Movement Mining - Computational Movement Analysis - page 50

or are “stable with respect to each other”. QTC represents such qualitative dynamic
behavior of a pair of moving point objects using a small set of symbols. The authors
illustrated their calculus for example for a predator-prey interaction in a 2DEuclidean
space (Van de Weghe et al. 2006 ) or for movement along a road network (Van de
Weghe et al. 2004 ; Bogaert et al. 2007 ). Gottfried ( 2011 ) argues in a very similar
way that decomposing motion pattern into qualitative features and employing formal
grammars has the advantage of being closer to human thinking and also suits often
noisy and imprecise data. In his piece, however, he does not consider velocity or
distance constraints, but investigates in contrast whether meaningful features can be
derived from directional information alone.
Whereas much previous work in GIScience has studied movement patterns in
isolation from their geographical embedding, more and more attention has recently
been given to context-aware movement analysis. Orellana explores various aspects
around pedestrian movement patterns as the result of the interactions between people
and their environment (Orellana 2012 ). Orellana andWachowicz ( 2011 ) andOrellana
et al. ( 2012 ) propose the use of a local indicator of spatial association (LISA, Anselin
1995 ), a measure for assessing local spatial autocorrelation to detect spatial clusters
of low speed vectors. Such suspension patterns explicitly do not search for stopping
behavior in individuals' trajectories but rather for collective behavior, as in potential
points of interest where many pedestrians stop (outlook, visitor center, picnic area).
When ordered in sequences , frequently visited sets of such stopping clusters allows
for the aggregation of visitor movement into flows.
Exploratory analysis and visual analytics . Introductory reviews for visual analytics
of spatio-temporal data in general can be found in Andrienko et al. ( 2010 ) and
specifically for movement data inAndrienko andAndrienko ( 2007 ). Given its explicit
spatio-temporal character, movement data has served in exploratory analysis and
visual analytics research as a signature case study. An extensive review of the wide
range of literature on exploratory analysis and visual analytics of movement data
would diverge too much from the path set out for this brief volume. Entry points for
selected aspects of exploratory analysis and visual analytics of movement data can be
found regarding generalization and aggregation (Andrienko and Andrienko 2011 ),
density estimation and related concepts (Downs and Horner 2010 , 2012 ), three-
dimensional space-time cubes, combined with additional techniques for structuring
the data (Rinzivillo et al. 2008 ; Demsar and Virrantaus 2010 ; Pelekis et al. 2012 ),
interactive combinations of visualization and clustering (Schreck et al. 2009 ), as well
as integrated visual analytics interfaces featuring linked and coordinated views of
spatial, temporal, and socio-economic characteristics (Zhang et al. 2013 ).
Evaluation . Surprisingly little work has been produced on evaluating proposedmove-
ment mining methods. As mentioned in Sect. 3.3 it is difficult to get hold of suitable
data featuring the semantic annotation needed for a thorough evaluation. However,
there is work where an explicit focus was put making sure that the methods were
sound and produced useful knowledge. One example for an assessing internal valid-
ity using a test data set can be found in Orellana andWachowicz ( 2011 ), where mined
suspension patterns (a stopping behavior) are compared with reference or “ground