Network Evaluation and Optimisation For ATR 72 in Indonesia

INTERNATIONAL UNIVERSITY LIAISON INDONESIA
BACHELOR’S THESIS
NETWORK EVALUATION AND OPTIMISATION FOR ATR 72 IN INDONESIA
By
Peter Oliver
11201701018
Presented to the Faculty of Engineering
In Partial Fulfillment of the Requirements for the Degree of
SARJANA TEKNIK
In
AVIATION ENGINEERING
FACULTY OF ENGINEERING
BSD City 15345

Indonesia
August 2021
APPROVAL PAGE
Peter Oliver
11201701018
Presented to the Faculty of Engineering
In Partial Fulfillment of the Requirements for the Degree of
SARJANA TEKNIK
In
AVIATION ENGINEERING
FACULTY OF ENGINEERING
Ir. Ananta Widjaja M.Sc _______________

Thesis Advisor Date
Dipl.-Ing. Sentot Wahjoe Goeritno, M.Si 17 / 02 / 2022

_______________
Dean of Faculty of Engineering and Life Science Date
i
EXAMINERS APPROVAL PAGE
Dr. Ir. Prianggada Indra Tanaya, M.M.E _______________

Examiner 1 Date
Neno Ruseno M.Sc 17/02/2022

_______________
Examiner 2 Date
ii
STATEMENT BY THE AUTHOR
I hereby declare that this submission is my own work and to the best of my
knowledge, it contains no material previously published or written by another person,
nor material which to a substantial extent has been accepted for the award of any
other degree or diploma at any educational institution, except where due
acknowledgement is made in the thesis.
Peter Oliver ___________________

Student Date
iii
ABSTRACT
In the passenger movement area, the Indonesian ATR 72 had a huge increase.
In 2013, 4.2 million people flew on 196 different routes. As a result, it's critical to
comprehend the ATR 72's structural network in Indonesia. Network analysis is required
to address any additional issues that may arise as a result of the expansion.
To visualize an exact relationship between a node and any other node in the
network, a modeling network is required. Characterizing the network will give you a
better understanding of it. The network characteristics or network metrics are detailed
information about each node in relation to the entire network. The average degree,
average weighted degree, diameter, average shortest path, graph density, average
clustering coefficient, and average neighbourhood overlap are all used to calculate
network metrics. The average clustering coefficient will be the most important statistic
for network optimization.
Due to the lack of a strong relationship between the two groups and between
the nodes within each group, the eastern section of the network is the best area to
optimize. With only 10 new network edges, the optimization version on Nabire Airport
manages to increase the network average clustering coefficient by 28.06% from its
initial value.
iv
COPY RIGHTPAGE
© Copyright 2021
by Peter Oliver
International University Liaison Indonesia
All rights reserved
v
DEDICATION
I would like to dedicate my work to my family and friends that has been giving
me endless support to finish this paper.
I also dedicate this paper to IULI and all the lecturer that has guided me
through the process of finishing my school and given me the inspiration for this paper.
And lastly, i dedicate this paper to God Almighty who provides me strength,
courage, wisdom, and healthy life to finally finish this paper.
vi
ACKNOWLEDGMENTS
Praise the Lord for all the grace and blessings given from Him that I have been
able to complete this thesis.
The final paper titled "Network Evaluation and Optimisation for ATR 72 in
Indonesia" is prepared and written to completing part of the academic qualification to
obtain the bachelor's degree of Engineering, faculty of Aviation Engineering,
International University Liaison Indonesia, Tangerang, Banten.
The writer acknowledges that without the guidance, support, and supervision
from various parties, this final paper would not have been completed in time.
Therefore, the writer wishes to deliver deepest appreciation to all these different
parties that were generous enough to help in the process of finishing this final paper.
This gratitude is directed to:
1. God, for His never ending blessings, who gave me strength to finish this
paper in time despite many obstacles that may have been in the way.
2. Dipl. Ing. Sentot Wahjoe Goeritno, M.Si, as Dean of Engineering faculty
3. Ir. Ananta Widjaja M.Sc, as Supervisor
4. My beloved family member
5. Varrent Natasha Surya
6. KUMDOS
Lastly, the writer aware that this paper is not perfect and can be improved in
this study. Thus, critics and suggestions from the reader will be useful and appreciated
for the writer. Hopefully this final paper will be beneficial for the readers.
vii
TABLE OF CONTENTS
APPROVAL PAGE ________________________________________________ i
EXAMINERS APPROVAL PAGE ______________________________________ii
STATEMENT BY THE AUTHOR _____________________________________ iii
ABSTRACT _____________________________________________________ iv
COPY RIGHTPAGE ________________________________________________v
DEDICATION ___________________________________________________ vi
ACKNOWLEDGMENTS ___________________________________________ vii
TABLE OF CONTENTS ___________________________________________ viii
LIST OF FIGURES ________________________________________________ xi
LIST OF TABLES _________________________________________________ xii
CHAPTER 1 ____________________________________________________ 1
INTRODUCTION ____________________________________________________ 1
1.1. General Statement of Problem Area ___________________________________ 1
1.2. Research Purpose __________________________________________________ 2
1.3. Research Problem __________________________________________________ 2
1.4. Significance of Study ________________________________________________ 2
1.5. Theoretical Perspective _____________________________________________ 2
1.6. Research Question _________________________________________________ 4
CHAPTER 2 ____________________________________________________ 5
LITERATURE REVIEW ________________________________________________ 5

2.1. Network Optimization ______________________________________________ 5
2.2. Air Transportation Network Around the World___________________________ 6
2.2.1. Worldwide Air Transportation Network _____________________________ 6
2.2.2. Brazilian Air Transportation Network _______________________________ 6
2.2.3. Indian Air Transportation Network _________________________________ 6
2.2.4. Italian Air Transportation Network _________________________________ 7
2.3. Small World Networks ______________________________________________ 7
2.4. Scale-Free Model __________________________________________________ 8
viii
CHAPTER 3 ___________________________________________________ 10
METHODOLOGY ___________________________________________________ 10
3.1. Data Collection ___________________________________________________ 10
3.2. Network Modelling ________________________________________________ 10
3.3. Network Characterise ______________________________________________ 12
3.3.1. In-degree and Out-degree _______________________________________ 12
3.3.2. Average Neighbourhood Overlap _________________________________ 12
3.3.3. Average Shortest Path __________________________________________ 13
3.3.4. Diameter _____________________________________________________ 14
3.3.5. Clustering Coefficient ___________________________________________ 14
3.3.6. Betweenness _________________________________________________ 15
3.3.7. Closeness Centrality ____________________________________________ 15
3.3.8. PageRank ____________________________________________________ 16
3.3.9. Graph Density _________________________________________________ 16
3.4. 2013 -2019 Data Comparison ________________________________________ 16
3.4.1. Community Structure ___________________________________________ 17
3.5. Analyze The Plausible Improvement __________________________________ 19
CHAPTER 4 ___________________________________________________ 20
RESULTS AND DISCUSSION __________________________________________ 20

4.1. Network Model and Characteristics ___________________________________ 20
4.1.1. Network Model and Characteristics of 2013_________________________ 20
4.2. Network Growth Analysis ___________________________________________ 64
4.2.1. Community Structure Growth ____________________________________ 64
4.2.2. Network Overview Analysis ______________________________________ 72
4.3. Network Optimization _____________________________________________ 73
4.3.1. TIM New Route Evaluation ______________________________________ 73
4.3.2. NBX New Route Evaluation ______________________________________ 75
CHAPTER 5 ___________________________________________________ 78
ix
SUMMARY AND RECOMMENDATION__________________________________ 78

5.1. Summary ________________________________________________________ 78
5.2. Recommendation _________________________________________________ 79
REFERENCES __________________________________________________ 80
x
LIST OF FIGURES
FIGURE 1 A RANDOM MOVING EDGES OF A WATTS-STROGATZ MODEL. THE NETWORK HAS 20 NODES AND 40 EDGES. (A)
UNSOPHISTICATED NETWORK (B) SMALL WORLD NETWORK (C) RANDOM NETWORK .............................................. 8
FIGURE 2 THE NETWORK INFRASTRUCTURE OF ATR 72 AROUND SULTAN HASANUDDIN INTERNATIONAL AIRPORT IN 2019 . 13
FIGURE 3 CLUSTERING COEFFICIENT OF GRAPH G IS 0.75 ........................................................................................ 15
FIGURE 4 RESOLUTION IN GEPHI MODULARITY SETTING .......................................................................................... 18
FIGURE 5 ATR 72 NETWORK AROUND INDONESIA IN 2013 ..................................................................................... 24
FIGURE 7 DEGREE COUNT GRAPH OF 2014 NETWORK. ........................................................................................... 29
FIGURE 10 COMMUNITY STRUCTURE OF ATR 72 NETWORK IN 2013 ........................................................................ 65
FIGURE 13 COMMUNITY STRUCTURE OF ATR 72 NETWORK IN 2016 ............................ ERROR! BOOKMARK NOT DEFINED.
FIGURE 17 COMMUNITY STRUCTURE FOCUS AREA TO BE OPTIMIZED .......................................................................... 71
FIGURE 18 MOSES KILANGIN AIRPORT CONNECTED NODES AFTER SIMULATION ........................................................... 74
FIGURE 19 NABIRE AIRPORT CONNECTED NODES AFTER SIMULATION ......................................................................... 76
xi
LIST OF TABLES
TABLE 1 ATR 72 NETWORK METRICES IN 2013 ................................................................................................... 21

TABLE 2 TOP 20 AIRPORTS OF 2013 NETWORK RANKED BY THEIR DEGREE ................................................................. 22
TABLE 3 TOP 20 AIRPORTS OF 2013 NETWORK RANKED BY THEIR WEIGHTED DEGREE .................................................. 23
TABLE 4 TOP 20 AIRPORTS OF 2013 NETWORK RANKED BY THEIR PAGERANK ............................................................. 25
TABLE 5 TOP 20 AIRPORTS OF 2013 NETWORK RANKED BY THEIR BETWEENNESS ........................................................ 26
TABLE 6 ATR 72 NETWORK METRICES IN 2014 ................................................................................................... 27
TABLE 7 TOP 20 AIRPORTS OF 2014 NETWORK RANKED BY THEIR BETWEENNESS ........................................................ 30
TABLE 8 TOP 20 AIRPORTS OF 2014 NETWORK RANKED BY THEIR DEGREE ................................................................. 32
TABLE 9 TOP 20 AIRPORTS OF 2014 NETWORK RANKED BY THEIR WEIGHTED DEGREE .................................................. 33
TABLE 10 TOP 20 AIRPORTS OF 2014 NETWORK RANKED BY THEIR PAGERANK .......................................................... 34
TABLE 11 ATR 72 NETWORK METRICES IN 2015 ........................................................................................... 35
TABLE 12 TOP 20 AIRPORTS OF 2015 NETWORK RANKED BY THEIR BETWEENNESS ...................................................... 37
TABLE 13 TOP 20 AIRPORTS OF 2015 NETWORK RANKED BY THEIR DEGREE ............................................................... 38
TABLE 14 TOP 20 AIRPORTS OF 2015 NETWORK RANKED BY THEIR WEIGHTED DEGREE ................................................ 39
TABLE 19 TOP 20 AIRPORTS OF 2016 NETWORK RANKED BY THEIR WEIGHED DEGREE.................................................. 45
xii
TABLE 36 ATR 72 NETWORK METRICES AFTER MOSES KILANGIN AIRPORT SIMULATION ............................ 75
xiii
CHAPTER 1
INTRODUCTION
1.1. General Statement of Problem Area

The 20th century was the beginning with the invention of aircrafts, ever since
then countless re-inventions have been made to optimize the potential to reach
almost anywhere. Air travel made the world as if it is shrinking and gave the ability to
connect easier. The connectivity of air transport also affecting the country’s economic
as it is facilitated to generate business worldwide, promotes tourisms, and provide
employments.
With the fast growing of aircraft innovation, some area does not have the
capability catch up, in the remote area having the infrastructure to support the
growth. Especially in Indonesia, a developing country with most of the area are still
underdeveloped. The most common jet engine aircraft, Boeing 737 requires at least
1,500 meters to land and 2,300 meters for take-off on MTOW, in which some airports’
infrastructure in Indonesia does not have. Therefore, ATR 72 a twin-engine turboprop
is commonly used to travel to regions inaccessible for jet engine aircraft.
Indonesian usage of ATR 72 had an immense growth from first introduced in

Indonesia. In 2013, 4.2 million passengers flew with 196 flight routes. In 2019, 7.8
million passengers flew with 464 flight routes. Within 6 years the passenger
movements have increased by 84% along with 268 new routes. Therefore,
understanding the structural network of the ATR 72 in Indonesia is essential.
Network analysis is required to undertake the further problem that might

occur along with the growth. As has been said, a better understanding, design, and
1/81
optimization for ATR 72 in Indonesian structural network are the main objectives of
this thesis research.
1.2. Research Purpose

The purpose of this research is on two key points:
a. To give a better understanding about ATR 72 structural network in
Indonesia.
b. To model an optimized network.
1.3. Research Problem

The problem for this research is on two key points:
a. There is not any or similar research in Indonesia attending ATR 72
structural network.
b. The immense growth in ATR 72 passenger movement for the past 8
years.
1.4. Significance of Study

The results from this research will allow airlines to make new routes based on
potential route modelled network. This allows passengers to travel much more
convenient and gain airlines profits from the travel.
1.5. Theoretical Perspective

The complex network structure is best described as a cobweb or web-like
structure enticed randomly. As an example, in World Wide Web, web pages are
connected if there is a link to one another. Both webpages are the nodes and the link
that connects them is the edge. Facebook is also a representative of a complex
network, with Facebook itself as the whole network, user are the nodes and any
2/81
relationships/friendships with one another are the edges. This is just a few of the
countless forms of network structures.
The network 𝒢 Airports is the nodes N and ATR 72 flight routes are the edges
E. To successfully perceive the whole network characteristics, modelling the right
scenario is required. There are two scenarios that will be used on this research, the
unweighted and the weighted. The weighted scenario is only used to plot the
passenger movements (link weight).
Characterizing the network help predicting the growth of the network. The
small world network is characterized as follows, there is a local clustering and have
shortcut link that reduce the path length. In brief, a network that has small network
characteristics most of its nodes are not neighbour or connected, but their neighbours
are likely to be a neighbour to one another, therefore from node A can reach node Z
with a few small hops. As in a scale free network, a network that follows the power law
degree distribution. In short, a scale free network has two characteristics, exponential
growth of nodes and preferential attachment (the probability of a node connecting to
a popular node is high).
Calculating the network metrics, connected components, average

connections, average weighed degree, diameter, average shortest path, graph density,
average clustering coefficient, average neighbourhood overlaps, betweenness, and
PageRank. Classification of community structure and resiliency analysis will be
calculated to mitigate any problem caused whenever either of the nodes closed. The
analysis will show how it affects any component and passengers’ movement, plotted
on a graph.
3/81
1.6. Research Question

Question # 1: How to understand the characteristics of structural
network of the ATR 72 in Indonesia?
Question # 2: How to optimize the current network based on current
passenger movements’ growth rate?
4/81
CHAPTER 2
LITERATURE REVIEW
Understanding the complex network behaviour is required to achieve the

objective of this research. Results are guaranteed when specific calculations and
evaluations are held. Evaluating similar work related to complex network structure will
shed some light on how the air transportation network behave in other countries
around the world and even the whole network structure in the world. Nonetheless,
the primary network models and network metrics should be beyond comprehension.
Therefore, the network can be optimized to the optimum version of itself.
2.1. Network Optimization

According to several authors regarding to the network optimization, the
purpose of network optimization is to use the existing data with an improved version
of the network with the goal to improve all the overall metrices of the network at the
minimum cost, in the air transportation network is minimizing the number of new
flight routes (edges).
Quoting (Wilson, 2020) “Network Optimization is a set of technologies and
methods that aim to improve the overall health of a network.”. the goal of a network
optimization is to improve all the overall metrices of the network. According to
(Technopedia, 2022) “The goal of any network optimization is with the given set of
constraints; ensuring an optimal network design with lowest cost structure and free
flow of data. Network optimization should be able to ensure optimal usage for system
resources, improve productivity as well as efficiency for the organization.”, improving
the productivity and efficiency on how to optimize the network is the method on to
optimize the network.
5/81
2.2. Air Transportation Network Around the World
2.2.1. Worldwide Air Transportation Network

There is some impactful literature with a similar or related topics with
optimising the air transportation network from both nationally and worldwide. The
worldwide air transportation network has been analysed by (Guimerà et al., 2005) for
its infrastructure impact and found that the worldwide air transportation network
follows two characteristics; it is a scale-free and small world network. Second, the
most connected cities do not give the assurance for the city to be the most central.
This anomaly was caused by the multi-community in the network, these communities
are also unable to be arranged merely by their geographical preferences. However,
geopolitical constraints must also take account into the bargain. Each city is given a
global role based on its pattern of intercommunity and intracommunity connections;
this will let to procure the scale-specific representation of the network.
2.2.2. Brazilian Air Transportation Network

Brazilian air transportation network exhibits small world characteristics, and
the national network follows the power-law distribution or also known as scale-free.
The author (Couto et al., 2015) built two networks classified by national, international
and both types of flights together. Virascopos Airport of Campinas City is found to be
the most central and connected airport on the national flights’ network. Using
resiliency analysis, the national flights’ network will be separated into six districts
subnetworks when an operational problem occurs in Virascopos Airport.
2.2.3. Indian Air Transportation Network

The air transportation network in India analysed by (Bagler, 2007) shows the
result of a small world network and truncated power-law degree distribution. The
network is found to be interconnected groups of airports and concentrated between
6/81
the large airports as hubs. However, the Indian air transportation network is found to
be offset from the traffic dynamics by having a disassortative mixing.
2.2.4. Italian Air Transportation Network

A slight difference of scale-free behaviour is found on the Italian air
transportation network that was analysed by (Guida & Maria, 2007), suggesting that
the growth mechanism might be different from the previously proposed. However, the
Italian air transportation network does not have any community within; this
occurrence is possible to have resulted from such smallness of clustering coefficient in
the network.
2.3. Small World Networks

A real-world network behaves like a random graph; however, the network
surprisingly has large clustering coefficients (Albert & Barabási, 2001). The traditional
or unsophisticated network’s clustering coefficient values depends on the nodes’
graphical position within the network. It is practically means; the probability of a node
connected to another node relies upon the geographical positions. A closer node will
have a larger probability of being connected with one another than the further node.
In a random graph network, the geographical position does not determine the
probability of a node connected to the nearest neighbouring node. Therefore, any
node will randomly connect to other random nodes depending on the randomness of
the network.
A small world network lies in between those networks, a network with regular
lattice then improved by moving a few of the existing edges to in between two non-
neighbouring nodes to create a short path or a shortcut in the network with the
probability of 𝑝, at random. However, an edge cannot move to an existing edge, in
other words, a double edge should not exist.
Watts and Strogatz proposed an ordered network and a random network
(Watts & Strogatz, 1998); the algorithm is as follows:
7/81
1. In an unsophisticated network order, every node is connected with

every two subsequent 𝐾 neighbouring nodes.
2. Randomizing network order randomly moves the edges with the
probability of 𝑝 with no double edge. By varying the 𝑝 between zero
until one. At the same time, one is the maximum value of randomness.
Figure 1 A random moving edges of a Watts-Strogatz model. The network has 20 nodes and 40 edges. (a)
unsophisticated network (b) small world network (c) random network
Watts and Strogatz also find the relation between clustering coefficient 𝐶 and
small path length 𝑙. With a significant interval of 𝑝, the average path length 𝑙(𝑝) is
almost equal to 𝑙(1), whereas in 𝐶(𝑝) ≫ 𝐶(1). Therefore, with a small number of 𝑝,
there is a massive drop of 𝑙(𝑝). However, the 𝐶(𝑝) will remain with a similar value or
even unchanged. The event has resulted in a clustered network; however, it has a
small average path length.
2.4. Scale-Free Model

Power-law distribution in network model was first introduced by (Barabási &
Albert, 1995), the vertex connectivities follows a scale-free power-law distribution. The
research model however does not include two main aspects from real world.
1. Assuming both models started with a fixed number of vertices which
then randomly connected or reconnected without varying the
number of vertices.
2. The network models assume the probability of any two vertices to
be connected is random and uniform.
8/81
However, in most of real world situation differs from the model; the real world
networks are open, and the number of vertices is increasing over time to the system
and the network performed a preferential connectivity. For example, the World Wide
Web (WWW) grows exponentially over time by the addition of new web pages, a less-
known actor is more likely to be casted as an extra or supporting role rather than the
main character. Therefore, the probability of a less-known actor is low compared to
the well-known actor to be casted as the main character.
9/81
CHAPTER 3
METHODOLOGY
This chapter explains how the research conducts, comprehended into four
steps. First, collecting data of ATR 72 flights in Indonesia from 2013 until 2019 for both
scheduled and non-scheduled flights. All the data is being projected into a model
network using Gephi then observed for each characteristic. Therefore, comparing each
network characteristic is required to determine the network growth that projects the
future network.
3.1. Data Collection

This research requires a bit of data to be filled, several of flights in the
network and a number of passengers in those flights of ATR 72 in Indonesia. All the
data were acquired from DEPHUB from 2013 until 2020. However, since COVID-19
spread from early 2020, calculating the anomaly data into the equation will create an
inconsistent result. Therefore, only the passenger’s numbers and the number of flights
data from 2013 until 2019 are used for modelling and calculating the network
characteristics. Even though the data acquired does not provide the detailed
information of each flight passenger, the total number of passengers in each flight
route yearly is sufficient to complete the evaluation to characterise and optimise the
network.
3.2. Network Modelling

A modelling network is required to visualising a precise relationship between
a node and any other node in the network. Being able to manipulate the data in real-
time is one of the crucial features required. Therefore, Gephi software is used in this
research. Gephi is specially programmed with many advantages in network
10/81
visualisation and analysis met all of the requirements to model the ATR 72 network’s
growth.
The network modelling needs two initial data sets, the list of nodes and the
edges connected in the network. These data sets are imported to Gephi in Comma-
separated values (CSV) files, which then will be manipulated in various ways to gain as
much information that can be exported. The layout is manipulated by using the Geo
Layout with the longitude and longitude of each airport as the reference for each node
position. Gephi will be able to provide the network statistics, filter the data to hide
undesired data in the network, and even give the appearance, colours and sizes to
each node and edges based on the statistics as pleased.
Network modelling is one way to determine the logical structure of a
database model. The network itself represents the objects in the system and the
relationships between each one. They are featured in a graph where the objects are
nodes 𝑣, and the relationships between objects are edges ε. In this network, airports
are represented as the nodes, and the flights’ routes are the links. A link is valid when
there is a connection between two airports regardless of the direction of flight.
Two types of graphs that are worth mentioning, the unweighted and weighted
links. As in the unweighted graph, the value of each link is equal to 1. However, in the
weighted graph, the value of each link might be different from one another. In this
research, two types of this graph will be used depending on the metrics calculated. The
weighted graph will only be used for the number of passengers on the flight.
Nevertheless, the links connecting also have two versions, un-directed and
directed links. Whereas, in the un-directed link, the connection between nodes does
not differentiate the source and target of the connection. Therefore, regardless of the
direction of flight, whether node 𝑢 to node 𝑣 or node 𝑣 to node 𝑢, it will count as the
same link. However, in the directed link, the weight of the link is countable if the
source and target are the same. A flight from node 𝑢 to node 𝑣 and node 𝑣 to node 𝑢
have their own separate links with an arrow to determine the source and target.
The network models are visualised using software Gephi and focused on the
national flight’s network of ATR 72 in Indonesia yearly from 2013 until 2019; this
11/81
network model will be the critical point to characterise the ATR 72 network’s
infrastructure. The main objective of modelling those networks is to predict the
possibility of finding new routes that will optimise the network behaviour.
3.3. Network Characterise

Characterising the network will give a better insight into the network. Detailed
information of each node to the whole network is the network characteristics or
network metrics.
Network metrics are computed from several variables, the average degree,
average weighted degree, diameter, average shortest path, graph density, average
clustering coefficient, average neighbourhood overlap. To characterise the network
metrics, (Newman, 2003; Barab´asi, 2002) methods are used.
3.3.1. In-degree and Out-degree

In-degree and out-degree are used in a directed graph. In-degree 𝑘in(𝑣) is the
incoming edges of node v, whereas the out-degree 𝑘out(𝑣) is the outgoing edges of
node v. The total from in-degree and out-degree link of node v is the total degree
𝑘(𝑣). the average degree < 𝑘 > of a graph 𝒢 is calculated as:
< 𝑘 > (𝒢) = ∑ 𝑘(𝑣)/|𝑉|.

∀𝑣∈𝑉
Equation 1
In the air transportation network, flight from Husein Sastranegara Airport

(BDO) to Halim Perdanakusuma International Airport (HLP) will be considered an in-
degree link of HLP and an out-degree link of BDO.
3.3.2. Average Neighbourhood Overlap

Neighbourhood overlap determines the percentage of neighbour overlap
between two nodes 𝑢 and 𝑣. The set of neighbourhoods of node 𝑢 to node 𝑣, 𝑁(𝑢)
and 𝑁(𝑣), respectively, the neighbourhood overlap 𝑛o(𝑢, 𝑣) is calculated as:
12/81
(𝑁(𝑢) ∩ 𝑁(𝑣))
𝑛o(𝑢, 𝑣) = .
(𝑁(𝑢) ∪ 𝑁(𝑣))
Equation 2
Therefore, the average neighbourhood overlap < 𝑛o> is calculated as:

(∑∀(𝑢,𝑣)∈𝜀 𝑛o (𝑢,𝑣))
< 𝑛o > = .
|𝜀|
Equation 3
3.3.3. Average Shortest Path

The shortest path 𝑙(𝑢, 𝑣) is best described as the lowest number of hops from
the set of paths from node 𝑢 to node 𝑣, 𝑃(𝑢, 𝑣). Therefore, with having the set of
every shortest path 𝐿, the average shortest path < 𝐿 > of graph 𝒢 is calculated as:
∑∀𝑙(𝑢,𝑣)∈𝐿 𝑙(𝑢, 𝑣)
< 𝑙 > (𝒢) = .
|𝐿|
Equation 4
Figure 1 gives the example of the shortest path from Sultan Aji Muhammad Sulaiman
Sepinggan International Airport (BPN) to A.A. Bere Tallo Airport (ABU), which is four
hops, from BPN to Sultan Hasanuddin International Airport (UPG) to Frans Seda
Airport (MOF) to El Tari Airport (KOE) then the last hop to ABU.
Figure 2 The network infrastructure of ATR 72 around Sultan Hasanuddin International Airport in 2019
13/81
3.3.4. Diameter
Diameter 𝑑 is the contrary of the shortest path, best described as the longest
path from the set of shortest paths between any pair within the network. The
diameter is calculated as:
𝑑(𝒢) = max 𝑙(𝑢, 𝑣).
∀(𝑢,𝑣)∈𝑉
Equation 5
3.3.5. Clustering Coefficient

Clustering coefficient 𝐶 is also known as network transitivity and is best
described as the probability of two neighbouring nodes in a network having the same
neighbour. If node 𝑢 is the neighbour of node 𝑣, and node 𝑣 is the neighbour of node
𝑞, hence, there is a chance for node 𝑢 to be the neighbour of node 𝑞. Connected tripes
is when a node has any links to any two other pairs of nodes, regardless of whether
those two other pairs of nodes are connected to one another. The clustering
coefficient is calculated as:
3 × 𝑛𝑢𝑚𝑏𝑒𝑟 𝑜𝑓 𝑡𝑟𝑖𝑎𝑛𝑔𝑙𝑒𝑠 𝑖𝑛 𝑡ℎ𝑒 𝑛𝑒𝑡𝑤𝑜𝑟𝑘
𝐶(𝒢) = .
𝑛𝑢𝑚𝑏𝑒𝑟 𝑜𝑓 𝑐𝑜𝑛𝑛𝑒𝑐𝑡𝑒𝑑 𝑡𝑟𝑖𝑝𝑒𝑠 𝑜𝑓 𝑣𝑒𝑟𝑡𝑖𝑐𝑒𝑠
Equation 6
The average clustering coefficient is simply calculated as:

∑𝑖 𝐶𝑖
<𝐶 >= .
𝑛
Equation 7
For example, the graph 𝒢(𝑢, 𝑣, 𝑞, 𝑡, 𝑟) has 1 triangle in the network and 4 tripes.
Therefore, the clustering coefficient is 0.75.
14/81
Figure 3 Clustering coefficient of graph G is 0.75
3.3.6. Betweenness
Betweenness 𝛽(𝑣) of a node 𝑣 is best described as the percentage of how
many times node 𝑣 is hopped on from the shortest path of any node to any other
node. If 𝐿(u, q) represent the set of all shortest paths from node 𝑢 to node 𝑞, and the 𝐿(u,
q)(𝑣) is the set of all shortest paths from node 𝑢 to node 𝑞 that passes through node 𝑣,
the betweenness is calculated as:
𝐿(u, q)
𝛽(𝑣) = ∑ .
𝐿(u, q)(𝑣)
𝑘≠𝑣,𝑘∈𝑉
Equation 8
3.3.7. Closeness Centrality

Closeness centrality 𝛾(𝑣) is best described as the average distance from node
𝑣 to any other nodes in the network 𝑘. 𝑙(𝑣, 𝑘), the length of the shortest path from
node 𝑣 to any other reachable nodes 𝑘, the closeness centrality is calculated as:
−1
𝛾(𝑣) = [ ∑ 𝑙(𝑣, 𝑘)] .

𝑘≠𝑣,𝑘∈𝑉
Equation 9
15/81
3.3.8. PageRank
PageRank is the ranks of nodes according to determine the importance of a
node in the network, essentially by counting the weight and number of connections of
the node and the centrality of the links. Therefore, it is safe to assume that a node with
more links will be ranked higher. The rank of node 𝑃(𝑣) is by summing the node ranks
that pointed to node 𝑣 divided by the number of links pointed out of node 𝑣. The
PageRank is calculated as:
𝑟(𝑃(𝑢))
𝑟(𝑃(𝑣)) = ∑𝑃(𝑢)∈𝐵𝑃(𝑢) .
|𝑃(𝑢)|
Equation 10
3.3.9. Graph Density

Graph density 𝐷 is best described as the ratio of how close the graph is to a
complete state; the complete graph has all the possible edges |𝜀| and the density
equals 1. Graph density is calculated as:
2 × |𝜀|
𝐷= .
(|𝑉| × (|𝑉| − 1))
Equation 11
3.4. 2013 -2019 Data Comparison
The comparison of the characteristics from 2013 growth until 2019 will
enlighten how to optimise the newest network. However, not necessarily compared to
each node, evaluating the overall network characteristics will give a better illustrate
how the network progresses over time. Each metric’s characteristics are evaluated
showing which metric’s characteristics progress from the beginning of data collected
until the last, affecting the network the most.
Knowing how the network progresses and what is affecting it will enable to
predict how the network will behave in the future and shows the best way to optimise
the network.
16/81
3.4.1. Community Structure

There is almost guaranteed in a real network that it has a community
structure, a group of nodes with strong links' density between themselves and have a
low connection to the other groups. A particular role forms these groups; it implies a
specific characteristic or interest that is shown only among the nodes in the group.
Therefore, perceiving the underlying reason for a community’s structure existence and
how the rules are set is important to network analysis.
James Moody conducted a study of the friendship network of children in a
U.S. school. The study was performed using the "spring embedding" algorithm
(Moody, 2001); the algorithm visualised linear springs are connected nodes in the
network, then relaxed using the first-order energy minimisation. The study can deduct
how the network’s community structure is formed; the first result comes when Moody
colours the nodes according to the race of the student representatives, like white,
black, and others. The second result deducted that the communities are also formed
by the division of ages (middle school and high school).
Another way to form the community structure is by using the Louvain
Method, created by (Blondel et al., 2008). This method uses the optimisation of
modularity 𝑄 to process the algorithm. In Louvain Method, the relative density of links
between the nodes in the group compared to the density links outside of the group,
the modularity 𝑄 is scaled from -0.5 to 1, a non-modular clustering and a fully modular
clustering, respectively. In the weighted network, the modularity is calculated as:
1 𝑘i𝑘j
𝑄= ∑ 𝐴ij − 𝛿(𝑐i, 𝑐j)
2𝑚 2𝑚
𝑖𝑗
Equation 12
𝐴ij = Edge weight between node 𝑖 and 𝑗

𝑘i and 𝑘j = Sum of weight attached to node 𝑖 and 𝑗, respectively
𝑚 = Sum of all edge weight in the graph
𝑐I and 𝑐j = Communities of the node 𝑖 and 𝑗, respectively
𝛿 = Kronecker delta function
17/81
The Kronecker delta function has the task to limit the summation for nodes which pairs
within the same community. If the argument is equal or the node 𝑖 and 𝑗 is connected,
then Kronecker delta function value 1, else the value is 0.
In the Gephi software, the resolution to be manipulated is to assess the sensitivity or

level of detail. Therefore, in a high resolution level the smaller group will join into the
larger group, and vice versa.
Figure 4 Resolution in Gephi modularity setting
18/81
3.5. Analyze The Plausible Improvement

The main purpose of analysing the network is only to improve what can be
optimised; therefore, a detail-oriented overview must be conducted. Evaluating the
2013-2019 data comparison results will determine the best approach to which metric
to focus on to improve and which group of community structures will be optimised.
Each of the metrics will have a unique way to be optimised; however,
increasing every metrics will optimise the air transportation network for ATR 72 in
Indonesia. Generally, to optimise a network, increasing the average degree < 𝑘 >,
average weighed degree < 𝑚 >, graph density 𝐷 and average clustering coefficient 𝐶;
shortening the diameter 𝑑 and average shortest path < 𝑙 > is the approach to be
taken. However, ATR 72 have their own characteristics to be considered as well; the
main disadvantage of ATR 72 compared to any jet propulsion engine is the flight
distance. ATR 72 only can fly for 1,500 km compared to Boeing 737-700 or Airbus
A320-200, which can fly up to 5,700 km (Dias & Lopes, 2020).
19/81
CHAPTER 4
RESULTS AND DISCUSSION
4.1. Network Model and Characteristics
4.1.1. Network Model and Characteristics of 2013

In 2013, Indonesia had 673 active airports throughout the islands; these
airports are computed from airports that handle international flights until the most
remote airports between the high mountains in the valley. However, ATR 72 was only
operated within 77 airports and 196 scheduled flights. Throughout 2013, ATR 72 was
able to be flown to deliver over 4.2 million passengers, which means it is around 81
thousand passengers are flown weekly.
The ATR 72’s network 𝒢 for 2013 generally has 77 nodes and 196 edges. The
network metrics are found in Table 1; in summary, the graph is separated into 2
connected components. The separation is in the north of the Sumatera region; there is
no connection between Silangit International Airport (DTB) with the nodes around.
This means it is impossible to fly and ready any airport in the western part of Indonesia
from the east side of DTB. Even though the network is separated, calculating the
network's diameter is still possible just by considering the larger part of the network; it
requires 15 transits or hops to reach from one end to the other end of the larger part
of the network. The network connectivity is deficient considering the total nodes in the
network; on average, each airport is connected to only 2 or 3 airports or 3.31% of the
total airports in the network.
20/81
Table 1 ATR 72 Network Metrices in 2013
Network Metric Value

Connected Component 2
Diameter 15
Average Degree 2.545 (3.31%)
Graph Density 0.033
Average Shortest Path 5.553
Average Weighted Degree 55,142.545
Average Clustering Coefficient 0.180
ATR 72 passengers are on average, have to travel 6 times (The Average Shortest
Path is equal to 5.553) to reach their destination. Within each of the travel routes, it is
expected to have 55 thousand passengers every year. The average clustering
coefficient is very low, with only 0.180. This indicates every airport has an 18% chance
of being connected to any other airport. Hence, explains the low graph density of the
network with only 3.3% of the maximum graph density (complete graph has every
possible edge).
From the model metrics, a detailed analysis of the network can be performed
to identify the main airports of the network and which role each has. The network is
expected to have a scale-free model on the degree count of the network. Only 66
nodes (85.71%) of the network have less than 10 connections to other nodes, which
from the left, only 2 nodes have more than 18 connections. Ngurah Rai International
Airport (DPS) has the highest degree with 19 connections. Therefore, the degree
cumulation is proven to abate the power-law theorem as the network has a lot of
nodes with low connectivity and a small number of nodes with high connectivity with
the role of a hub Table 2 until Table 5 shows the top 20 airports based on the
betweenness centrality, degree count, weighted degree, and PageRank. Based on the
degree count table, it shows the two largest hubs are DPS and Kualanamu
International Airport (KNO). Based on the weighted degree table, DPS is claiming the
21/81
number 2 position on the top 2 airports, after Juanda International Airport (SUB) with
807 thousand passengers in a year.
Table 2 TOP 20 airports of 2013 network ranked by their degree
AIRPORT (IATA) BETWEENNESS DEGREE WEIGHTED PAGERANK

DEGREE
DPS 887.767 19 807120 0.041
KNO 20.000 18 344196 0.009
UPG 1601.000 14 668712 0.052
PNK 1302.233 14 201411 0.033
AMQ 690.667 14 239918 0.025
MDC 1182.000 13 153684 0.028
BTH 658.000 13 158138 0.040
PKN 618.167 12 110851 0.027
BPN 1050.500 11 185668 0.027
SRG 1004.033 10 188703 0.021
SUB 222.633 10 873863 0.023
BDJ 1575.000 8 174330 0.019
SOQ 851.000 8 69198 0.015
DJJ 256.000 8 217910 0.020
TRK 250.000 8 67175 0.021
FKQ 61.833 8 45570 0.011
KNG 14.000 8 51525 0.014
AAP 92.000 7 74261 0.014
LUV 80.667 7 146510 0.018
SMQ 488.567 6 52198 0.014
22/81
Table 3 TOP 20 airports of 2013 network ranked by their weighted degree

DEGREE
SUB 222.633 10 873863 0.023
DPS 887.767 19 807120 0.041
UPG 1601.000 14 668712 0.052
LOP 2.400 5 505197 0.012
KDI 115.000 4 378960 0.016
KNO 20.000 18 344196 0.009
AMQ 690.667 14 239918 0.025
GNS 0.000 4 230746 0.003
DJJ 256.000 8 217910 0.020
JOG 190.967 6 215350 0.015
KTG 10.233 6 210226 0.013
PNK 1302.233 14 201411 0.033
SRG 1004.033 10 188703 0.021
BPN 1050.500 11 185668 0.027
BDJ 1575.000 8 174330 0.019
WMX 0.000 4 171761 0.011
BTH 658.000 13 158138 0.040
MDC 1182.000 13 153684 0.028
LUV 80.667 7 146510 0.018
BEJ 395.500 6 140620 0.014
Ngurah Rai International Airport is once again shown to be the most influential
airport in the ATR 72 network as it appears in the second place of the top 20 airports
ranked by their PageRank, with 0.011 differences with Sultan Hasanuddin International
Airport. The differences are managed to be gained by Sultan Hasanuddin International
Airport as the airport is located in the most central of the network, holding the sole
node to connect the east and the west side of the network as all of the nodes in the
network will pass through the airport to cross as in Figure 5. Therefore, the
23/81
exceptionally high betweenness of the Sultan Hasanuddin International Airport is

justified.
Figure 5 ATR 72 network around Indonesia in 2013
The main purpose of using ATR 72 is to gain connectivity, determining how the
analysis will be headed; however, Sultan Hasanuddin International Airport managed to
gain the highest Page Rank centrality. These points were gained by the immensely high
betweenness, high connectivity. Therefore, the most important airport in the 2013
Network is Sultan Hasanuddin International Airport, located in Makassar, South
Sulawesi.
24/81
Table 4 TOP 20 airports of 2013 network ranked by their pagerank

DEGREE
UPG 1601.000 14 668712 0.052
DPS 887.767 19 807120 0.041
BTH 658.000 13 158138 0.040
PNK 1302.233 14 201411 0.033
MDC 1182.000 13 153684 0.028
PKN 618.167 12 110851 0.027
BPN 1050.500 11 185668 0.027
AMQ 690.667 14 239918 0.025
SUB 222.633 10 873863 0.023
SRG 1004.033 10 188703 0.021
TRK 250.000 8 67175 0.021
DJJ 256.000 8 217910 0.020
BDJ 1575.000 8 174330 0.019
LUV 80.667 7 146510 0.018
ENE 35.000 5 57287 0.017
KDI 115.000 4 378960 0.016
JOG 190.967 6 215350 0.015
SOQ 851.000 8 69198 0.015
KOE 1.500 4 61956 0.014
AAP 92.000 7 74261 0.014
25/81
Table 5 TOP 20 airports of 2013 network ranked by their betweenness

DEGREE
UPG 1601.000 14 668712 0.052
BDJ 1575.000 8 174330 0.019
KBU 1480.000 4 92080 0.012
PNK 1302.233 14 201411 0.033
MDC 1182.000 13 153684 0.028
BPN 1050.500 11 185668 0.027
LUW 1008.000 3 47673 0.012
SRG 1004.033 10 188703 0.021
DPS 887.767 19 807120 0.041
SOQ 851.000 8 69198 0.015
NTX 721.000 4 35586 0.012
AMQ 690.667 14 239918 0.025
BTH 658.000 13 158138 0.040
PKN 618.167 12 110851 0.027
SMQ 488.567 6 52198 0.014
BEJ 395.500 6 140620 0.014
NBX 328.000 6 48440 0.012
DJJ 256.000 8 217910 0.020
TRK 250.000 8 67175 0.021
SUB 222.633 10 873863 0.023

In 2014, the total number of airports and the number of flight routes had
several removals, and new placement of edges and even a new node, 25 existing
routes in 2013 were removed and adding of total 114 new routes throughout the
network. With the changes in 2014, the total number of airports and the number of
flight routes of ATR 72 were increased by 10 nodes and 89 edges. The number of edges
26/81
was increased immensely, with over 45% of the initial edges in 2013. With the
increment of edges around the network, the total number of passengers who had
flown in 2014 were able to be increased by slightly over 12% from 2013.
The ATR 72 network metrics in 2014 are listed in Table 6; the network is still
separated into 2 parts. However, the separation is no longer in the north of Sumatra.
Silangit International Airport is now connected to the neighbouring node, Kualanamu
International Airport. The network separation results from the removal of edges
between Natuna Airport (NTX) and Supadio International Airport (PNK), as visualised in
Figure 6. The removal might be subsequently to the low passenger movements
between the airports; only around 100 passengers were flown between the two
airports weekly. This removal brings the 2013 larger part of the network smaller.
Hence, the diameter of the network is smaller compared to the previous graph with 12
hops. The network connectivity rose by 0.731 points from the previous year; this brings
up the percentages of the connectivity of each airport respected to the total number
of airports flown by ATR 72 in 2014 to 3.77%. The increment of average degree also
affects the graph density positively, increasing the number to 3.8% of the complete
graph with perfect edges.
Table 6 ATR 72 Network Metrices in 2014

Diameter 12
Graph Density 0.038
27/81
The ATR 72 passengers in 2014 are, on average, have to transit 4 times or fly 5
times to reach their final destination. This number is improved from the previous year
by around 0.5 points. However, the average number of passengers flown from each
route yearly decreased by around 500 pax. The decline might correspond to the very
high increase of edges from the previous year. The average clustering coefficient was
increased by almost double the initial point in the previous year. This means the
possibility of a node being connected with any random node is 31.5%.
The model metrics allow us to identify the main airports of the network along
with the role each of the nodes corresponds to. By plotting the degree count of the
network, the results are as expected to be following the power law, only 10 of the 87
nodes have connections over half of the maximum connection in the network and a
very high number of nodes with only 2 connections.
28/81
Degree Count in 2014 Network

35
30
30
25
20
15 12
9
10 7
5
5 2 2 3 3 3
1 2 0 1 0 1 1 1 0 2 1 0 1
0 0
0
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25
-5
Figure 7 Degree count graph of 2014 network.
Table 7 until Table 10 shows the highest tier of airports based on their betweenness,
number of connections, weighted degree, and PageRank. Based on table 7, Syukuran
Aminuddin Amir Airport (LUW) ranked the highest, passing the previously ranked
number 1 in 2013, Sultan Hassanuddin International Airport, with the point of
2636.833. It is 1000 points higher than the previously ranked number 1. This might be
the result not only from Syukuran Aminuddin Amir Airport is positioned to be the
bridge from the centre part of the network to the eastern part of the network, but also
has another connection to the neighbouring nodes besides Syukuran Aminuddin Amir
Airport as in the previous year the only connection of Syukuran Aminuddin Amir
Airport was sole to Sultan Hassanuddin International Airport and Sam Ratulangi
International Airport. The new connection gave a perfect shortcut from airports
around Kalimantan to hops through Syukuran Aminuddin Amir Airport to the eastern
part of the network.
29/81

DEGREE
LUW 2636.833 6 83934 0.012
MDC 2495.500 16 179715 0.034
UPG 2476.786 22 686942 0.039
GTO 2392.000 4 50851 0.009
DPS 2076.416 25 863615 0.034
BPN 1309.190 22 276793 0.025
AMQ 1209.500 19 277553 0.031
SOQ 1198.333 9 90683 0.017
PLW 915.190 6 10914 0.011
KOE 829.833 23 153688 0.045
PNK 756.365 18 282131 0.023
SRG 726.810 14 275912 0.016
NBX 700.000 7 124105 0.013
TTE 601.667 6 74658 0.012
DJJ 576.000 8 189762 0.018
SUB 545.132 20 629466 0.023
KBU 335.870 6 167659 0.009
BDJ 302.330 12 248454 0.014
LOP 252.652 10 439107 0.015
FKQ 180.333 9 87711 0.012
30/81
Table 8 and Table 9 shows similar results, with Ngurah Rai International Airport ranked
number 1 with 25 degrees and 863 thousand passengers movement every year.
Followed by a new airport in the ATR 72 network, El Tari Airport (KOE), with 23
degrees, El Tari Airport did not manage to gain enough passenger movements yearly to
be in the top 20. Table 10 shows the top 20 airports based on the PageRank of each
node; the new node in the network, El Tari Airport, managed to gain 0.045 points
followed by Sultan Hasanuddin International Airport with a 0.006 point difference.
Based on the four tables shown, the most important airport in 2014 is Ngurah
Rai International Airport, as the respected airport managed to gain a very high point on
all betweenness, connectivity, weight degree, and PageRank tables.
31/81

DEGREE
DPS 2076.416 25 863615 0.034
KOE 829.833 23 153688 0.045
UPG 2476.786 22 686942 0.039
BPN 1309.190 22 276793 0.025
SUB 545.132 20 629466 0.023
AMQ 1209.500 19 277553 0.031
PNK 756.365 18 282131 0.023
MDC 2495.500 16 179715 0.034
SRG 726.810 14 275912 0.016
TRK 4.833 14 93076 0.015
AAP 3.000 13 80811 0.015
BDJ 302.330 12 248454 0.014
SMQ 83.478 12 70177 0.014
PKN 38.967 12 84077 0.014
LOP 252.652 10 439107 0.015
BTH 38.000 10 109857 0.013
KNO 30.000 10 294854 0.015
KTG 4.900 10 225297 0.012
MLK 0.000 10 34318 0.011
NNX 0.000 10 38301 0.011
32/81

DEGREE
DPS 2076.416 25 863615 0.034
UPG 2476.786 22 686942 0.039
SUB 545.132 20 629466 0.023
LOP 252.652 10 439107 0.015
KDI 148.000 4 318524 0.010
KNO 30.000 10 294854 0.015
PNK 756.365 18 282131 0.023
AMQ 1209.500 19 277553 0.031
BPN 1309.190 22 276793 0.025
SRG 726.810 14 275912 0.016
BDJ 302.330 12 248454 0.014
LBJ 98.833 9 238551 0.014
KTG 4.900 10 225297 0.012
JOG 3.571 4 224286 0.006
GNS 0.000 2 200542 0.004
DJJ 576.000 8 189762 0.018
MDC 2495.500 16 179715 0.034
BEJ 0.333 8 177955 0.009
KBU 335.870 6 167659 0.009
BMU 0.000 6 159312 0.010
33/81
Table 10 TOP 20 airports of 2014 network ranked by their PageRank

DEGREE
KOE 829.833 23 153688 0.045
UPG 2476.786 22 686942 0.039
MDC 2495.500 16 179715 0.034
DPS 2076.416 25 863615 0.034
AMQ 1209.500 19 277553 0.031
BPN 1309.190 22 276793 0.025
PNK 756.365 18 282131 0.023
SUB 545.132 20 629466 0.023
DJJ 576.000 8 189762 0.018
SOQ 1198.333 9 90683 0.017
SRG 726.810 14 275912 0.016
LUV 74.500 5 130410 0.016
LOP 252.652 10 439107 0.015
TRK 4.833 14 93076 0.015
AAP 3.000 13 80811 0.015
KNO 30.000 10 294854 0.015
BDJ 302.330 12 248454 0.014
LBJ 98.833 9 238551 0.014
PKN 38.967 12 84077 0.014
SMQ 83.478 12 70177 0.014
34/81

2015 is when there is a significant drop of ATR 72 usage in Indonesia; from the
previous year, even with a bit of increase of nodes in the network to 94 nodes, a total
of 26 edges were removed from the network. Which leave 257 edges in the network.
The network itself also got separated into 3 parts of networks; the previously
connected nodes between Silangit International Airport with Ferdinand Lumban
Tobing Airport must be removed due to the shallow interest of passengers flying with
this route. Only 33 passengers had flown using this route in the previous year. The
diameter of the network remains the same, with 12 hops within the network. The
average degree is lower by 0.55 points from the previous year (only 2.9% of the total
nodes).
Table 11 ATR 72 NETWORK METRICES IN 2015

Diameter 12
Graph Density 0.029
35/81
The network graph density also got shrank to 0.029, 0,009 points lower than
the previous year. It is expected due to the increase of nodes along with the drop off of
edges in the network. However, the average shortest path does not increase as much,
only a 1.4% increase from 2014, considering the other metrics increase. The drop of
average weighted degree also does not have a significant number 4.2% decrease from
2014. However, the average clustering coefficient had an enormous drop from 0.315
to 0.175, almost reduced by half of its initial point.
Table 12 until Table 15 shows the rank of nodes in the network based on 4
metric values: betweenness, degree, weighted degree, and PageRank. As expected,
with few network changes, UPG, Sam Ratulangi International Airport (MDC), and LUW
are placed on the top three of airports ranked by their betweenness, respect. The
Syukuran Aminuddin Amir Airport are back in third place due to the break off of the
shortcut connection to Kalimantan Island's airports. Therefore, aircraft must fly
through Sultan Hasanuddin International Airport, which brings the airport to rank
number 1. Table 13 shown that Ngurah Rai International Airport is no longer the
number one in terms of connections; the shortened number of edges seems to affect
Ngurah Rai International Airport significantly, from the previous year with 25 degrees
36/81
into 18 degrees in 2015. Therefore, Sultan Hasanuddin International Airport does not
affect as many claims as to the number one in the degree metric.

DEGREE
UPG 3370.143 21 571452 0.034
MDC 2536.000 16 233756 0.031
LUW 2066.500 8 128710 0.017
DPS 1701.514 18 719680 0.027
BDJ 1664.271 12 211153 0.016
BPN 1298.067 14 251130 0.022
AMQ 1202.833 18 335735 0.029
LOP 1196.200 8 363893 0.014
SOQ 1144.167 9 92753 0.013
SRG 954.052 16 318918 0.020
TTE 808.833 7 80794 0.014
GTO 671.500 6 78515 0.011
NBX 634.000 9 122224 0.015
KOE 599.000 17 279811 0.035
PNK 525.333 12 282155 0.019
DJJ 442.000 8 221603 0.017
SUB 437.862 16 485231 0.018
BEJ 427.000 6 148439 0.011
LBJ 380.667 7 301243 0.011
KBU 321.376 6 139918 0.008
37/81

DEGREE
KOE 3370.143 21 571452 0.034
UPG 1202.833 18 335735 0.029
UPG 1701.514 18 719680 0.027
AMQ 599.000 17 279811 0.035
DPS 2536.000 16 233756 0.031
KOE 954.052 16 318918 0.020
MDC 437.862 16 485231 0.018
SRG 42.000 14 378617 0.025
SUB 1298.067 14 251130 0.022
KNO 525.333 12 282155 0.019
BPN 1664.271 12 211153 0.016
PNK 88.667 10 66942 0.014
BDJ 109.276 10 58686 0.013
KNG 634.000 9 122224 0.015
PKN 1144.167 9 92753 0.013
NBX 194.250 9 70041 0.011
SOQ 32.000 8 102150 0.020
FKQ 42.000 8 74996 0.020
BTH 299.000 8 80791 0.017
PLM 442.000 8 221603 0.017
38/81

DEGREE
DPS 1701.514 18 719680 0.027
UPG 3370.143 21 571452 0.034
SUB 437.862 16 485231 0.018
KNO 42.000 14 378617 0.025
LOP 1196.200 8 363893 0.014
AMQ 1202.833 18 335735 0.029
SRG 954.052 16 318918 0.020
LBJ 380.667 7 301243 0.011
PNK 525.333 12 282155 0.019
KOE 599.000 17 279811 0.035
BPN 1298.067 14 251130 0.022
GNS 0.000 2 238632 0.004
JOG 2.405 4 233908 0.006
MDC 2536.000 16 233756 0.031
DJJ 442.000 8 221603 0.017
BDJ 1664.271 12 211153 0.016
KTG 5.567 6 194693 0.009
ENE 2.500 4 162326 0.008
BMU 0.000 4 157614 0.007
BEJ 427.000 6 148439 0.011
However, even with a significant drop in connections, Ngurah Rai

International Airport still has the most passengers flown with around 720 thousand
passengers, followed by Sultan Hasanuddin International Airport with less than 150
thousand differences as shown in Table 14. El Tari Airport, with very low betweenness
and passenger, has flown yearly, managed to gain the top of PageRank as shown in
Table 15 with 0.035 points, followed by Sultan Hasanuddin International Airport with
0.001 difference. Therefore, it is to be concluded that in 2015, the most important
39/81
airport was Sultan Hasanuddin International Airport, as the airport managed to hold
part of the network together when smaller bridges were crumbling until they
disappeared.

DEGREE
KOE 599.000 17 279811 0.035
UPG 3370.143 21 571452 0.034
MDC 2536.000 16 233756 0.031
AMQ 1202.833 18 335735 0.029
DPS 1701.514 18 719680 0.027
KNO 42.000 14 378617 0.025
BPN 1298.067 14 251130 0.022
PLW 298.000 7 51358 0.021
SRG 954.052 16 318918 0.020
BTH 32.000 8 102150 0.020
PLM 42.000 8 74996 0.020
PNK 525.333 12 282155 0.019
SUB 437.862 16 485231 0.018
TRK 299.000 8 80791 0.017
DJJ 442.000 8 221603 0.017
LUW 2066.500 8 128710 0.017
BDJ 1664.271 12 211153 0.016
NBX 634.000 9 122224 0.015
LUV 150.000 6 134579 0.015
PGK 28.000 6 116550 0.015

2016 is the year of revival of the ATR 72 network in Indonesia. 17 new nodes
are inserted into the network, with 70 new edges totalling the edges in the network
into 327 edges. These new edges manage to connect three separated components into
40/81
one large and complete network. The first connection is on the western part of the
network, between Pinang Kampai Airport (DUM) and Silangit International Airport. The
second connection is between Husein Sastranegara Airport (BDO) and Radin Inten II
International Airport (TKG). As shown in Figure 9, with these two connections, the
eastern part of the network is now able to hop through to the western part of the
network. Hence, a higher diameter is found in the network compared to the 2015
network with 15 hops. The average degree is increased; therefore, on average, every
node is connected to almost always 3 connections, rarely 2 connections. With the
increase of the number of nodes and edges, the graph density is decreased from the
previous year; therefore, the connection is still higher even with the decrease of graph
density.
41/81

Diameter 20
Graph Density 0.027
The increase of average shortest path is also due to the previously separated
network into one giant network; therefore, each network can now be taken into
calculation, increasing the required hops to reach any nodes from any nodes. The
number of passengers flown on average of each route is also increased tremendously,
around a 15% increment of passengers flown yearly from the previous year. The
average clustering coefficient also gains an incredible amount of points, with a 31.2%
rise from 2015.
Table 17 until Table 20 shows the top 20 airports ranked by their
betweenness, degree, weighted degree, and PageRank. The airports worth noting are
Ngurah Rai International Airport, Sultan Hasanuddin International Airport, and
Kualanamu International Airport. As Ngurah Rai International Airport manages to gain
the number one position based on the average weighted degree. Ngurah Rai
International Airport has over 740 thousand passengers movement this year, a slight
increase from last year passenger movement.
42/81

DEGREE
UPG 5273.729 32 595125 0.043
MDC 4817.111 21 357014 0.031
BDO 4017.900 8 227844 0.011
LUW 3917.663 10 182524 0.012
TKG 3914.000 6 211838 0.010
SRG 3799.190 14 397157 0.017
PLM 3641.000 8 233434 0.013
AMQ 2783.000 20 404130 0.027
BTH 2744.000 13 103172 0.025
DPS 2551.073 21 743957 0.030
BDJ 2528.404 11 256853 0.015
SOQ 2207.000 8 117417 0.012
BPN 2196.116 14 352191 0.019
KNO 1889.000 20 577702 0.045
KOE 1818.900 25 517787 0.033
NBX 1687.333 11 144379 0.014
SUB 1656.320 17 666804 0.020
PGK 1598.000 6 154225 0.012
PLW 1588.450 12 154939 0.016
TTE 1394.000 8 152890 0.012
43/81

DEGREE
UPG 5273.729 32 595125 0.043
KOE 1818.900 25 517787 0.033
MDC 4817.111 21 357014 0.031
DPS 2551.073 21 743957 0.030
AMQ 2783.000 20 404130 0.027
KNO 1889.000 20 577702 0.045
SUB 1656.320 17 666804 0.020
DJJ 1169.000 15 220642 0.013
SRG 3799.190 14 397157 0.017
BPN 2196.116 14 352191 0.019
BTH 2744.000 13 103172 0.025
PLW 1588.450 12 154939 0.016
BDJ 2528.404 11 256853 0.015
NBX 1687.333 11 144379 0.014
LUW 3917.663 10 182524 0.012
PNK 852.092 10 362008 0.015
FKQ 515.833 10 84025 0.012
LBJ 462.714 10 323190 0.013
TRK 460.333 10 81012 0.017
KNG 64.333 10 71114 0.014
44/81
Table 19 TOP 20 airports of 2016 network ranked by their weighed degree

DEGREE
DPS 2551.073 21 743957 0.030
SUB 1656.320 17 666804 0.020
UPG 5273.729 32 595125 0.043
KNO 1889.000 20 577702 0.045
KOE 1818.900 25 517787 0.033
LOP 789.955 9 486523 0.011
AMQ 2783.000 20 404130 0.027
SRG 3799.190 14 397157 0.017
PNK 852.092 10 362008 0.015
MDC 4817.111 21 357014 0.031
BPN 2196.116 14 352191 0.019
LBJ 462.714 10 323190 0.013
GNS 218.000 4 320763 0.009
JOG 306.514 4 291063 0.006
BDJ 2528.404 11 256853 0.015
KTG 4.000 6 252342 0.008
BMU 474.622 6 242649 0.008
PLM 3641.000 8 233434 0.013
BDO 4017.900 8 227844 0.011
DJJ 1169.000 15 220642 0.013
Able to maintain the leverage as the main bridge of the network, Sultan
Hasanuddin International Airport still manages to increase connection from the initial
very high degree to even higher, with 32 connections. Therefore, Sultan Hasanuddin
International Airport managed to gain the number 2 rank of the top airports based on
PageRank with 0.043 points, just 0.002 point differences with the number one,
Kualanamu International Airport. The respected airport manages to reach the top of
PageRank due to the very low connectivity of the neighbouring nodes, as the equation
45/81
of PageRank highly depended on the connectivity of the neighbouring nodes of the

calculated node, therefore gaining a high number of PageRank is easily achieved when
the neighbouring nodes have a very low connection. Kualanamu International Airport
neighbour has the maximum connections of 2 nodes (only 1 of 10 nodes, the rest have
1 connection).
How Kualanamu International Airport operates is aligned with the purpose of
ATR 72, which is to gain connectivity to remote airports. Therefore, even if the most
important airport is not Kualanamu International Airport, creating a small hub to reach
every small region in ATR 72 network analysis. As being said, the most important
airport in the ATR 72 network in 2016 is the Ngurah Rai International Airport, followed
by Sultan Hasanuddin International Airport.
46/81

DEGREE
KNO 1889.000 20 577702 0.045
UPG 3193.645 32 595125 0.043
KOE 1543.767 25 517787 0.033
MDC 1844.975 21 357014 0.031
DPS 4799.025 23 744064 0.030
AMQ 1878.745 20 404130 0.027
BTH 2744.000 13 103172 0.025
SUB 1365.915 17 666804 0.020
BPN 1660.534 14 352191 0.019
SRG 4354.836 14 397157 0.017
TRK 455.678 10 81012 0.017
PLW 1008.670 12 154939 0.016
BDJ 1743.760 11 256853 0.015
PNK 808.551 10 362008 0.015
KNG 2949.554 12 71221 0.014
NBX 1651.817 11 144379 0.014
ENE 238.474 9 194065 0.013
PLM 3641.000 8 233434 0.013
LBJ 535.383 10 323190 0.013
DJJ 1169.000 15 220642 0.013
47/81

In 2017, the network managed to grow even further; the 2017 network have
120 nodes and 382 edges. The network remains as one whole network with no
separation in any area. The increment of nodes and edges manage to diminish the
number of hops required to reach one end from another end of the network; it is now
reduced by 2 sets of hops. With the tremendous increase of new edges compared to
the increment of nodes, the average degree can climb over 3 points, which means
every node is connected on average to 3 or more other nodes. Although the points are
put into percentage over the total number of nodes in the network, the average
degree does not increase or decrease from 2016; it remains the same with 2.65%.
Which explains the graph density remains similar from the previous year with 0.027 to
0.026.

Diameter 13
Graph Density 0.026
The average shortest path went down by 0.722 points to 5.217, which means
the passengers from any node flying to any other nodes are, on average, needed to
hop 5 times to reach their final destination; the number is similar to 2014 metrics.
However, the 2014 network war-separated into 2 components, which means in 2017,
with the network growth, it is now able to maintain the initial branches reach. The
average weighted degree is increased by approximately 10% from the 2016 average
weighted degree. The increment of edges also affects the average clustering
coefficient bring it up to 0.243 points.
48/81
Table 22 until table 25 lists the highest-ranked airports in 2017 by their

betweenness, degree, weighted degree, and PageRank. Airports worth noting are
Sultan Hassanuddin International Airport, Ngurah Rai International Airport, and
Kualanamu International Airport. Sultan Hassanuddin International Airport managed to
secure three spots on the first rank based on the betweenness, connection, and
PageRank. Notice how Ngurah Rai International Airport drops down to rank 16 in the
betweenness table; the main reason for the drop is because of the break of connection
with Kaimana Airport, the main bridge to the east part of the network. The point of
betweenness of Ngurah Rai International Airport is just a third of the initial value,
whilst Sultan Hassanuddin International Airport went up by 137% with just one year
gap. The number of connections of Sultan Hassanuddin International Airport remains
almost the same with just one new connection. This means many nodes depend on
Sultan Hassanuddin International Airport to go across from the west side of the
network to the eastern part of the network.
49/81

DEGREE
UPG 7566.413 33 685985 0.041
TKG 4623.133 10 223008 0.011
AMQ 4408.683 32 417699 0.030
KDI 3453.017 8 123476 0.010
BDO 3165.592 14 267256 0.015
DJB 2795.000 14 183887 0.017
MLG 2683.494 6 59562 0.007
MDC 2620.300 25 350214 0.030
LUW 2613.671 8 187867 0.010
SOC 2088.473 10 113206 0.011
KNO 2038.833 24 706341 0.036
SUB 1820.933 16 667550 0.019
PLM 1741.500 17 285921 0.021
KOE 1693.285 23 725394 0.029
NBX 1667.483 10 230075 0.013
DPS 1584.922 17 763634 0.018
MOF 1584.868 6 119453 0.007
DJJ 1213.117 12 315259 0.014
PKY 977.561 10 109202 0.011
BPN 965.876 14 378234 0.016
50/81

DEGREE
UPG 7566.413 33 685985 0.041
AMQ 4408.683 32 417699 0.030
MDC 2620.300 25 350214 0.030
KNO 2038.833 24 706341 0.036
KOE 1693.285 23 725394 0.029
DPS 1584.922 17 763634 0.018
PLM 1741.500 17 285921 0.021
SUB 1820.933 16 667550 0.019
SOQ 448.483 16 157853 0.017
BPN 965.876 14 378234 0.016
SRG 869.007 14 370150 0.015
BDO 3165.592 14 267256 0.015
DJB 2795.000 14 183887 0.017
DJJ 1213.117 12 315259 0.014
BDJ 853.505 12 298262 0.013
PLW 805.842 12 195568 0.017
TRK 749.662 12 45440 0.016
BTH 328.167 11 122903 0.012
PNK 589.689 10 433435 0.013
NBX 1667.483 10 230075 0.013
There are no significant changes in the nodes' weighted degree; Ngurah Rai
International Airport remains the number one position with no changes from the
previous year. However, the changes are noticeable from number 2 until number 4; a
total of 25% increase occurred from El Tari Airport, Kualanamu International Airport,
and Sultan Hassanuddin International Airport. All three airports managed to overtake
Juanda International Airport as it was always in the top 10 in terms of the weighted
degree, even claiming the number 1 in 2013.
51/81

DEGREE
DPS 1584.922 17 763634 0.018
KOE 1693.285 23 725394 0.029
KNO 2038.833 24 706341 0.036
UPG 7566.413 33 685985 0.041
SUB 1820.933 16 667550 0.019
LOP 516.916 8 538348 0.010
PNK 589.689 10 433435 0.013
AMQ 4408.683 32 417699 0.030
BPN 965.876 14 378234 0.016
SRG 869.007 14 370150 0.015
LBJ 288.357 8 353710 0.010
MDC 2620.300 25 350214 0.030
KTG 3.000 6 324326 0.007
DJJ 1213.117 12 315259 0.014
GNS 9.000 4 309628 0.006
BDJ 853.505 12 298262 0.013
PLM 1741.500 17 285921 0.021
BMU 392.846 6 280571 0.007
JOG 7.333 4 267975 0.005
BDO 3165.592 14 267256 0.015
Even though Ngurah Rai International Airport claims the number one position
based on the weighted degree, Sultan Hassanuddin International Airport still manage
to claim not far off the number one with only around 80 thousand passenger
movement yearly. Therefore, it is distinguishable in which Sultan Hassanuddin
International Airport is the most important airport in 2017, not only one aspect it
superior in, but in four out of four.
52/81

DEGREE
UPG 7566.413 33 685985 0.041
KNO 2038.833 24 706341 0.036
MDC 2620.300 25 350214 0.030
AMQ 4408.683 32 417699 0.030
KOE 1693.285 23 725394 0.029
PLM 1741.500 17 285921 0.021
SUB 1820.933 16 667550 0.019
DPS 1584.922 17 763634 0.018
PLW 805.842 12 195568 0.017
SOQ 448.483 16 157853 0.017
DJB 2795.000 14 183887 0.017
TRK 749.662 12 45440 0.016
BPN 965.876 14 378234 0.016
BDO 3165.592 14 267256 0.015
SRG 869.007 14 370150 0.015
DJJ 1213.117 12 315259 0.014
BDJ 853.505 12 298262 0.013
PNK 589.689 10 433435 0.013
NBX 1667.483 10 230075 0.013
BTH 328.167 11 122903 0.012
53/81

In 2018, the network expands further, with addition of 9 new nodes and 42
new edges bringing the network nodes and edges to 129 and 424, respectively. The
network remains as one whole network with no separation on any area. However, the
increment does not affect the network diameter nor the average degree. Even though
the number of average degree is increasing, when compared to the total number of
nodes, the value has decreased by 0.11%. The decrement of average degree shown is
effect on graph density, the number of graph density decreased by 0.001.

Diameter 13
Graph Density 0.025
The increment of number of edges is profoundly affecting the average

shortest path, the number went down by 0.153 or 2.93% from last year point. The
average clustering coefficient decreased by 0.021, which means, the increment of
nodes does not trail with enough number of edges.
Table 27 until table 30 are the list of highest ranked airports in 2018 by their
betweenness, degree, weighted degree, and PageRank. There are 4 airports worth
noting, Sultan Hassanuddin International Airport, Ngurah Rai International Airport,
Pattimura Airport, El Tari Airport. Sultan Hassanuddin International Airport and El Tari
Airport placed first and second respectively on top 20 airports of 2018 ranked by their
betweenness, degree and PageRank.
54/81

DEGREE
UPG 7976.081 36 781694 0.035
AMQ 5329.943 34 438927 0.035
BUW 4200.291 6 265006 0.006
TKG 3726.932 13 236874 0.011
SOC 3659.785 12 72342 0.011
SUB 2882.823 23 721273 0.021
DJB 2699.639 18 229825 0.018
PKY 2595.049 10 165483 0.010
PNK 2421.014 10 618193 0.012
MDC 2037.718 18 361961 0.021
KNO 2033.465 23 652547 0.029
PGK 1917.881 8 171333 0.009
PLW 1843.271 18 204999 0.020
BPN 1810.398 18 506809 0.018
LBJ 1748.536 16 372001 0.015
NBX 1670.611 10 266133 0.010
KOE 1669.102 26 880133 0.029
BTH 1481.184 22 147364 0.024
LUW 1481.118 6 166553 0.007
DJJ 1392.075 13 470140 0.015
55/81

DEGREE
UPG 7976.081 36 781694 0.035
AMQ 5329.943 34 438927 0.035
KOE 1669.102 26 880133 0.029
SUB 2882.823 23 721273 0.021
KNO 2033.465 23 652547 0.029
BTH 1481.184 22 147364 0.024
SRG 1134.192 20 491701 0.018
BPN 1810.398 18 506809 0.018
PLM 1089.986 18 397692 0.019
MDC 2037.718 18 361961 0.021
DJB 2699.639 18 229825 0.018
PLW 1843.271 18 204999 0.020
DPS 625.610 16 725858 0.015
LBJ 1748.536 16 372001 0.015
LOP 1370.076 14 521481 0.014
DJJ 1392.075 13 470140 0.015
BDJ 1317.305 13 327097 0.013
TKG 3726.932 13 236874 0.011
PKU 128.792 13 156043 0.014
BDO 570.465 12 329311 0.011
56/81
El Tari Airport manage to place the highest in terms of their weighted degree,
increased by around 150 thousand passengers in one year. Sultan Hassanuddin
International Airport also manage to gain more passenger by around 100 thousand
passengers surpassed Ngurah Rai International Airport by 55 thousand passengers.

DEGREE
KOE 1669.102 26 880133 0.029
UPG 7976.081 36 781694 0.035
DPS 625.610 16 725858 0.015
SUB 2882.823 23 721273 0.021
KNO 2033.465 23 652547 0.029
PNK 2421.014 10 618193 0.012
LOP 1370.076 14 521481 0.014
BPN 1810.398 18 506809 0.018
SRG 1134.192 20 491701 0.018
DJJ 1392.075 13 470140 0.015
AMQ 5329.943 34 438927 0.035
KTG 209.968 7 409763 0.006
PLM 1089.986 18 397692 0.019
LBJ 1748.536 16 372001 0.015
MDC 2037.718 18 361961 0.021
BDO 570.465 12 329311 0.011
BDJ 1317.305 13 327097 0.013
GNS 7.500 4 288689 0.005
JOG 44.424 4 274931 0.004
NBX 1670.611 10 266133 0.010
57/81
UPG is still the most important node in the network, with huge betweenness,
very high connectivity, already able to overtake the number of passengers movement
of the centre of tourism in Indonesia, Ngurah Rai International Airport, and sharing the
top of PageRank with Pattimura Airport (AMQ).

DEGREE
UPG 7976.081 36 781694 0.035
AMQ 5329.943 34 438927 0.035
KNO 2033.465 23 652547 0.029
KOE 1669.102 26 880133 0.029
BTH 1481.184 22 147364 0.024
MDC 2037.718 18 361961 0.021
SUB 2882.823 23 721273 0.021
PLW 1843.271 18 204999 0.020
PLM 1089.986 18 397692 0.019
BPN 1810.398 18 506809 0.018
SRG 1134.192 20 491701 0.018
DJB 2699.639 18 229825 0.018
TTE 495.216 11 247900 0.017
LBJ 1748.536 16 372001 0.015
DPS 625.610 16 725858 0.015
DJJ 1392.075 13 470140 0.015
SOQ 298.750 12 198666 0.014
PKU 128.792 13 156043 0.014
LOP 1370.076 14 521481 0.014
KNG 477.883 11 66387 0.013
58/81

The number of nodes and edges continue to grow in 2019; the total number
of nodes and edges are now 132 and 464, respectively. Not a significant change from
the previous year, with around 10% increase of edges, the network diameter is able to
be shortened by 2 hops, leaving only 11 hops to travel from one end to the other. The
average degree is increased by a little less than 10%; however, when the average
degree is put into percentage over the total number of nodes in the network, the
increment is just 0.12% from the previous year. As well as the graph density, it does
not increase as much to 2.7% of the complete graph with full edges.

Diameter 11
Graph Density 0.027
The number of hops required to travel to any destination is smaller than the
previous year, with only 4.844 hops required. However, with the increase of nodes and
edges, most of the new routes do not carry a lot of passengers, which lead to a
decrease in the average weighted degree. The cause of the drop also corresponds to
the drop of the passenger movement from every existing node pre-2019. The total
drop is quite significant, it is almost a 20% drop from the previous year. As well as the
average clustering coefficient decreased to 0.196.
Table 32 until Table 35 contain the list of the top 20 airports in 2019 based on
the betweenness, connections, weighted degree, and PageRank. The airports worth
being observed are Sultan Hassanuddin International Airport, Ngurah Rai International
Airport, El Tari Airport, and Kualanamu International Airport. Maintaining a very high
59/81
betweenness and connections are still Sultan Hassanuddin International Airport with
no other node with a close point to the respected airport.

DEGREE
UPG 8852.575 38 700277 0.027
SUB 5963.055 27 578715 0.019
KOE 1451.932 26 720192 0.027
PLM 723.081 26 348040 0.015
BDO 6343.586 24 485596 0.018
BPN 1824.620 24 458957 0.011
AMQ 6207.205 23 349224 0.025
KNO 2147.143 22 432411 0.032
BTH 1312.782 22 135570 0.019
DPS 1205.377 21 484370 0.019
DJB 4028.148 20 208778 0.013
PLW 1083.061 20 185693 0.013
MDC 1537.968 18 301188 0.019
SRG 1290.450 17 356881 0.013
LBJ 3812.002 16 300168 0.014
BDJ 811.314 16 254676 0.010
TKG 492.666 14 200562 0.011
AAP 489.145 14 81669 0.006
PNK 608.675 13 403942 0.012
DJJ 1229.967 12 530019 0.016
60/81

DEGREE
UPG 8852.575 38 700277 0.027
SUB 5963.055 27 578715 0.019
KOE 1451.932 26 720192 0.027
PLM 723.081 26 348040 0.015
BDO 6343.586 24 485596 0.018
BPN 1824.620 24 458957 0.011
AMQ 6207.205 23 349224 0.025
KNO 2147.143 22 432411 0.032
BTH 1312.782 22 135570 0.019
DPS 1205.377 21 484370 0.019
DJB 4028.148 20 208778 0.013
PLW 1083.061 20 185693 0.013
MDC 1537.968 18 301188 0.019
SRG 1290.450 17 356881 0.013
LBJ 3812.002 16 300168 0.014
BDJ 811.314 16 254676 0.010
TKG 492.666 14 200562 0.011
AAP 489.145 14 81669 0.006
PNK 608.675 13 403942 0.012
DJJ 1229.967 12 530019 0.016
From Table 34, every airport has a drop in passenger movement; however,
the most noticeable drop is on Ngurah Rai International Airport. The airport lost over
220 thousand passengers in 2019 compared to 2018. Even with the all-around
dropping of passenger movement, El Tari Airport had dropped by around 160
thousand passengers still managed to maintain the number one position.
As previously explained, Kualanamu International Airport can gain a high point
in PageRank due to the low connection of the neighbouring nodes; therefore,
Kualanamu International Airport still managed to increase their point to 0.032, an
61/81
increment of 0.003 from the previous year. This concludes that the most important
airport in the 2019 ATR 72 network is Sultan Hassanuddin International Airport, with
all-around number 1 and number 2 positions in all four metrics.

DEGREE
KOE 1451.932 26 720192 0.027
UPG 8852.575 38 700277 0.027
SUB 5963.055 27 578715 0.019
DJJ 1229.967 12 530019 0.016
BDO 6343.586 24 485596 0.018
DPS 1205.377 21 484370 0.019
BPN 1824.620 24 458957 0.011
KNO 2147.143 22 432411 0.032
PNK 608.675 13 403942 0.012
LOP 670.185 12 385004 0.011
SRG 1290.450 17 356881 0.013
AMQ 6207.205 23 349224 0.025
PLM 723.081 26 348040 0.015
MDC 1537.968 18 301188 0.019
LBJ 3812.002 16 300168 0.014
WMX 14.267 6 293533 0.009
KTG 32.286 6 292683 0.005
BDJ 811.314 16 254676 0.010
SOQ 865.000 12 230581 0.014
TTE 608.032 12 211579 0.015
62/81

DEGREE
KNO 2147.143 22 432411 0.032
UPG 8852.575 38 700277 0.027
KOE 1451.932 26 720192 0.027
AMQ 6207.205 23 349224 0.025
DPS 1205.377 21 484370 0.019
BTH 1312.782 22 135570 0.019
MDC 1537.968 18 301188 0.019
SUB 5963.055 27 578715 0.019
BDO 6343.586 24 485596 0.018
DJJ 1229.967 12 530019 0.016
PLM 723.081 26 348040 0.015
TTE 608.032 12 211579 0.015
NBX 2121.067 12 186177 0.014
SOQ 865.000 12 230581 0.014
LBJ 3812.002 16 300168 0.014
DJB 4028.148 20 208778 0.013
PLW 1083.061 20 185693 0.013
SRG 1290.450 17 356881 0.013
PNK 608.675 13 403942 0.012
BPN 1824.620 24 458957 0.011
63/81
4.2. Network Growth Analysis

To optimise the current network, two types of analysis is required after
knowing the characteristics of the network growth; the first analysis is to analyse the
community structure growth, second, to overview the network metrics. The first
analysis is to determine which region should be focused on to optimise. Therefore,
fixating the focal point to improve will give a more explicit step to act on. The second
analysis is to gain information on which metrics is to concentrate on, which metrics will
give a better way to optimise the network with the limited information available.
4.2.1. Community Structure Growth

The community structure growth is essentially an analysis of how
communities in a network behave over time by naturally increasing or
decreasing the number of nodes, edges, connectivity, and edges' weight. All of
the variables will determine each node's similarity, which can then be
controlled on how the group is made by varying the resolution to gain how
similar the communities to be. Using a low resolution or high modularity, have
a thicker connection within the nearby nodes and a sparse connection with the
node belonging to another community. Although it is not always best to group
the nodes with high modularity, it depends on what to do with the community
afterwards.
In 2013, the community structure was grouped into 7 modules, as
shown in Figure 10, where each of the modules has its own hub which reaches
out to all around the nodes within the group, it is quite often to see more
than one hub in the same community, this is depended on how the resolution
is set at the beginning of the calculation. In the 2013 purple module, two hubs
are shown, Supadio International Airport and Sultan Aji Muhammad Sulaiman
Sepinggan International Airport. Both airports are located in Kalimantan
Island, on the west and east sides of the island. However, they are grouped
into the same module because of the connection between their neighbouring
nodes. Nonetheless, the corresponding airports are connected to one
64/81
another, strengthening the bond between the airports and this bond is valued
greatly to the grouping process. It is a different case on the dark green
community, which contain one large hub and one smaller hub, Pattimura
Airport and Sentani Airport (DJJ), respectively. Both airports are not
connected directly with one another. However, Sentani Airport is not big
enough to be a sole hub in its own community in the set resolution. Hence,
the community are made broader and more sparse than other modules in the
network.
Figure 10 Community structure of ATR 72 network in 2013
In 2014, there were no significant changes in the community besides

the alteration of some nodes into another group due to new connections to
nearby existing or new nodes, as shown in Figure 11. The community is still
grouped into 7 modules in which two of the communities stand on their own
without any bridges to other communities, which is the light green
community and pink community. However, it is worth noting that General
Ahmad Yani International Airport (SRG), Husein Sastranegara Airport and El
Tari Airport are getting bigger, which potentially be an individual hub with
their own community.
65/81
In 2015, the alteration of community structure is now more apparent

and noticeable, as shown in Figure 12. The communities are still grouped into
7 modules. However, the previously orange community in 2014 is now part of
the light blue community; this is potentially due to the stronger bond
between the bridges and can be considered part of the community itself. The
community alternation is also found in the previously purple community in
2014, which is now separated into two modules, yellow and light brown
communities. The alterations are occurred due to the two large hub (General
Ahmad Yani International Airport and Husein Sastranegara Airport), which is
now able to stand on their own, and the connection between the separated
communities are now too sparse to be in the same characteristic or set of
rules between the nodes in the community.
66/81
In 2016, the community was grouped into 7 modules, as shown in

Figure 13, just about the same as the previous community. Although, the
previously orange community in 2015 are now separated into two part and
then merged with another stronger community. The west part of the
Kalimantan Island community is merged to the west part of Javanese Island,
and the eastern part of Kalimantan Island is merged to the South Sulawesi
community. The separation is due to the nonexistence of a vital hub in the
Kalimantan community, which can hold the community together. Therefore,
Javanese community manage to snatch several Kalimantan airports to be part
of the community, the separation and merge between three community are
caused also by the growth of passenger movement between two main hub of
Javanese community. General Ahmad Yani International Airport (SRG) and
Juanda Airport (SUB) to the west side of Kalimantan airports, Iskandar Airport
(PKN), Supadio International Airport (PNK) and Iskandar Airport (PKN). The
North Sulawesi small community is separated from the previous 2015 light
blue community and making their own community in the North Sulawesi
region due to lack of bond between the airports in the North Sulawesi airport
with the East side of Indonesia.
67/81
In 2017, the number of communities was reduced to 6 modules; as

shown in Figure 14, the previous two separated communities in Sumatera
Island are now merged into one large community with one large hub and one
smaller hub, which holds the community sparse. The merging of Sumatera
community is supported by the immense growth of new edges between the
previously green and yellow community in 2016, hence the community grow
stronger and manage to create a big community. The other alteration in the
community structure is between the previously light blue community and the
west part of the orange community in 2016. The west part of the previously
68/81
sparse orange community in 2016 is separated and merged with the light blue
community forming another sparse community, which is now a green
community. The merge is caused to several new edges between the region,
even though these new edges are not a major alteration to the network, it
somehow creates a similarity between the nodes therefore, however weak
the bond, it is still considered as a singular community with the same
behaviour.
In 2018, there are 2 newly formed communities due to significant
alterations, as shown in Figure 15, which were formed between the previously
merged communities in 2017. The first alteration is found in the previously
orange community; the northern part of the community is separated and
form a tiny community. This small community might happen as well, even in a
high-resolution community structure; if the bond of the nodes is not strong
enough to band together with other communities, they are forced to form a
small community with a weak bond. The second alteration found on the newly
merged green community in 2017 is now separated into two new
communities of yellow and light green. The separation and merging of
communities on the eastern part of the network is most likely to result from
the community’s lack of a large hub. Therefore, whenever there is a new
random and robust connection, the group might be separated and merged
69/81
into another group.

In 2019, the communities remained the same with 8 modules. However,
the communities have 2 alterations as well as shown in Figure 16; the first
alteration is on the Javanese Island, the previously orange community has
merged with the purple community; however, the west part of the purple
community got separated with Kualanamu International Airport as the central
hub of the tiny new brown community. The second alteration is between the
previously dark green and yellow community; the east part of the dark green
community is separated from the rest of the original community and merged
to the previously yellow community. These alterations are also caused to the
nonexistence of a vital hub to hold the community together and several new
edges between the West of Javanese airports with the South of Sumatera
airports. With a stronger bond between the region and, the community
characteristics shifted and leaving the North of Sumatera airports which has a
weak bond with the rest of Sumatera airports. The brown community only has
6 edges with the purple community compared to the West of Javanese
airports bond with the South of Sumatera airports.
70/81
To conclude, the best way to create a community is to have a strong

node as the central hub; the community should be in the same region or area
(as Indonesia is an archipelagic country). Optimisation of the network can be
started by solving the separation and merging problem that was occurring on
the eastern part of the network. Therefore, creating a new hub that will
connect the neighbouring communities is the main priority, as shown in Figure
17.
Figure 17 Community structure focus area to be optimized
71/81
4.2.2. Network Overview Analysis

Every network can be optimised by shortening the number of hops required
to reach the passenger's destination and the network's diameter, increasing the
average degree, average weighted degree, graph density, and average clustering
coefficient. However, each of the metrics is connected to the others differently for any
network. Therefore, finding which metric affects the other the most will be the best
way to optimise the network. One of the ways to find the most prominent metric is by
evaluating the network goal, which is the most appropriate method in ATR 72
Network. In the ATR 72 Network, the one metric which has the maximum effect on the
network will be average coefficient clustering followed by graph density.
ATR 72 Network has an important goal: to gain connectivity to reach the most
remote area. Therefore, communities grouping is essential; enhancing the bond
between the nodes in the group is the priority. Hence, the average coefficient
clustering will pilot how the other metrics move. After clearing out how the
communities can be optimised, the whole network will gain a positive impact;
naturally, the same number of nodes in the network with an increasing number of
edges will increase every other metrics value, which will improve the network.
Therefore, with 6 metrices in the network, each of the metrices will play the
role of:
1. Increasing the Average Degree will optimize the network, however,
does not play the critical role of optimization the network.
2. Increasing the Average Weighted Degree will optimize the network,
however, to accurately predict this metric requires more detailed
passenger flight’s information.
3. Shortening the Diameter is not the primary objective as it is in
contradiction with the purpose of ATR 72.
4. Shortening the Average Shortest Path is not the primary objective as it
is in contradiction with the purpose of ATR 72.
5. The Graph Density will the second primary focused metric for
optimizing the network.
72/81
6. The Average Clustering Coefficient will be the primarily focused metric

for optimizing the network.
4.3. Network Optimization

Network optimisation is performed by using the result from network growth
analysis. The community structure growth indicates the eastern area of the network is
the optimum area to be optimised by focusing on increasing the average clustering
coefficient according to the network overview analysis. Figure 17 shows the two
communities of orange and dark green has the potential to be merged. Focusing on
these two communities, both have similar characteristics: very low graph density,
which can be solved by merging these two groups by creating a new vital hub that
intertwines with both communities. There are two suitable nodes to be turned into the
hub: Moses Kilangin Airport (TIM) and Nabire Airport (NBX).
4.3.1. TIM New Route Evaluation

Moses Kilangin Airport has a solid base for potentially being a hub; the airport
itself is positioned right in between the two communities. Therefore, branching out to
reach every node in the community will be easier than nodes positioned on the edges
of the community. The airport itself has quite a big apron compared to several of its
neighbour nodes, which can hold more aircraft in its vicinity.
Moses Kilangin Airport initially only has 3 connections: Dumatubun Airport
(LUV), Wamena Airport (WMX), and Nabire Airport. In this simulation, another 7 edges
to nearby nodes will be imported to the network connecting Moses Kilangin Airport
with:
1. Nop Goliat Dekai Airport (DEX)
2. Oksibil Airport (OKL)
3. Tanah Merah Airport (TMH)
73/81
4. Mopah International Airport (MKQ)

5. Sentani Airport (DJJ)
6. Dobo Airport (DOB)
7. Sudjarwo Tjondronegoro Airport (ZRI)
The total nodes connected to Moses Kilangin Airport are 10 nodes which are
shown in Figure 18.
Figure 18 Moses Kilangin Airport connected nodes after simulation
With only 7 additional edges to the network, the overall network metrics have
improved significantly. Of overall 464 original edges, adding 1.5% new edges has
improved 0.025 points of average clustering coefficient, which is a 12.76% increase.
The overall network metrics have improved, and the community itself, the original
orange and dark green communities, have merged into one large community with a
strong hub in Moses Kilangin Airport.
74/81
Table 36 ATR 72 NETWORK METRICES AFTER MOSES KILANGIN AIRPORT SIMULATION

Diameter 11
Graph Density 0.028
4.3.2. NBX New Route Evaluation

The other airport that can be a strong hub that will connect both communities
of orange and dark green is Nabire Airport, which is located nearby from Moses
Kilangin Airport, right in the between of the two potentially merged communities. The
airport has an even bigger apron than Moses Kilangin Airport, whereas, in Moses
Kilangin Airport, the apron has around 10,000m2, and in Nabire Airport, the apron is
around twice as large. Therefore, Nabire Airport can have more aircraft even though
both airports have only 1 runway.
Nabire Airport already has 6 connections to the nearby airports, which
are to Pattimura Airport, Frans Kaisepo Airport (BIK), Sentani Airport, Moses Kilangin
Airport, Domine Edward Osok Airport (SOQ), and Rendani Airport (MKW). In this
simulation, another 10 edges to nearby nodes will be imported to the network
connecting Nabire Airport with:
1. Wamena Airport (WMX)
2. Nop Goliat Dekai Airport (DEX)
3. Oksibil Airport (OKL)
4. Dumatubun Airport (LUV)
5. Tanah Merah Airport (TMH)
6. Mopah International Airport (MKQ)
75/81
7. Dobo Airport (DOB)

8. Sudjarwo Tjondronegoro Airport (ZRI)
9. Babo Airport (BXB)
10. Kaimana Airport (KNG)
The total number of nodes connected to Nabire Airport is 16 nodes which are
shown in Figure 19.
Figure 19 Nabire Airport connected nodes after simulation
With 10 additional edges to the network, the overall network metrics have
improved significantly. Of overall 464 original edges, adding 2.16% new edges has
improved 0.055 points of average clustering coefficient, which is a 28.06% increase.
The overall network metrics have improved, and the community itself, the original
orange and dark green communities, have merged into one large community with a
strong hub in Nabire Airport.
76/81

Diameter 11
Graph Density 0.028
77/81
CHAPTER 5
SUMMARY AND RECOMMENDATION
5.1. Summary
The Indonesian air transportation network is analysed in this research,
specified only to flights operated with ATR 72. The purpose is to understand the ATR
72 structural network in Indonesia and modelling the optimized version of the
network. To achieve satisfactory result, finding the network characteristics and
determine which region is eligible to be the suitable area for optimisation.
The Indonesian ATR 72 network has small world network properties, and the
airports connecting the networks follow the power-law degree distribution or scale-
free model. Based on the network modelling and characterising results from 2013 until
2019, the most important airport is concluded to be Sultan Hassanuddin International
Airport. In 2019, passengers were expected to travel 5 connecting flights to reach their
final destination.
The eastern part of the network is the suitable region to be optimised due to
the lack of a strong bond between the two groups and between the nodes inside the
group. The metric to be focused on optimising the network is also found from
analysing the metrics' growth and evaluating the aircraft ATR 72 characteristics.
Therefore, the focused metric is the average clustering coefficient.
After the two modelling experiments of Moses Kilangin Airport and Nabire
Airport are expected to enhance the average clustering coefficient exceptionally by
only adding a minimal number of edges to the network. However, one exceeds more
than the other, the optimization version on Nabire Airport, with only 10 new edges to
the network, manages to increase 28.06% from the initial value of the network average
78/81
clustering coefficient. The previously two troubled groups are now bonded with a
strong hub connecting the whole network together.
5.2. Recommendation
1. A better and detailed analysis can be achieved by adding another data of the
source and destination of the passengers' tickets are brought. Therefore, the
network can be modelled and analysed more thoroughly.
2. Resiliency analysis can be performed to find the network robustness under
topology change or bad weather, which lead to airport closed.
3. Time analysis can be performed to find the time interval when the airports are
performed with maximum load. The interruptions will create a cascade impact
on the neighbouring airport. Therefore, mitigating any interruption within the
time interval is crucial. Re-routing the flights to the neighbouring airport in
case of a particular airport must be closed for any reason.
79/81
REFERENCES
Albert, R., & Barabási, A.-L. (2001). Statistical Mechanics of Complex Networks.
Bagler, G. (2007). Analysis of the Airport Network of India as a complex weighted
network.
Barabási, A.-L., & Albert, R. (1995). Emergence of Scaling in Random Networks. In Mat.
Res. Soc. Symp. Proc (Vol. 74). www.sciencemag.org
Blondel, V. D., Guillaume, J.-L., Lambiotte, R., & Lefebvre, E. (2008). Fast unfolding of
communities in large networks. https://doi.org/10.1088/1742-
5468/2008/10/P10008
Couto, G. S., da Silva, A. P. C., Ruiz, L. B., & Benevenuto, F. (2015). Structural properties
of the Brazilian air transportation network. Anais Da Academia Brasileira de
Ciencias, 87(3), 1653–1674. https://doi.org/10.1590/0001-3765201520140155
Dias, M. de O., & Lopes, R. de O. A. (2020). Case on Leading Commercial Aircraft in the
Brazilian Domestic Aviation Market. International Journal of Academic Research in
Accounting, Finance and Management Sciences, 10(1).
https://doi.org/10.6007/IJARAFMS/v10-i1/7064
Guida, M., & Maria, F. (2007). Topology of the Italian airport network: A scale-free
small-world network with a fractal structure? Chaos, Solitons and Fractals, 31(3),
527–536. https://doi.org/10.1016/j.chaos.2006.02.007
Guimerà, R., Mossa, S., Turtschi, A., Amaral, L. A. N., & Wachter, K. W. (2005). The
worldwide air transportation network: Anomalous centrality, community
structure, and cities’ global roles. www.pnas.orgcgidoi10.1073pnas.0407994102
Moody, J. (2001). Race, School Integration, and Friendship Segregation in America.
American Journal of Sociology, 107(3). https://doi.org/10.1086/338954
Watts, D. J., & Strogatz, S. H. (1998). Collective dynamics of ‘small-world’ networks.
Nature, 393(6684). https://doi.org/10.1038/30918
Technopedia. (2022). Network Optimization. Retrieved from Technopedia:
https://www.techopedia.com/definition/31667/network-optimization
80/81
Wilson, M. (2020, 11 19). Network Optimization – A Quick Guide to Get You Started!
Retrieved from PC & NETWORK DOWNLOADS:
https://www.pcwdld.com/network-optimization-quick-guide#wbounce-modal
81/81

Network Evaluation and Optimisation For ATR 72 in Indonesia

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Network Evaluation and Optimisation For ATR 72 in Indonesia

Uploaded by

Copyright:

Available Formats

INTERNATIONAL UNIVERSITY LIAISON INDONESIA

NETWORK EVALUATION AND OPTIMISATION FOR ATR 72 IN INDONESIA

BSD City 15345

NETWORK EVALUATION AND OPTIMISATION FOR ATR 72 IN INDONESIA

Ir. Ananta Widjaja M.Sc _______________

Dipl.-Ing. Sentot Wahjoe Goeritno, M.Si 17 / 02 / 2022

EXAMINERS APPROVAL PAGE

Dr. Ir. Prianggada Indra Tanaya, M.M.E _______________

Neno Ruseno M.Sc 17/02/2022

STATEMENT BY THE AUTHOR

Peter Oliver ___________________

APPROVAL PAGE ________________________________________________ i

EXAMINERS APPROVAL PAGE ______________________________________ii

STATEMENT BY THE AUTHOR _____________________________________ iii

COPY RIGHTPAGE ________________________________________________v

ACKNOWLEDGMENTS ___________________________________________ vii

TABLE OF CONTENTS ___________________________________________ viii

LIST OF FIGURES ________________________________________________ xi

LIST OF TABLES _________________________________________________ xii

LITERATURE REVIEW ________________________________________________ 5

RESULTS AND DISCUSSION __________________________________________ 20

SUMMARY AND RECOMMENDATION__________________________________ 78

TABLE 1 ATR 72 NETWORK METRICES IN 2013 ................................................................................................... 21

1.1. General Statement of Problem Area

Indonesian usage of ATR 72 had an immense growth from first introduced in

Network analysis is required to undertake the further problem that might

1.2. Research Purpose

1.3. Research Problem

1.4. Significance of Study

1.5. Theoretical Perspective

Calculating the network metrics, connected components, average

1.6. Research Question

Understanding the complex network behaviour is required to achieve the

2.1. Network Optimization

2.2. Air Transportation Network Around the World

2.2.1. Worldwide Air Transportation Network

2.2.2. Brazilian Air Transportation Network

2.2.3. Indian Air Transportation Network

2.2.4. Italian Air Transportation Network

2.3. Small World Networks

1. In an unsophisticated network order, every node is connected with

2.4. Scale-Free Model

3.1. Data Collection

3.2. Network Modelling

3.3. Network Characterise

3.3.1. In-degree and Out-degree

< 𝑘 > (𝒢) = ∑ 𝑘(𝑣)/|𝑉|.

In the air transportation network, flight from Husein Sastranegara Airport

3.3.2. Average Neighbourhood Overlap

Therefore, the average neighbourhood overlap < 𝑛o> is calculated as:

3.3.3. Average Shortest Path

3.3.5. Clustering Coefficient

The average clustering coefficient is simply calculated as:

Figure 3 Clustering coefficient of graph G is 0.75

3.3.7. Closeness Centrality

𝛾(𝑣) = [ ∑ 𝑙(𝑣, 𝑘)] .

3.3.9. Graph Density

3.4. 2013 -2019 Data Comparison

3.4.1. Community Structure

𝐴ij = Edge weight between node 𝑖 and 𝑗

In the Gephi software, the resolution to be manipulated is to assess the sensitivity or

Figure 4 Resolution in Gephi modularity setting

3.5. Analyze The Plausible Improvement