JP2018195244A - Od traffic prediction device, method, and program - Google Patents

Abstract

To accurately predict OD traffic data when external information having correlation with the OD traffic data is given.SOLUTION: An operation part 10 accepts OD traffic data and external information of each period. A parameter estimation part 16 resolves a data tensor representing the OD traffic data into the product of a core tensor of the period and a plurality of factor matrixes, resolves a matrix representing the external information of each time zone in the period into the product of a scale parameter and the plurality of factor matrixes, and resolves a matrix representing external information regarding a place into the product of the scale parameter and the plurality of factor matrixes for each period. The prediction part 22 predicts the OD traffic data of the period and the time zone of a prediction object based on the core tensor of each period and the plurality of factor matrixes obtained by the parameter prediction part 16.SELECTED DRAWING: Figure 3

Description

  The present invention relates to an OD traffic prediction apparatus, method, and program for predicting OD traffic data.

  OD traffic is the flow of people, things, information, etc. from Origin to Destination, for example, traffic volume and trade volume between regions (city, station, etc.).

OD traffic is treated as “related data” in the context of machine learning. The relation data is data representing a “relation” defined between a plurality of objects or a plurality of types of data, and can be represented using a “graph” (see FIG. 9A). Graph G = {V, E} is a set of vertices

And a set of edges corresponding to the pair of vertices

Defined by In the case of OD traffic data, the vertex corresponds to the traffic volume between each region, and the side corresponds to the traffic volume between the two regions. In machine learning literature, data having such a graph structure is usually described using a multidimensional array. First, i and j are introduced as indexes representing nodes. The relationship between the i-th node and the j-th node is written as x _ij . If i is regarded as a row index and j is regarded as a column index, a set of x _ij can be represented by one matrix X (see FIG. 9B). For example, traffic data between N regions can be expressed as an N × N matrix. The problem of estimating unobserved x _ij from other observed data is called “link prediction”.

OD traffic varies with time. For example, it can be considered that there is much movement during morning and evening rush hours, and little movement at midnight. Tensors are used to describe graphs that change over time. Assuming that the number of time zones is T, the data with time information added is an N × N × T third-floor tensor.

(See FIG. 10).

Each element x _ijt = is “the number of people who moved from region i to j in time zone t”. Also, when considering the day of the week in addition to the time zone, the 4th floor tensor of N x N x T x 7

The data is expressed by (see FIG. 11A). in this case

Each element _x′ijdt = is “the number of people who moved from the region i to j in the day of the week d and the time zone t”. By increasing the tensor dimension in this way, a third-order or higher polynomial relation can be expressed (see FIG. 11B). Tensor decomposition such as CP decomposition or Tucker decomposition is used for analysis of data represented by tensors. Tensor decomposition decomposes a data tensor into the form of a matrix product and gives a low-dimensional representation of the data (see FIG. 12).

DUNLAVY, Daniel M .; KOLDA, Tamara G .; ACAR, Evrim.Temporal link prediction using matrix and tensor factorizations.ACM Transactions on Knowledge Discovery from Data (TKDD), 2011, 5.2: 10. ERMI, Beyza; ACAR, Evrim; CEMGIL, A. Taylan. Link prediction via generalized cou-pled tensor factorisation.arXiv preprint arXiv: 1208.6231, 2012. NOMIKOS, Paul; MACGREGOR, John F. Multi-way partial least squares in monitoring batch processes.Chemometrics and intelligent laboratory systems, 1995, 30.1: 97-108.

  However, the conventional tensor decomposition does not consider autoregressiveness (relationship between the value of the previous time and the value of the current time), and is not suitable for modeling a graph that develops with time. For example, the OD traffic (the number of trips between the start bike station and the end bike station) in the bike share system increases with time as the bike share system spreads. Such OD traffic data needs to be modeled in consideration of time evolution. In addition, if the basic tensor decomposition method is used as it is, information on external factors (external information) cannot be handled. OD traffic, such as the number of trips in the bike sharing system and the number of people traveling between regions, may vary depending on external factors such as the weather and surrounding events. There are a plurality of models describing the time evolution of the graph (see Non-Patent Document 1), but these models cannot take in external information as they are. As a method for considering external information, a tensor decomposition with auxiliary information (Non-Patent Document 2) and a method using external information as a regression variable (Non-Patent Document 3) have been proposed. However, these models cannot take into account time evolution.

  Thus, with the conventional method, time development and external information cannot be considered simultaneously. For this reason, there is a problem that the influence of external factors is large and the graph that develops with time cannot be appropriately modeled, and the accuracy of link prediction is lowered.

  The present invention has been made in view of the above circumstances, and an OD traffic prediction apparatus, method, and program capable of accurately predicting OD traffic data when external information having correlation with OD traffic data is given. The purpose is to provide.

  In order to achieve the above object, an OD traffic prediction device according to the present invention has a correlation between OD (Origin-Destination) traffic data in each time zone and each OD traffic data in each time zone. An external unit for receiving the external information and the external information about the place, and for each period, a data tensor representing the OD traffic data of each time period at the time period received by the operation unit is a core tensor of the period , A factor parameter relating to the time zone, and a factor parameter including a factor matrix relating to the location, and a tensor or matrix representing external information of each time zone at the time period as a scale parameter for the external information of the time zone And a plurality of factor matrices including a factor matrix related to a time zone among the plurality of factor matrices. A parameter estimation unit that decomposes a tensor or matrix representing external information related to the location into a product of a scale parameter for the external information related to the location and a plurality of factor matrices including the factor matrix related to the location among the plurality of factor matrices Based on the core tensor for each period obtained by the parameter estimation unit, the factor matrix for the time zone, and the plurality of factor matrices including the factor matrix for the location, And a predicting unit for predicting the OD traffic data.

  Also, in the OD traffic prediction method according to the present invention, the operation unit correlates with OD (Origin-Destination) traffic data in each time zone in each time period and the OD traffic data in each time zone. Information, and external information related to the location, the parameter estimation unit, for each time period, a data tensor representing the OD traffic data of each time zone at the time period received by the operation unit, as the core tensor of the time period , A factor parameter relating to the time zone, and a factor parameter including a factor matrix relating to the location, and a tensor or matrix representing external information of each time zone at the time period as a scale parameter for the external information of the time zone And a plurality of factor matrices including a factor matrix related to a time zone among the plurality of factor matrices. A tensor or matrix representing external information related to the location is decomposed into a product of a scale parameter for the external information related to the location and a plurality of factor matrices including a factor matrix related to the location among the plurality of factor matrices, and a prediction unit Is based on the core tensor of each period obtained by the parameter estimation unit, the factor matrix related to the time zone, and the plurality of factor matrices including the factor matrix related to the location, and the time and time zone to be predicted OD traffic data is predicted.

  Moreover, the program of this invention is a program for functioning a computer as each part which comprises said OD traffic prediction apparatus.

  As described above, according to the OD traffic prediction apparatus, method, and program of the present invention, for each period, the data tensor representing the OD traffic data in each time zone in the period is the core tensor of the period. A tensor or matrix that is decomposed into a product with a plurality of factor matrices and represents external information of each time zone at the time period, a plurality of factor matrices including a scale parameter for the external information of the time zone and a factor matrix related to the time zone And the tensor or matrix representing the external information about the location is decomposed into a product of a scale parameter for the external information about the location and a plurality of factor matrices including a factor matrix for the location, and the core of each period Based on a tensor and the plurality of factor matrices including a factor matrix for the time zone and a factor matrix for the location. Thus, by predicting the OD traffic data of the prediction target time and time zone, it is possible to accurately predict the OD traffic data when external information having a correlation with the OD traffic data is given. .

It is a figure for demonstrating the tensor simultaneous decomposition | disassembly extended by introducing time evolution. It is a figure for demonstrating the time development of a core tensor. It is a block diagram of the OD traffic prediction apparatus in an embodiment of the present invention. It is a figure which shows an example of OD traffic log | history. It is a figure which shows an example of external information. It is a figure which shows the algorithm of parameter learning. It is a flowchart which shows the learning process routine of the OD traffic prediction apparatus in embodiment of this invention. It is a flowchart which shows the OD traffic prediction processing routine of the OD traffic prediction apparatus in embodiment of this invention. (A) The figure which shows the graph expression of OD flow, (B) It is the figure which shows the OD flow matrix. It is a figure which shows the tensor expression of OD flow. It is a figure which shows the variation of the tensor expression of OD flow. It is a figure for demonstrating tensor decomposition.

  Hereinafter, embodiments of the present invention will be described in detail with reference to the drawings.

<Overview>
In the embodiment of the present invention, a technique for simultaneously handling the influence of external factors and the time evolution is used by introducing autoregression into the simultaneous tensor decomposition method. The tensor simultaneous decomposition method simultaneously decomposes a tensor (data tensor) representing data and a tensor (or matrix) representing external information while sharing a factor matrix, and captures the indirect relationship between data and external information. (See FIG. 1). At that time, the data tensor is approximated by a product of a small tensor called a core tensor and a plurality of factor matrices.

  In the embodiment of the present invention, autoregressiveness of the core tensor is assumed. Specifically, the probability that a core tensor at a certain time transitions to another core tensor at the next time step is modeled and learned from data (see FIG. 2).

  Therefore, it is possible to predict with high accuracy the OD traffic data that is greatly influenced by external factors and that develops with time. In addition, by modeling the time evolution of periodicity, a relatively long-term prediction can be made.

  Hereinafter, a case will be described in which near-future OD traffic is predicted based on OD traffic data that is a history of OD (Origin-Destination) traffic and external information correlated with the OD traffic data.

<Configuration of OD Traffic Prediction Device According to Embodiment of the Present Invention>
Next, the configuration of the OD traffic prediction apparatus according to the embodiment of the present invention will be described. As shown in FIG. 3, the OD traffic prediction apparatus 100 according to the embodiment of the present invention includes a CPU, a RAM, and a ROM that stores programs and various data for executing processing routines to be described later. Can be configured with a computer. The OD traffic prediction device 100 is based on OD (Origin-Destination) traffic data in each time zone in each time period, external information in each time zone in each time period, and external information related to a location, which is correlated with the OD traffic data. Thus, the OD traffic of the prediction target time, time zone, and place is predicted. Functionally, the OD traffic prediction device 100 includes an operation unit 10, an OD history storage unit 12, an external information storage unit 14, a parameter estimation unit 16, a parameter storage unit 18, as shown in FIG. A search unit 20, a prediction unit 22, and an output unit 24 are provided.

  The operation unit 10 accepts various operations from the user for data stored in an OD history storage unit 12 and an external information storage unit 14 which will be described later. The various operations include operations for registering, correcting, and deleting information stored in the OD history storage unit 12 and the external information storage unit 14.

  The input means of the operation unit 10 may be anything such as a keyboard, mouse, menu screen, or touch panel. The operation unit 10 can be realized by a device driver for input means such as a mouse or control software for a menu screen.

  The search unit 20 receives information on the time (week), day of the week, time zone, location, and time difference that are subject to prediction. The input unit of the search unit 20 may be anything such as a keyboard, mouse, menu screen, or touch panel. The search unit 20 can be realized by a device driver for input means such as a mouse or control software for a menu screen.

The OD history storage unit 12 stores OD traffic history information that can be analyzed by the apparatus, reads out the OD traffic history information according to a request from the parameter estimation unit 16, and transmits the information to the parameter estimation unit 16. The OD traffic history information is, for example, the number of transitions between people in the region, `` People are (1) what day (2) what day of the week (3) what time (4) where (5) where (6) which (7) How much has it moved with a time difference of about 7 ”(see FIG. 4). Such data is a tensor consisting of five axes corresponding to (2) day of week d, (3) time zone t, (4) start point location ID i, (5) end point location ID j, and (6) time difference l. Can express. _Let each component of the tensor be y _{i, j, l, d, t} . If the data of multiple weekly amount is present, the data set of the w th week is defined as Y ^w. Y ^w is the set of y ^w _{i, j, l, d, t}

It is. Here, y ^w _{i, j, l, d, t} is the number of people who moved with a time difference l from place i to j in the w week, d day of the week, and time zone t. D is the number of days, T is the number of time zones, N is the number of places, and L is the number of time differences to consider. If the number of weeks is W, the set of all data is

It is. Note that the OD history storage unit 12 may be provided outside the OD traffic prediction apparatus 100, and may be configured by a Web server that holds a Web page, a database server that includes a database, or the like.

The external information storage unit 14 stores external information that can be analyzed by the apparatus, reads external information according to a request from the parameter estimation unit 16, and transmits the information to the parameter estimation unit 16. External information is data related to external factors affecting OD traffic, such as weather (temperature, precipitation) and geographic information (geographic distance between regional pairs). As an example, let us consider a situation in which there is information on the temperature (see FIG. 5) for each week, day of the week, and time zone, the precipitation for each week, day of the week, and time zone, and the geographical distance between regions. Day w of week w, temperature p ^w _{dt of} time zone t, precipitation of day w of week ^w , q ^w _dt of precipitation of time zone t, and geographical distance between places i and j z ^w _{i, j} . Set of elements p ^w _dt

A matrix consisting of P ^w and a set of elements q ^w _dt

A matrix consisting of Q ^w and the set of elements z ^w _ij

Let Z ^w be a matrix consisting of. Each set of matrices

Write. Here, when the geographical distance z ^w _{i, j} between the places i and j does not change in each week, the geographical distance z ^w _{i, j} between the places i and _j is a common value in each week w. The external information storage unit 14 may be provided outside the OD traffic prediction device 100, and may be configured by a Web server that holds a Web page, a database server that includes a database, or the like.

Based on the information stored in the OD history storage unit 12 and the external information storage unit 14, the parameter estimation unit 16 extracts a low-dimensional representation of these information and estimates the time evolution. The procedure will be described using the above example. Consider applying CP decomposition to a tensor representing OD traffic history. CP decomposition is a technique that approximates a tensor with a product of a small tensor called a core tensor and a factor matrix. The goal of this embodiment is to find a set of core tensor and factor matrix that reproduces the tensor. The data tensor y ^w _{i, j, l, d, t} of OD traffic is decomposed as follows.

here

Is the factor matrix and λ ^w _k is the diagonal component of the core tensor. In the case of CP decomposition, the core tensor has only a diagonal component. K is the rank number of CP decomposition, and it is determined based on prior knowledge or determined by cross validation. Each set of factor matrix elements

And the set of diagonal components of the core tensor

Write. Similarly, the matrices Z, P, and Q representing external information are decomposed as follows using Θ ⁽¹⁾ , Θ ⁽²⁾ , Θ ⁽⁴⁾ , and Θ ⁽⁵⁾ .

Here, γ ^z _k , γ ^p _k , and γ ^q _k are scale parameters. Each set of scale parameters

far. Differences in scale between external information and OD traffic data are absorbed by this parameter. By sharing the factor matrix of the data tensor Y ^{w and} the factor matrix of the external information matrix P ^w , Q ^w , Z ^w , tensor decomposition considering external information becomes possible. Assume that the core tensor evolves over time to account for autoregressiveness. In the present embodiment, as an example, it is assumed that the value of this week's core tensor is determined by linear transformation of the value of the last week's core tensor.

  In addition, the likelihood of this model can be written as follows.

Where! Represents the factorial. The goal of this embodiment is to set parameters {Θ ⁽¹⁾ , Θ ⁽²⁾ , Θ ⁽³⁾ , Θ ⁽⁴⁾ , Θ ⁽⁵⁾ , Γ ^z , Γ that minimize the likelihood L ^p , Γ ^q , Λ}. Any method can be used for parameter optimization, but each parameter is updated alternately by updating the parameters to be updated so as to minimize the likelihood L by fixing other parameters than the parameters to be updated. (For example, Non-Patent Document 4) is widely used. The parameter learning procedure is shown in FIG.

[Non-Patent Document 4]: M. Welling and M. Weber. Positive tensor factorization. Pattern Recognition Letters, 22 (12): 1255-1261, 2001.

  The estimated parameter set is stored in the parameter storage unit 20. The parameter storage unit 20 may be anything as long as such information is stored and can be restored. For example, it is stored in a specific area of a database or a general-purpose storage device (memory or hard disk device) provided in advance.

Accordingly, the parameter estimation unit 16 performs, for each week w, each time zone t in the week w based on the OD traffic data stored in the OD history storage unit 12 and the external information stored in the external information storage unit 14. Data tensor Y ^w _t representing the OD traffic data of the week, the core tensor λ ^w _k of the week w, the factor matrix θ ⁽¹⁾ _ik , θ ⁽²⁾ _jk , θ ⁽³⁾ _lk, and the factor related to the day Decomposing the matrix θ ⁽⁴⁾ _dk and the factor matrix θ ⁽⁵⁾ _tk related to the time zone into a matrix P representing the external information of each time zone in the week w, and the scale parameter γ ^p _k for the external information , The factor matrix θ ⁽⁴⁾ _dk for the day and the factor matrix θ ⁽⁵⁾ _tk for the time zone are decomposed into products, and the matrix Q representing the external information for each time zone in the week w γ ^q _k and, factor matrix for the day θ ⁽⁴⁾ _dk and, time Decomposed to the product of the factor matrix theta ⁽⁵⁾ _tk relates, the external information Z relating to the location, the scale parameter gamma ^z _k to external information about the location, the factor matrix about the location theta ⁽¹⁾ _ik, and theta ⁽²⁾ _jk It decomposes into the product of

Specifically, the OD traffic data of each time zone in each week w, the external information of each time zone in each week w, and the location calculated from the result of decomposing the data tensor and the matrix representing the external information Each model value of external information

The core tensor λ ^w _k of each week w, the factor matrices θ ⁽¹⁾ _ik , θ ⁽²⁾ _jk , θ ⁽³⁾ _lk, and the factor matrix θ ^{(4 )} Repeat updating _dk , factor matrix θ ⁽⁵⁾ _tk for time zone, scale parameters γ ^p _k , γ ^q _k for external information, and scale parameter γ ^z _k for external information for location.

The parameter estimation unit 16 further estimates a parameter α _k related to time evolution obtained from the core tensor λ ^w _k of each week w.

The prediction unit 22 relates to the prediction target week, day of the week, time zone, location, and time difference received by the operation unit 10, the core tensor λ ^w _{k of} each week w stored in the parameter storage unit 18, and the location Based on the factor matrix θ ⁽¹⁾ _ik , θ ⁽²⁾ _jk , θ ⁽³⁾ _lk , the factor matrix θ ⁽⁴⁾ _dk for the day, and the factor matrix θ ⁽⁵⁾ _tk for the time zone OD traffic data for the week, day of week, time zone, location, and time difference. At this time, using the parameter α _k related to time evolution, OD traffic data related to the time, time zone, location, and time difference of the prediction target is predicted.

  For example, in the case of the above-described example, the number of people who move from the location i to the j in the time zone t in the day d and the time zone t in the next week (w + 1 week) can be predicted by the following equation.

  The output unit 24 outputs the OD traffic data predicted by the prediction unit 22 as a result. Here, output is a concept including display on a display, printing on a printer, sound output, transmission to an external device, and the like. The output unit 24 may be considered as including or not including an output device such as a display or a speaker. The output unit 24 may be implemented by output device driver software, or output device driver software and an output device.

<Operation of OD Traffic Prediction Device According to Embodiment of the Present Invention>
Next, the operation of the OD traffic prediction apparatus 100 according to the embodiment of the present invention will be described.

<Learning processing routine>
First, when OD traffic history information is input from the operation unit 10, the OD traffic prediction device 100 stores the OD traffic history information in the OD history storage unit 12, and when external information is input from the operation unit 10, Information is stored in the external information storage unit 14. And the OD traffic prediction apparatus 100 performs the learning process routine shown in FIG.

First, in step S100, the core tensor λ ^w _{k for} each week w, the factor matrices θ ⁽¹⁾ _ik , θ ⁽²⁾ _jk , θ ⁽³⁾ _lk for the _location , the factor matrix θ ⁽⁴⁾ _dk for the day, A factor matrix θ ⁽⁵⁾ _tk related to the time zone, scale parameters γ ^p _k and γ ^q _k for external information, and a scale parameter γ ^z _k for external information related to location are initialized.

In step S102, a factor matrix θ ⁽¹⁾ _ik , θ ⁽²⁾ _jk , θ ⁽³⁾ _lk related to _location , a factor matrix θ ⁽⁴⁾ _dk related to day, a factor matrix θ ⁽⁵⁾ _tk related to time zone, Based on the scale parameters γ ^p _k , γ ^q _k for the external information and the scale parameter γ ^z _k for the external information about the location, the core tensor of each week w so as to minimize the likelihood shown in the above equation (9). Update λ ^w _k .

In step S104, based on the core tensor λ ^w _{k for} each week w, the scale parameters γ ^p _k and γ ^q _k for external information, and the scale parameter γ ^z _k for external information related to location, the above equation (9) is shown. Factor matrix θ ⁽¹⁾ _ik , θ ⁽²⁾ _jk , θ ⁽³⁾ _{lk for location} , factor matrix θ ⁽⁴⁾ _dk for day, and factor matrix θ ^{( 5)} Update _tk .

In step S105, the core tensor λ ^w _{k for} each week w, the factor matrix θ ⁽¹⁾ _ik , θ ⁽²⁾ _jk , θ ⁽³⁾ _lk for the _location , the factor matrix θ ⁽⁴⁾ _dk for the day, and the time zone The scale parameters γ ^p _k and γ ^q _k for the external information and the scale for the external information related to the location so as to minimize the likelihood shown in the above equation (9) based on the factor matrix θ ⁽⁵⁾ _tk regarding Update the parameter γ ^z _k .

  In step S106, it is determined whether or not a predetermined convergence determination condition is satisfied. If the convergence determination condition is not satisfied, the process returns to step S102. On the other hand, if the convergence determination condition is satisfied, step S106 is performed. Proceed to S108.

  As the convergence determination condition, it may be used that the estimated change amount of each parameter is equal to or less than a threshold value or that a predetermined number of repetitions has been reached.

In step S108, the core tensor λ ^w _k of each week w finally updated in step S102 to step S105, the factor matrices θ ⁽¹⁾ _ik , θ ⁽²⁾ _jk , θ ⁽³⁾ _{lk regarding the location} , Factor matrix θ ⁽⁴⁾ _dk for day, factor matrix θ ⁽⁵⁾ _tk for time zone, scale parameters γ ^p _k , γ ^q _k for external information, scale parameter γ ^z _k for external information about location, and each week The parameter α _k related to the time evolution estimated from the core tensor λ ^w _{k of} ^w is stored in the parameter storage unit 18 and the learning processing routine is terminated.

<OD traffic prediction processing routine>
Next, the OD traffic prediction processing routine shown in FIG. 8 will be described.

The above learning processing routine is executed, the core tensor λ ^w _k for each week w, the factor matrix θ ⁽¹⁾ _ik , θ ⁽²⁾ _jk , θ ⁽³⁾ _lk, and the factor matrix for the day are stored in the parameter storage unit 20. θ ⁽⁴⁾ _dk , factor matrix for time zone θ ⁽⁵⁾ _tk , scale parameters γ ^p _k , γ ^q _k for external information, scale parameter γ ^z _k for external information about location, parameter α for time evolution _k is stored, and when the information regarding the prediction target week, day of week, time zone, location, and time difference is input, the OD traffic prediction device 100 executes the OD traffic prediction processing routine shown in FIG.

  In step S <b> 110, the operation unit 10 receives information regarding the prediction target week, day of the week, time zone, location, and time difference.

In step S112, the core tensor λ ^w _k of each week w stored in the parameter storage unit 18, the factor matrix θ ⁽¹⁾ _ik , θ ⁽²⁾ _jk , θ ⁽³⁾ _lk for the _location, and the factor matrix for the day _{Read out} θ ⁽⁴⁾ _dk , factor matrix θ ⁽⁵⁾ _tk related to time zone, and parameter α _k related to time evolution.

  In step S114, based on the information on the prediction target week, day of the week, time zone, location, and time difference, and each parameter read in step S112, the prediction target week, time zone Predict OD traffic data regarding location, time, and time differences.

  In step S116, the output unit 24 outputs OD traffic data regarding the prediction target week, time zone, place, and time difference predicted in step S114 as a result, and ends the OD traffic prediction processing routine.

  As described above, according to the OD traffic prediction apparatus according to the embodiment of the present invention, for each week w, the data tensor representing the OD traffic data of each time zone in the week w is represented by the core tensor of the week w. And a matrix of a plurality of factor matrices, and a matrix representing the external information of each time zone in the week w, a scale parameter for the external information of the time zone, and a plurality of factor matrices including a factor matrix related to the time zone, The matrix representing the external information about the location is decomposed into a product of a scale parameter for the external information about the location and a plurality of factor matrices including the factor matrix about the location, the core tensor for each week w, and the time zone And a plurality of factor matrices including a factor matrix relating to location, and an OD traffic data of a week, a time zone, and a location to be predicted. By predicting the data, the OD traffic data when applying an external information having a correlation with OD traffic data can be predicted accurately.

  Note that the present invention is not limited to the above-described embodiment, and various modifications and applications are possible without departing from the gist of the present invention.

  For example, in the above embodiment, the case where a matrix representing external information is used has been described as an example. However, the present invention is not limited to this, and a tensor representing external information may be used.

  Further, the above-described OD traffic prediction apparatus 100 has a computer system inside, but the “computer system” includes a homepage providing environment (or display environment) if a WWW system is used. Shall be.

  Further, in the present specification, the embodiment has been described in which the program is installed in advance. However, the program can be provided by being stored in a computer-readable recording medium or provided via a network. It is also possible to do.

DESCRIPTION OF SYMBOLS 10 Operation part 12 OD history storage part 14 External information storage part 16 Parameter estimation part 18 Parameter storage part 18 History storage part 20 Search part 22 Prediction part 24 Output part 100 OD traffic prediction apparatus

Claims (7)

An operation unit for receiving OD (Origin-Destination) traffic data in each time zone in each time period, external information in each time zone in each time period, and external information related to the location, which is correlated with the OD traffic data;
For each time period, a data tensor representing the OD traffic data of each time zone at the time period received by the operation unit includes a plurality of core tensors of the time period, a factor matrix relating to the time zone, and a factor matrix relating to the location. Decomposed into a product with a factor matrix,
A tensor or matrix representing external information of each time zone in the time period is a product of a scale parameter for the external information of the time zone and a plurality of factor matrices including a factor matrix related to the time zone among the plurality of factor matrices. Disassembled into
A parameter estimation unit that decomposes a tensor or matrix representing external information related to the location into a product of a scale parameter for the external information related to the location and a plurality of factor matrices including the factor matrix related to the location among the plurality of factor matrices When,
Based on the core tensor of each period obtained by the parameter estimation unit, the factor matrix related to the time zone, and the plurality of factor matrices including the factor matrix related to the location, the OD of the prediction target time and time zone A prediction unit for predicting traffic data;
OD traffic prediction device including The parameter estimation unit further estimates a parameter relating to time evolution obtained from the core tensor of each period,
The OD traffic prediction apparatus according to claim 1, wherein the prediction unit predicts OD traffic data of a prediction target period and time zone using the parameter related to time evolution.   The parameter estimator calculates the OD traffic data of each time zone at each time period, each time zone at each time period, calculated from the result of decomposing the data tensor and the result of decomposing a tensor or matrix representing the external information. A plurality of factor matrices including a core tensor for each period, a factor matrix for the time period, and a factor matrix for the location, so as to optimize the likelihood of the model value of the external information and the external information regarding the location; The OD traffic prediction apparatus according to claim 1, wherein the scale parameter for the external information in the time zone and the scale parameter for the external information regarding the place are repeatedly updated. The operation unit receives OD (Origin-Destination) traffic data for each time zone at each time period, external information for each time zone at each time period, and external information related to the location, which is correlated with the OD traffic data,
For each time period, the parameter estimation unit represents a data tensor representing the OD traffic data of each time zone at the time period received by the operation unit, a core tensor of the time period, a factor matrix related to the time period, and a factor related to the location Decompose into products with multiple factor matrices, including matrices,
A tensor or matrix representing external information of each time zone in the time period is a product of a scale parameter for the external information of the time zone and a plurality of factor matrices including a factor matrix related to the time zone among the plurality of factor matrices. Disassembled into
Decomposing a tensor or matrix representing external information about the location into a product of a scale parameter for the external information about the location and a plurality of factor matrices including a factor matrix related to the location among the plurality of factor matrices;
Based on the core tensor of each period obtained by the parameter estimation unit, the factor matrix related to the time zone, and the plurality of factor matrices including the factor matrix related to the place, the prediction unit is configured to calculate the time to be predicted and An OD traffic prediction method for predicting OD traffic data in a time zone. In the estimation by the parameter estimator, a parameter relating to time evolution obtained from the core tensor of each period is further estimated,
The OD traffic prediction method according to claim 4, wherein the prediction unit predicts OD traffic data of a prediction target time period and a time zone by using the parameter related to the time evolution.   In the estimation by the parameter estimation unit, the OD traffic data of each time zone in each time period, calculated from the result of decomposing the data tensor and the result of decomposing the tensor or matrix representing the external information, The plurality of times including a core tensor for each period, a factor matrix for the time zone, and a factor matrix for the location so as to optimize the likelihood of the model value of the external information for each time zone and the external information about the location The OD traffic prediction method according to claim 4 or 5, wherein updating of the factor matrix, the scale parameter for the external information of the time zone, and the scale parameter for the external information of the place is repeated.   The program for functioning a computer as each part which comprises the OD traffic prediction apparatus of any one of Claims 1-3.