Extracting Places from Traces of Locations

1
Extracting Places from Traces of Locations

• Paper Authors
• Jong Hee Kang
• Benjamin Stewart
• William Welbourne
• Gaetano Borriello

PowerPoint Author Michael Cook
2
Michael Cook

• 4th Year Computer Science (Junior)
• Co-oping at Synovus
• Interests
• Databases
• Networking
• Web Development
• Twin brother

3
The Problem

• Location aware systems today are limiting
• Place An area of importance to a user
• Usage Example
• Cell phone goes to silent mode when entering a
  classroom

4
Ideal Situation

• Requires little user interaction
• All important places are located
• No false positives
• Works for indoor and outdoor places

5
Tracking User Movement

• Place Lab access points
• Works indoors

6
Popular Clustering Algorithms
K-means
Gaussian mixture model
Large amounts of computation
7
Time-Based Clustering

• Streaming computation
• Small clusters ignored
• Time threshold and distance threshold can be
  changed

8
Time-Based Clustering Result
9
Changing Distance and Time
10
Changing Distance and Time
d30m t300sec
d50m t300sec
d300m t600sec
11
Frequently Visited Places

• Not much time is spent at the place, but
  frequently visited
• Different time threshold needed
• How to differentiate the place and in-transit
  motion?

12
Future Work

• Automatic labeling of places
• Can use users calendar
• Learn proper distance and time thresholds
  automatically

13
Critique

• Easy to read and understand
• Cool idea with practical applications

• WiFi hotspots not always available
• Trying to do too much at once
• Long duration places
• Short duration, frequent places

14
Questions?