Information Technology Reference
In-Depth Information
truth” data. This reference data comes in the form of identified points of interest
that the individuals are likely to visit and thereby express the suspension pattern:
check-points in an outdoor game, as well as points of interest and crucial infrastruc-
ture in a recreational park. Cross-checking which of the expected points of interest
are actually detected and which are missed allows an indication of error of omission
and commission.
In problem-driven research domains the thorough evaluation of proposed meth-
ods is more common than in somewhat theoretical computer science or GIScience,
also because in ecology a thorough evaluation is often a
condicio sine qua non
for
publication. Given its roots in behavioral science, movement ecology occasionally
produces movement with rich semantic annotation, painstakingly captured by human
observers in the field. For instance in Shamoun-Baranes et al. (
2012
) the classifica-
tion of oystercatcher behavior through supervised classification trees on location
and sensor data was cross-checked with simultaneous visual observations. Similarly,
Guilford et al. (
2009
) evaluate a machine learning based behavior classification (akin
segmentation and labeling) through cross-validation of two simultaneously recorded
sensor streams.
3.5 Concluding Remarks
The introduction of data mining concepts into GIScience with the goal of a better
understanding of movement processes has clearly led to significant progress in struc-
turing notoriously messy movement data. In an area dominated by a static view of
the world inherited from cartography, the arrival of a flexible toolset allowing the
search for patterns, trends, and similarities not only in space, but explicitly in space,
time, and attributes was much needed and hence is to be warmly welcomed. Since
movement data is inherently spatio-temporal, recording the location of an object
at potentially thousands of time stamps can rapidly flood and “fill-up” maps, the
GIScience signature analysis metaphor. Here, data mining's approach of concep-
tualizing and formalizing patterns and rules, akin search templates, that then can
be searched for by efficient algorithms, offers analytical tools complementing the
conventional GIScience tool box.
Many areas have contributed to establishing data mining and knowledge discov-
ery in databases as a key toolsets of CMA.
GIScience
has contributed its theory of
representing and abstracting both the moving entities as well as the spaces embed-
dingmovement. GIScience' theory onmodeling spatio-temporal phenomena, entities
and processes of the natural and built environment have made a significant contribu-
tion to the conceptualization of movement mining patterns and rules. A sometimes
underestimated key contribution lies in GIScience' expertise in integrating multi-
source and multi-scale data, preprocessing and transforming uncertain and noisy
geodata to make it ready for the data mining algorithms. The
computational geome-
try
community has contributed in many ways to movement mining in CMA. Whereas
early collaborations investigated movement patterns such as flock, convoy, leader-
