Information Technology Reference
In-Depth Information
There are two versions of the data: a set coded from the lead sentences only
(57,000 events), and a set coded from full stories (304,000 events)”
[20].
The dataset covers the connection of states within the Gulf region and the
Arabian peninsula over almost 20 years. It is annotated with timestamps thus
making it perfect for temporal network analysis. For the current comparison, we
used the dataset coded from the lead sentences, for which the networks are also
available preprocessed both in monthly and daily granularity [21].
Fig. 4.
Degree distribution evolution for the cumulative network constructed for the
Gulf dataset. The degree distribution P(k) is shown on the z axis, while how the
distribution changes over time can be read from left to right on the picture on the
left hand side. The picture on the right hand side shows slices of the 3D chart in four
selected points on a log-log scale.
Results and Discussion. The network created from the dataset contains 174
nodes and 57,131 edges. We made simulations with identically parameterized ver-
sions of the introduced dynamic network models for comparison. The appendix
contains the concrete results for different structural properties for the dataset
both in daily and monthly granuality as a reference. The properties of the cu-
mulative network does not extend a significant difference between the daily and
the monthly versions, hence in the following we consider the monthly version for
the discussion.
The first important thing to note is that the structure of the degree distribu-
tion is similar to the preferential elementary dynamic networks. They show the
same spread out structure that characterizes the DoubleCPA and CPA models
(illustrated on Figure 4).
