Database Reference
In-Depth Information
the missing data for these 2 days, all the points between 9 p.m. Monday to 10 p.m.
Wednesday will be at home. So it is better to mark those points during the long
missing period as invalid points. And when conducting pattern mining methods, we
will only consider the valid points. When designing data mining methods, we should
pay attention to the issue of incomplete, noisy, and unevenly sampled data. Ideally,
a pattern mining method should take the raw data as input or even handle the raw
data with uncertainties.
2.3
Background Information
Few moving objects move in free space. Vehicles, obviously, need to follow the road
network. Planes and boats need to follow more or less the scheduled paths. Animals,
which live in a more free space, are also confined to embedding landscape, such as
rivers, mountains and the food resources.
When considering the background information, the mining tasks become more
challenging. For example, the distance between two cars cannot be calculated simply
by Euclidean distance. Similarly for animals, if there is a mountain or a big river
between two animals, they could be actually far away from each other. Consider-
ing background information will result in more complex distance calculation and
correspondingly require different data mining methods.
For domain experts to interpret the discovered patterns, it is important to consider
the underlying geography in order to understand where, when and ultimately why the
entities move the way they do. Grazing sheep, for example, may perform a certain
movement pattern only when they are on a certain vegetation type. Homing pigeons
may show certain flight patterns only when close to a salient landscape feature such
as a rive or a highway. And, the movement patterns expressed by tracked vehicle will
obviously be very dependent on the environment the vehicle is moving in, be it in a
car park, in a suburb or on a highway. Thus, patterns have to be conceptualized that
allow linking of the movement with the embedding environment.
3
Individual Periodic Pattern
One most common activity in moving objects is the
periodic behavior
. A periodic
behavior can be loosely defined as the repeating activities at certain locations with
regular time intervals. For example, bald eagles start migrating to South America in
late October and go back to Alaska around mid March.
Periodic behaviors provide an insightful and concise explanation over the long
moving history. For example, animal movements could be summarized using several
daily
and
yearly
periodic behaviors. Periodic behaviors are also useful for compress-
ing movement data [
3
,
28
,
38
]. Moreover, periodic behaviors are useful in future
movement prediction [
17
], especially for a distant querying time. At the same time,
