what-when-how
In Depth Tutorials and Information
process, which is a statistical process involving a number of random variables
depending on a variable parameter (which is usually time). he probabilistic model
typically uses different hypotheses such as the individual interest factor, the group
behavior factor, and the time lapse factor.
A dynamic social network can handle large dynamic multinode, multilink net-
works with different levels of uncertainty. A meta-group is used to define the main
concept of a dynamic network. Using the hypotheses from the probabilistic model,
a more accurate model can be created. Dynamic modeling is important to get an
accurate representation of a network.
he small-world model uses the hypothesis that the chain of social acquain-
tances required to connect on arbitrary person anywhere in the world is gener-
ally short. We will explain the small-world phenomenon experiments by Stanley
Milgram. We will also examine the two theorems that Kleinberg uses to study the
ability of decentralized network control.
Large-scale models (better known as sociotechnical systems) are an approach to
complex organizational work design that recognizes the interaction between people
and technology in work places. he term also refers to the interaction between
complex society infrastructure and human behavior.
5.2 ProbabilisticModels[2]
Social network models can benefit from a probabilistic approach. Using probabil-
ity, a network can predict the parameters, actors, and actions of these actors in the
future. his prediction can be very advantageous to the stability and eiciency of
the network, and it can be performed in a variety of different ways.
In one example, a learned function can be implemented as a stochastic pro-
cess. In this scenario, the test will be a realization of the stochastic process, and
a multistep prediction will be used to test the data. Considering a social group
structure, the actions of the actors or the actors' paths into the future can be pre-
dicted. Based on these paths into the future, an evolving social group structure
can be constructed. hese predicted groups can be compared with the observed
groups via the test data. he distribution of group sizes can be used to measure
performance.
his example shows how prediction models can be integrated with learning
models. However, the learning process is time consuming since there are so many
possible combinations of actors that can be used to establish groups. Adding a pre-
diction model to this will only increase the time needed. Considering that there are
N
actors and
K
groups, at each time step there are 2
NK
/
K
! possible actors' combina-
tions that can be used to create and set up groups. herefore, for
T
time steps, the
complexity of finding the optimal group path is
O
(
T
×
2
NK
/
K
) or
O
(
T
×
2
K
(
N
−
log
K
)
).
Based on the time complexity shown here, it is easy to see how the learning process
is very time consuming.
