Movement Mining - Computational Movement Analysis - page 45

1 500

2
3
4
5
6
7
8
9
10

2
3
4
5
6
8

0.9

0.8 400

0.7

0.6

300

0.5

0.4

200

0.3

0.2 100

0.1

0 0

0 0.2 0.4 0.6 0.8 1

0.0 0.2 0.4 0.6 0.8 1.0

c/p

c/p

Fig. 3.10 Balancing eoo versus eoc in decentralized flock mining, simulated movement
data ( P12 . Laube et al. 2011a ) (Republished from Laube, P., Duckham, M., Palaniswami,
M., Deferred Decentralized Movement Pattern Mining for Geosensor Networks, International
Journal of Geographical Information Science , 25(2), pp. 273-292, 2011, Taylor & Francis,
DOI:10.1080/13658810903296630)

3.3.2 Credibility

Method evaluation should also make sure that a method meets the needs of its users,
assuring a certain notion of acceptance and suitability with domain experts. The work
summarized in this topic strongly supports the argument that data mining and hence
movement mining methods must prove useful to users in an application context.
Clearly, testing suggested methods with real data supports credibility. Real data
emerging an application context was used in Merki and Laube ( P16 . 2012 , tracked
students in an outdoor game), Dodge et al. ( P14 . 2012 , hurricanes, couriers), and
Laube et al. ( P12 . 2011a , cows in a smart farming study). Earlier work by Laube and
Purves ( 2006 ) investigated, for example, the relevance of movement patterns based
on the notion of interestingness . I argue that the interestingness measures proposed
by Silberschatz and Tuzhilin ( 1996 ) and Geng and Hamilton ( 2006 ) also build a
useful starting point for assessing the credibility of patterns in movement mining.
Objective interestingness measures depend solely on the structure of the pat-
tern and the underlying data. The most commonly known objective measure for
the quality, strength or interestingness of data mining rules are support and con-
fidence given for association rules. Support is generally defined as the frequency
of a pattern in a data set, while confidence expresses the prediction strength of
the rule (Mohammad and Nishida 2010 ). In Bleisch et al. ( P20 . 2014 ) support
and confidence measures were adapted for mining candidate causal relationships
between movement events and environmental states, such as “allows”, “initiates”
or “terminates”. The study developed a sequence mining approach relating environ-
mental states (“high water temperature”, “high river flow”) with movement events
of fish moving in a river network (“upstream movement”). For example, in 84