Information Technology Reference
In-Depth Information
knowledge
patterns
integrated
data
transformed
data
preprocessed
data
target data
multiple &
heterogeneous
data sources
Fig. 5.1
The KDD process and its core step data mining in a world of big geo-data, adapted from
(Fayyad et al.
1996
)
Mayer-Schönberger and Cukier (
2013
) emphasize three core challenges around
big data. Firstly, the authors argue that the need to sample is an artifact of an era
dominated by information scarcity. Overcoming this information scarcity requires a
fundamental shift in analytical thinking, where analysis not primarily bases on sam-
ples but moves more and more towards analyzing entire (statistical) populations. For
example, the group around Rein Ahas has, as of 2013, access to mobile phone logs
of almost the entire Estonian population through building up cooperations with all
major service providers. Secondly, traditional ways of reasoning and analysis must be
complemented with new approaches inferring knowledge frommore comprehensive
but at the same time much more uncertain and noisy, in short, messy data. Mayer-
Schönberger and Cukier (
2013
) argue that gaining access to more comprehensive
data sets allows shedding some of the rigid exactitude required when analyzing con-
ventional samples. Clearly, positional uncertainty will remain an issue with tracking
data for many years to come. Compared to the long and intense history of research
on uncertainty, it is surprising to see so few articles addressing this pivotal problem
in CMA. Thirdly, correlations found in big data may not allow us to understand
why
something is happening (causality), but exploit the correlations found for the equally
important alert
that
something is happening.
5.2 Bridging the Semantic Gap
CMA aims at bridging the gap between low-level movement data and the high-level
conceptual schemes required for understanding movement processes, just as it must
be the goal for GIScience in general (Galton
2005
). Through the work summarized
