Mining Interesting Locations and Travel Sequences from GPS
Download
Report
Transcript Mining Interesting Locations and Travel Sequences from GPS
Mining Interesting Locations
and Travel Sequences from GPS
Trajectories
defense by Alok Rakkhit
Overview
Take GPS data of Multiple users to find travel
sequences and interesting locations
That data can be used to help understanding of
surrounding area and provide travel
recommendations
Interesting places get many users and experienced
users
Interest and user experience are region-related
Knowledge of Beijing doesn’t imply knowledge of New York
Architecture
Take raw data and
determine locations (stay
points) and travel paths
Combine users location
histories and create TreeBased Hierarchical Graph
Apply Hypertext Induced
Topic Search to TBHG to
infer location interest and
user experience
Create travel
recomendations based on
the data
Data
107 users tracked from May
2007 to October 2008
Beijing plus 36 other Chinese
cities, as well as cities in US,
South Korea, and Japan
Stay point detection thresholds
prevent inclusion of irrelevant
locations (stopping at traffic
lights, users’ homes and
offices)
Clustering using density-based
algorithm, OPTICS
Evaluation
29 subjects, living in Beijing for over 6 years answered
evaluation questions about the locations and paths
Baselines for locations were rank by count and by frequency
Baseline for paths were rank by count, by interests, and by
experience
Results
GeoLife
Significance
Project GeoLife
Advantages of hierarchical system in
understanding location data at multiple scales
of resolution
Integrate social networking into real time
location information
Apply location/user relationship to improve
recommendation systems