JP6855359B2 - Congestion prediction model creation device - Google Patents

Description

The present invention relates to a traffic congestion prediction model creating device that creates a traffic congestion prediction model used for predicting traffic congestion.

A technique for acquiring traffic information and predicting traffic congestion at a desired point based on the acquired traffic information is known (see, for example, Patent Document 1). Patent Document 1 describes that a traffic congestion prediction model is created based on past traffic congestion data, and current traffic information is input to the traffic congestion prediction model to predict traffic congestion.

Japanese Unexamined Patent Publication No. 2004-272408

There are various factors in traffic congestion, and in the conventional congestion prediction, there is a concern that the predicted congestion and the actual congestion may deviate due to unexpected factors. For example, in the method described in Patent Document 1, a traffic jam prediction model is created from past traffic jam data, that is, data of a car that is already in a running state, and a car that is not in the running state at the time of sampling the past traffic jam data is considered. It has not been. For this reason, when predicting traffic congestion only from past congestion data, the longer the time from the current time to the predicted time, the more the congestion prediction is caused by a vehicle that is not currently in the driving state but is in the driving state at the predicted time. The error may increase.

On the other hand, if a traffic congestion prediction model is constructed in consideration of various factors of traffic congestion, the error of the traffic congestion prediction can be reduced and the accuracy of the traffic congestion prediction can be improved. However, in this case, the amount of calculation required to build the congestion prediction model may increase dramatically.

The present invention has been made in view of the above problems, and provides a traffic jam prediction model creating device capable of improving the accuracy of traffic jam prediction and suppressing an increase in the amount of calculation required for constructing a traffic jam prediction model. The purpose.

In order to achieve the above object, the traffic jam prediction model creating device according to the present invention is a traffic jam prediction model creating device that creates a traffic jam prediction model used for predicting traffic congestion, and is a traffic jam prediction model creating device according to the position and attribute of a person. The population data acquisition unit that acquires the population data of multiple classified groups, the congestion data acquisition unit that acquires the congestion data indicating the congestion status according to the position, and the congestion data acquisition unit that acquires the congestion data according to the position, and after normalizing the population data for each group, each group Contribution to the traffic congestion data at the traffic congestion prediction position from multiple groups by performing regression analysis using the normalized population data of the above as the explanatory variable and the traffic congestion data at the traffic congestion prediction position as the objective variable. degrees selects population data of relatively high group, and a traffic jam prediction model constructing section for constructing the traffic jam prediction model by congestion data and the based rather regression analysis on selected population data traffic jam predicted position.

In the traffic jam prediction model generating apparatus according to the present invention is to construct a traffic jam prediction model by based rather regression analysis and population data and traffic congestion data. By normalizing the population data for each group and then using the normalized population data for each group, it is possible to deal with various factors of traffic congestion. For example, population data also correlates with the number of vehicles that are not currently in driving but will be in driving at the predicted time. Therefore, if the traffic jam is predicted using the above-mentioned traffic jam prediction model, the error of the traffic jam prediction can be reduced and the accuracy of the traffic jam prediction can be improved. Further, using regression analysis, select the population data of contribution higher group for congestion data in the traffic jam prediction position, building a traffic jam prediction model by a selected population data congestion data and the based rather regression analysis. Therefore, while ensuring the accuracy of the traffic jam prediction, it is possible to suppress an increase in the amount of calculation required for constructing the traffic jam prediction model. That is, by performing the traffic jam prediction using the traffic jam prediction model constructed by the model construction unit, it is possible to both improve the degree of traffic jam prediction and suppress the increase in the amount of calculation.

According to the present invention, it is possible to provide a traffic jam prediction model creating device that can improve the accuracy of traffic jam prediction and suppress an increase in the amount of calculation required for constructing a traffic jam prediction model.

It is a figure for demonstrating the function of the traffic jam prediction model creation apparatus which concerns on one Embodiment. It is a figure which shows the composition example of the population data. It is a figure which shows the composition example of the traffic jam data. It is a figure which shows the composition example of the output of the traffic jam prediction result when it is predicted that there is a traffic jam or there is no traffic jam. It is a figure which shows the display example of the traffic jam prediction result when it is predicted that there is a traffic jam or there is no traffic jam. It is a flowchart which shows the process executed by the traffic jam prediction model creation apparatus which concerns on one Embodiment. It is a flowchart which shows the process executed by the traffic jam prediction model creation apparatus which concerns on one Embodiment. It is a figure which shows the hardware structure of the traffic jam prediction model creation apparatus which concerns on one Embodiment.

Hereinafter, embodiments of the congestion prediction model creating device according to the present invention will be described in detail together with the drawings. In the description of the drawings, the same elements are designated by the same reference numerals, and duplicate description will be omitted.

FIG. 1 shows a traffic jam prediction model creating device 100 according to the present embodiment. The congestion prediction model creating device 100 creates a congestion prediction model used for predicting traffic congestion based on the learning population data and the learning congestion data. The traffic congestion prediction model creating device 100 uses the created traffic congestion prediction model to predict traffic congestion by inputting population data. In the following, a congestion prediction model will be constructed based on the original data sampled within a predetermined period (hereinafter referred to as "sampling period"), and the population data for each regional mesh of the day will be input as the input for the day. The case of predicting traffic congestion will be described as an example, but the present invention is not limited to this.

Population data is data showing how many people existed in an area at a certain time. In this embodiment, the population data is a plurality of groups classified according to the sampling time, the position of the person (such as a tabby, the area mesh), and the attribute of the person (for example, gender, age, place of residence, etc.). Shows the population belonging to each of. The sampling time is the time when the data was sampled. The position and attribute of a person are the position and attribute of a person whose data is to be sampled (hereinafter, simply referred to as "sampling target person"). For example, the population data is generated from the location information that is periodically sampled in advance of the mobile terminal and the registration information about the attributes of the user of the mobile terminal, which are stored in an external database or the like. In the present embodiment, as shown in FIG. 2, the population data is composed of the items of "time", "location identification number", "gender", "age", "residential place", and "population". ..

"Time" indicates the date and time when the data was sampled. The "position identification number" indicates a number that identifies the position of the sampling target person. For example, "JIS X 0410 area mesh code" is used as the "position identification number". For the "position identification number", a regional mesh code such as a secondary mesh, a tertiary mesh, or a half regional mesh may be used, or a prefecture code or a city / ward / town / village code of the place of residence may be used. You may. In the present embodiment, since the half area mesh is used as the "position identification number", the position identification numbers are classified into about 1.5 million types.

“Gender” indicates the gender of the sampling target. "Age" indicates the age group of the person to be sampled. In the present embodiment, the "age" is 20 years old or younger, 21-25 years old, 26-30 years old, 31-35 years old, 36-40 years old, 41-45 years old, 46-50 years old, 51-55 years old, 56 years old. It is classified into 13 types: ~ 60 years old, 61-65 years old, 66-70 years old, 71-75 years old, 76 years old and over. “Residence” indicates the residence of the person to be sampled. For the "place of residence", a regional mesh code such as a secondary mesh, a tertiary mesh, or a half area mesh may be used, or a prefecture code or a city / ward / town / village code of the place of residence is used. May be good. In the present embodiment, as the "residential place", the city, ward, town, and village codes classified into about 2000 are used.

In the present embodiment, the population data is divided into a plurality of groups according to the "position identification number" corresponding to the position of the person and the "gender", "age", and "place of residence" corresponding to the attribute of the person. being classified. Therefore, the population data can be classified into groups of about 1.5 million x 2 x 13 x about 2000. That is, the space in which the combination of the position and the attribute of a person is a variable theoretically extends to about 80 billion dimensions.

The traffic jam data is data showing the traffic jam situation according to the position, and is stored in advance in an external database that manages the traffic jam information. For example, in the traffic jam data, as shown in FIG. 3, the "traffic jam occurrence date", "start time", and "end" are collected for each section (so-called link) in which roads (for example, major roads nationwide) are divided in advance. It consists of items such as "time", "maximum congestion length", and "congestion peak time". In the present embodiment, the above-mentioned pre-divided section (hereinafter, simply referred to as “position”) is the target of prediction. In the "traffic jam occurrence date", the date when the traffic jam occurred at the target position is indicated. The "start time" indicates the time when the traffic jam occurred at the target position. The "end time" indicates the time when the traffic jam ended at the target position. The "maximum congestion length" indicates the maximum congestion length between the start time and the end time of the congestion that occurred at the target position. "Traffic congestion peak time" indicates the time when the maximum congestion length is reached.

The learning population data acquisition unit 111 acquires the learning population data (hereinafter, referred to as “learning population data”) used for constructing the congestion prediction model from the outside of the congestion prediction model creating device 100. The learning population data acquisition unit 111 acquires the learning population data sampled within a predetermined sampling period as the original data used for the congestion prediction. The learning population data is, for example, the past population data of the whole country recorded in an external database, and is divided into a plurality of groups classified according to the sampling time, the position of the person, and the attribute of the person. In the present embodiment, the learning population data acquisition unit 111 acquires only the learning population data sampled at noon every day. The learning population data storage unit 112 stores the learning population data input to the learning population data acquisition unit 111.

The learning congestion data acquisition unit 113 acquires learning congestion data (hereinafter, referred to as “learning congestion data”) used for constructing the congestion prediction model from the outside of the congestion prediction model creating device 100. The learning congestion data acquisition unit 113 acquires the learning congestion data at the congestion prediction position sampled within a predetermined sampling period as the original data used for the congestion prediction. The learning congestion data indicates, for example, the congestion status according to the congestion prediction position for predicting the traffic congestion among the past congestion data recorded in the external database. The learning congestion data storage unit 114 stores the learning congestion data input to the learning congestion data acquisition unit 113.

The congestion prediction model construction unit 115 constructs a congestion prediction model used for predicting traffic congestion for each position with reference to the learning population data storage unit 112 and the learning congestion data storage unit 114. The congestion prediction model construction unit 115 includes a variable selection unit 116, a dimension compression unit 117, and a prediction regression equation creation unit 118.

Hereinafter, the traffic jam prediction model construction unit 115 describes the "traffic jam occurrence date", "start time", "end time", "maximum traffic jam length", and "traffic jam peak time" of the learning traffic jam data for a certain traffic jam prediction position. An example of constructing a congestion prediction model based on each of the items will be described. Specifically, the traffic jam prediction model construction unit 115 calculates a traffic jam prediction model that predicts the presence or absence of traffic jam on the day to be traffic jam prediction target based on the learning traffic jam data for each "traffic jam occurrence date" for the traffic jam prediction position. To construct. Further, the traffic jam prediction model building unit 115 builds a traffic jam prediction model that predicts the start time of the traffic jam based on the "start time" of the traffic jam prediction position. In addition, the traffic jam prediction model building unit 115 builds a traffic jam prediction model that predicts the end time of the traffic jam based on the "end time" of the traffic jam prediction position. In addition, the traffic jam prediction model construction unit 115 builds a traffic jam prediction model that predicts the maximum traffic jam length based on the "maximum traffic jam length". In addition, the traffic jam prediction model building unit 115 builds a traffic jam prediction model that predicts the traffic jam peak time based on the “traffic jam peak time”. That is, the traffic jam prediction model construction unit 115 builds five traffic jam prediction models for each traffic jam prediction position.

The traffic jam prediction model construction unit 115 builds a traffic jam prediction model using the numerical values obtained based on the values of the above-mentioned items of the traffic jam data for learning at the traffic jam prediction position. For example, when constructing a congestion prediction model that predicts the presence or absence of congestion, the congestion prediction model construction unit 115 determines that there was congestion at the congestion prediction position in the learning congestion data within the above sampling period. The numerical value indicating the presence or absence of traffic jam in (the date when there is the traffic jam data for learning about the predicted traffic jam position) is set to "1", and the date when it is judged that there is no traffic jam in the predicted traffic jam position in the learning traffic jam position (the predicted traffic jam position). The numerical value indicating the presence or absence of traffic congestion on the date when there is no traffic congestion data for learning is treated as "0". When constructing a traffic jam prediction model that predicts the start time of traffic jam, the traffic jam prediction model building unit 115 numerically calculates the value of the "start time" of the training traffic jam data at the traffic jam prediction position within the sampling period by a well-known method. To become. When constructing a traffic jam prediction model that predicts the end time of traffic jam, the traffic jam prediction model building unit 115 numerically calculates the value of the "end time" of the training traffic jam data at the traffic jam prediction position within the sampling period by a well-known method. To become. When constructing a congestion prediction model that predicts the maximum congestion length of congestion, the congestion prediction model construction unit 115 uses the value of the "maximum congestion length" of the learning congestion data at the congestion prediction position within the sampling period as the above numerical value. handle. When constructing a traffic jam prediction model for predicting a traffic jam peak time, the traffic jam prediction model building unit 115 numerically calculates the value of the "traffic jam peak time" of the learning traffic jam data at the traffic jam prediction position within the sampling period by a well-known method. To become.

The variable selection unit 116 is a learning population data (explanation) of a group related to traffic congestion from the learning population data (for example, the national learning population data for each regional mesh) stored in the learning population data storage unit 112. Variable) is selected. The processing performed by the variable selection unit 116 will be specifically described below.

First, the variable selection unit 116 normalizes the learning population data within the sampling period for each group. For example, the variable selection unit 116 converts the population in each group of the learning population data into a value having an average of 0 and a variance of 1 in each group with respect to all the times within the sampling period. By normalizing the learning population data, the rate of increase or decrease in the population can be appropriately determined according to the size of the normal population of each group.

Hereinafter, the normalized population data for learning is simply referred to as "population data for learning". A numerical value indicating the presence or absence of traffic congestion, a numerical value indicating "start time", a numerical value indicating "end time", a numerical value indicating "maximum traffic congestion length", and a numerical value indicating "congestion peak time" are simply "congestion for learning". It is called "data".

The variable selection unit 116 performs logistic regression using the learning population data for each group as an explanatory variable and the learning congestion data as an objective variable. The variable selection unit 116 performs logistic regression using a plurality of sets of sampling periods, with the learning population data and the learning congestion data showing information on the same day as one set for each congestion prediction model to be constructed. For example, the variable selection unit 116 performs logistic regression using a numerical value indicating the "start time" as an objective variable. The variable selection unit 116 selects an explanatory variable having a large absolute value of the coefficients up to the upper nth position in the result of logistic regression. n is a parameter and can be arbitrarily changed by the user. n is set to, for example, several hundreds. The variable selection performed by the variable selection unit 116 is not limited to the method using the logistic regression described above. For example, parametric regression such as lasso regression may be used, or selection may be made according to the value of the correlation coefficient.

The dimension compression unit 117 performs coordinate conversion by PLS regression using the learning population data of the group selected by the variable selection unit 116 as an explanatory variable and the learning congestion data as an objective variable. The dimension compression unit 117 uses a plurality of sets of sampling periods, with the learning population data and the learning congestion data selected by the variable selection unit 116 showing information on the same day as one set for each congestion prediction model to be constructed. Then, PLS regression is performed. For example, the dimension compression unit 117 performs PLS regression using a numerical value indicating the "start time" as an objective variable. As a result, the dimensional compression unit 117 performs dimensional compression on the explanatory variable having the higher degree of contribution to the objective variable. m is a parameter and can be arbitrarily changed by the user. m is set to, for example, about 5 to 100. As a result, the explanatory variables that behave similarly with the explanatory variables selected by the variable selection unit 116 are combined and converted into low-dimensional data. Therefore, in the prediction regression equation creation unit 118, the accuracy deteriorates due to the curse of dimensionality. Can be prevented. The regression used in the dimensional compression unit 117 is not limited to PLS regression. For example, PCA, NMF, or the like may be used instead of PLS regression.

The prediction regression equation creation unit 118 creates a prediction regression equation (congestion prediction model) used for congestion prediction based on the learning population data and the learning congestion data that have been dimensionally compressed by the dimension compression unit 117. The prediction regression equation creation unit 118 creates a congestion prediction model using the variable converted by the dimension compression unit 117 as an explanatory variable and the training congestion data as an objective variable. The prediction regression equation creation unit 118 sets a plurality of sampling periods for each congestion prediction model to be constructed, with the values of variables converted by the dimension compression unit 117 showing information on the same day and the congestion data for learning as one set. Create a traffic jam prediction model using. For example, the prediction regression equation creation unit 118 predicts the start time of congestion based on the learning population data subjected to dimension compression by the dimension compression unit 117 and the numerical value indicating the "start time" (prediction regression equation (prediction regression equation creation unit 118). Create a congestion prediction model). That is, the traffic jam prediction model construction unit 115 creates a traffic jam prediction model for predicting the traffic jam value from the low-dimensional data. Specifically, the prediction regression equation creation unit 118 performs k-nearest neighbor regression using a weighted average based on the distance of the nearest neighbor sample.

The number of nearest neighbor points is a parameter and can be arbitrarily changed by the user. The number of nearest neighbor points is set to, for example, about 2 to 100. The regression used in the prediction regression equation creation unit 118 is not limited to the k-nearest neighbor regression. Instead of k-nearest neighbor regression, for example, a parametric method such as linear regression or SVR regression may be used, or a nonparametric method such as a decision tree may be used. Since the relationship between population data and congestion data is non-linear, it is possible to predict congestion with higher accuracy than using parametric methods by performing regression analysis using non-parametric methods such as k-nearest neighbor regression. You can create a traffic jam prediction model.

The traffic jam prediction model storage unit 119 stores the traffic jam prediction model constructed by the traffic jam prediction model construction unit 115. In the present embodiment, the traffic jam prediction model storage unit 119 includes a traffic jam prediction model that predicts the presence or absence of traffic jam on the day that is the target of traffic jam prediction, a traffic jam prediction model that predicts the start time of traffic jam, and the end of traffic jam for each traffic jam prediction position. It stores a traffic jam prediction model that predicts the time, a traffic jam prediction model that predicts the maximum traffic jam length, and a traffic jam prediction model that predicts the peak traffic time. The traffic jam prediction model storage unit 119 stores the group selected by the variable selection unit 116 and the information for performing the coordinate conversion performed by the dimension compression unit 117 for each traffic jam prediction model.

As described above, the congestion prediction model construction unit 115 performs regression analysis using the learning population data for each group as an explanatory variable and the learning congestion data as an objective variable. As a result, the traffic jam prediction model construction unit 115 selects the learning population data of the group having a relatively high contribution to the traffic jam data at the traffic jam prediction position from the plurality of groups. The traffic jam prediction model construction unit 115 builds a traffic jam prediction model by regression analysis based on the selected population data and the traffic jam data at the traffic jam prediction position.

The traffic jam prediction position acquisition unit 120 acquires the traffic jam prediction position from the outside of the traffic jam prediction model creation device 100. The traffic jam prediction position acquisition unit 120 may acquire the traffic jam prediction position input by the user by an input device (not shown).

The forecasting population data acquisition unit 121 acquires the forecasting population data used for forecasting traffic congestion. The forecasting population data is population data at a predetermined date and time, and is, for example, the latest population data among the available population data, population data at a predetermined time on the day, and the like. In the present embodiment, the forecasting population data acquisition unit 121 acquires the population data sampled at noon on the day when the traffic congestion is predicted.

The traffic congestion prediction unit 122 uses the traffic congestion prediction model stored in the traffic congestion prediction model storage unit 119 and the data acquired by the traffic congestion prediction position acquisition unit 120 and the prediction population data acquisition unit 121 to detect traffic congestion. Predict. The traffic jam prediction unit 122 acquires a traffic jam prediction model corresponding to the traffic jam prediction position acquired by the traffic jam prediction position acquisition unit 120 from the traffic jam prediction model storage unit 119. In the present embodiment, the traffic congestion prediction unit 122 inputs the forecast population data (for example, the population data of the day), and uses the traffic congestion prediction model acquired from the traffic congestion prediction model storage unit 119 to use the traffic congestion prediction model at the traffic congestion prediction time. Outputs the congestion prediction result of the predicted position. Congestion prediction model that predicts the presence or absence of congestion on the target day, congestion prediction model that predicts the start time of congestion, congestion prediction model that predicts the end time of congestion, congestion prediction model that predicts the maximum congestion length of congestion , And by applying the forecasting population data to each of the congestion prediction models that predict the congestion peak time, each congestion prediction is performed. For example, the traffic congestion prediction unit 122 selects a group corresponding to the group of the learning population data selected by the variable selection unit 116 from a plurality of groups in the prediction population data, and is performed by the dimension compression unit 117. Performs the same coordinate transformation as the coordinate transformation, and substitutes it into the regression equation created by the prediction regression equation creation unit 118.

The traffic congestion prediction unit 122 obtains a predicted value by, for example, applying the forecasting population data to a traffic congestion prediction model that predicts the presence or absence of traffic congestion on the day of the traffic congestion prediction target. The predicted value is a numerical value of 0 to 1 obtained from the regression formula created by the prediction regression formula creating unit 118. The traffic congestion prediction unit 122 determines that there is a traffic jam when the obtained predicted value is equal to or higher than a predetermined threshold value, and determines that there is no traffic jam when the predicted value is less than a predetermined threshold value.

The traffic congestion prediction unit 122 determines that there is a traffic jam, a traffic jam prediction model that predicts the start time of the traffic jam, a traffic jam prediction model that predicts the end time of the traffic jam, a traffic jam prediction model that predicts the maximum traffic jam length, and a traffic jam prediction model. Predicted values are obtained by applying the population data for prediction to the traffic jam prediction model that predicts the peak time of traffic jam. FIG. 4A shows an output example of the prediction result from the traffic congestion prediction unit 122 when it is determined that there is congestion. For example, as shown in FIG. 4A, a value indicating "congestion" is output for the item "presence or absence of congestion", and "start time", "end time", and "maximum congestion length" are output. , And for the item of "traffic jam peak time", the derived results are output respectively.

When it is determined that there is no traffic jam, the traffic jam prediction unit 122 does not predict the "start time", "end time", "maximum traffic jam length", and "traffic jam peak time". FIG. 4B shows an output example of the prediction result from the traffic congestion prediction unit 122 when it is determined that there is no congestion. For example, as shown in FIG. 4B, for the item of "presence or absence of traffic jam", a value indicating "no traffic jam" is output, and "start time", "end time", and "end time", which are not predicted, are output. The items of "maximum traffic jam length" and "traffic jam peak time" are NULL.

The traffic congestion prediction result display unit 123 displays the traffic congestion prediction result predicted by the traffic congestion prediction unit 122 on the display 124. FIG. 5A shows a display example displayed on the display 124 when it is determined that there is a traffic jam and the value shown in FIG. 4A is output from the traffic jam prediction unit 122. FIG. 5B shows a display example displayed on the display 124 when it is determined that there is no traffic congestion.

Next, the operation of the traffic jam prediction model creating device 100 according to the present embodiment will be described with reference to FIG. FIG. 6 is a flowchart showing a process performed by the traffic jam prediction model creating device 100.

In step S101, the traffic jam prediction model creating device 100 acquires the traffic jam prediction position by the traffic jam prediction position acquisition unit 120. Subsequently, the traffic jam prediction model creating device 100 proceeds to the process in step S102.

In step S102, the traffic jam prediction model creating device 100 acquires the forecasting population data by the forecasting population data acquisition unit 121. Subsequently, the traffic jam prediction model creating device 100 proceeds to the process in step S103.

In step S103, the traffic jam prediction model creating device 100 determines whether or not the traffic jam prediction model corresponding to the traffic jam prediction position acquired in step S101 exists in the traffic jam prediction model storage unit 119. When the traffic jam prediction model corresponding to the traffic jam prediction position acquired in step S101 exists in the traffic jam prediction model storage unit 119 (YES in step S103), the traffic jam prediction model creating device 100 proceeds to the process in step S105. When the traffic jam prediction model corresponding to the traffic jam prediction position acquired in step S101 does not exist in the traffic jam prediction model storage unit 119 (NO in step S103), the traffic jam prediction model creating device 100 proceeds to the process in step S104.

In step S104, the traffic jam prediction model creating device 100 performs the traffic jam prediction model construction process described later. Subsequently, the traffic jam prediction model creating device 100 proceeds to step S105.

In step S105, the traffic jam prediction model creating device 100 includes the traffic jam prediction position acquired in step S101, the forecast population data acquired in step S102, and the traffic jam prediction model stored in the traffic jam prediction model storage unit 119. Therefore, the traffic congestion at the congestion prediction position is predicted. Subsequently, the traffic jam prediction model creating device 100 proceeds to the process in step S106.

In step S106, the traffic congestion prediction model creating device 100 displays the traffic congestion prediction result performed in step S105 on the display 124. When the traffic jam prediction model creating device 100 finishes step S106, the entire process is finished.

Next, the congestion prediction model construction process in step S104 will be described with reference to FIG. 7. FIG. 7 is a flowchart showing a traffic jam prediction model construction process performed by the traffic jam prediction model creation device 100.

In step S111, the congestion prediction model creating device 100 determines whether the learning population data exists in the learning population data storage unit 112. When the learning population data exists in the learning population data storage unit 112 (YES in step S111), the congestion prediction model creating device 100 proceeds to the process in step S114. When the learning population data does not exist in the learning population data storage unit 112 (NO in step S111), the congestion prediction model creating device 100 proceeds to the process in step S112.

In step S112, the congestion prediction model creating device 100 acquires the learning population data sampled within the sample period predetermined by the learning population data acquisition unit 111, and proceeds to the process in step S113.

In step S113, the congestion prediction model creating device 100 stores the learning population data acquired in step S112 in the learning population data storage unit 112, and proceeds to the process in step S114.

In step S114, the traffic jam prediction model creating device 100 determines whether the learning traffic jam data of the traffic jam prediction position acquired in step S101 exists in the learning traffic jam data storage unit 114. When the learning congestion data of the congestion prediction position acquired in step S101 exists in the learning congestion data storage unit 114 (YES in step S114), the congestion prediction model creating device 100 proceeds to the process in step S117. The congestion prediction model creating device 100 proceeds to step S115 when the learning congestion data of the congestion prediction position acquired in step S101 does not exist in the learning congestion data storage unit 114 (NO in step S114).

In step S115, the congestion prediction model creating device 100 acquires the learning congestion data at the congestion prediction position acquired in step S101, which is sampled within a predetermined sampling period by the learning congestion data acquisition unit 113. The process proceeds to step S116.

In step S116, the congestion prediction model creating device 100 stores the learning congestion data acquired in step S115 in the learning congestion data storage unit 114, and proceeds to the process in step S117.

In step S117, the congestion prediction model creation device 100 acquires learning population data within the sampling period from the learning population data storage unit 112, and learns from the learning congestion data storage unit 114 at the congestion prediction position within the sampling period. Get traffic jam data. Subsequently, the traffic jam prediction model creating device 100 proceeds to the process in step S118.

In step S118, the traffic jam prediction model creating device 100 selects the learning population data of the group having a relatively high contribution to the traffic jam data at the traffic jam prediction position by performing regression analysis in the variable selection unit 116. Subsequently, the traffic jam prediction model creating device 100 proceeds to step S119.

In step S119, the congestion prediction model creating device 100 performs dimensional compression by performing regression analysis in the dimensional compression unit 117. Subsequently, the traffic jam prediction model creating device 100 proceeds to the process in step S120.

In step S120, the congestion prediction model creating device 100 creates a prediction regression equation, which is a congestion prediction model, based on the learning population data and the learning congestion data that have been dimensionally compressed by the dimension compression unit 117. Subsequently, the traffic jam prediction model creating device 100 proceeds to the process in step S121.

In step S121, the congestion prediction model creating device 100 stores the congestion prediction model constructed in the three steps (step S118, step S119, and step S120) of variable selection, dimension compression, and creation of the prediction regression equation. Store in unit 119. When the traffic jam prediction model creating device 100 finishes step S121, the traffic jam prediction model building process ends.

Next, the operation and effect of the traffic congestion prediction model creating device 100 according to the present embodiment will be described.

The traffic jam prediction model creating device 100 builds a traffic jam prediction model by regression analysis based on the population data and the traffic jam data. By using population data, it is possible to deal with various factors of traffic congestion. Therefore, it is possible to reduce the error of the congestion prediction. For example, population data also correlates with the number of vehicles that are not currently in driving but will be in driving at the predicted time. Since the traffic congestion prediction model creating device 100 predicts traffic congestion based on the above-mentioned congestion prediction model and the population data for prediction used for predicting traffic congestion, it is possible to reduce the error of the congestion prediction and improve the accuracy of the congestion prediction. ..

Further, as described above, in the present embodiment, there are groups according to the "position identification number" corresponding to the position of the person and the "gender", "age", and "place of residence" corresponding to the attribute of the person. To classify. In this embodiment, there are about 1.5 million "position identification numbers", 2 "genders", 13 "ages", and about 2000 "residential areas". In this case, the population data can be divided into groups of about 1.5 million x 2 x 13 x about 2000. That is, the space in which the combination of the position and the attribute of a person is a variable theoretically extends to about 80 billion dimensions. Most of the population data of these groups has a very small effect on the congestion of the predicted target position and becomes noise for prediction.

The traffic jam prediction model creation device 100 uses regression analysis to select population data of a group having a high contribution to the traffic jam data at the traffic jam prediction position, and predicts traffic jam by regression analysis based on the selected population data and the traffic jam data. Build a model. That is, the traffic jam prediction model creating device 100 selects the population data (explanatory variable) of the group related to the traffic jam. Therefore, while ensuring the accuracy of the traffic jam prediction, it is possible to suppress an increase in the amount of calculation required for constructing the traffic jam prediction model. For example, since people in Okinawa have almost no effect on road congestion in Tokyo, the amount of calculation required to build a congestion prediction model can be reduced by selecting population data other than the group in which the position of the person is Okinawa. As described above, in the traffic jam prediction model creating device 100, improvement of the traffic jam prediction accuracy and suppression of an increase in the amount of calculation can be achieved at the same time.

The attributes of a person in the population data include the place of residence of the person (sampling target). In this case, it is possible to predict the congestion caused by the person returning to the place of residence, that is, the so-called U-turn rush. For example, from the population of people living in Tokyo and Kanagawa at the outlet mall in Kisarazu at noon, it is possible to predict the traffic congestion in the evening of Aqualine.

The traffic jam prediction model creating device 100 acquires the traffic jam prediction position, and when the traffic jam prediction model corresponding to the acquired traffic jam prediction position does not exist in the traffic jam prediction model storage unit 119, the traffic jam corresponding to the acquired traffic jam prediction position. Build a predictive model. That is, since the congestion prediction model is constructed as needed, the amount of calculation for constructing the congestion prediction model can be reduced. In addition, a congestion prediction model can be constructed using the latest learning population data.

Even if the number of population increases / decreases is the same, it is considered that the larger the size of the population, the lower the relationship between the number of population increases / decreases and traffic congestion. For example, it is assumed that an area with a normal population of 10,000 and an area with a normal population of 500 have different degrees of influence on the occurrence of traffic congestion when an increase or decrease of 500 people occurs. The variable selection unit 116 converts the population in each group of the learning population data into a value in which the average is 0 and the variance is 1 in each group with respect to all the times within the sampling period. In this way, by normalizing the learning population data within the sampling period for each group, the rate of increase or decrease in the population according to the size of the normal population of each group can be appropriately determined. Therefore, by performing variable selection after normalizing the learning population data as described above, it is possible to further improve the accuracy of traffic congestion prediction in consideration of the size of the normal population of each group.

The block diagram used in the description of the above embodiment shows a block of functional units. These functional blocks (components) are realized by any combination of hardware and / or software. Further, the means for realizing each functional block is not particularly limited. That is, each functional block may be realized by one physically and / or logically coupled device, or directly and / or indirectly by two or more physically and / or logically separated devices. (For example, wired and / or wireless) may be connected and realized by these a plurality of devices.

For example, the congestion prediction model creating device according to the embodiment of the present invention may function as a computer that performs the processing according to the present embodiment. FIG. 8 is a diagram showing an example of a hardware configuration of a traffic jam prediction model creating device according to an embodiment of the present invention. The congestion prediction model creating device 100 described above may be physically configured as a computer device including a processor 1001, a memory 1002, a storage 1003, a communication device 1004, an input device 1005, an output device 1006, a bus 1007, and the like.

In the following description, the word "device" can be read as a circuit, a device, a unit, or the like. The hardware configuration of the traffic congestion prediction model creating device 100 may be configured to include one or more of the devices shown in the figure, or may be configured not to include some of the devices.

For each function of the congestion prediction model creating device 100, by loading predetermined software (program) on hardware such as the processor 1001 and the memory 1002, the processor 1001 performs an operation, and communication by the communication device 1004 and the memory 1002 are performed. And by controlling the reading and / or writing of data in the storage 1003.

Processor 1001 operates, for example, an operating system to control the entire computer. The processor 1001 may be composed of a central processing unit (CPU) including an interface with a peripheral device, a control device, an arithmetic unit, a register, and the like. For example, the variable selection unit 116, the dimension compression unit 117, the prediction regression equation creation unit 118, the traffic congestion prediction unit 122, the congestion prediction result display unit 123, and the like may be realized by the processor 1001.

Further, the processor 1001 reads a program (program code), a software module, and data from the storage 1003 and / or the communication device 1004 into the memory 1002, and executes various processes according to these. As the program, a program that causes a computer to execute at least a part of the operations described in the above-described embodiment is used. For example, the variable selection unit 116, the dimension compression unit 117, the regression equation creation unit 118 for prediction, the traffic congestion prediction unit 122, and the traffic congestion prediction result display unit 123 are stored in the memory 1002 and realized by a control program that operates on the processor 1001. It may be realized in the same manner for other functional blocks. Although it has been described that the various processes described above are executed by one processor 1001, they may be executed simultaneously or sequentially by two or more processors 1001. Processor 1001 may be mounted on one or more chips. The program may be transmitted from the network via a telecommunication line.

The memory 1002 is a computer-readable recording medium, and is composed of at least one such as a ROM (Read Only Memory), an EPROM (Erasable Programmable ROM), an EPROM (Electrically Erasable Programmable ROM), and a RAM (Random Access Memory). May be done. The memory 1002 may be referred to as a register, a cache, a main memory (main storage device), or the like. The memory 1002 can store a program (program code), a software module, or the like that can be executed to carry out the method according to the embodiment of the present invention.

The storage 1003 is a computer-readable recording medium, for example, an optical disk such as a CD-ROM (Compact Disc ROM), a hard disk drive, a flexible disk, an optical magnetic disk (for example, a compact disk, a digital versatile disk, a Blu-ray). It may consist of at least one (registered trademark) disk), smart card, flash memory (eg, card, stick, key drive), floppy (registered trademark) disk, magnetic strip, and the like. The storage 1003 may be referred to as an auxiliary storage device. The storage medium described above may be, for example, a database, server or other suitable medium containing memory 1002 and / or storage 1003.

The communication device 1004 is hardware (transmission / reception device) for performing communication between computers via a wired and / or wireless network, and is also referred to as, for example, a network device, a network controller, a network card, a communication module, or the like. For example, the learning population data acquisition unit 111, the learning congestion data acquisition unit 113, the congestion prediction position acquisition unit 120, the prediction population data acquisition unit 121, and the like may be realized by the communication device 1004.

The input device 1005 is an input device (for example, a keyboard, a mouse, a microphone, a switch, a button, a sensor, etc.) that receives an input from the outside. The output device 1006 is an output device (for example, a display, a speaker, an LED lamp, etc.) that outputs to the outside. The input device 1005 and the output device 1006 may have an integrated configuration (for example, a touch panel).

Further, each device such as the processor 1001 and the memory 1002 is connected by a bus 1007 for communicating information. Bus 1007 may be composed of a single bus, or may be composed of different buses between devices.

Further, the congestion prediction model creation device 100 is hardware such as a microprocessor, a digital signal processor (DSP: Digital Signal Processor), an ASIC (Application Specific Integrated Circuit), a PLD (Programmable Logic Device), and an FPGA (Field Programmable Gate Array). It may be configured to include, and a part or all of each functional block may be realized by the hardware. For example, processor 1001 may be implemented on at least one of these hardware.

Although the present invention has been described in detail above, it is clear to those skilled in the art that the present invention is not limited to the embodiments described herein. The present invention can be implemented as modifications and modifications without departing from the spirit and scope of the invention as defined by the claims. Therefore, the description of the present specification is for the purpose of exemplification and does not have any limiting meaning to the present invention.

For example, in the above-described embodiment, the traffic congestion prediction model creating device 100 functions as a traffic congestion prediction device by creating a traffic congestion prediction model and predicting traffic congestion using the traffic congestion prediction model. However, a configuration may be configured in which the creation of the traffic congestion prediction model and the prediction of traffic congestion are performed by different devices.

The learning population data and the forecasting population data may be real population data or fictitious population data.

In the above-described embodiment, the configuration in which the learning population data acquisition unit 111 acquires the learning population data from the outside and stores it in the learning population data storage unit 112 has been described, but the present invention is not limited to this. For example, the learning population data storage unit 112 stores the learning population data in advance, and the learning population data acquisition unit 111 acquires the learning population data from the learning population data storage unit 112 and is a congestion prediction model construction unit. The data may be passed to 115.

In the above-described embodiment, the configuration in which the learning congestion data acquisition unit 113 acquires the learning congestion data from the outside and stores it in the learning congestion data storage unit 114 has been described, but the present invention is not limited to this. For example, the learning congestion data storage unit 114 stores the congestion data in advance, and the learning congestion data acquisition unit 113 acquires the learning congestion data from the learning congestion data storage unit 114 and causes the congestion prediction model construction unit 115. It may be configured to pass the data.

In the above-described embodiment, the learning population data acquisition unit 111 acquires the population data sampled at noon on each day, but may acquire the population data every few days, or at a plurality of times on each day. Population data sampled in may be obtained. When the learning population data acquisition unit 111 acquires population data sampled at a plurality of times on each day, the number of groups in which the population data can be classified further increases. Even in this case, since the traffic jam prediction model creating device 100 selects the population data of the group related to the traffic jam, the increase in the amount of calculation required for constructing the traffic jam prediction model can be suppressed.

The sampling time of the population data acquired by the prediction population data acquisition unit 121 may correspond to the sampling time of the population data acquired by the learning population data acquisition unit 111. Therefore, the population data acquired by the forecasting population data acquisition unit 121 is not limited to the population data sampled at noon.

The traffic congestion prediction model construction unit 115 described above predicts the traffic congestion on the prediction date, but other than that, the traffic congestion prediction model for predicting the traffic congestion within a predetermined time width is constructed. It may be configured. In this case, the variable selection unit 116, the dimension compression unit 117, and the prediction regression equation creation unit 118 use the congestion data according to the predetermined time width as the learning congestion data.

In the above-described embodiment, the congestion prediction model construction unit 115 performs dimensional compression by the dimensional compression unit 117. However, the congestion prediction model construction unit 115 does not have to have the dimensional compression unit 117, and the dimensional compression may not be performed.

In the above-described embodiment, the variable selection unit 116 performs variable selection after normalizing the learning population data for each group. However, normalization of learning population data does not have to be performed.

The traffic jam prediction model creating device 100 may build a traffic jam prediction model corresponding to the acquired traffic jam prediction position each time the traffic jam prediction position is acquired. This makes it possible to build a congestion prediction model using the latest learning population data each time.

Each aspect / embodiment described herein includes LTE (Long Term Evolution), LTE-A (LTE-Advanced), SUPER 3G, IMT-Advanced, 4G, 5G, FRA (Future Radio Access), W-CDMA. (Registered Trademarks), GSM (Registered Trademarks), CDMA2000, UMB (Ultra Mobile Broadband), IEEE 802.11 (Wi-Fi), IEEE 802.16 (WiMAX), IEEE 802.20, UWB (Ultra-WideBand), It may be applied to systems utilizing Bluetooth®, other suitable systems and / or next-generation systems extended based on them.

The order of the processing procedures, sequences, flowcharts, etc. of each aspect / embodiment described in the present specification may be changed as long as there is no contradiction. For example, the methods described herein present elements of various steps in an exemplary order, and are not limited to the particular order presented.

The input / output information and the like may be stored in a specific location (for example, memory) or may be managed by a management table. Input / output information and the like can be overwritten, updated, or added. The output information and the like may be deleted. The input information or the like may be transmitted to another device.

The determination may be made by a value represented by 1 bit (0 or 1), by a true / false value (Boolean: true or false), or by comparing numerical values (for example, a predetermined value). It may be done by comparison with the value).

The data described herein may be represented using any of a variety of different techniques. For example, the data that can be referred to throughout the above description may be represented by voltage, current, electromagnetic waves, magnetic fields or magnetic particles, light fields or photons, or any combination thereof.

Further, the data, parameters, etc. described in the present specification may be represented by absolute values, relative values from predetermined values, or other corresponding information. .. For example, the radio resource may be indexed. The names used for the above parameters are not limited in any way.

The terms "determining" and "determining" as used herein may include a wide variety of actions. "Judgment" and "decision" are, for example, judgment, calculation, computing, processing, deriving, investigating, looking up (for example, table). , Searching in a database or another data structure), ascertaining can be considered as a "judgment" or "decision". Also, "judgment" and "decision" are receiving (for example, receiving information), transmitting (for example, transmitting information), input (input), output (output), and access. (Accessing) (for example, accessing data in memory) may be regarded as "judgment" or "decision". In addition, "judgment" and "decision" mean that the things such as solving, selecting, choosing, establishing, and comparing are regarded as "judgment" and "decision". Can include. That is, "judgment" and "decision" may include considering some action as "judgment" and "decision".

The phrase "based on" as used herein does not mean "based on" unless otherwise stated. In other words, the statement "based on" means both "based only" and "at least based on".

As long as "include", "including", and variations thereof are used within the scope of the present specification or claims, these terms are as comprehensive as the term "include". Intended to be targeted. Moreover, the term "or" as used herein or in the claims is intended not to be an exclusive OR.

In the whole of the present disclosure, if the context clearly does not indicate the singular, it shall include more than one.

100 ... Congestion prediction model creation device, 111 ... Learning population data acquisition unit, 113 ... Learning congestion data acquisition unit, 115 ... Congestion prediction model construction unit, 121 ... Prediction population data acquisition unit, 122 ... Traffic congestion prediction unit.

Claims (3)

It is a traffic jam prediction model creation device that creates a traffic jam prediction model used for traffic jam prediction.
A population data acquisition unit that acquires population data of multiple groups classified according to the position and attributes of people,
The traffic jam data acquisition unit that acquires traffic jam data indicating the traffic jam status according to the position,
After normalizing the population data for each group, regression analysis is performed using the normalized population data for each group as an explanatory variable and the congestion data at the congestion prediction position for predicting traffic congestion as an objective variable. Therefore, the population data of the group having a relatively high contribution to the traffic jam data at the traffic jam prediction position is selected from the plurality of groups, and the selected population data and the traffic jam data at the traffic jam prediction position are selected. and a congestion prediction model constructing section for constructing the traffic jam prediction model by based rather regression analysis DOO, traffic jam prediction model generating apparatus. The traffic congestion prediction model creating device according to claim 1, wherein the attribute includes the place of residence of the person. The forecasting population data acquisition department, which acquires population data used for forecasting traffic congestion,
The traffic congestion prediction model creating device according to claim 1 or 2, further comprising a traffic congestion prediction unit that predicts traffic congestion based on the congestion prediction model and population data used for predicting traffic congestion.

Images