Action pattern analysis device, action pattern analysis method, and action pattern analysis program

• US 9,497,584 B2
• Filed: 01/12/2012
• Issued: 11/15/2016
• Est. Priority Date: 01/14/2011
• Status: Active Grant

• Current Assignee: NEC Corporation
• Original Assignee: NEC Corporation
• Inventors: Nakahara, Fumitaka
• Primary Examiner(s): Heiber, Shantell L
• Application Number:
  US13/977,386
• Publication Number:
  US 20130273937A1
• Time in Patent Office: 1,769 Days
• US Class Current: 1/1
• CPC Class Codes: G01S 19/13 Receivers
• CPC Class Codes: G01S 19/24 Acquisition or tracking or demodulation of signals transmitted by the system synchronisation aspects of direct sequence spread spectrum modulation H04B1/7073
• CPC Class Codes: G01S 19/35 Constructional details or hardware or software details of the signal processing chain
• CPC Class Codes: G01S 19/42 Determining position
• CPC Class Codes: G06F 16/287 Visualization; Browsing
• CPC Class Codes: G06F 16/29 Geographical information databases
• CPC Class Codes: H04W 4/02 Services making use of location information

Abstract

A position information plot means 81 plots a position information log containing a user'"'"'s measurement position and measurement date/time on a multidimensional space defined by numerical information indicating the measurement position and a time. A slay point cluster extraction means 82 weights the position information log so that a Euclidean distance in a time direction relative to a position information space is easily determined as short, and clusters the weighted position information logs thereby to extract a stay point where the user frequently stays. A moving path cluster extraction means 84 weights a position information log at a non-stay point so that a Euclidean distance in a time direction relative to a position information space is easily determined as long, and clusters the weighted position information logs thereby to extract a user'"'"'s moving path.

Claims

1. An action pattern analysis device comprising:

hardware including a processor;

a stay point cluster extraction unit, implemented at least by the hardware, for clustering position information logs containing a user'"'"'s measurement position and measurement time thereby to extract a stay point where the user stays;

a non-stay point position information log extraction unit, implemented at least by the hardware, for extracting a set of position information logs from which a position information log extracted as the stay point is removed;

a moving path cluster extraction unit, implemented at least by the hardware, for clustering the extracted set of position information logs plotted in a multidimensional space defined by numerical information indicating the measurement position and a time based on a Euclidean distance on the multidimensional space thereby to extract a user'"'"'s moving path; and

a position information plot unit, implemented at least by the hardware, for plotting a position information log containing the user'"'"'s measurement position and the measurement time on the multidimensional space defined by the numerical information indicating the measurement position and the time,

wherein the moving path cluster extraction unit weights the extracted set of position information logs by applying a first weight to the Euclidean distance in a time direction in a position information space defined by numerical information indicating the measurement position on the multidimensional space, and clusters the position information logs weighted by applying the first weight,

wherein the position information plot unit plots a position information log containing values indicating latitude and longitude as a user'"'"'s measurement position on a 3D space defined by latitude, longitude and time,

wherein the stay point cluster extraction unit weights the position information log by applying a second weight to the Euclidean distance in the time direction relative to a latitude/longitude plan defined by latitude and longitude in the 3D space, and

wherein the moving path cluster extraction unit weights the extracted set of position information logs by applying the first weight to the Euclidean distance in the time direction relative to the latitude/longitude plan defined by latitude and longitude in the 3D space.

2. An action pattern analysis device comprising

hardware including a processor;

a moving vector function value calculation unit, implemented at least by the hardware, for calculating a time different in measurement time contained in position information logs and a distance between measurement positions as differential values between two adjacent position information logs sorted in an order of measurement time contained in the position information logs containing a user'"'"'s measurement position and measurement time;

an attribute determination unit, implemented at least by the hardware, for determining, based on the time difference in measurement time and the distance between the measurement positions, whether an attribute of the position information log is a stay attribute indicating a stay point where the user frequently stays or a non-stay attribute indicating a non-stay point on a user'"'"'s moving path;

a position information plot unit, implemented at least by the hardware, for plotting the position information log on a multidimensional space defined by numerical information indicating the measurement position and a time;

a stay point cluster extraction unit, implemented at least by the hardware, for weighting the position information log by applying a first weight to a Euclidean distance in a time direction relative to a position information space defined by the numerical information indicating the measurement position in the multidimensional space, and for clustering position information logs determined as having the stay attribute among the position information logs weighted by applying the first weight thereby to extract the stay point; and

a moving path cluster extraction unit, implemented at least by the hardware, for clustering the position information log plotted in the multidimensional space based on a Euclidean distance on the multidimensional space thereby to extract a user'"'"'s moving path,

wherein the moving path cluster extraction unit weights the extracted set of position information log by applying a second weight to the Euclidean distance in the time direction in a position information space defined by numerical information indicating the measurement position on the multidimensional space, and clusters the position information log weighted by applying the second weight.

3. The action pattern analysis device according to claim 2,

wherein the moving vector function value calculation unit calculates a differential value of a travelling direction angle indicating a travelling direction between measurement positions as a differential value between two position information logs,

the position information plot unit plots a record in which a travelling direction angle is added to a position information log containing a user'"'"'s measurement position and measurement time on a multidimensional space defined by numerical information indicating the measurement position, a time and the travelling direction angle,

and the moving path cluster extraction unit clusters a record in which a travelling direction angle is added to a position information log determined as having a moving path attribute thereby to extract a user'"'"'s moving path.

4. The action pattern analysis device according to claim 2, further comprising:

a travelling direction distance coefficient calculation unit, implemented at least by the hardware, for calculating, per position information log, a travelling direction distance coefficient which is smaller as a value of an angle formed by a travelling direction from a measurement position indicated by each of the position information logs and a direction connecting measurement positions indicated by two position information logs based on the two position information logs,

wherein the moving path cluster extraction unit performs weighting by multiplying the Euclidean distance by a travelling direction distance coefficient calculated by the travelling direction distance coefficient calculation unit, and clusters position information logs determined as having the non-stay attribute among the position information logs weighted by using the travelling direction distance coefficient thereby to extract a user'"'"'s moving path.

5. An action pattern analysis method comprising:

clustering position information logs containing a user'"'"'s measurement position and measurement time thereby to extract a stay point where the user stays;

extracting a set of position information logs from which a position information log extracted as the stay point is removed;

clustering the extracted set of position information logs plotted in a multidimensional space defined by numerical information indicating the measurement position and a time based on a Euclidean distance on the multidimensional space thereby to extract a user'"'"'s moving path, and

plotting a position information log containing a user'"'"'s measurement position and measurement time on a multidimensional space defined by numerical information indicating the measurement position and a time,

wherein the clustering the extracted set of position information logs comprises weighting the extracted set of position information logs by applying a first weight to the Euclidean distance in a time direction in a position information space defined by numerical information indicating the measurement position on the multidimensional space, and clustering the position information logs weighted by applying the first weight,

wherein the plotting the position information log comprises plotting a position information log containing values indicating latitude and longitude as a user'"'"'s measurement position on a 3D space defined by latitude, longitude and time,

wherein the clustering the position information logs comprises weighting the position information log by applying a second weight to a Euclidean distance in a time direction relative to a latitude/longitude plan defined by latitude and longitude in the 3D space, and

wherein the clustering the extracted set of position information logs comprises weighting the extracted set of position information logs by applying the first weight to the Euclidean distance in the time direction relative to the latitude/longitude plan defined by latitude and longitude in the 3D space.

6. A non-transitory computer readable information recording medium storing an action pattern analysis program that, when executed by a processor, performs a method for:

clustering position information logs containing a user'"'"'s measurement position and measurement time thereby to extract a stay point where the user frequently stays;

extracting a set of position information logs from which a position information log extracted as the stay point is removed;

clustering the extracted set of position information logs plotted in a multidimensional space defined by numerical information indicating the measurement position and a time based on a Euclidean distance on the multidimensional space thereby to extract a user'"'"'s moving path, and

plotting a position information log containing a user'"'"'s measurement position and measurement time on a multidimensional space defined by the numerical information indicating the measurement position and the time,

wherein the clustering the extracted set of position information logs comprises weighting the extracted set of position information logs by applying a first weight to the Euclidean distance in a time direction in a position information space defined by numerical information indicating the measurement position on the multidimensional space, and clustering the position information logs weighted by applying the first weight,

wherein the plotting the position information log comprises plotting a position information log containing values indicating latitude and longitude as a user'"'"'s measurement position on a 3D space defined by latitude, longitude and time,

wherein the clustering the position information logs comprises weighting the position information log by applying a second weight to the Euclidean distance in the time direction relative to a latitude/longitude plan defined by latitude and longitude in the 3D space, and

wherein the clustering the extracted set of position information logs comprises weighting the extracted set of position information logs by applying the first weight to the Euclidean distance in the time direction relative to the latitude/longitude plan defined by latitude and longitude in the 3D space.