Abstract

The air travel industry which is a highly regulated industry collects a lot of data. For the

airline industry, delays and cancellations can result in additional costs to the airlines. With the

use of data through better visualisation, these data can be used to discover key insights and

present information in ways that are easier to be understood by the users. In this paper,

various visualisation idioms were taken into consideration in analysing the delay and

cancellations of the flights in the United States of America in 2015. Each of these

visualisation idioms designed using Tableau in a structured manner based on a structured

visualisation framework from data and task abstraction to the final visualisation using

appropriate marks and channels. These visualisation idioms serve their purpose with

respective strengths and weaknesses. Future works have also been identified to further

leverage the power of visualisation from greater benefits among the users. In conclusion,

insights derived from these visualisations will enable its user to gain better insights into the

delay and cancellations of the air travel industry.

VISUALIZATION OF FLIGHT DELAYS AND CANCELLATION

Introduction

Data visualization is representing any kind of data or information, it can be in many

forms available, such as graph, bar, or any visualization format available. Minard’s Map,

Playfair’s map, Beck’s map, and Snow’s map are some of the examples of famous

visualization forms. Visualization is also known as understanding from diagrams (Munzner,

2014). It is to form a mental model for humans since it is easier for humans to understand

data by looking at its visualization. However, according to Munzner, computer-based

visualization systems provide a visual representation of datasets designed to help people carry

out tasks more effectively (Munzner, 2014). Visualization can be defined as a relationship

between the data and images (Brehmer, 2016). It is crucial to have visualization upon big

data because the problem arises for humans to interpret larger trends by not looking at

pictures, that is how visualization will help.

There are two classical domains of visualization, known as scientific visualization and

information visualization (Brehmer, 2016). Scientific visualization is, using physical data that

are spatial and quantitative, it is usually used in communicating, searching, exploring, and

sense-making (Torsten, n.d.). Meanwhile, for information visualization, it used abstract data

that s non-spatial and qualitative. It is also used in searching, exploring, and sense-making,

however, information visualization emphasis more on interaction with and interconnections

among data.

Data visualization has become so important. A bigger dataset now can be analyzed

and visualized. Graphic or visualization form that comes into action is becoming handy in

extracting information from the dataset (Torsten, n.d.). Not to forget that visualization can

testify much more effectively than words or numbers (Munzner, 2014). Data visualization
VISUALIZATION OF FLIGHT DELAYS AND CANCELLATION

makes copious amounts of numbers and statistics accessible to both business holders and

their audience (Nguyen & Tran, 2017). The same things applied to flight delays and

cancellations of major airlines in the united states in 2015. Visualization can help to

distinguish which flight is always cancelled or delays and find out the reason behind it. Hence

decision-making can be made to improve their services.

The paper is structured as follows. Section 2 explains the domain background. Section

3 describes data and task abstraction, Section 4 provides an overview of related work, Section

5 discusses the details of methods and tools used. Section 6 is the analysis of the project;

Section 7 is the discussion and future work on this project and Section 7 concludes.

Domain Background

Air travel has become a very competitive industry over the past years before the

pandemic brought the industry to a halt. In these highly regulated and competitive industries,

airlines compete with each other not only on pricing but the quality of their services including

minimizing delays. Delays and cancellations can be very costly to each airline as the

additional cost would be incurred. These costs can be in terms of quantitative and qualitative.

Costs which can be quantified include compensation payment to customers, higher airport

fee, extra allowances for pilots and cabin crews. Costs that are qualitative include loss of

customers for future bookings, aircraft downtime, and the opportunity cost of time and

productivity for the travelers.

Therefore, the airline industry must keep the delays and cancellations to a minimum.

Certainly, some cancellations are not avoidable due to poor weather conditions which are

beyond the control of the airlines. However, airlines have control over the flight schedule for

them to strategize the best solution such as minimizing travel in areas prone to storm at a
VISUALIZATION OF FLIGHT DELAYS AND CANCELLATION

certain time. With changing situations both on the ground and in the air, airlines need to react

quickly to minimize delay while always maintaining the safety of their crew and passengers

at all times.

With the advancements of big data analytics, the task becomes easier as millions of

rows of data can be analyzed instantly for quick decision making. With better visualisation, a

domain expert can extract trends and insights which may not be visible in the standard reports

currently used by the airlines. These insights can be used by airlines to promote their

competitive advantage but also allow their management to understand the root cause of

problems and identify solutions to fix them.

On-Time Performance (OTP) is an indicator that measures the punctuality of

transportation. In the aviation industry, an airline departure or arrival which is considered on

time has a departure or arrival that occurs within 15 minutes of the scheduled time (OAG,

n.d.).

Data and Task Abstraction

Data Abstraction

Data abstraction is about the description of the data (Brehmer, 2016). It is also a

simplified representation of the whole dataset by the reduction of a particular body of data. In

other words, data abstraction is a translation from a specific domain to the vocabulary of

visualization. Visualization design and evaluation involve mapping between domain

problems or interests and appropriate visual encoding and interaction design choices

(Munzner, 2014). There are different types of data collection methods available, such as
VISUALIZATION OF FLIGHT DELAYS AND CANCELLATION

sensors, logs, experiments, human-generated data, surveys, and many more. In this paper, we

obtained our dataset from Kaggle. It was generated by the US Bureau of Transportation

Statistics. It covers all the records of flight delays of major airlines in the United States

throughout the year 2015. The type of data used in this dataset is number and it is in a table.

Not to forget that it is a semantics dataset which means, it is a real-world meaning data. Data

types used in this dataset are varied, there are categorical like origin city name, quantitative

like distance, and we also have geography, interval(location), and time. The dataset type

used is in tables. Each data item in a new row and each column contains attributes. Attributes

are something that can be measured, observed, or logged. The attributes dataset semantics are

continuous and discrete. The table below is an example of attributes in this dataset with its

description. There are 323 rows in this dataset. In this dataset, 5 main pieces of information

were collected, which were the time, data related to time, such as year, months, day of the

months, day of the week, and the flight date. The second main information is the airline. The

information that was being collected from this is the carrier code, the tail number, and the

flight number. The next information is the origin. This is about location. The airport ID, the

state, the city name, and the world area code for the origin. Then, destination. This is most

likely the destination airport, the identification number assigned by US DOT to identify a

unique airport, the destination city name, state, and others that related to the destination.

There are more, such as departure performance, arrival performance, cancellations and

diversions, flight summaries, cause of delay, gate return to the airport, and diverted airport

information.

Table 1 The Attributes of the Dataset With its Description

ATTRIBUTES TYPES DESCRIPTION
VISUALIZATION OF FLIGHT DELAYS AND CANCELLATION

Categorical Categorical variables represent types of data that may be

divided into groups such as city names.
Quantitative Quantitative data is the type of data whose value is measured in
the form of numbers or counts such as distance.
Geography Geographic data is the series of information related to the
location such as city, state, country, and continent.
Interval Interval is a type of data that is measured along a scale
Time Data that is related to time.

Task Abstraction

Abstraction is an effective way to hide the details of the subject under study, hence it

provides a good view of a domain as it translates domain-specific situation into an abstract

visualization task. Task abstraction allows us to understand why data is visualized. The first

step of task abstraction is to collect the user’s task which is also known as questions to be

addressed. In this paper, the questions to be addressed by the visualizations were in reference

to the reference paper which focuses on the delay and cancellation of the flights from each

airline available. Hence, in domain language, the goals are to compare the airlines in terms of

their consistency to be on time and to discover which airline has the least delay in the year

2015 to be presented to the public. Additionally, we would like to discover more the reason

or situations that cause the flights delays and cancellation.

In terms of Munzer’s task abstraction, this is categorized as their Discover under

Analyze action. There are two choices under analyze, which are consume and produce.

Under consume category, two activities we were involved in were discover and present. The

goal of discover is to generate or confirming a new hypothesis or to unearth hidden patterns

that we will discover from this study. Present on the hand is to communicate the known facts

and explain the known phenomena discovered. In our case, we use the visualizations to
VISUALIZATION OF FLIGHT DELAYS AND CANCELLATION

discover pattern of flights delays and cancellation in addition to other discoveries, as well as

to present them. Next is the lower level actions of search which goal is to find for elements of

interest within the visualization idiom, (Guo, Pang, Han, & Shan, 2012). In early stage of

discovery mode, not knowing the location within the idiom but knowing the target, our search

is known as a locate. There are three high-level targets, which are the trends, outliers, and

features. Trends is the one who shows patterns wither increases, decreases, peaks, and

troughs & plateaus. The attributes are specific properties that are visually encoded. The

available techniques that fulfill these criteria are discussed in the next sections.

Related Work

Based on our reading, two ways of airline performance comparison is either

visualizing a single derived value that represents multiple attributes or multiple visualizations

for multiple attributes and tasks.

A study by Keerthana, Viswa, and Deivarani is on the presentation of the airlines with

the most consistent service to the public in terms of having the least delay or cancellation (K,

TN, & Deivarani, 2019). The method used by these authors was through deriving single value

result using Analysis of Variance (ANOVA) of multiple selected attributes (Table 2) from a

dataset prepared by the US Department of Transportation’s Bureau of Transportation

Statistics. In ANOVA, the variance between groups was calculated for the purpose of

accepting or rejecting the hypotheses made. The null hypothesis assumed the flight delays

and cancellations are least occurring and the alternative hypothesis was that flight delays and

cancellations occur frequently. There was a single visualization to represent the ANOVA

result as shown in Figure 1, which was a bar chart with the airlines' names positioned on the
VISUALIZATION OF FLIGHT DELAYS AND CANCELLATION

horizontal axis and ANOVA F-values on the vertical axis. In the chart, the length of the bar

corresponds to the F value of each airline which if it is lower than the F critical value, the

airline is said to have the least flight delays and cancellation (null hypothesis accepted), on

the other hand, any value higher than F-critical value means the airlines are not consistently

good in term of flight delays and cancellation (reject the null hypothesis). F-critical value is

not directly shown on the visualization used but to be extracted from the paper. The use of

ANOVA as an airline benchmark can clearly show whether the individual airline consistently

good in managing delay or not, but the comparison between airlines is more ambiguous

should the user figure out the next best airline if the best airline is not available, or which

airline to avoid. Nevertheless, it provides a single value result which may be a simpler and

easier result to remember.

Table 2 Selected Attributes for ANOVA

Selected Attributes
Year Airline
Flight number Tail Number
Origin Airport Destination Airport
Departure Delay Elapsed Time
Scheduled Time Air Time
Distance Arrival Delay
Diverted Flight
VISUALIZATION OF FLIGHT DELAYS AND CANCELLATION

Figure 1 ANOVA F-values for each airline

On the other hand, a paper by Xi, Xiaoyu, and Miaoqiong uses a different approach to

address the same problem of presenting airline performance. Instead of relying on a single

derived value to represent airlines’ consistency, they used multiple visualizations to capture

different angles of contributing factors to the performance of an airline (Xi, Xiaoyu, &

Miaoqiong). Through multiple visualizations accessed from a single dashboard, the

visualizations users can explore in detail each attribute that affects the airline’s consistency in

managing flight delays or cancellations. Examples of idioms used by the authors are shown in

Table 3. Comparing the two approaches, visualization using ANOVA result allows the user

to get a direct answer of which airline consistently in a good performance but they would not

know in detail reasons for the stellar performance should they enquire so. On the other hand,

multiple visualizations allow the user to find out more about the factors that contribute

towards airline performance in managing flight delays and cancellations which they can

apply suited to different flight situations, but more information to be remembered may lead to

a cognitive burden.
VISUALIZATION OF FLIGHT DELAYS AND CANCELLATION

Table 3 Multiple Vis Idioms Used to Represent Different Factors Contribution to Airline Performance
Visualization Idiom Description
Choropleth To show air traveler’s density
map by country

Node-link To visualize the average

diagram arrival delay of each route for
the selected departure and
arrival airport.
Represent centrality measure

Bar chart To visualize overall delay by

airline
VISUALIZATION OF FLIGHT DELAYS AND CANCELLATION

Radar chart To visualize multiple delay

reasons and time by airline

Bubble chart Visualization of an overview

of all cancelled airlines

Methods and Tools

Several aspects were considered in choosing the visualization idioms for this project.

Firstly, we designed the task from the dataset we have. The 2015 statistics of the major

airlines in the United States gave us an idea to work on presenting the delay information. We

chose to focus our objective on the flight delays and cancellations because we feel like in

other airlines (not just in the United States), one of the satisfactory factors is the timeliness of

the flight departure and arrival. No one wants to be bothered with having one’s flight delayed

or cancelled (which only happens when there is a serious emergency). This is because it will

interfere with one’s planning. Thus, having a flight delayed is considered a pretty big issue

for travelers.
VISUALIZATION OF FLIGHT DELAYS AND CANCELLATION

Besides, the inspiration behind the visualization idioms that we are presenting stems

from the desire to present information that will benefit users from various backgrounds. We

want to design an efficient visualization idiom that would present information that is cohesive

and understandable. To achieve this, we will be using Tableau, a visualization software.

Tableau Software as Visualization Tool

Tableau is used for this project to develop visualization idioms that will fulfill the

objective as mentioned previously. Tableau is a visualization tool that is user friendly and

with minimal learning period. Tableau has built-in visuals that aid in flattening everyone’s

learning curve and there is a limitless exploration of data provided which will not interrupt

the analysis flow (Tableau, n.d.). Although Tableau is a paid software, they offer free trials

for up to 14 days. However, we were able to get to explore Tableau for up to a year when

signing up using our students’ account. Tableau is also deployable on a cloud service so this

makes it easy for people who have limited space in their local machine.

Having an efficient visualization idiom requires some preparation for the data before

proceeding with the visualization itself. Based on the objective that we want to achieve from

this project, there are a few steps that we did as follows.

Data Collection

This dataset was downloaded from the Kaggle website whereby the user who

uploaded the dataset have obtained it from the United States Department of Transportation
VISUALIZATION OF FLIGHT DELAYS AND CANCELLATION

(DOT) Bureau of Transportation Statistics. The link to this dataset is as shown

“[Link]

Data Cleaning

An important step before proceeding is the data cleaning process. The data cleaning is

a crucial step so we can prepare the data to avoid having biased results, infer wrong insights,

and conclude false findings. In this dataset, our main focus would be on the delay and time

attributes as these are what we want to visualize and improve upon.

Missing values

We first checked for any presence of missing values. There are a large number of

missing values present in the dataset. As we investigated further, these missing values mostly

imply that there were no delays and/or cancellation/diverted flights at that time. As most

person who was in charge for the records would only record for the occurrence of delay by

assigning the amount of time and using codes to imply if cancellation happens and note the

reason for it. Despite the presence of null values in this dataset, we decided not to do much

with it. Initially, we planned to replace these null values with ‘0’. However, after further

inspection, there are some columns with the value ‘0’. As we want to avoid misunderstanding

further, thus, we decided to keep the null values as they are while keeping in mind that the

null values represent the circumstances that did not happen.

Replacing values

We also replaced some values in the dataset. For the ‘CANCELLATION_REASON’

attribute, the raw data was filled with codes of ‘A’, ‘B’, ‘C’, and ‘D’. For visualization

purposes and even for further analysis, these codes will make it difficult for us as we have no
VISUALIZATION OF FLIGHT DELAYS AND CANCELLATION

idea what the codes mean. Thus, these codes are replaced with their respective reason as

shown in the table below (bold and italicized).

Table 4 Codes of the Cancellation Reason of the Flight

Code Reason
A Air Carrier. Delays related to the aircraft such as baggage loading and fuelling.
Extreme Weather. The condition of the weather prohibits the flight to be on
B
schedule.
National Aviation System (NAS). Delays related to the aviation system and
C
system control such as air traffic control.
D Security. Delays caused by security such as immigration issues at the checkpoint.

Merge some attributes and create a new column

In the raw dataset, there is no date. Instead, the date was divided by day, month, and

year, each in its column. This would make it harder for us to study the timeline as this creates

more variables and it will look messy if we attempt to do the visualization. One of our

proposed visualization idioms is studying the travel trend on monthly basis. Merging these

three attributes as a single date would assist the job in producing neater visualization which is

in line with our objective, to create efficient and understandable visualizations that cater to

various users.

Binning values into two labels

Our main focus is the arrival and departure delay, which is not shown (dataset) in a

way that we would like it to be. The ‘ARRIVAL_DELAY’ and ‘DEPARTURE_DELAY’

attributes contain values in minutes, whereby the records were in terms of how late and early

the flights are. If the flight departs or arrives late, the value would be positive. However, if

the flight departs or arrives earlier, the value is negative which indicates that the flight is

ahead of time.
VISUALIZATION OF FLIGHT DELAYS AND CANCELLATION

Data Preparation

In this step, each of us with our assigned visualization idioms would be preparing the

data that would fit the visualization we are assigned to. Since we are working with different

visualizations, thus, the data was not ‘cleaned’ much as the preparation to create the

visualization would depend on the type of visualizations each of us is working on.

Table 5 Preparation of Data for Visualization Implementation on Tableau

Visualization Idiom Data Prepared
Treemap Sum of cancelled flights
Average of:
Heatmap  Arrival delay in minutes
 Departure delay in minutes
In term of percentage of:
 The number of flights delayed
divided by the total flights offered
Bar Chart by the individual airline.
 The number of flights cancelled
divided by the total flights offered
by the individual airline.
In term of percentage of:
 Numbers of flights delayed divided
Choropleth Map by total flights out from/into a state.
 Numbers of flights cancelled divided
by total flights out from/into a state
Average of:
 Arrival delay in minutes
Scatterplots
 Departure delay in minutes
By type or airline and tail number
Average arrival delay and average departure
Line chart
delay in each month of 2015

Visualization Implementation on Tableau

The implementation of Tableau to create the visualizations we proposed took a little

time for us to get used to, mainly in terms of the interface. As Tableau is very user-friendly,

not much time was spent on learning how to utilize it. The ‘drag-and-drop’ feature helps us to

select which data we want to work on for columns and rows. There is also a ‘Show Me’
VISUALIZATION OF FLIGHT DELAYS AND CANCELLATION

feature which showcases different visualization idioms for us to work with such as a table,

bar chart, line chart, heatmap, scatter plot, box-and-whiskers plot, and many more. Tableau

also highlights some options of the visualization idioms that would work for the data we

selected.

The ‘Marks’ area is an area where we can customize marks and channels of our

visualization idioms. In a way, this area is where we can enhance the expressiveness of our

visualizations. There are many colour options that we can choose from simply by ‘Edit

Colour’ and Tableau even suggests a colour map that is suitable for our visualizations by the

option ‘Automatic’. This makes it very convenient for us, and other users as well. We can

also add some details to our visualizations as well so we can make the visualizations more

informative.

Result and Analysis

The results and analysis for each of the visualization that we have created will be

explained more thoroughly in terms of our domain problem and the reflection on the ways

these visualizations help us to understand it. We will also discuss further the different

approaches that we took when creating them by stating the strengths and weaknesses.

1. Stacked Bar Chart: Flight Delay for Each Airlines by Minutes

VISUALIZATION OF FLIGHT DELAYS AND CANCELLATION

Figure 2 Stacked Bar Chart of reasons for flight delayed, in minutes.

Stacked bar chart is the extension of bar chart from extending one categorical value

by two. In each bar in stacked bar chart, there are 3 information can be evaluated. In this case,

in a single stacked bar chart, we can see minute delay for each flight and the cause of the

delay. Orange colours represents air system delay. Pink is late aircraft delay, turquoise

represents weather delay and green is security delay.

Air systems delay usually related to the status within the airline’s control. Late aircraft

delay is because of the late arrival of the earlier flight that usually used the same plane will be

leaving next. It is the same as domino affect where the late, causing the passengers at the

other airport to departure late as well. In several situations weather conditions were severe,

either it was forecasted or presented. For instance, lightning storms, blizzards, lightning

storms and tornados. Lastly is because of security delay. Out of all, security delay is one of

the most critical delay types. It is the result of terminal or evacuation.

VISUALIZATION OF FLIGHT DELAYS AND CANCELLATION

From the visualization of the cause of flight delay for each airline by minutes, can be

seen that WN shows the highest delay in minutes compared to the other airlines. When it was

investigated, WN offers the most flight of the year. That is why the value is high compared to

the other airlines. The least delay airlines show by VN with less than 100k minutes of delay

in a year. There is a huge gap of delay between VN, NK, HA, and F9 with, WN, DL and EV.

These two groups offer huge different numbers of flight in 2015. That is how the other group

show so much less in delay compared to the other group. This might be because of the price.

VN, NK, HA and F9 might be a flight for business class and WN, DL and EV for economic

class.

There are major disadvantages of this stacked bar chart which are, it is harder to read.

The more segments for each bar have, it will be more difficult to read the result. This is

because they are not aligned on the base line. But, by using stacked bar chart, from this

visualization formed, we can see the relationship of the small categories with the main idea of

the visualization.

2. Bar Chart: Visualization of Flight Delays and Cancellation Ratio of each Airline

A user's questions on which airline has the least amount, or the most amount of delay

or cancellation (task abstraction, compare in a low level query) can be answered through a

bar chart which offers simple and understandable visualization. In the bar chart shown in

Figure 3, three flight categories are represented by three different color marks; orange mark

indicates fights that are delayed before departure, blue indicates flight arrival delay and red

represents flight cancellation. One of the strengths in visualization through bar charts is often

intuitive with no further explanation needed. In this case, the length of the bar corresponds to

the ratio of flights delayed or cancelled against the total number of flights offered by each
VISUALIZATION OF FLIGHT DELAYS AND CANCELLATION

airline. The rationale to derive ratio value is due to some airlines might have a higher count

number of delayed or cancelled flights than others but only so because they offer more

flights, this can be seen similarly from the visualization of stacked bar chart that showed

Southwest Airline has a significantly long time of flights delayed but only because it has the

most flights. Hence, the ratio between flights delayed or cancelled against total flights offered

per each airline gives a fair view to the users of how much the flights and the airlines offered

are actually delayed or cancelled. In visualization, however, the ratio is converted to a

percentage by multiplying it to 100 for even easier information to be grasped by the user as

percentage value is more commonly used in daily life.

A user might be more interested to find the extreme value first i.e., least delayed

airline or most delayed airline before finding information on other less extreme value

information which can be addressed by the bar chart. By rearranging the chart from the

minimum value number to the maximum, the user can quickly extract that information as it

manipulates the highly effective position mark. Comparing one bar to another bar that is not

directly next to it is also easy. However, it can be seen in the idiom that the sorting of airlines

is not separated whereby sorting the airline from the least departure flight delay to the highest

does not correspond to arrival flight delay or flight cancellation which means the user has to

reiteratively sort multiple times for each event to be viewed (Appendix Figure 1). In this data

for both departure and arrival delay, the airline with the least delayed flights is Hawaiian

Airlines (HA) and the opposite end is Spirit Airlines (NK). For flight cancellations, Hawaiian

Airlines again has the least flights cancelled, while Envoy Air (MQ) has the highest flight

cancellation percentage.
VISUALIZATION OF FLIGHT DELAYS AND CANCELLATION

The vis shown represents the data for the whole 2015 year which arguably should

convey enough information to the user but should they inquire or discover more from the vis,

they can choose to use the date slider at the bottom of the vis to discover the pattern of flight

delays and cancellation given the vis is available to the user in dynamic form. Another

alternative for this is to animate the static vis with the changing date to show the delay or

cancellation pattern. Additionally, the user can view more detail information by hovering the

mouse on each of the bars. Just a slight issue with the vis is when the user hovers the mouse

to view more information, it shown the attribute with the name that may not be the most

obvious information to the user. To justify this, the naming convention is named such a way

for ease of the vis designer to separate between tens of attributes with similar names but for

different task purposes. This is one of the limitations of Tableau where the alias name cannot

be given for all instances.

VISUALIZATION OF FLIGHT DELAYS AND CANCELLATION

Figure 3 Flight Delays and Cancellation Ratio for each Airline, color mark separated the categories.

3. Choropleth Map: Visualization of Flights Cancellation and Delays, by state

Choropleth map can address the question whether a neighbouring state can have

different numbers of flights delays or cancellation, or to confirm hypothesis such that a

bigger state would have more flight delays and cancellation. The downside of choropleth

however, the layering of visualization is quite limited as it can interfere with the information

shown. Additionally, for some states, they are quite small on the screen and easy to be

ignored which is not the case for other idioms, such as bar chart.
VISUALIZATION OF FLIGHT DELAYS AND CANCELLATION

In this visualization, we visualized the percentage of flight delays and cancellation

against the total number of flights leaving/landing from/into a state. The data represented in

choropleth map can be easily visualized with other idioms such as bar chart. However, the

effectiveness of choropleth map lies within the ability to actually connecting the information

visualized (percentage of flights delayed or cancelled) with the spatial position of the data.

This way, any trend can be easily observed without relying on mental model of which states

are neighbors to a state, which can be useful for those who are not familiar with spatial

relation of the data. For choropleth map, we visualized three separate maps; departure flight

delay, arrival flight delay, and flights cancellation. All three maps are juxtaposed in Tableau

Dashboard so the user can easily compare between the flight situations as shown in Figure 4.

Date slider is provided to enable the user to select any particular date of interest for further

discovery, such as holiday season flight trend in each state.

Based on the total flights number for the whole 2015-year, one interesting discovery

is generally the states in north are less in term of percentage to have departure delay as

compared to the states in the south. For arrival delay however, its distribution among states

generally does not seems in favour to any direction as one state in the north can have 15%

arrival flights delay while neighbouring states having 19% flights arrival delay. Nearing the

holiday season, again the states in the north seems to be having lower percentage of departure

flight delay than states in south, east coast, and west coast. Unfortunately, in arrival flight

delay no clump of states are likely to have less delay. For flight cancellation during the

holiday season, an interesting observation is that more flights are cancelled in certain region

as shown in the Figure 5, suggesting an unusual phenomenon may occur that causes this

situation. The curiosity can be addressed if the user navigates to the cancellation reason vis

and selecting the date interested.

VISUALIZATION OF FLIGHT DELAYS AND CANCELLATION

From the vis, choropleth can help a passenger to make informed decision before

booking any flights. For example, should an international passenger to have connecting

flights through the US, he may be better to land in the southern states if the flight he took

before the connecting flight is less reliable since flight departing from the southern state

usually depart later than its scheduled time.

Figure 4 Flight Delays and Cancellation, by states

Figure 5 Flight Delays and Cancellation approaching holiday season

VISUALIZATION OF FLIGHT DELAYS AND CANCELLATION

4. Scatterplots – Average Departure Delay vs Average Arrival Delay for each airline

Scatterplot idiom is used to explore for trends or outliers between two quantitative

value attributes. The average departure delay is plotted against the average arrival delay for

each airline using Tableau as shown in Figure 6. Both average departure delay and average

arrival delay are quantitative value variables. The mark type for the scatterplots is a point.

The average departure delay and average arrival delay is encoded using the vertical and

horizontal position channel while the type of airline is encoded using colour as the channel.

Figure 6 Scatterplot of average arrival delay vs departure delay

Analysing the above scatterplots, we can see that the average departure delay and the

average arrival delay are positively correlated. Based on the above, one airline AS has an

average arrival delay of -0.98 minutes which means that it arrival is ahead of schedule by

0.98 minutes despite its average departure delay is 1.79 minutes.

VISUALIZATION OF FLIGHT DELAYS AND CANCELLATION

Figure 7 Scatterplot with superimposed regression line

A regression line is then superimposed on the scatterplots using the Trend Line in

Tableau as shown in Figure 7. Nevertheless, there a number of airlines which are quite far

off from the regression line such asx DL, UA, F9 and NK.

4.1 Scatterplots – Average Departure Delay vs Average Arrival Delay for each tail number

(unique aircraft)

Tail number is like the car registration number which are unique to each aircraft.
VISUALIZATION OF FLIGHT DELAYS AND CANCELLATION

Figure 8 Scatter plot of average delays (arrival, departure) for each aircraft

In these scatterplots, we explore what is how is relationship between the average

departure delay and average arrival delay for each unique aircraft. This scatterplot mark is

point while use the vertical and horizontal position channel to indicate the average departure

delay and average arrival delay. Meanwhile, the shape channel is used to for the different tail

number and the colour channel is used for the type of airlines.

Based on the scatterplot shown in Figure 8, we can see that most aircraft identified by

the unique tail numbers are mostly within 100 minutes of the average departure delay and

average arrival delay. There are some outliers whereby a number of aircrafts have

significantly longer delay time as compared to the rest with over 160 minutes of average

departure delay and average arrival delay such as the N79011, N128UA, N843MH, N121UA

and N37018. Most of these aircraft are from the UA airlines as identified in red. This may

indicate that these aircrafts especially for UA airlines should be avoided as they are very

likely to cause delay in their flights. Data on the age of the aircraft can be useful for further

analysis.
VISUALIZATION OF FLIGHT DELAYS AND CANCELLATION

Due to many tail numbers, the shapes cannot really distinguish different tail numbers.

Evidently both N128UA and N121UA seemed to have the same “+” shape. With additional

data such as the age of the aircraft, we can further filter the aircraft into a smaller number for

better analysis. With reduced number, we would have sufficient shapes to distinguish

different types of tail number. Alternatively, we can conduct this analysis based on each

airline. Therefore, the number of tail number would be reduced and more distinguishable.

4.2 Scatterplot - Average departure delay and average arrival delay against the average
distance for each airline.

Figure 9 Scatter plot of distance against delays of all airlines

Vertical and horizontal spatial position channel indicate the average departure delay

or average arrival delay and the average distance. The mark is point on the scatterplot. The

channel colour is used to identify the airlines.

There is no relationship of the average departure delay and average arrival delay with

the average distance. Even when superimposing a regression line, there is a weak positive

correlation between average departure delay and average distance while there is a weak

negative correlation between average arrival delay and average distance for each airline.
VISUALIZATION OF FLIGHT DELAYS AND CANCELLATION

5. Line Chart – Average Departure Delay and Average Arrival Delay against the Time in
months

Figure 10 Line Chart of average delay by months

The line chart above uses line as the mark. The vertical and horizontal position is the

channel in this line chart. In each line chart, there is on quantitative value attribute namely the

average departure delay or the average arrival delay and one ordered key attribute which is

the date in months.

The line chart shows a trend of the average departure delay and average arrival delay

in 2015. It can be observed that the average departure delay and average arrival delay are the

highest in June 2015.

The average departure delay is the lowest in September 2015 at 4.824 minutes while the

average arrival delay is the lowest in October 2015 at -0.78 minutes. Negative average arrival

delay means that on average the airlines were 0.78 ahead of the schedule.
VISUALIZATION OF FLIGHT DELAYS AND CANCELLATION

6. Treemap: A Visualization Idiom of Cancellation Reasons of Each Airline

A treemap is a form of visualization that shows the hierarchical structure with a

containment, meaning that there is no node-link, but rather the information is enclosed within

a space, usually in a mixture of squares and rectangles. It consists of one quantitative attribute

at the leaf nodes. A treemap is encoded by the containment feature of an area mark. The size

of the squares or rectangles represents the quantitative attribute. A colour channel also

enhances this idiom further so users can distinguish the information categorically.

Different Approaches in Creating a Treemap Visualization

For this project, we took into account the frequency of cancellation and the respective

cancellation reason as we wanted to see which reason affects the decision to cancel the flights

more. At first, we only wanted to look at the overall cancellation reasons while disregarding

categorical attributes such as an airline. However, due to the huge difference between the

most occurring cancellation reason (extreme weather at 48,851 counts) with the least

occurring cancellation reason (security at 22 counts), this idea was discarded. The figure for

this visualization is shown in the Appendix Figure 2. Clearly, ‘Security’ which consists of a

small value is not visible at all – unless the cursor hovers. Thus, this will make it difficult to

convey the information that we want users to see.

Then, we included the airline to further segregate the cancellation reasons by the type

of airline the cancellation happened. However, the outcome was still problematic, as shown

in Figure 3 in Appendix. The huge difference between the counts of extreme weather and

security cancellation reasons is most likely the culprit for this. The security count is so less

that it is not visible at all as well (see Figure 4 in Appendix).

VISUALIZATION OF FLIGHT DELAYS AND CANCELLATION

Ultimately, we changed the leaf nodes of the treemap from airlines to cancellation

reasons. This change became more presentable as we can see the separation of the

cancellation reasons for each airline rather than seeing the separation of each of the

cancellation reason by airlines. Figure 11 shows the completed treemap idiom along with its

legend.

Figure 11 Treemap Visualization of Cancellation Reasons for Each Airline

Note. The legend shows how the airlines are categorized using colours and the mark area is

determined by the size of rectangles by taking the sum of each cancellation reason that

occurred in the airline respectively.

Analysis of The Treemap Visualization

Users can easily get the idea which reason affected the decision to cancel the flight

the most – which is extreme weather. This is understandable because airlines want to avoid

any unwanted events due to meteorological consequences. The next reason which affects the

cancellation decision is also straightforward from the vis idiom, which is air carrier. A
VISUALIZATION OF FLIGHT DELAYS AND CANCELLATION

cancellation caused by an air carrier is probably an unavoidable circumstance of which the

aircraft is possibly not in a suitable condition to operate – mechanical malfunction or

problems. Airlines have no choice but to cancel the flight if this happens, and similarly, it is

to avoid unwanted events during departure, airtime, and arrival. Another distinct reason to

cancel a flight is the National Aviation System. This is related to the airline system mainly

from the air traffic controls. There can be some unavoidable events from the air traffic

controls that cause this cancellation to happen. There is another reason that leads to the

cancellation of flights, which occurred the least – security. This involves matters regarding

immigration issues or any related issues that occur at the security checkpoint. In most

airlines, a lot of flights are cancelled due to prolonged delays.

Shifting gears, to understand this visualization further, we can simply hover the cursor

to the legend and select a specific airline that we want to assess. There is also a filter area to

filter information on the cancellation reasons. For example, since the ‘Security’ reason is very

few, it is hardly visible on the treemap visualization. We can use the filter so we can only see

the airlines which had their flights cancelled due to security issues (as shown in Figure 5 from

Appendix).

When comparing between airlines, WN has the highest number of cancellations (see

Figure 6 from Appendix) while HA has the least number of cancellations (see Figure 7 from

Appendix). It is shown that extreme weather is the top reason for cancellation for WN. WN is

the code for Southwest Airlines in Dallas, Texas (the headquarters). It is the world’s largest

carrier offering low costs to passengers. If we want to compare it with the airline available in

Malaysia, we can think of it as AirAsia. Being the world’s largest low-cost airline simply

means it is busy and offers the greatest number of flights. Thus, this makes sense that it has
VISUALIZATION OF FLIGHT DELAYS AND CANCELLATION

the highest number of cancellations. Besides that, the airline with the least number of

cancellations is HA – 170 times caused by the air carrier and one time caused by extreme

weather. HA is the code for Hawaiian Airlines, based in Honolulu, Hawaii. It is mentioned

that HA has been ranked first for reliability in terms of service as compared to the other 15

carriers in the United States (Star Advertiser, 2013). It is also mentioned that HA has 8.4%

more in terms of on-time performance as compared to the average 85.7% of the airline

industry.

Strengths and Weaknesses of Treemap

In terms of its pros, treemap is great at visualizing the overall hierarchical structure of

the cancellation reasons by airlines. Users can immediately capture the idea that each colour

channel represents the airline while the individual rectangular spaces (mark area) represent

the reasons – as it is also highlighted on the visualization with text labels. The size channel of

the leaf nodes also provides the idea of which airline has the most and least frequency of

cancellations.

However, the weaknesses can be seen in terms of the size channel of the leaf nodes.

As we can see from the visualizations, security reason has the least frequency, thus, the

display for this is difficult to see with our naked eye. Publishing this visualization physically

will then be difficult for users to comprehend the information. The only way to tackle this

would be to publish it on an online medium where users can hover the cursor to the visuals

and get details from it or simply by clicking on the legend and filter area.

7. Heatmap: Visualization of Average Arrival and Departure Delays by Weeks in

2015
VISUALIZATION OF FLIGHT DELAYS AND CANCELLATION

Heatmap is a type of visualization which involves two attributes in categorical form

and one quantitative attribute. It is usually represented as an area mark with a colour channel

(mainly in the diverging form of a colour map). For this project, we decided to create a

heatmap to simplify or summarize the average delay time for both arrival and departure

across the whole year of 2015 among airlines. As designers, we want users to see the overall

summary of how each airline is performing. Hence, we divided our categorical data in terms

of weeks and airlines with the delay time as our quantitative attributes. We also included the

frequency of cancellation as a piece of extra information for users. Figure 12 shows the

heatmap created using Tableau.

Figure 12 Heatmap Visualization of Both Arrival and Departure Delay in 2015 by Weeks
VISUALIZATION OF FLIGHT DELAYS AND CANCELLATION

Note. The top visual shows the average arrival delay and the bottom one shows the average

departure delay.

Analysis of The Heatmap Visualization

Based on Figure 12, the colour map used for both average arrival and departure delay

time is the same. This is to ease the comparison between average arrival and departure delay.

There is also a date slider on the right side to allow users to slide to the timeframe they want.

The colour map chosen here is diverging blue-orange. The diverging colour shows that the

flight arrived or departed earlier than its scheduled time. Another piece of information that

we included on the right is the lowest and highest average delay time of both arrival and

departure. This gives instant information to users that the on average, the flight arrived 2.73

minutes earlier and 17.71 minutes later than scheduled. Similarly, for departure, the flight

departed on average 1.55 minutes earlier and 18.8 minutes later.

Other information we can extract from here is that AS, the code for Alaska Airlines,

shows the majority of early arrival for most days. Interestingly, AS departed slightly later

than the scheduled time but was able to arrive earlier. Some of the possible reasons for this

might be the destination is short or the aircraft has good performance in terms of speed.

Hawaiian Airlines (HA) also has good performance in the average delay time compared to

others, as mentioned also in Star Advertiser (2013). We can see that HA is pretty much

consistent, where at most, their average delay time is slightly over the scheduled time (as

seen in Figure 8 on Appendix). Also, another interesting finding, the day which they had a

slight delay happened on Friday. There is not much information on why this happened. The
VISUALIZATION OF FLIGHT DELAYS AND CANCELLATION

only valid explanation would be departure delay in most cases would result in arrival delay.

However, if we want to pinpoint the actual cause for this, more data exploratory analysis

might help us to uncover this question.

Next, from here we can also see that both Frontier Airlines (F9) and Spirit Airlines

(NK) have denser blue saturations. This indicates that these two airlines often have arrival

delays in 2015. Similarly, for the heatmap showing average departure delay, four airlines can

be seen with dense blue saturations – JetBlue Airways (B6), Frontier Airlines (F9), Spirit

Airlines (NK), and United Airlines (UA). These four airline companies – all low-cost airlines,

can benefit a lot from this heatmap visualization so they can strategize better in the future so

they can be as efficient as HA and AS.

The date slider provided can aid users in finding information from a certain period.

For instance, if any of the airline companies want to see the performance for a certain, the

date slider can help them go to the specific period. As shown in Figure 9 from the Appendix,

the average arrival is 6.71 minutes earlier and 38.73 minutes late. Meanwhile, as for the

departure, the average time is 3.25 minutes earlier and 39.5 minutes later than the scheduled

time. This date slider will help airline companies to track the performance of their airlines

and solve this problem or strategize better in the future as this will also reflect on how

satisfied their customers are.

Different Approach, Strengths, and Weaknesses of The Heatmap Visualization

When we first did the heatmap, we were using a single colour saturation as the

channel. However, we realized that this did not convey the information effectively. The

heatmap visualization for this can be seen in Figure 10 from Appendix. As shown, we could

not distinguish much difference especially for HA where the colour looks similar across the
VISUALIZATION OF FLIGHT DELAYS AND CANCELLATION

weeks. This made us shifted to diverging colour instead. The single colour saturation would

only cause confusion which is the opposite of what we want to achieve with this project. As

seen in the main visualization in Figure 12 previously, it is much distinct in terms of colour

and conveys the information better.

As mentioned, a heatmap is great as a summary of the overall event. Here, we can

easily extract information and compare the performance of each airline across 2015 by

weeks. It is simple and yet it gives a punch of valuable information for users. However, the

explanation behind this information cannot be easily extracted simply from this visualization.

For example, we saw how HA had the most delay on Friday but we cannot infer why this

happened just from this vis alone. Similarly, for AS, despite the slight departure delay they

were able to arrive earlier. We could only make assumptions why this happened instead of a

solid and trusted piece of information to confirm the situation. Heatmap should be paired

with other visualizations to help explain the whole story.

Discussion and Future Work

Lessons Learned from Project

Before proceeding with this project, even way back before we learn about data

visualization, it always occurred to us that visualization is a creative field. However, as we

went deeper with this course, we realized that it is more than just being ‘creative’. Data

visualizations involve some thought processes and rules so that as an efficient data scientist,

we will be able to convey the information we want to various users. We also learned that data

visualization is not simply about conveying information, though the main focus is that, there

is so much more to data visualization than we realized.

VISUALIZATION OF FLIGHT DELAYS AND CANCELLATION

For example, in the case of our stacked bar chart versus bar chart, our initial focus

was more on conveying information, but as we learned more and move forward, there is an

interesting insight that we could get from it, simply by deriving some data. Indeed, data is

vast and visualization helps in terms of our cognitive ability to map the idea or the insight so

that it is visible.

Other than that, we learned a lot from the utilization of Tableau. We were so used to

prepare data using either Jupyter Notebook or Microsoft Excel that we thought we have a lot

to do during the implementation of this project in terms of data preparation. However, we

learned that Tableau is indeed a versatile visualization software that we do not need to handle

raw data so much because Tableau already has some features to help us with that. As some of

us are still new in coding, having such user-friendly software to work with has helped us a

lot.

Future Works Using Flights Data

The dataset we have chosen was a great dataset to work with. Plus, this is already

considered big data with almost 6 million items. It is a great start for all of us to work with

such a huge amount of data – more data, more valuable insights, as they say. Most of the

visualizations that we presented are simple and easy to understand to complement our

objective which is to create visualizations where multiple users from various backgrounds

can understand.

However, if we could have more time, we would like to explore the dataset further

and/or dive deeper to understand the whole story or background of what happened with the

aviation industry in the United States in 2015. We believe that data can also tell stories if we

can do a whole exploratory data analysis. In terms of delays, there are multiple reasons for a
VISUALIZATION OF FLIGHT DELAYS AND CANCELLATION

delay to happen. Based on the choropleth map that we created, it would be interesting to see

which type of delay happens the most in a particular state or region. For example, the state of

Florida is more prone to the typhoon as it is located near the coastal area of the Atlantic

Ocean and the Gulf of Mexico, it will be interesting to see how the weather affects the

airlines connected to or from there.

Another thing that would be interesting to look at is to study the time of delays. We

would want to see how each airline decides to cancel a certain flight, as in how long the delay

is before they cancel the flight. With this insight, perhaps airline companies can recommend a

better system to compensate passengers if their flight is cancelled due to prolonged delay.

Besides, we also think that we can gain some insights on how correlated the performance of

aircraft is with the delays. This is also one of the things we would like to investigate further if

we have more time and more information on the aircraft used. The only information about the

aircraft that we have from the dataset is just the tail number. If we could get more information

on other metrics to measure the performance of the aircraft, it would be a good piece of

information so airlines can always maintain or upgrade (if needed) to ensure they are on par.

In terms of the destination, studying the connectivity between and to/from the airports

would be informative as well. Interchanging flights also affect the delays, for example, the

aircraft used is different, thus gaining some insights from that would be beneficial as well.

Airports and airline companies can utilize this information to provide a good experience for

passengers who have interchanging flights – especially for international flights that would

require long waiting hours. Instead of letting the passengers stranded at the airport, perhaps

the airport or the airline companies can provide something that would ease them while
VISUALIZATION OF FLIGHT DELAYS AND CANCELLATION

waiting for their next flight – i.e. a resting area, baggage lockers so passengers do not have to

drag or carry their baggage all the time worrying about having their items stolen, and more.

With all of this information, the most common question to be asked would be which

airline is the best. A lot of people mainly focus on the airline as a whole – best customer

service, infrequent delay, great facilities, and so on. If we could have data from previous

years, we would like to see how each airline is performing over the years. This information is

also beneficial to the airline companies so they can see how they have been performing as

well as their competitors. This creates a healthy competition between them so they can strive

to provide only the best for their passengers.

Conclusion

The aim to present the discovery of overall airlines performance throughout 2015 via

visualization idioms in terms of flights delays and cancellation was achieved, with several

visualizations aimed to improve understanding the delays and cancellation from different

angles. As it was intentioned from the beginning to present the data with multiple

visualizations idioms, a single conclusion to point out which airline has the definite best

performance was not the one and only goal but additionally to ensure the user can understand

more in detail the stories behind individual airlines’ performance, for example Hawaiian

Airline generally do have the least number of flights delay and cancellation but as with other

airlines, it does not travels to all destination which force the traveler to choose other airlines.

Hence, the multiple angles conclusions that provide in depth information that can be made

are:
VISUALIZATION OF FLIGHT DELAYS AND CANCELLATION

1. Hawaiian Airlines has the smallest percentage of flight delays while Spirit has the

highest percentage. Envoy Airways has the highest flight cancellation among other

airlines.

2. North states have less departure delay as compared to the rest of the states, but arrival

delay in each state is not varied much.

3. No single day in a week consistently having less delay than any other days.

4. If delays occurred, the most common reason for the delay is extreme weather.

Security delays contribute insignificant amount of delay for each airline.

5. Small number of flights with certain airplane consistently having more delayed flights

than other airplanes, those flights are associated with UA.

6. Summer has the peak flight delays and fall has the lowest delays.

In designing the visualizations idioms, we try as the best of our knowledge to adhere to the

guiding principles to avoid relaying false or non-existence information due to our selections

of method. For the project, we learned how easy it is with Tableau to transform the ideas or

questions that we have into observable results. And from the visualizations we did, we saw

how different vis are making the data easily understandable such in choropleth map or bar

chart. We also learned how the marks and channels used can change what we perceived from

the vis.
VISUALIZATION OF FLIGHT DELAYS AND CANCELLATION

References

References
Brehmer, M. M. (2016). Why visualization?: task abstraction for analysis and design. Retrieved from
Open Library University of British Columbia:
[Link]

Guo, H., Pang, J., Han, L., & Shan, Z. (2012). Flight Data visualization for simulation & evaluation: A
general framework. Proceedings - 2012 5th International Symposium on Computational
Intelligence and Design., 497-502.

K, K., TN, H. V., & Deivarani, S. (2019). Prediction of Flight Delays and Cancellation. International
Journal of Engineering Research and Management.

Munzner, T. (2014). Visualization Analysis and Design. Visualization Analysis and Design. New York: A
K Peters/CRC Press.

Nguyen, H. T., & Tran, P. V. (2017). Multidimensional cube for representing flight data in
visualization-based system for tracking flyer. 2016 International Conference on Control,
Automation and Information Sciences, (pp. 120-125).

Star Advertiser. (2013, January 11). Hawaiian Airlines ranks first in punctuality, fewest cancellations.
Retrieved from Star Advertiser: [Link]
news/hawaiian-airlines-ranks-first-in-punctuality-fewest-cancellations/

Tableau. (n.d.). Retrieved from Tableau: [Link]

Torsten, V. (n.d.). Data + Dataset Types/Semantics Tasks.

Xi, H., Xiaoyu, S., & Miaoqiong, W. (n.d.). Airline Analysis Dashboard - An Analysis of Airline On-time
Performance and Airport Connectivity.
VISUALIZATION OF FLIGHT DELAYS AND CANCELLATION

Appendix

1. Barchart

Figure 1 Sorting the chart multiple times for multiple events

VISUALIZATION OF FLIGHT DELAYS AND CANCELLATION

2. Treemap

Figure 2 Treemap Visualization of All Cancellation Reasons

VISUALIZATION OF FLIGHT DELAYS AND CANCELLATION

Note. There are four cancellation reasons (extreme weather, air carrier, National Aviation

System, and security). It is clearly shown here that ‘Security’ reason is not visible at all

because the overall cancellation caused by security issues in the year 2015 is only 22.

Figure 3 Treemap Visualization of All Cancellation Reasons by Airlines

Note. We can see here the colour representing ‘Security’ as shown in the legend is not visible

in the treemap idiom.

Figure 4 Treemap Visualization of ‘Security’ Cancellation Reason

VISUALIZATION OF FLIGHT DELAYS AND CANCELLATION

Note. Even after clicking ‘Security’ from the legend area, it is still not visible to be seen.

Figure 5 Filtered Cancellation Reason: Security

Figure 6 Airline with The Highest Number of Cancellations

VISUALIZATION OF FLIGHT DELAYS AND CANCELLATION
VISUALIZATION OF FLIGHT DELAYS AND CANCELLATION

Figure 7 Airline with The Least Number of Cancellations

VISUALIZATION OF FLIGHT DELAYS AND CANCELLATION

Heatmap

Figure 8 Average Arrival and Departure Delay Time of Hawaiian Airlines (HA)
VISUALIZATION OF FLIGHT DELAYS AND CANCELLATION

Figure 9 Date Slider Function to Filter a Certain Time Period

VISUALIZATION OF FLIGHT DELAYS AND CANCELLATION

Figure 10 First Try of The Heatmap Visualization

Do not delete this because it’s troublesome to keep opening the link to refer to proper APA

guidance.
VISUALIZATION OF FLIGHT DELAYS AND CANCELLATION
VISUALIZATION OF FLIGHT DELAYS AND CANCELLATION
VISUALIZATION OF FLIGHT DELAYS AND CANCELLATION
VISUALIZATION OF FLIGHT DELAYS AND CANCELLATION