Behavior prediction apparatus, behavior prediction method, and behavior prediction program

• US 20060179014A1
• Filed: 09/13/2005
• Published: 08/10/2006
• Est. Priority Date: 02/09/2005
• Status: Active Grant

• Current Assignee: Kabushiki Kaisha Toshiba (Toshiba Corporation)
• Original Assignee: Kabushiki Kaisha Toshiba (Toshiba Corporation)
• Inventors: Yoshida, Takufumi, Koiso, Takashi
• Granted Patent: US 7,778,863 B2
• Time in Patent Office: Days
• US Class Current: 706/12
• CPC Class Codes: G06Q 30/02 Marketing; Price estimation or determination; Fundraising
• CPC Class Codes: G06Q 30/0202 Market predictions or forecasting for commercial activities
• CPC Class Codes: G06Q 30/0204 Market segmentation

Abstract

According to an aspect of the present invention, there is provided with a behavior prediction apparatus including: an area attribute database which stores area attributes of each of a plurality of areas in an activity area of customers; an inter-area attribute database which stores inter-area attributes between two areas; a behavior history database which stores behavior history of the customers in the activity area; a movement frequency database generator which generates a movement frequency database representing a movement frequency between the two areas on the basis of the individual behavior history database; and a movement frequency model estimator which estimates a value of a parameter in a movement frequency model having area attributes and inter-area attributes as its input and a movement frequency as its output.

Claims

1. A behavior prediction apparatus comprising:

an area attribute database which stores area attributes of each of a plurality of areas in an activity area of customers;

an inter-area attribute database which stores inter-area attributes between two areas;

a behavior history database which stores behavior history of the customers in the activity area;

a movement frequency database generator which generates a movement frequency database representing a movement frequency between the two areas on the basis of the behavior history database; and

a movement frequency model estimator which estimates a value of a parameter in a movement frequency model having area attributes and inter-area attributes as its input and a movement frequency as its output.

2. The behavior prediction apparatus according to claim 1, wherein the movement frequency database generator generates a normalized movement frequency database as the movement frequency database.

3. The behavior prediction apparatus according to claim 1, further comprising:

a prediction database which stores area attributes and inter-area attributes for an activity area as a subject of prediction; and

a movement frequency prediction unit which inputs area attributes of each of two areas and inter-area attributes between the two areas in the prediction database to the movement frequency model, and acquires a movement frequency between the two areas.

4. The behavior prediction apparatus according to claim 3, wherein the movement frequency model is a regression model having a movement frequency database as its output, and

the movement frequency prediction unit generates a prediction movement frequency database which represents the movement frequency between the two areas in the activity area as the subject of the prediction, by using the regression model.

5. The behavior prediction apparatus according to claim 3, wherein the movement frequency model outputs probability distribution of movement frequency between the two areas, and

the movement frequency prediction unit generates a prediction movement frequency database which represents the movement frequency between the two areas in the activity area as the subject of the prediction, based on the probability distribution.

6. The behavior prediction apparatus according to claim 5, wherein the movement frequency prediction unit refers to the probability distribution output from the movement frequency model, and predicts a movement frequency corresponding to a maximum probability, or an average movement frequency based on the probability distribution, as the movement frequency between the two areas in the activity area as the subject of the prediction.

7. The behavior prediction apparatus according to claim 1, further comprising:

an individual attribute database which stores individual attributes of each of individuals; and

a behavior selection model estimator which estimates a value of a parameter in a behavior selection model to obtain behavior selection of the individual from the individual attributes of the individual, area attributes, inter-area attributes, a movement frequency, and behavior history of the individual ending at a time point.

8. The behavior prediction apparatus according to claim 7, wherein the behavior selection model estimator estimates a value of a parameter in a next area selection model to obtain an area as destination of next movement of the individual, as the behavior selection model.

9. The behavior prediction apparatus according to claim 7, further comprising a behavior selection prediction unit which inputs individual attributes of an individual as a subject of prediction, area attributes and inter-area attributes in an activity area as a subject of the prediction, a movement frequency in the activity area as the subject of the prediction, and behavior history of the individual ending at a time point in the activity area as the subject of the prediction to the behavior selection model, and predicts behavior selection of the individual as the subject of the prediction.

10. The behavior prediction apparatus according to claim 9, wherein the behavior selection model uses area attributes of an area where the individual as the subject of the prediction is located and inter-area attributes including this area, as the area attributes and the inter-area attributes in the activity area as the subject of the prediction, and uses a movement frequency of inter-area having at least this area as a shift origin or a shift destination, as the movement frequency in the activity area as the subject of the prediction.

11. The behavior prediction apparatus according to claim 9, wherein the behavior selection model is a regression model having behavior selection as its output, and

the behavior selection prediction unit predicts behavior selection of the individual from the output of the behavior selection model.

12. The behavior prediction apparatus according to claim 9, wherein the behavior selection model is a model that outputs probabilities with which respective behavior selections corresponding to the behavior selection model are selected by the individual, and

the behavior selection prediction unit predicts behavior selection of the individual from the probabilities output by the behavior selection model.

13. The behavior prediction apparatus according to claim 12, wherein

the behavior selection prediction unit compares probabilities output from the behavior selection model, and predicts behavior selection corresponding to a maximum probability as behavior selection of the individual.

14. The behavior prediction apparatus according to claim 9, wherein the behavior selection prediction unit predicts transition of behavior selection of the individual as the subject of the prediction by using a simulation model, and successively records the behavior selection of the individual as behavior history.

15. The behavior prediction apparatus according to claim 14, further comprising a moving trail data generator which generates moving trail data indicating movement history of the individual as the subject of the prediction, from the behavior history.

16. The behavior prediction apparatus according to claim 15, further comprising a moving trail analysis unit which conducts a moving trail analysis by using the moving trail data of a plurality of individuals.

17. A behavior prediction method comprising:

reading out data from a behavior history database which stores behavior history of customers in an activity area;

generating a movement frequency database representing a movement frequency between two areas in the activity area;

reading out data from an area attribute database which stores area attributes of each of a plurality of areas in the activity area, an inter-area attribute database which stores inter-area attributes between two areas, and the movement frequency database; and

estimating a value of a parameter in a movement frequency model having area attributes and inter-area attributes as its input and a movement frequency as its output.

18. The behavior prediction method according to claim 17, further comprising:

reading out data from an individual attribute database which stores individual attributes of each of individuals; and

estimating a value of a parameter in a behavior selection model to obtain behavior selection of the individual from the individual attributes of the individual, area attributes, inter-area attributes, a movement frequency, and behavior history of the individual ending at a time point.

19. A behavior prediction program to make a computer execute:

reading out data from a behavior history database which stores behavior history of customers in an activity area;

generating a movement frequency database representing a movement frequency between two areas in the activity area;

reading out data from an area attribute database which stores area attributes of each of a plurality of areas in the activity area, an inter-area attribute database which stores an inter-area attributes between two areas, and the movement frequency database; and

estimating a movement frequency model having area attributes and inter-area attributes as its input and a movement frequency as its output.

20. The behavior prediction program according to claim 19, to make the computer further execute:

reading out data from an individual attribute database which stores individual attributes of each of individuals; and

estimating a value of a parameter in a behavior selection model to obtain behavior selection of the individual from the individual attributes of the individual, area attributes, inter-area attributes, a movement frequency, and behavior history of the individual ending at a time point.