Page 236 - Contributed Paper Session (CPS) - Volume 2

CPS1845 Devni P.S. et al.
Bayesian networking applications in terms of minimizing the risk of natural
disasters: floods (Zhang, Yang, & Wang, 2002), tsunamis (Blaser, Ohrnberger,
Riggelsen, Babeyko, & & Scherbaum, 2011), earthquakes (Bayraktarli, Ulfkjaer,
Yazgan, & Faber, 2005) (Bayraktarli, Baker, & Faber, 2011) (Li, Wang, & Leung,
2010) (Sari, Rosadi, Effendie, & Danardono, 2018).
The formation of a BN model is a complex and time-consuming task. It is
difficult to get a complete and consistent model. Usually, there are two
methods for entering probability values into opportunity nodes of the
Bayesian network model. The first method is to consult with experts for
probability values and put them in the model. The second method is to get
probability values from statistical data (Druzdzel & Flynn, 2002). To simplify
work while struggling with BN, we use R, we can use an algorithm in
programming R. The R package which is famous for BN is called "bnlearn".
This package contains various algorithms for BN structure learning, parameter
learning, and inference. In addition, we also use GeNIe to enhance the
appearance of the network. Using CPT output obtained from R, a network will
be built using GeNIe. It is a reasoning machine used for graphical probabilistic
models and provides functionality for making diagnoses. Users can also do
Bayesian inference in the model and they can calculate the impact of
observing the subset value of the model variable on the remaining variable
probability distribution based on real-time data.

2. Methodology
Bayesian Network (BN) is an explicit description of depending directly
between a set of variables. This description is in the form of a Directed Acyclic
Graph (DAG) and a set of Node Probability Tables (NPT) (Zhou, Fenton, & Neil,
2014). A directed graph, also called a BN structure, consists of a set of nodes
and arcs. The formation of BN is divided into two, namely the construction of
structures and Conditional Probability Tables (CPT). BN structure is a DAG that
represents a pattern from a set of data. Graph representation can be done by
identifying concepts of information that are relevant to the problem.
Furthermore, the concept is called set variables. The set is then represented as
nodes in the graph. The influence between variables is stated explicitly using
edge on a graph. To get a beneficial relationship between nodes is done using
expert knowledge or algorithms.
(The chain rule for Bayesian networks). Suppose BN is the Bayesian
network above = { , ⋯ , }. Then the BN determines the unique joint
1

probability distribution () given by the multiplication of all the conditional
probability tables specified in BN:

() = ∏ ( |pa( )),

=1
225 | I S I W S C 2 0 1 9