Title:
HUMAN FLOW PREDICTING DEVICE, METHOD, AND PROGRAM
Document Type and Number:
Japanese Patent JP2019095895
Kind Code:
A
Abstract:
To estimate a human flow differing for each cluster classified by user properties using accepting aggregate data of human position information and movement history data relating to a small number of users as inputs, and accurately predict the amount of human flow for the respective clusters.SOLUTION: A human flow predicting device 90 comprises: a clustering unit 7 for classifying a plurality of users into multiple kinds of clusters from movement history data and calculating, for each cluster, each of the rate of human outflow amount and the rate of human inflow amount at each time of day at each observation point; a parameter estimation unit 10 for estimating, for each cluster, the parameter and latent variable of transition probability so as to optimize an objective function on the basis of the aggregate data on outflow amount, the aggregate data on inflow amount, and data on the number of observed persons, as well as the rate of human outflow amount and the rate of human inflow amount; and a predicted human inflow amount calculation unit 13 for calculating a predicted inflow amount at a target observation point at a prediction time of day with regard to a target observation point for the target observation point, target cluster, and prediction time of day accepted for input by a search unit 5.SELECTED DRAWING: Figure 1
Inventors:
TANAKA SUKENORI
TODA HIROYUKI
TODA HIROYUKI
Application Number:
JP2017222734A
Publication Date:
June 20, 2019
Filing Date:
November 20, 2017
Export Citation:
Assignee:
NIPPON TELEGRAPH & TELEPHONE
International Classes:
G06F16/00; G06Q10/04
Domestic Patent References:
| JP2015076005A | 2015-04-20 |
Foreign References:
| WO2014155828A1 | 2014-10-02 |
Attorney, Agent or Firm:
Patent Service Corporation Taiyo International Patent Office