Professional Documents
Culture Documents
Network Evaluation and Optimisation For ATR 72 in Indonesia
Network Evaluation and Optimisation For ATR 72 in Indonesia
BACHELOR’S THESIS
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
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
i
NETWORK EVALUATION AND OPTIMISATION FOR ATR 72 IN INDONESIA
ii
NETWORK EVALUATION AND OPTIMISATION FOR ATR 72 IN INDONESIA
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.
iii
NETWORK EVALUATION AND OPTIMISATION FOR ATR 72 IN INDONESIA
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
NETWORK EVALUATION AND OPTIMISATION FOR ATR 72 IN INDONESIA
COPY RIGHTPAGE
© Copyright 2021
by Peter Oliver
International University Liaison Indonesia
All rights reserved
v
NETWORK EVALUATION AND OPTIMISATION FOR ATR 72 IN INDONESIA
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
NETWORK EVALUATION AND OPTIMISATION FOR ATR 72 IN INDONESIA
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
NETWORK EVALUATION AND OPTIMISATION FOR ATR 72 IN INDONESIA
TABLE OF CONTENTS
ABSTRACT _____________________________________________________ iv
DEDICATION ___________________________________________________ vi
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
viii
NETWORK EVALUATION AND OPTIMISATION FOR ATR 72 IN INDONESIA
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
CHAPTER 5 ___________________________________________________ 78
ix
NETWORK EVALUATION AND OPTIMISATION FOR ATR 72 IN INDONESIA
REFERENCES __________________________________________________ 80
x
NETWORK EVALUATION AND OPTIMISATION FOR ATR 72 IN INDONESIA
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 6 ATR 72 NETWORK AROUND INDONESIA IN 2014 ..................................................................................... 28
FIGURE 7 DEGREE COUNT GRAPH OF 2014 NETWORK. ........................................................................................... 29
FIGURE 8 ATR 72 NETWORK AROUND INDONESIA IN 2015 ..................................................................................... 36
FIGURE 9 ATR 72 NETWORK AROUND INDONESIA IN 2016 ..................................................................................... 41
FIGURE 10 COMMUNITY STRUCTURE OF ATR 72 NETWORK IN 2013 ........................................................................ 65
FIGURE 11 COMMUNITY STRUCTURE OF ATR 72 NETWORK IN 2014 ........................................................................ 66
FIGURE 12 COMMUNITY STRUCTURE OF ATR 72 NETWORK IN 2015 ........................................................................ 67
FIGURE 14 COMMUNITY STRUCTURE OF ATR 72 NETWORK IN 2017 ........................................................................ 68
FIGURE 13 COMMUNITY STRUCTURE OF ATR 72 NETWORK IN 2016 ............................ ERROR! BOOKMARK NOT DEFINED.
FIGURE 15 COMMUNITY STRUCTURE OF ATR 72 NETWORK IN 2018 ........................................................................ 69
FIGURE 16 COMMUNITY STRUCTURE OF ATR 72 NETWORK IN 2019 ........................................................................ 71
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
NETWORK EVALUATION AND OPTIMISATION FOR ATR 72 IN INDONESIA
LIST OF TABLES
xii
NETWORK EVALUATION AND OPTIMISATION FOR ATR 72 IN INDONESIA
TABLE 34 TOP 20 AIRPORTS OF 2019 NETWORK RANKED BY THEIR WEIGHTED DEGREE ................................................ 62
TABLE 35 TOP 20 AIRPORTS OF 2019 NETWORK RANKED BY THEIR PAGERANK .......................................................... 63
TABLE 36 ATR 72 NETWORK METRICES AFTER MOSES KILANGIN AIRPORT SIMULATION ............................ 75
xiii
NETWORK EVALUATION AND OPTIMISATION FOR ATR 72 IN INDONESIA
CHAPTER 1
INTRODUCTION
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.
1/81
NETWORK EVALUATION AND OPTIMISATION FOR ATR 72 IN INDONESIA
optimization for ATR 72 in Indonesian structural network are the main objectives of
this thesis research.
2/81
NETWORK EVALUATION AND OPTIMISATION FOR ATR 72 IN INDONESIA
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).
3/81
NETWORK EVALUATION AND OPTIMISATION FOR ATR 72 IN INDONESIA
4/81
NETWORK EVALUATION AND OPTIMISATION FOR ATR 72 IN INDONESIA
CHAPTER 2
LITERATURE REVIEW
5/81
NETWORK EVALUATION AND OPTIMISATION FOR ATR 72 IN INDONESIA
6/81
NETWORK EVALUATION AND OPTIMISATION FOR ATR 72 IN INDONESIA
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.
7/81
NETWORK EVALUATION AND OPTIMISATION FOR ATR 72 IN INDONESIA
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.
8/81
NETWORK EVALUATION AND OPTIMISATION FOR ATR 72 IN INDONESIA
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
NETWORK EVALUATION AND OPTIMISATION FOR ATR 72 IN INDONESIA
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.
10/81
NETWORK EVALUATION AND OPTIMISATION FOR ATR 72 IN INDONESIA
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 EVALUATION AND OPTIMISATION FOR ATR 72 IN INDONESIA
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.
12/81
NETWORK EVALUATION AND OPTIMISATION FOR ATR 72 IN INDONESIA
(𝑁(𝑢) ∩ 𝑁(𝑣))
𝑛o(𝑢, 𝑣) = .
(𝑁(𝑢) ∪ 𝑁(𝑣))
Equation 2
Equation 3
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
NETWORK EVALUATION AND OPTIMISATION FOR ATR 72 IN INDONESIA
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
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
NETWORK EVALUATION AND OPTIMISATION FOR ATR 72 IN INDONESIA
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
15/81
NETWORK EVALUATION AND OPTIMISATION FOR ATR 72 IN INDONESIA
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
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
NETWORK EVALUATION AND OPTIMISATION FOR ATR 72 IN INDONESIA
Equation 12
17/81
NETWORK EVALUATION AND OPTIMISATION FOR ATR 72 IN INDONESIA
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.
18/81
NETWORK EVALUATION AND OPTIMISATION FOR ATR 72 IN INDONESIA
19/81
NETWORK EVALUATION AND OPTIMISATION FOR ATR 72 IN INDONESIA
CHAPTER 4
RESULTS AND DISCUSSION
20/81
NETWORK EVALUATION AND OPTIMISATION FOR ATR 72 IN INDONESIA
Diameter 15
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
NETWORK EVALUATION AND OPTIMISATION FOR ATR 72 IN INDONESIA
number 2 position on the top 2 airports, after Juanda International Airport (SUB) with
807 thousand passengers in a year.
22/81
NETWORK EVALUATION AND OPTIMISATION FOR ATR 72 IN INDONESIA
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
NETWORK EVALUATION AND OPTIMISATION FOR ATR 72 IN INDONESIA
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
NETWORK EVALUATION AND OPTIMISATION FOR ATR 72 IN INDONESIA
25/81
NETWORK EVALUATION AND OPTIMISATION FOR ATR 72 IN INDONESIA
26/81
NETWORK EVALUATION AND OPTIMISATION FOR ATR 72 IN INDONESIA
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.
Diameter 12
27/81
NETWORK EVALUATION AND OPTIMISATION FOR ATR 72 IN INDONESIA
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
NETWORK EVALUATION AND OPTIMISATION FOR ATR 72 IN INDONESIA
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
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
NETWORK EVALUATION AND OPTIMISATION FOR ATR 72 IN INDONESIA
30/81
NETWORK EVALUATION AND OPTIMISATION FOR ATR 72 IN INDONESIA
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
NETWORK EVALUATION AND OPTIMISATION FOR ATR 72 IN INDONESIA
32/81
NETWORK EVALUATION AND OPTIMISATION FOR ATR 72 IN INDONESIA
33/81
NETWORK EVALUATION AND OPTIMISATION FOR ATR 72 IN INDONESIA
34/81
NETWORK EVALUATION AND OPTIMISATION FOR ATR 72 IN INDONESIA
Diameter 12
35/81
NETWORK EVALUATION AND OPTIMISATION FOR ATR 72 IN INDONESIA
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
NETWORK EVALUATION AND OPTIMISATION FOR ATR 72 IN INDONESIA
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.
37/81
NETWORK EVALUATION AND OPTIMISATION FOR ATR 72 IN INDONESIA
38/81
NETWORK EVALUATION AND OPTIMISATION FOR ATR 72 IN INDONESIA
39/81
NETWORK EVALUATION AND OPTIMISATION FOR ATR 72 IN INDONESIA
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.
Table 15 TOP 20 airports of 2015 network ranked by their PageRank
40/81
NETWORK EVALUATION AND OPTIMISATION FOR ATR 72 IN INDONESIA
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
NETWORK EVALUATION AND OPTIMISATION FOR ATR 72 IN INDONESIA
Diameter 20
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
NETWORK EVALUATION AND OPTIMISATION FOR ATR 72 IN INDONESIA
43/81
NETWORK EVALUATION AND OPTIMISATION FOR ATR 72 IN INDONESIA
44/81
NETWORK EVALUATION AND OPTIMISATION FOR ATR 72 IN INDONESIA
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
NETWORK EVALUATION AND OPTIMISATION FOR ATR 72 IN INDONESIA
46/81
NETWORK EVALUATION AND OPTIMISATION FOR ATR 72 IN INDONESIA
47/81
NETWORK EVALUATION AND OPTIMISATION FOR ATR 72 IN INDONESIA
Diameter 13
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
NETWORK EVALUATION AND OPTIMISATION FOR ATR 72 IN INDONESIA
49/81
NETWORK EVALUATION AND OPTIMISATION FOR ATR 72 IN INDONESIA
50/81
NETWORK EVALUATION AND OPTIMISATION FOR ATR 72 IN INDONESIA
51/81
NETWORK EVALUATION AND OPTIMISATION FOR ATR 72 IN INDONESIA
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
NETWORK EVALUATION AND OPTIMISATION FOR ATR 72 IN INDONESIA
53/81
NETWORK EVALUATION AND OPTIMISATION FOR ATR 72 IN INDONESIA
Diameter 13
54/81
NETWORK EVALUATION AND OPTIMISATION FOR ATR 72 IN INDONESIA
55/81
NETWORK EVALUATION AND OPTIMISATION FOR ATR 72 IN INDONESIA
56/81
NETWORK EVALUATION AND OPTIMISATION FOR ATR 72 IN INDONESIA
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.
Table 29 TOP 20 airports of 2018 network ranked by their weighted degree
57/81
NETWORK EVALUATION AND OPTIMISATION FOR ATR 72 IN INDONESIA
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).
Table 30 TOP 20 airports of 2018 network ranked by their PageRank
58/81
NETWORK EVALUATION AND OPTIMISATION FOR ATR 72 IN INDONESIA
Diameter 11
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
NETWORK EVALUATION AND OPTIMISATION FOR ATR 72 IN INDONESIA
betweenness and connections are still Sultan Hassanuddin International Airport with
no other node with a close point to the respected airport.
Table 32 TOP 20 airports of 2019 network ranked by their betweenness
60/81
NETWORK EVALUATION AND OPTIMISATION FOR ATR 72 IN INDONESIA
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
NETWORK EVALUATION AND OPTIMISATION FOR ATR 72 IN INDONESIA
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.
Table 34 TOP 20 airports of 2019 network ranked by their weighted degree
62/81
NETWORK EVALUATION AND OPTIMISATION FOR ATR 72 IN INDONESIA
63/81
NETWORK EVALUATION AND OPTIMISATION FOR ATR 72 IN INDONESIA
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.
65/81
NETWORK EVALUATION AND OPTIMISATION FOR ATR 72 IN INDONESIA
66/81
NETWORK EVALUATION AND OPTIMISATION FOR ATR 72 IN INDONESIA
67/81
NETWORK EVALUATION AND OPTIMISATION FOR ATR 72 IN INDONESIA
68/81
NETWORK EVALUATION AND OPTIMISATION FOR ATR 72 IN INDONESIA
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
NETWORK EVALUATION AND OPTIMISATION FOR ATR 72 IN INDONESIA
70/81
NETWORK EVALUATION AND OPTIMISATION FOR ATR 72 IN INDONESIA
71/81
NETWORK EVALUATION AND OPTIMISATION FOR ATR 72 IN INDONESIA
72/81
NETWORK EVALUATION AND OPTIMISATION FOR ATR 72 IN INDONESIA
73/81
NETWORK EVALUATION AND OPTIMISATION FOR ATR 72 IN INDONESIA
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
NETWORK EVALUATION AND OPTIMISATION FOR ATR 72 IN INDONESIA
Diameter 11
75/81
NETWORK EVALUATION AND OPTIMISATION FOR ATR 72 IN INDONESIA
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
NETWORK EVALUATION AND OPTIMISATION FOR ATR 72 IN INDONESIA
Diameter 11
77/81
NETWORK EVALUATION AND OPTIMISATION FOR ATR 72 IN INDONESIA
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
NETWORK EVALUATION AND OPTIMISATION FOR ATR 72 IN INDONESIA
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
NETWORK EVALUATION AND OPTIMISATION FOR ATR 72 IN INDONESIA
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
NETWORK EVALUATION AND OPTIMISATION FOR ATR 72 IN INDONESIA
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