Movement Mining - Computational Movement Analysis - page 52

ship, the focus has recently shifted towards trajectory similarity, simplification and
aggregation, as well as segmentation. The database community has made significant
contributions to CMA with respect to storing, managing and querying movement
data in specialized moving object databases (MOD) on the one hand, and by increas-
ingly producing relevant research on data mining applications for movement data.
The concept of semantic enrichment is another important concept towards struc-
turing movement data streams for a better understanding of the underlying dynamic
processes. Finally, visual analytics happily adopted the problem of movement analy-
sis, allowing the efficient integration of all the above contributions in interactive
analytics environments.
Hence, data mining helped CMAmaking significant progress in seeking structure
in movement data, but semantic annotation of the found patterns remains difficult.
Isolated analysis of the geometric footprint of movement is far from understand-
ing movement behavior. Given the complexity of human and animal behavior, more
and more researchers acknowledge that it may simply be too difficult to understand
complex behavior just by studying its mere spatio-temporal footprint. It is thus lit-
tle surprising that more and more work aims at capturing multi-sensor data, where
location data is complemented by sensors that simultaneously record other attributes
indicating the observed activity, such as acceleration, heart rate, or other physical
properties of the monitored individual. Clearly, combining different sensor read-
ings, where the location is only one variable, opens up exciting research avenues,
advancing movement mining towards activity mining.
Despite initial work on ontological foundations of movement processes, there is
little agreement in sight on a set of basic operations and patterns. For a start, the
application domains interested in computational movement analysis seem to be so
diverse, the phenomena they all study so variable that patterns or rules developed
for one application simply have no relevance in another. Similarly, since the applica-
tion problems are so diverse and the supply of new and interesting problems seems
endless, the majority of movement mining methods remain custom-built prototypes
tailored to one specific problem. A wider agreement on a basic set of movement
analysis problems and their possible solutions—leading towards a theory—is rare
so far.
Promoting data mining tools in GIScience for CMA clearly led to a much needed
shift of perspective. On the downside, so far CMA as a developing area remains
dominated by tool-driven researchers. In most cases it's not applications people
requesting the involvement of methods people, but its methods people searching
for interesting applications. This brings the danger of an analytical process that
tries to reshape the real world to fit preexisting solutions. Database experts see the
world of moving things in queries, computational geometry experts in data structures
and algorithms, visual analytics experts in multi-views, parallel coordinate plots
and space-time cubes. “If all you have is a hammer, everything looks like a nail”.
Community efforts such as the 2012 workshop on “Progress in movement analysis”
held in Zurich, explicitly seeking research emerging collaborations between methods
experts and domain specialists, acknowledge this danger.