Mining in Geographic Data
Methods used for mining spatial and spatiotemporal data
Motivation
Spatial data describe objects, events or other
features on or near the surface of the Earth. It
typically combines location and attribute
information, for example, characteristics of related
objects, events, or phenomena with temporal
information. The location provided may be static in
the short term, such as a piece of equipment, an
earthquake event, or the location of a poor child. It
may also be dynamic, such as moving vehicle or
the spread of an infectious disease.
Such data are usually large in volume and are
collected from various sources, e. g. by satellite
imagery, phone data, or from social media. We
gathered several approaches which proccess
these kind of data with great results.
Location prediction
Frequent pattern mining allows us to discover
common patterns in the dataset. It helps us
characterize the data, predict the behavior of
existing objects, and even detect anomalies when
one of the objects starts to behave out of order.
Geographic customer segmentation
People living close to each other often
have similar buying patterns. Detecting these
patterns is useful from a marketing
perspective. Association rules mining or clustering
techniques can help in revealing these patterns.
Trajectory clustering
The point of clustering trajectory data is to find
groups of trajectories that share similar or even
identical paths. An example of trajectory
clustering can be seen in a framework which
groups together hurricanes with similar
trajectories.
Cluster analysis of Atlantic Tropical Storms and Hurricanes
Crime hot-spot analysis in Washington DC
Hot-Spot analysis
Hot-Spot is an area with a high concentration of
observed events. Hot-Spots can be used to
identify outbreaks of diseases. Also can be used to
find areas with high criminality activity or traffic
accidents. This area uses mainly point clustering
algorithms.
Evolution of a convoy in spatio temporal dataset
Convoy clustering
Convoy clustering is also based on trajectories. It
tracks big groups of object moving together, such
as flocks of animals or convoys of cars.
Andrej Dravecký, Ka Man Pang, Lucie Novotná, Petr Matonoha
Weather prediction