Learning a user's activity preferences from GPS traces and known nearby venues

• US 7,984,006 B2
• Filed: 09/18/2007
• Issued: 07/19/2011
• Est. Priority Date: 09/18/2007
• Status: Active Grant

• Current Assignee: Xerox Corporation (Xerox Holdings Corp.)
• Original Assignee: Palo Alto Research Center, Inc. (Xerox Holdings Corp.)
• Inventors: Price, Robert R.
• Primary Examiner(s): Gaffin; Jeffrey A
• Assistant Examiner(s): Wong; Lut
• Application Number:
  US11/857,425
• Publication Number:
  US 20090073033A1
• Time in Patent Office: 1,400 Days
• Field of Search: 706/46
• US Class Current: 706/46
• CPC Class Codes: G06Q 30/02 Marketing; Price estimation or determination; Fundraising

Abstract

A method for inferring activities to a user is provided. The system receives at least one location trace and corresponding contextual information. The system then derives a set of venues based on a venue database, wherein a respective hypothetical visit is associated with the contextual information corresponding to the location trace. The system derives a set of activity types associated with a context based on the venues, the corresponding context indicated by the location trace, and a venue-to-activity mapping. In addition, the system receives a user query context and identifies a number of activity types of which the associated contextual information is similar to the user query context. The system further weights a respective identified activity type based on its associated context'"'"'s similarity to the user query context, normalizes weights associated with each identified activity type, and produces an activity-type probability distribution, thereby facilitating inferring activities associated with the user.

Claims

1. A method for inferring activities associated with a user, the method comprising:

receiving a cluster of locations that indicates a location trace of the user during a period of time and corresponding context;

smoothing the location trace, which involves statistically removing a first subset of locations from the location cluster, and interpolating a second subset of locations into the location cluster;

deriving a set of venues based on the location cluster from a venue database, wherein a visit to a respective venue is identified as a hypothetical visit responsive to the venue being located within a location range which includes the received location cluster and being associated with the corresponding context;

deriving a set of activity types associated with the venues from a venue-to-activity mapping;

associating attributes of the venues to the activity types based on the corresponding context;

identifying a subset of the activity types of which the associated attributes are similar to a query context;

assigning a weight to each identified activity type based on similarity between its attributes and the query context; and

producing a probability distribution of various activity types based on the subset of the activity types and their weights, wherein the probability distribution predicts likelihoods that the user engages in each identified activity type associated with the query context and the location.

2. The method of claim 1,

wherein the context indicated by the location trace includes one or more of:

a location;

an indication of a time of a day or a time range;

an indication of a day of a week or a range of days of a week; and

an indication of a weather condition.

3. The method of claim 1,

wherein the query context, comprises:

a location;

a time;

a day of a week; and

a weather condition.

4. The method of claim 1,

wherein the location trace includes a set of coordinates with corresponding time stamps.

5. The method of claim 1,

wherein deriving the set of venues comprises:

identifying a location based on one or more locations indicated by the location trace within a time period; and

identifying a number of venues in the vicinity of the identified location based on the venue database.

6. The method of claim 1,

wherein the venue database indicates a number of venues and their corresponding coordinates.

7. The method of claim 1,

further comprising pooling activity-type-to-context mappings associated with multiple, similar users by explicitly clustering similar users or by using latent-variable methods to implicitly cluster similar users.

8. A non-transitory computer readable storage medium storing instructions which when executed by a computer cause the computer to perform a method for inferring activities associated with a user, the method comprising:

receiving a cluster of locations that indicates a location trace of the user during a period of time and corresponding context;

smoothing the location trace, which involves statistically removing a first subset of locations from the location cluster, and interpolating a second subset of locations into the location cluster;

deriving a set of venues based on the location cluster from a venue database, wherein a visit to a respective venue is identified as a hypothetical visit responsive to the venue being located within a location range which includes the received location cluster and being associated with the corresponding context;

deriving a set of activity types associated with the venues from a venue-to-activity mapping;

associating attributes of the venues to the activity types based on the corresponding context;

identifying a subset of the activity types of which the associated attributes are similar to a query context;

assigning a weight to each identified activity type based on similarity between its attributes and the query context; and

producing a probability distribution of various activity types based on the subset of the activity types and their weights, wherein the probability distribution predicts likelihoods that the user engages in each identified activity type associated with the query context and the location.

9. The computer readable storage medium of claim 8,

wherein the context indicated by the location trace includes one or more of:

a location;

an indication of a time of a day or a time range;

an indication of a day of a week or a range of days of a week; and

an indication of a weather condition.

10. The computer readable storage medium of claim 8,

wherein the query context comprises:

a location;

a time;

a day of a week; and

a weather condition.

11. The computer readable storage medium of claim 8,

wherein the location trace includes a set of coordinates with corresponding time stamps.

12. The computer readable storage medium of claim 8,

wherein deriving the set of venues comprises:

identifying a location based on one or more locations indicated by the location trace within a time period; and

identifying a number of venues in the vicinity of the identified location based on the venue database.

13. The computer readable storage medium of claim 8,

wherein the venue database indicates a number of venues and their corresponding coordinates.

14. The computer readable storage medium of claim 8,

wherein the method further comprises pooling activity-type-to-context mappings associated with multiple, similar users by explicitly clustering similar users or by using latent-variable methods to implicitly cluster similar users.

15. A computer system for recommending leisure activities to a user, the computer system comprising:

a memory;

a processor;

a location trace receiving mechanism that receives a cluster of locations that indicates a location trace of the user during a period of time and corresponding context;

a location trace smoothing mechanism that smoothes the location trace, which involves statistically removing a first subset of locations from the location cluster, and interpolating a second subset of locations into the location cluster;

a location trace preprocessing module coupled to the processor that;

derives a set of venues based on the location cluster from a venue database, wherein a visit to a respective venue is identified as a hypothetical visit;

responsive to the venue being located within a location range which includes the received location cluster and being associated with the corresponding context;

derives a set of activity types associated with the venues from a venue-to-activity mapping; and

associates attributes of the venues to the activity types based on the context; and

an inferring mechanism coupled to the processor that;

identifies a subset of the activity types of which the associated attributes are similar to a query context;

assigns a weight to each identified activity type based on similarity between its attributes and the query context; and

produces a probability distribution of various activity types based on the subset of the activity types and their weights, wherein the probability distribution predicts likelihoods that the user engages in each identified activity type associated with the query context and the location.

16. The computer system of claim 15,

wherein the context indicated by the location trace includes one or more of:

a location;

an indication of a time of a day or a time range;

an indication of a day of a week or a range of days of a week; and

an indication of a weather condition.

17. The computer system of claim 15,

wherein the query context comprises:

a location;

a time;

a day of a week; and

a weather condition.

18. The computer system of claim 15,

wherein the location trace includes a set of coordinates with corresponding time stamps.

19. The computer system of claim 15,

wherein while deriving the set of venues, the location trace preprocessing module is further configured to:

identify a location based on one or more locations indicated by the location trace within a time period; and

identify a number of venues in the vicinity of the identified location based on the venue database.

20. The computer system of claim 15,

wherein the venue database indicates a number of venues and their corresponding coordinates.

21. The computer system of claim 15,

further comprising a pooling mechanism configured to pool activity-type-to-context mappings associated with multiple, similar users by explicitly clustering similar users or by using latent-variable methods to implicitly cluster similar users.