Automatic extraction of user mobility behaviors and interaction preferences using spatio-temporal data

Patent metadata

Field	Value
Publication number	US-10831827-B2
Application number	US-201615088227-A
Country	US
Kind code	B2
Filing date	Apr 1, 2016
Priority date	Apr 1, 2016
Publication date	Nov 10, 2020
Grant date	Nov 10, 2020

Abstract

A user trajectory graph may be constructed based on spatio-temporal data. A mobility pattern may be extracted from the user trajectory graph. Users may be clustered into groups, wherein the users in a same group possess similar feature values in the mobility pattern, and the users in different groups have different feature values, to identify personas and location sets. A distribution model may be constructed that models user timing and location preference, wherein an outcome indicates a preference for a particular time bin on a particular day for a particular location.

First claim

We claim:

1. A computer-implemented method to automatically determine user mobility behaviors and interaction preferences using spatio-temporal data and engage users, the method performed by a processor, comprising:
receiving spatio-temporal data comprising location data as a function of time, app interaction data as a function of time, and geographic map data comprising geographic coordinates;
constructing a user trajectory graph based on the spatio-temporal data;
determining mobility patterns based on the user trajectory graph;
deriving features from the mobility patterns;
clustering users into groups of mobility levels, wherein the users in a same group have similar feature values associated with the features, and the users in different groups have different feature values, whether a feature value is similar or different determined based on meeting a similarity threshold, wherein cluster centroids are separated with unique signatures, wherein at least one cluster contains users with lower mobility, with lower diversity, and with largest duration at a location type compared to users in other clusters, and at least another cluster contains users with higher mobility, with larger diversity, and with shortest duration at the location type compared to users in other clusters;
identifying personas and location sets based on the clustering, the personas including at least degrees of mobility during temporal units of time;
constructing a distribution model, wherein an outcome indicates a preference for a particular time bin on a particular day for a particular location of a user engaging an app, wherein the distribution model uses bins with weights, the weights being updated based on recency of events mapped to the bins, wherein a bin is weighed based on an exponential function of a negative value of a time elapsed since an occurrence of an event mapped to the bin divided by a decay factor; and
monitoring user device's location and responsive to detecting that the user device is at the particular location at a particular time estimated as a preferred time and location of the user engaging the app as indicated by the outcome, pushing content to the user device.

2. The method of claim 1 , wherein the trajectory graph comprises nodes representing locations a user has visited and edges representing travel between the nodes, the edges having associated edge weights that indicate likelihood of the travel between the nodes, the nodes each having a node weight indicating frequency of visits to the respective nodes, wherein the trajectory graph is a directed cyclic graph, and the nodes comprise time duration labels.

3. The method of claim 1 , wherein the features comprise user location-time signature comprising aggregated time spent at a given location and at time bins of weekday and weekend days.

4. The method of claim 3 , wherein the time bins comprise hour bins representing groups of hours.

5. The method of claim 3 , wherein the features further comprise:
total number of unique locations visited;
number and type of unique locations visited in weekday and weekend routines, respectively;
stay time at user's home as a percentage of total stay time across all locations;
average stay time at user's home;
number of visits to user's home;
average distance from home to other locations with direct visits;
distance from home to work;
number and type of the unique locations that are only visited during weekdays; and
number and type of the unique locations that are only visited during weekends.

6. The method of claim 3 , wherein the features are clustered to determine location types comprising at least home and work.

7. The method of claim 3 , wherein the distribution model comprises at least one of a time-preference model indicating a preferred time and location for opening a mobile app.

8. A computer program product to automatically determine user mobility behaviors and interaction preferences using spatio-temporal data and engage users, the computer program product comprising a computer readable storage medium having program instructions embodied therewith, wherein the computer readable storage medium is not a transitory signal per se, the program instructions readable by a device to cause the device to perform a method comprising:
receiving spatio-temporal data comprising location data as a function of time, app interaction data as a function of time, and geographic map data comprising geographic coordinates;
constructing a user trajectory graph based on the spatio-temporal data;
determining mobility patterns based on the user trajectory graph;
deriving features from the mobility patterns;
clustering users into groups of mobility levels, wherein the users in a same group have similar feature values associated with the features, and the users in different groups have different feature values, whether a feature value is similar or different determined based on meeting a similarity threshold, wherein cluster centroids are separated with unique signatures, wherein at least one cluster contains users with lower mobility, with lower diversity, and with largest duration at a location type compared to users in other clusters, and at least another cluster contains users with higher mobility, with larger diversity, and with shortest duration at the location type compared to users in other clusters;
identifying personas and location sets based on the clustering, the personas including at least degrees of mobility during temporal units of time;
constructing a distribution model, wherein an outcome indicates a preference for a particular time bin on a particular day for a particular location of a user engaging an app, wherein the distribution model uses bins with weights, the weights being updated based on recency of events mapped to the bins, wherein a bin is weighed based on an exponential function of a negative value of a time elapsed since an occurrence of an event mapped to the bin divided by a decay factor; and
monitoring user device's location and responsive to detecting that the user device is at the particular location at a particular time estimated as a preferred time and location of the user engaging the app as indicated by the outcome, pushing content to the user device.

9. The computer program product of claim 8 , wherein the trajectory graph comprises nodes representing locations a user has visited and edges representing travel between the nodes, the edges having associated edge weights that indicate likelihood of the travel between the nodes, the nodes each having a node weight indicating frequency of visits to the respective nodes, wherein the trajectory graph is a directed cyclic graph and the nodes comprise time duration labels.

10. The computer program product of claim 8 , wherein the features comprise user location-time signature comprising aggregated time spent at a given location and at time bins of weekday and weekend days.

11. The computer program product of claim 10 , wherein the time bins comprise hour bins representing groups of hours.

12. The program product of claim 10 , wherein the features further include at least one of:
total number of unique locations visited;
number and type of unique locations visited in weekday and weekend routines, respectively;
stay time at user's home as a percentage of total stay time across all locations;
average stay time at user's home;
number of visits to user's home;
average distance from home to other locations with direct visits;
distance from home to work;
number and type of the unique locations that are only visited during weekdays; and
number and type of the unique locations that are only visited during w

Assignees

• IBM

Inventors

• Gifford Wesley M
• Li Ying
• Liu Rong
• Sheopuri Anshul

Classifications

• G06Q30/02 Primary

  Marketing; Price estimation or determination; Fundraising · CPC title

• G06F16/9024 Primary

  Graphs; Linked lists (G06F16/9027 takes precedence) · CPC title

• G06N20/00

  Machine learning · CPC title