Mining Semantics from Trajectories
Speaker: Kexin Xie
When: 2009-09-03 15:00:00
Venue: 78-420
Host: Xiaofang Zhou
Abstract:Abstract: Semantic patterns are people’s activity sequences such as “drinking in a bar after work”, “eating at a restaurant after 2 hours of shopping”. These patterns mined can be very useful in a very wide range of applications to provide convenient and useful services. As the increasing popularity of GPS-enabled mobile devices, a huge amount of tra jectories which show people’s movement behaviours have been acquiring. People travel to different places for different activities, such as fishing on a lake, studying in the library etc. Those activities can be extracted by analysing the stops and moves from the tra jectory combined the background geographic information (e.g. lakes, libraries, hotels). Studies on mining movement patterns from tra jectories have been received many attentions in the recent years. In these works, tra jectories are treated as sequences of temporally annotated sampled points, and thus people’s movement patterns can be mined. These works deal with tra jectory patterns in the sense of geographical movements only. However, the integration of tra jectories with the relevant geographic information, which shows richer information about the users’ semantic patterns, are seldom addressed. Motivated by this, we propose to study ways to mine semantics from tra jectories in this research project. We take the approach to first join trajectories with geographic locations that carry semantic meanings (i.e. places activities may took place) to find activity sequences, and then mines semantic patterns from such activity sequences. There are two research problems addressed in this report. The first one is how to find activity sequences from large set of trajectories with large set of geographic background data set. Since the activity sequence is uncertain, i.e. we are not exactly sure which activity took place, the second challenge is how to mine patterns from sequences with uncertainty. Since in the first stage, we introduce uncertainty and at the second stage we deal with such uncertainty and try to minimise the impact. The ultimate goal of this research project is combine those two phases to increase the efficiency and effectiveness of the data mining process.
Biography:
Type: PhD Confirmation
Contact: