal device 4 also have the same configuration as the computer 20. For example, in the operation information providing device 2, the control unit 21 controls the operation information providing process, and in the movement information providing device 3, the control unit 21 controls the traffic demand providing process. Further, in the terminal device 4, the control unit 21 performs a process of evaluating the traffic flow (traffic flow evaluation application). Each part in the device is selected according to the function and purpose of use of each device. For example, in the traffic demand forecasting device 1, the operation information providing device 2, and the moving information providing device 3, the display unit 22 and the operation unit 24 can be deleted.

[Traffic demand calculation processing]
Next, the process of calculating the traffic demand used for the traffic flow evaluation in the traffic flow evaluation system 10 will be described.

FIG. 3 is a flowchart showing a procedure example of the traffic demand calculation process executed by the traffic demand forecasting device 1 in response to the request from the terminal device 4. First, the terminal device 4 starts communication with the traffic demand forecasting device 1 by executing the traffic evaluation application. Then, the terminal device 4 uses the forecast target city or OD (sometimes abbreviated as "section"), the predicted date and time, the OD (occurrence / concentration) traffic volume, and each means of transportation and route. Prediction conditions such as travel time, distance, and usage fee are transmitted to the traffic demand prediction device 1 to request the prediction of traffic demand (S1). In the present specification, the intercity or section to be predicted may be referred to as a “target area”.

The traffic demand forecasting device 1 receives the request and the forecasting condition from the terminal device 4 via the communication interface 101 in the reception processing unit 102, and interprets the request content (S11). Next, the prediction model selection unit 103 searches the basic model storage unit 110 based on the prediction conditions (S12), and determines whether or not there is a basic model of the corresponding intercity or section (target section) (S13).

Here, when the prediction model selection unit 103 determines that there is a basic model of the corresponding intercity or section (YES in S13), the prediction model selection unit 103 selects it as a prediction model for predicting the traffic demand between the cities or sections to be predicted. (S14). On the other hand, when the prediction model selection unit 103 does not have a basic model between the corresponding cities or sections (NO in S13), the prediction model selection unit 103 reads the basic model between other cities or sections from the basic model storage unit 110, and the city to be predicted It is selected as a forecast model for predicting traffic demand between sections or sections (S15). The details of the process in step S15 will be described later.

Next, the traffic demand calculation unit 104 sets a necessary prediction variable (hereinafter, may be abbreviated as “variable”) from the prediction conditions in the prediction model selected by the prediction model selection unit 103 (S16). Next, the traffic demand calculation unit 104 calculates the traffic demand between the cities or the section based on the prediction model in which the variable is set (S17). Then, the transmission processing unit 105 transmits the prediction result of the traffic demand to the terminal device 4 via the communication interface 101 (S18).

The terminal device 4 acquires a prediction result from the traffic demand prediction device 1 (S2), executes a traffic flow evaluation application using the prediction result, and compares the prediction result with the current road capacity, transportation capacity, and the like. Understand traffic issues and execute applications that implement traffic measures and transportation plans (S3).

As described above, the traffic demand forecasting device 1 of the present embodiment has a function of selecting a basic model corresponding to the prediction condition from the basic model storage unit 110, and when the basic model corresponding to the prediction target section cannot be obtained. A prediction model selection unit 103 that selects a basic model of a section different from the prediction target section as a prediction model of the prediction target section is provided.

That is, in the present embodiment, a basic model for predicting traffic demand is prepared in advance for each section, and for a prediction target section for which a basic model does not exist, the basic model for other sections is substituted to predict traffic demand. This makes it possible to forecast traffic demand for more areas, including sections where the basic model does not exist and sections where it is difficult to obtain traffic data.

[Selection of basic model in other regions]
Next, in the traffic demand calculation process (step S15) of FIG. 3, a process of selecting a basic model of another region by the prediction model selection unit 103 will be described.

FIG. 4 is a flowchart showing a procedure example of the prediction model selection process by the prediction model selection unit 103. FIG. 4 shows a process of selecting a basic model of an area having a similar traffic environment as the target area as a prediction model in the process (S15) of selecting a basic model of another area as a prediction model of the target area.

The prediction model selection unit 103 indicates each area corresponding to each basic model stored in the basic model storage unit 110 when the basic model corresponding to the intercity or section of the prediction target cannot be obtained (NO in S13). Regional data representing regional characteristics such as the transportation network, population distribution, and average movement route of the above are read from the regional data storage unit 112 (S21).

Then, the prediction model selection unit 103 calculates features such as the number of nodes, link lengths, intersection angles, average widths, population distribution ranges, and average densities that constitute the transportation network and movement routes in each region (S22). ), Evaluate the similarity of the feature quantities in each region (S23). Then, the prediction model selection unit 103 selects a basic model of an area similar to the target area as a prediction model of the target section (S24). It can be said that the feature quantity representing the regional characteristics indicates the traffic environment of the region. As a method for calculating the similarity between regions executed in step S23, a general similarity evaluation method such as a cosine scale or a Kullback suspected distance is used. After the process of step S24 is completed, the process proceeds to step S16.

As described above, the traffic demand forecasting device 1 of the present embodiment is based on the regional data of each region of the regional data storage unit 112 when the basic model of the forecast target section cannot be obtained from the basic model storage unit 110. Prediction model selection unit 103 that compares the regional characteristics of the prediction target section with the regional characteristics of the section different from the prediction target section, and selects the basic model of the area including the sections with similar regional characteristics as the prediction model of the prediction target section. To be equipped. The prediction model selection unit 103 uses at least the transportation network and route in each region as the feature quantity of the regional characteristic, and based on the feature quantity of the regional characteristic, the similarity between the prediction target section and the section different from the prediction target section. To evaluate.

According to such a configuration, by selecting a basic model of another area having a similar traffic environment as a prediction model of a prediction target area, it is possible to further improve the prediction accuracy of traffic demand in a section where the basic model does not exist. is there.

By the way, in the process shown in FIG. 4, the basic model of the area similar to the target area is selected as the prediction model of the target area, but the present invention is not limited to this example. For example, as will be described later, in the prediction model calculation unit 106, statistics such as the number of adoptions (cumulative distribution) of prediction variables constituting the basic model among a plurality of ODs existing in each region and the distribution of weight parameters of each prediction variable Therefore, a prediction variable and a range of each weight parameter thereof may be set for each region to newly generate a basic model and store it in the basic model storage unit 110 (see FIG. 5 described later). In that case, in step S24, a newly calculated basic model of another area similar to the target area is selected as the basic model of the target area.

Further, in the process shown in FIG. 4, the prediction model selection unit 103 calculates the regional feature amount, but the present invention is not limited to this example. When the prediction model calculation unit 106 generates the prediction model, the feature amount of the area where the prediction target section exists may be calculated and stored in the area data storage unit 112 (details will be described with reference to FIG. 5). ).

[Prediction model calculation and correction processing]
FIG. 5 is a flowchart showing an example of a procedure for calculating and correcting a traffic demand forecast model in the traffic demand forecasting device 1. FIG. 5 shows a process of calculating a prediction model from traffic data when there is traffic data but not a prediction model, and appropriately modifying the calculated prediction model by the corresponding basic model.

The prediction model calculation unit 106 obtains traffic data (transportation means and route, date and time of use, travel time, travel charge, operation status, traffic status, user attributes, etc.) that is the travel record between the traffic data storage unit 111 and the target OD. Read from the traffic data storage unit 111 (S31).

Next, the prediction model calculation unit 106 learns from the traffic data on what basis the traffic user moving between the target ODs selects the transportation means and the route, and calculates the prediction model (S32). In this embodiment, for example, a multinomial logit regression model or a multiple regression model is assumed as a forecast model for predicting traffic demand. The predictive variable Xi, which is the basis (factor) for selecting a means of transportation and a route, is selected from the date and time of use and the travel time acquired as traffic data. In the following, the prediction variable Xi may be abbreviated as "variable Xi".

For example, it is assumed that the prediction model between predetermined ODs is constructed by the multiple regression model shown in the following equation (1). The traffic volume V (Oa, Da) moving between the ODs by the transportation means a (or the route a) is obtained by the sum of products of the predictive variable Xia of the transportation means a (or the route a) and its weight parameter Wi. In step S32, a large number of traffic data read from the traffic data storage unit 111 are learned, a variable Xi that is an important factor in selecting a transportation means and a route when moving between the ODs is selected, and a weight parameter Wi Is calculated.

V (Oa, Da) = W ₁・ X ₁ a + W ₂・ X ₂ a + ・ ・ ・ + Wi ・ Xia
・ ・ ・ ・ (1)

V (Oa, Da): Traffic volume moving from O zone to D zone by transportation means a (or route a) Xia: Value of variable item i when using transportation means a (or route a) Wi: Prediction item i weight parameter

Next, in the model diagnosis unit 107, the basic model of another section in the city (region) including the target OD is read from the basic model storage unit 110 (S33), and the prediction model calculated in step S32 is read. Compare with (S34). Then, the model diagnosis unit 107 diagnoses whether or not the calculated prediction model is appropriate (whether or not it is a valid model) (S35).

For example, the model diagnosis unit 107 predicts whether the prediction item i (for example, usage fee) constituting the basic model is selected as the prediction variable of the calculated prediction model (corresponding to the condition (1) in FIG. 8). It is evaluated whether the range of each weight parameter Wi of the model does not exceed the distribution range of the weight parameter Wi of the basic model (corresponding to the condition (2) of FIG. 8). Judgment on predictive variables may be limited to those important to the prediction rather than all predictor variables.

Then, when the calculated prediction model satisfies these conditions, the model diagnosis unit 107 determines that a valid model has been generated (YES in S35), and targets the calculated prediction model. It is stored in the basic model storage unit 110 as a basic model between ODs (S37). This calculated basic model is used when forecasting traffic demand between target ODs.

If any of the range of the variable item i and the weight parameter Wi of the calculated prediction model does not satisfy the above conditions, the model diagnosis unit 107 determines that a valid model has not been generated (NO in S35).

Next, the model correction unit 108 receives the diagnosis result of the model diagnosis unit 107, corrects the calculated prediction model with the above basic model (S36), and stores the corrected prediction model as the basic model between the target ODs. It is stored in the unit 110 (S37). After the process of step S37 is completed, the process of this flowchart ends.

The basic model read in step S33 and the basic model used for correction in step S36 may be the same or different. For example, in step S36, the basic model of the region with higher similarity may be used.

[Modification of prediction model by basic model of similar area]
FIG. 6 is a flowchart showing a procedure example of the model correction process by the model correction unit 108 in step S36 in the prediction model calculation and correction process of FIG. FIG. 6 shows a model modification process (S36) executed by the model modification unit 108 that more reflects the characteristics of the area (of the traffic environment).

When the calculated prediction model is not evaluated as a valid model (NO in S35), the model correction unit 108 includes the area where the predetermined OD interval exists and each basic stored in the basic model storage unit 110. Regional data representing regional characteristics such as the transportation network, population distribution, and average movement route of each region corresponding to the model is read from the regional data storage unit 112 (S41).

Then, the model correction unit 108 calculates the feature quantities such as the number of nodes, the link length, the intersection angle, the average width, the range of population distribution, and the average density that constitute the transportation network and the movement route in each area (S42). , Evaluate the similarity of the feature quantities in each region (S43). Then, the model correction unit 108 modifies the prediction model between predetermined ODs with a basic model of an area similar to the area where the predetermined ODs exist (S44). After the process of step S44 is completed, the process proceeds to step S37. As a method for calculating the similarity of regions, a general similarity evaluation method such as a cosine scale or a Kullback suspected distance is used. As a method for evaluating the similarity between regions performed in step S43, a general similarity evaluation method such as a cosine scale or a Kullback suspected distance is used.

There are two methods for modifying the prediction model. The first method is a method of reflecting a part or all of the features (prediction variable Xi) of the basic model in the calculated prediction model. When reflecting all the prediction variables, the calculated prediction model may be replaced with the basic model. In the second method, among the statistics (distribution in this example) of the weight parameter Wi of the prediction model, those that are out of the range significantly for the prediction of the weight parameter Wi of the basic model are adjusted within the range.

As described above, the traffic demand forecasting device 1 of the present embodiment predicts the traffic demand according to the prediction conditions based on the traffic data storage unit 111 that stores the traffic data indicating the movement record of an arbitrary section and the traffic data. The prediction model calculation unit 106 that calculates the prediction model, the model diagnosis unit 107 that compares the calculated prediction model with the basic model of the section different from the prediction conditions, and diagnoses the validity of the prediction model, and the validity of the prediction model. When it is diagnosed that there is no (or low), the model correction unit 108 that corrects the prediction model by using the basic model is provided.

According to such a configuration, the validity of the prediction model generated based on the traffic data is evaluated by using the basic model of other sections, and the invalid (low) prediction model is corrected based on the basic model. .. Therefore, it is possible to forecast traffic demand with higher accuracy for all sections (or regions) such as sections where a valid forecast model cannot be generated.

The model correction unit 108 compares the regional characteristics of the prediction target section with the regional characteristics of the section different from the prediction target section based on the regional data of each region stored in the regional data storage unit 112, and compares the regional characteristics with the regional characteristics. Use the basic model of the region containing similar sections to modify the predictive model diagnosed as invalid (or low). Further, the model correction unit 108 uses the transportation network and the movement record in each area as the feature amount of the regional characteristic, and evaluates the similarity between the prediction target section and the section different from the prediction target section based on the feature amount.

[Details of basic model calculation process]
Next, a process of calculating a new basic model from a plurality of basic models will be described.
FIG. 7 is a flowchart illustrating a procedure example of a process of calculating a new basic model from a plurality of basic models in the traffic demand forecasting device 1.

First, the prediction model calculation unit 106 reads a basic model among a plurality of ODs existing in an arbitrary area from the basic model storage unit 110 (S51). Then, the prediction model calculation unit 106 reads the variable Xi and the weight parameter Wi of each basic model (S52). In this example, the basic model between a plurality of ODs is read, but it is possible to create a new basic model from the basic model between one OD.

Next, the prediction model calculation unit 106 calculates the number of times of adoption (cumulative distribution) for each variable item i from the plurality of basic models (S53). Then, the prediction model calculation unit 106 selects the variable item in which the number of times of adoption (frequent value) is equal to or higher than the predetermined value th among the variable items i, or the variable item in which the number of times of adoption (frequent value) is ranked high, for traffic in the area. It is selected as an important variable item i'(hereinafter referred to as "important item i'") that affects demand (S54).

Next, the prediction model calculation unit 106 calculates a statistic (for example, distribution) of the weight parameter Wi'for each important item i'in each basic model (S55). Then, the prediction model calculation unit 106 sets a significant parameter range for the prediction given with high frequency based on a predetermined deviation for each weight parameter Wi'(S56).

Next, the prediction model calculation unit 106 outputs the calculated basic model to the basic model storage unit 110 as a feature of the new basic model in the area together with the effective range of the important item i'and its weight parameter Wi' (S57). ). After the process of step S57 is completed, the process of this flowchart ends.

[Predictive variables and weight parameters]
Here, the prediction variable Wi and the weight parameter Wi will be described.
FIG. 8 is a diagram for explaining the features of the basic model newly created from the plurality of basic models. In FIG. 8, the prediction variable Xi and the weight parameter Wi of the basic model created from the plurality of basic models of the A model to the Z model are shown. The graph on the left side of FIG. 8 (condition 1) shows an example of the relationship between the prediction variable Xi and the number of times of adoption, and the graph on the right side of FIG. 8 (condition 2) shows an example of the relationship between the weight parameter Wi (weight value) and the frequency (distribution). There is.

In the example of FIG. 8, of the variables _X 1 to Xn, adopted count variables _X 2 to X ₄ is not less than the predetermined value th is set as an important feature amount. Further, the range indicated by the double-headed arrow in the distribution of the weight parameter Wi is set as a range useful for prediction.

As described above, the traffic demand forecasting device 1 of the present embodiment includes a plurality of sections, reflecting the respective statistics of the variables constituting the basic model of the plurality of sections in an arbitrary area and the weight parameters for each variable. There is no validity between the prediction model calculation unit 106 that generates a new basic model of the intended region and the model diagnosis unit 107 that diagnoses the validity of the prediction model using the above new basic model of any region. It is provided with a model correction unit 108 that corrects a predicted model diagnosed as (or low) by using the new basic model in an arbitrary region.

With such a configuration, by generating a new basic model from a plurality of basic models existing in an arbitrary area, a basic model that becomes a standard of the area can be obtained. Therefore, it is possible to stably predict the traffic demand in a section or region where a prediction model does not exist with a prediction accuracy of a certain level or higher.

[Details of general-purpose model calculation processing]
Next, a process of calculating a general-purpose model as a new basic model from a plurality of basic models existing in a group composed of a plurality of regions will be described. Here, the similarity of the structure (for example, traffic environment) of the area (city) is evaluated, classified into a plurality of area groups, and the conditions (1) and (2) of FIG. 8 are set in each area group.

FIG. 9 is a flowchart for explaining a procedure example of a process of calculating a general-purpose model reflecting the similarity of traffic environments in a plurality of regions as a traffic demand forecast model in the traffic demand forecasting device 1.

First, the prediction model calculation unit 106 reads regional data representing regional characteristics such as the transportation network, population distribution, and average movement route of a plurality of regions from the regional data storage unit 112 (S61).

Next, the prediction model calculation unit 106 calculates features such as the number of nodes, link lengths, intersection angles, average widths, population distribution ranges, and average densities that make up the transportation network and movement routes for each region (S62). ), Evaluate the similarity of the features of each region (S63). Next, the prediction model calculation unit 106 generates a regional group (regional group) in a plurality of similar regions based on the evaluation result of similarity (S64).

Next, the prediction model calculation unit 106 reads the basic model of each region from the basic model storage unit 110 for each region group, and reads the variable Xi and the weight parameter Wi of each basic model (S65). Then, the prediction model calculation unit 106 calculates the statistics of the plurality of read basic models, and uses the conditions (1) and (2) of FIG. 8 as a general-purpose basic model (general-purpose model) of the region group. It is generated and set, and stored in the basic model storage unit 110 (S66). The process of generating the basic model for this region group is executed by the process of steps S53 to S56. When the generation process of the basic model in step S67 is completed, the process of this flowchart is completed. For each type of similar regional group, the process of FIG. 8 may be executed to generate a basic model (general-purpose model) for each of the plurality of regional groups.

The model correction unit 108 can correct the prediction model diagnosed as invalid (low) by using the basic model (general-purpose model) of the regional group.

FIG. 10 is a diagram for explaining a basic model of a regional group. As shown in FIG. 10, for example, a regional group is composed of similar regions 1 to 3 that are similar to each other (corresponding to S64), and a new basic model (general-purpose model) of the regional group is created from the basic models 1-3 of each region. Generate (corresponds to S65 and S66).

As described above, the traffic demand forecasting device 1 of the present embodiment is for each region group (region group) composed of a plurality of regions having similar regional characteristics, and for each variable and each variable constituting the basic model of the forecast target section. The prediction model calculation unit 106 that generates a general-purpose model that is the basic model of the region group, reflecting each statistic of the weight parameter, and the prediction model that is diagnosed as invalid (or low) are used for the region group. A model correction unit 108 for correction using a general-purpose model is provided.

With such a configuration, a basic model (general-purpose model) that is not biased to a specific area can be obtained by calculating a basic model for each regional group composed of areas with similar traffic environments. Therefore, it is possible to stably predict the traffic demand in a section or region where a prediction model does not exist with a prediction accuracy of a certain level or higher.

According to the first embodiment described above, it is difficult to predict the traffic demand because an appropriate model for predicting the traffic demand cannot be generated because the traffic data necessary for predicting the traffic demand cannot be obtained. It will be possible to predict traffic demand using basic models of different sections or regions for sections and regions, as well as between ODs where new cities and roads for which past traffic data cannot be obtained are newly constructed. Therefore, in the first embodiment described above, the traffic volume of all sections or regions is evaluated based on the predicted traffic demand without passing through the period for acquiring traffic data, and traffic measures for each section or region are performed. Transportation services can be deployed in a short period of time.

<2. Second embodiment>
FIG. 11 is a diagram showing a traffic demand forecasting system to which the traffic demand forecasting device according to the second embodiment of the present invention is applied. In the traffic flow evaluation system 10A of FIG. 11, components having substantially the same function or configuration as the traffic flow evaluation system 10 of FIG. 1 are designated by the same reference numerals, and duplicate description will be omitted.

In the first embodiment described above, the traffic demand forecasting device 1 executes the forecasting model generation process, but the forecasting model generation function may be separated from the traffic demand forecasting device 1. In this case, for example, as shown in FIG. 11, the traffic demand forecasting device 1A includes a communication interface 101, a reception processing unit 102, a prediction model selection unit 103, a traffic demand calculation unit 104, a transmission processing unit 105, and a basic model storage unit 110. And can be configured by the regional data storage unit 112.

Further, the prediction model generation device 6 includes a communication interface 101, a prediction model calculation unit 106, a model diagnosis unit 107, a model correction unit 108, a transmission processing unit 105, a basic model storage unit 110, a traffic data storage unit 111, and a regional data storage unit. It can be configured by the unit 112. That is, the traffic demand forecasting system 30 is configured by the traffic demand forecasting device 1A and the forecasting model generating device 6.

In the traffic demand forecasting device 1A, the prediction model selection unit 103 receives the traffic demand prediction request and the prediction condition from the terminal device 4, and determines whether or not the basic model storage unit 110 has a basic model corresponding to the prediction condition. .. Here, when the prediction model selection unit 103 determines that the basic model corresponding to the prediction condition does not exist in the basic model storage unit 110, the prediction model selection unit 103 instructs the prediction model generation device 6 to generate the prediction model (or the basic model). ..

The prediction model generation device 6 generates a model in response to a prediction model (or basic model) generation instruction and prediction conditions from the traffic demand prediction device 1A, and a communication network 5 or a recording medium (for example, a USB memory) is generated from the transmission processing unit 105. The generated model is provided to the traffic demand forecasting device 1A via the above. At this time, the model diagnosis unit 107 diagnoses the validity of the created prediction model, and if it is not valid (low), the basic model stored in the basic model storage unit 110 by the model correction unit 108 is used. After modifying the forecast model, it is transmitted to the traffic demand forecasting device 1A.

Then, in the traffic demand forecasting device 1A, the traffic demand calculation unit 104 predicts the traffic demand according to the prediction conditions by using the prediction model (basic model) received from the prediction model generation device 6. Further, in the traffic demand forecasting device 1A, the reception processing unit 102 stores the received prediction model (basic model) in the basic model storage unit 110.

With such a configuration, the traffic demand forecasting device 1A can receive the forecasting model from a plurality of forecasting model generating devices 6. On the other hand, the prediction model generation device 6 can provide a prediction model for a plurality of traffic demand prediction devices 1A (only one is described in FIG. 11). Therefore, it is possible to easily expand the system scale according to the forecast target section of traffic demand and the scale between cities.

It is desirable that the same data be stored in the basic model storage unit 110 and the regional data storage unit 112 of the traffic demand forecasting device 1A and the forecasting model generation device 6. For example, the traffic demand forecasting device 1A may provide the forecasting model generating device 6 with the latest data such as a basic model. On the contrary, the latest data may be provided from the forecast model generation device 6 to the traffic demand forecast device 1A.

<3. Modification example>
In the first and second embodiments described above, each process has been described with each process unit as a "program" (for example, a prediction model selection unit 103, etc.) as a subject (acting subject), but the program is a processor (for example, an operation subject). , CPU21a), the defined processing is performed while using the memory and the communication port (communication control device), so that the description may be based on the processor. Further, the process disclosed with the program as the subject may be a process performed by various computers or the like. Part or all of the program may be implemented on dedicated hardware or may be modularized. Various programs may be installed on each computer by a program distribution server or storage media. Further, the present embodiment may be implemented by software running on a general-purpose computer, or may be implemented by dedicated hardware or a combination of software and hardware.

Further, based on the instruction of the program code, the OS (operating system) or the like running on the computer performs a part or all of the actual processing, and the processing enables the functions of the above-described embodiment to be realized. May be good. Further, after the program code read from the storage medium is written in the memory on the computer, the CPU of the computer or the like performs a part or all of the actual processing based on the instruction of the program code, and by the processing. The functions of the above-described embodiments may be realized.

Further, by distributing the program code of the software that realizes the functions of the above-described embodiment via the network, it is stored in a storage means such as a hard disk or memory of the system or device, or in storage of a CD-RW, CD-R, or the like. It may be stored in a medium, and the computer (or CPU or MPU) of the system or device may read and execute the program code stored in the storage means or the storage medium at the time of use.

It should be understood that the processes and techniques described herein are not essentially associated with any particular device and can be implemented with any suitable combination of components. In addition, various types of devices for general purpose can be used according to what is described herein. You may find it useful to build a dedicated device to carry out the steps of the method described here. In addition, various inventions can be formed by appropriately combining the plurality of components disclosed in the embodiments. For example, some components may be removed from all the components shown in the embodiments. In addition, components across different embodiments may be combined as appropriate. The present invention has been described in the context of specific examples, but these are for illustration purposes only, not for limitation in all respects. Many combinations of hardware, software, and firmware suitable for carrying out the present invention are possible. For example, the described software can be implemented in a wide range of programs or scripting languages such as assembler, C / C ++, perl, Shell, PHP, Java®.

Furthermore, the present invention is not limited to the above-described embodiments, and it goes without saying that various other application examples and modifications can be taken as long as the gist of the present invention described in the claims is not deviated. is there.

For example, the above-described embodiment describes the configuration of the device and the system in detail and concretely in order to explain the present invention in an easy-to-understand manner, and is not necessarily limited to the one including all the components described. In addition, it is possible to replace a part of the configuration of one embodiment with a component of another embodiment. It is also possible to add components of another embodiment to the configuration of one embodiment. It is also possible to add, delete, or replace other components with respect to a part of the configuration of each embodiment.

1,1A ... Traffic demand forecasting device, 10, 10A ... Traffic flow evaluation system, 21 ... Control unit, 30 ... Traffic demand forecasting system, 101 ... Communication interface, 102 ... Reception processing unit, 103 ... Prediction model selection unit, 104 ... Traffic demand calculation unit, 105 ... Transmission processing unit, 106 ... Forecast model calculation unit, 107 ... Model diagnosis unit, 108 ... Model correction unit, 110 ... Basic model storage unit, 111 ... Traffic data storage unit, 112 ... Regional data storage unit

Claims (9)

A model storage unit that stores the basic model used for forecasting traffic demand for each section,
A prediction condition acquisition unit that acquires prediction conditions including at least a prediction target section and a movement route within the prediction target section,
It has a function of selecting a basic model corresponding to the prediction condition from the model storage unit, and when a basic model corresponding to the prediction target section is not obtained, a basic model of a section different from the prediction target section is used. A prediction model selection unit selected as a prediction model for the prediction target section,
A traffic demand forecasting device including a traffic demand calculation unit that calculates traffic demand according to the forecasting conditions using the selected forecasting model. In addition, it is equipped with a regional data storage unit that stores regional data including at least regional transportation networks and household information.
When the basic model of the prediction target section is not obtained from the model storage unit, the prediction model selection unit uses the area data of each area of the area data storage unit to obtain the area of the prediction target section. The traffic according to claim 1, wherein the characteristics are compared with the regional characteristics of a section different from the forecast target section, and the basic model of the region including the sections having similar regional characteristics is selected as the prediction model of the forecast target section. Demand forecaster. The prediction model selection unit uses at least the traffic network and route in each area as the feature amount of the regional characteristic, and based on the feature amount of the regional characteristic, the prediction target section and the section different from the prediction target section The traffic demand forecasting device according to claim 2, which evaluates the similarity. Furthermore, a traffic data storage unit that stores traffic data indicating the movement record of an arbitrary section,
A prediction model calculation unit that calculates a prediction model that predicts traffic demand according to the prediction conditions based on the traffic data,
A model diagnosis unit that compares the calculated prediction model with a basic model of a section different from the prediction conditions and diagnoses the validity of the prediction model.
The traffic demand forecasting device according to claim 1, further comprising a model modification unit that modifies the forecast model by using the basic model when it is diagnosed that the forecast model is not valid. In addition, it is equipped with a regional data storage unit that stores regional data including at least regional transportation networks and household information.
The model correction unit compares the regional characteristics of the prediction target section with the regional characteristics of a section different from the prediction target section based on the regional data of each region stored in the regional data storage unit. The traffic demand forecasting device according to claim 4, which modifies the forecasting model diagnosed as invalid by using the basic model of the region including sections having similar regional characteristics. The model correction unit uses the traffic network and the movement record in each area as the feature amount of the regional characteristic, and based on the feature amount, the similarity between the prediction target section and the section different from the prediction target section is determined. The traffic demand forecasting device according to claim 5 to be evaluated. The prediction model calculation unit reflects the respective statistics of the variables constituting the basic model of the plurality of sections in the arbitrary area and the weight parameters for each variable, and is new to the arbitrary area including the plurality of sections. Generate a basic model and
The model diagnostic unit diagnoses the validity of the prediction model using the new basic model in the arbitrary region.
The traffic demand forecasting device according to claim 4, wherein the model modification unit modifies the forecast model diagnosed as invalid by using the new basic model in the arbitrary area. The prediction model calculation unit reflects the statistics of the variables constituting the basic model of the prediction target section and the weight parameters for each variable for each region group composed of a plurality of regions having similar regional characteristics. , Generate a general-purpose model that is the basic model of the regional group,
The traffic demand forecasting device according to claim 4, wherein the model modification unit modifies the forecast model diagnosed as invalid by using the general-purpose model of the regional group. It is a traffic demand forecasting system consisting of a traffic demand forecasting device and a forecasting model generator.
The traffic demand forecasting device is
A prediction condition acquisition unit that acquires prediction conditions including at least a prediction target section and a movement route within the prediction target section,
A prediction model selection unit that selects a basic model that meets the prediction conditions,
A traffic demand calculation unit that calculates traffic demand according to the forecast conditions using the selected forecast model, and a traffic demand calculation unit.
A communication processing unit that notifies the prediction model generation device of the prediction conditions and receives the prediction model from the prediction model generation device.
The predictive model generator is
A model storage unit that stores the basic model used for forecasting traffic demand for each section,
A traffic data storage unit that stores traffic data indicating the movement record of an arbitrary section,
A forecast model calculation unit that calculates a forecast model that predicts traffic demand according to the forecast conditions notified from the traffic demand forecast device based on the traffic data.
A model diagnosis unit that compares the calculated prediction model with the basic model in a section different from the prediction conditions stored in the model storage unit and diagnoses the validity of the prediction model.
When it is diagnosed that the prediction model is not valid, a model correction unit that modifies the prediction model by using the basic model, and a model correction unit.
A communication processing unit that receives the forecast condition from the traffic demand forecasting device and transmits the modified forecasting model to the traffic demand forecasting device.
The traffic demand calculation unit of the traffic demand forecasting device is a traffic demand forecasting system that calculates traffic demand according to the forecasting conditions by using the modified forecasting model received from the forecasting model generation device.

Images