Scribd
Upload
7 views9 pages

Network Visualization of Airline Flights

The document discusses network visualization of airline flight data. It loads airline data, creates a networkx graph with airports as nodes and flights as edges weighted by distance. It then visualizes the graph, finding that airport JFK has the most number of long distance flights.

Uploaded by

danial.genuineinside
Copyright
© All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
7 views9 pages

Network Visualization of Airline Flights

The document discusses network visualization of airline flight data. It loads airline data, creates a networkx graph with airports as nodes and flights as edges weighted by distance. It then visualizes the graph, finding that airport JFK has the most number of long distance flights.

Uploaded by

danial.genuineinside
Copyright
© All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd

11/26/23, 11:09 PM Lab Activity - Network Visualization.

ipynb - Colaboratory

NetworkX example

Import the networkx library

import networkx as nx

Initialize the graph. Graph() is for undirected and DiGraph() is for directed graph

G = [Link]()
#G = [Link]()

Add nodes from a list / array

nodes = ["A","B","C"]
G.add_nodes_from(nodes)

Add an edge from A to B

G.add_edge("A","B")

Draw the graph

import [Link] as plt

[Link]()
[Link](G)

Draw with node label

[Link](G, with_labels=True)

[Link] 1/9
11/26/23, 11:09 PM Lab Activity - Network [Link] - Colaboratory

Airlines Network

Import libraries and dataset

import numpy as np # linear algebra


import pandas as pd # data processing and data structure

Load the dataset

data = pd.read_csv('/airlines_network_optimization.csv')

Data Overview

from [Link] import drive


[Link]('/content/drive')

Mounted at /content/drive

[Link]

(99, 16)

Display the first 5 records in the dataset

[Link](5)

year month day dep_time sched_dep_time dep_delay arr_time sched_arr_time a

0 2013 2 26 1807 1630 97 1956 1837

1 2013 8 17 1459 1445 14 1801 1747

2 2013 2 13 1812 1815 -3 2055 2125

3 2013 4 11 2122 2115 7 2339 2353

4 2013 8 5 1832 1835 -3 2145 2155

Drawing network graph of flight distance

Create a new dataframe by selecting only origin, dest and distance columns

data_distance = [Link](['origin', 'dest', 'distance'])

Display the new dataframe

data_distance

[Link] 2/9
11/26/23, 11:09 PM Lab Activity - Network [Link] - Colaboratory

origin dest distance

0 EWR MEM 946

1 LGA FLL 1076

2 EWR SEA 2402

3 JFK DEN 1626

4 JFK SEA 2422

... ... ... ...

94 LGA TPA 1010

95 EWR LAX 2454

96 JFK BOS 187

97 EWR SJU 1608


There are several records where origin and dest are the same. Group them together and calculate the mean distance for that group if any.
98 LGA IAH 1416

99 rows × 3=columns
data_distance data_distance.groupby(['origin', 'dest'], as_index=False).mean()
data_distance

[Link] 3/9
11/26/23, 11:09 PM Lab Activity - Network [Link] - Colaboratory
25 JFK DEN 1626.0

26 JFK DTW 509.0

27 JFK FLL 1069.0

28 JFK IAD 228.0

29 JFK JAX 828.0

30 JFK LAX 2475.0

31 JFK MCO 944.0

32 JFK PBI 1028.0

33 JFK SEA 2422.0

34 JFK SJU 1598.0

35 JFK SRQ 1041.0

36 JFK TPA 1005.0

37 LGA ATL 762.0

38 LGA BHM 866.0

39 LGA CLT 544.0

40 LGA CMH 479.0

41 LGA CVG 585.0

42 LGA DCA 214.0

43 LGA DEN 1620.0

44 LGA DTW 502.0

45 LGA FLL 1076.0

46 LGA IAD 229.0

47 LGA IAH 1416.0

48 LGA MCO 950.0


Draw network graph based on the data above. Identify which airport (EWR, JFK or LGA) has the most number of long distance flight.
49 LGA MDW 725.0

50 LGA MIA 1096.0


import networkx as nx
51 LGA MSP 1020.0
graphD = [Link]()
52 LGA ORD 733.0

53 using
Add nodes LGA origin
PBIand dest
1035.0
column
54 LGA RDU 431.0
graphD.add_nodes_from(data_distance['origin'])
55 LGA STL 888.0
graphD.add_nodes_from(data_distance['dest'])
56 LGA TPA 1010.0

Add edges with distance as the weight

for index, row in data_distance.iterrows():


graphD.add_edge(row['origin'], row['dest'], weight=row['distance'])

Draw the graph

import [Link] as plt


[Link]()

[Link](graphD, with_labels=True)

[Link] 4/9
11/26/23, 11:09 PM Lab Activity - Network [Link] - Colaboratory

Lets draw the label for the edges

pos = nx.spring_layout(graphD)
edge_labels=dict([((u,v,),d['weight']) for u,v,d in [Link](data=True)])

[Link](graphD, pos, with_labels=True)


nx.draw_networkx_edge_labels(graphD,pos,edge_labels=edge_labels)

[Link] 5/9
11/26/23, 11:09 PM Lab Activity - Network [Link] - Colaboratory

{('EWR', 'BOS'): Text(0.34717333823612767, -0.2352327433966388, '200.0'),


('EWR', 'CVG'): Text(-0.1219596621367476, -0.38961664996052137, '569.0'),
('EWR', 'DFW'): Text(-0.3352747551485213, -0.16334520274627184, '1372.0'),
('EWR', 'IAD'): Text(-0.4815249181096476, 0.21080692310871185, '212.0'),
('EWR', 'IAH'): Text(-0.39974887175863405, 0.3754992830540069, '1400.0'),
('EWR', 'IND'): Text(-0.3925309166976173, 0.29091877548460665, '645.0'),
('EWR', 'LAX'): Text(-0.22071915879980544, 0.49696270993680997, '2454.0'),
('EWR', 'MCO'): Text(-0.31042212792148116, 0.4512580304058955, '937.0'),
('EWR', 'MEM'): Text(0.18116175190146444, -0.3384410415138626, '946.0'),
('EWR', 'MIA'): Text(0.1408098692387888, 0.5667378937823518, '1085.0'),
('EWR', 'MSP'): Text(0.44368252609923947, 0.26010802335761557, '1008.0'),
('EWR', 'MSY'): Text(0.09562273141941097, -0.39327628931425823, '1167.0'),
('EWR', 'ORD'): Text(-0.4596279157574, 0.039352349420861556, '719.0'),
('EWR', 'PBI'): Text(0.4426389053408897, -0.1859033634485659, '1023.0'),
('EWR', 'RDU'): Text(-0.052004077748965226, 0.55643929320759, '416.0'),
('EWR', 'RSW'): Text(-0.3366231787243114, -0.3022052720406718, '1068.0'),
('EWR', 'SEA'): Text(0.03636006922501661, 0.5958029595968523, '2402.0'),
('EWR', 'SFO'): Text(0.38978342790312553, 0.4602155197542636, '2565.0'),
('EWR', 'SJU'): Text(0.4267774486062334, 0.35746535010220065, '1608.0'),
('EWR', 'TPA'): Text(0.25126850639905984, 0.5506072167197091, '997.0'),
weight ('JFK',
= [[Link][e]['weight'] for e in [Link]]
'ATL'): Text(0.2876553328498047, 0.42935093497598203, '760.0'),
('JFK', 'BOS'): Text(0.343383746063162, -0.27049083812446567, '187.0'),
('JFK', 'CLE'): Text(-0.24275899472649543, -0.3729213133172455, '425.0'),
[Link](graphD, pos, node_color='#A0CBE2',
('JFK', 'CLT'): edge_color=weight,
Text(0.46832732639200114, width=2,
0.05166425538083779, edge_cmap=[Link], with_labels=True)
'541.0'),
('JFK', 'DCA'): Text(0.4935188094349753, -0.04091672247641189, '213.0'),
('JFK', 'DEN'): Text(-0.11732291598550598, -0.33087911722892427, '1626.0'),
('JFK', 'DTW'): Text(0.2714231708413726, -0.343012173859437, '509.0'),
Question 1: Which airport (EWR, JFK or LGA) has the most number of long distance
('JFK', 'FLL'): Text(-0.005340912572215486, -0.40922454363260924, '1069.0'),
('JFK', 'IAD'): Text(-0.48531451028261324, 0.17554882838088495, '228.0'),
flight ('JFK', 'JAX'): Text(-0.42476279309307163, 0.0982398779840032, '828.0'),
('JFK', 'LAX'): Text(-0.2245087509727711, 0.4617046152089831, '2475.0'),
('JFK', 'MCO'): Text(-0.3142117200944468, 0.41599993567806864, '944.0'),
('JFK', 'PBI'): Text(0.438849313167924, -0.22116145817639277, '1028.0'),
('JFK', 'SEA'): Text(0.03257047705205096, 0.5605448648690253, '2422.0'),
Drawing network graph of flight delays
('JFK', 'SJU'): Text(0.42298785643326775, 0.3222072553743738, '1598.0'),
('JFK', 'SRQ'): Text(-0.4437820730565589, -0.2273168209141458, '1041.0'),
('JFK', 'TPA'): Text(0.2474789142260942, 0.5153491219918823, '1005.0'),
('LGA', 'ATL'): Text(0.30345675245929205, 0.40154240248432727, '762.0'),
Create('LGA',
a new dataframe by selecting origin, dest and
'BHM'): Text(-0.146988835704565, dep_delay columns '866.0'),
0.5046050572741247,
('LGA', 'CLT'): Text(0.4841287460014885, 0.023855722889183036, '544.0'),
('LGA', 'CMH'): Text(0.44111023942248967, -0.15351661614217302, '479.0'),
data_delayed
('LGA',= 'CVG'):
[Link](['origin', 'dest', 'dep_delay'])
Text(-0.10994783470022594, -0.452683277180003, '585.0'),
('LGA', 'DCA'): Text(0.5093202290444626, -0.06872525496806664, '214.0'),
('LGA', 'DEN'): Text(-0.10152149637601868, -0.35868764972057904, '1620.0'),
data_delayed
('LGA', 'DTW'): Text(0.2872245904508599, -0.37082070635109177, '502.0'),
('LGA', 'FLL'): Text(0.010460507037271824, -0.437033076124264, '1076.0'),
('LGA', 'IAD'):
origin dest Text(-0.4695130906731259,
dep_delay 0.1477402958892302, '229.0'),
('LGA', 'IAH'): Text(-0.38773704432211237, 0.3124326558345253, '1416.0'),
('LGA',
0 EWR'MCO'):
MEM Text(-0.2984103004849595,
97 0.3881914031864139, '950.0'),
('LGA', 'MDW'): Text(-0.4538752269072912, -0.12703456079419814, '725.0'),
('LGA',
1 LGA'MIA'):
FLL Text(0.15282169667531048,
14 0.5036712665628702, '1096.0'),
('LGA', 'MSP'): Text(0.45569435353576115, 0.19704139613813396, '1020.0'),
2 EWR
('LGA', SEA Text(-0.4476160883208783,
'ORD'): -3 -0.02371427779862008, '733.0'),
('LGA', 'PBI'): Text(0.45465073277741136, -0.2489699906680475, '1035.0'),
3 JFK DEN 7
('LGA', 'RDU'): Text(-0.039992250312443564, 0.49337266598810836, '431.0'),
('LGA', 'STL'): Text(0.5161229109133811, 0.11963212566402007, '888.0'),
4 JFK SEA -3
('LGA', 'TPA'): Text(0.2632803338355815, 0.4875405895002275, '1010.0')}
... ... ... ...

94 LGA TPA -6

95 EWR LAX 0

96 JFK BOS -1

97 EWR SJU 1

98 LGA IAH -4

99 rows × 3 columns

Normalize the data to remove negative values. Negative means the flight depart before schedule. Let's just take only rows with dep_delay > 0

data_delayed =data_delayed[data_delayed['dep_delay'] > 0]

data_delayed

[Link] 6/9
11/26/23, 11:09 PM Lab Activity - Network [Link] - Colaboratory

origin dest dep_delay

0 EWR MEM 97

1 LGA FLL 14

3 JFK DEN 7

16 JFK TPA 16

24 EWR TPA 19

25 EWR MSY 2

27 JFK DTW 199

35 LGA MSP 2

36 LGA DTW 2

43 EWR IAH 4

44 JFK CLT 9

45 LGA STL 2

54 EWR IAD 2

55 LGA PBI 27

62 JFK LAX 21

64 LGA CVG 2

65 EWR CVG 68

67 EWR IAH 5

68 LGA DTW 14

69 LGA ORD 37

71 LGA TPA 86

72 JFK PBI 48

80 EWR MSP 9

83 EWR LAX 17
Some 84
record has
JFK no value in dep_delay
MCO 235 column. Drop those record(s).

88 JFK SJU 1
data_delayed = data_delayed.dropna()
89 JFK LAX 119

90
data_delayed JFK ATL 48

97 EWR SJU 1

[Link] 7/9
11/26/23, 11:09 PM Lab Activity - Network [Link] - Colaboratory

origin dest dep_delay

0 EWR MEM 97

1 LGA FLL 14

3 JFK DEN 7

16 JFK TPA 16

24 EWR TPA 19

25 EWR MSY 2

27 JFK DTW 199

35 LGA MSP 2

36 LGA DTW 2

43 EWR IAH 4
Draw network graph
44 JFK CLT 9

import45 LGA as STL


networkx nx 2

54 EWR IAD 2
graph = [Link]()
55 LGA PBI 27

graph.add_nodes_from(data_delayed['origin'])
62 JFK LAX 21

64 LGA CVG 2
for index, row in data_delayed.iterrows():
65 EWR CVG 68 row['dest'], weight=row['dep_delay'])
graph.add_edge(row['origin'],
67 EWR IAH 5

68 LGA DTW 14
weight = [[Link][e]['weight'] for e in [Link]]
69 LGA ORD 37

[Link](graph,
71 LGAnx.spring_layout(graph),
TPA 86 node_color='#FA8072', edge_color=weight, width=2, edge_cmap=[Link], with_labels=True)

72 JFK PBI 48

80 EWR MSP 9

83 EWR LAX 17

84 JFK MCO 235

88 JFK SJU 1

89 JFK LAX 119

90 JFK ATL 48

97 EWR SJU 1

Question 2: Which airport (EWR, JFK or LGA) has the most number of departure delayed
flight?

airport JGK has the most number of depatured delayed flight.

Question 3: Draw a network graph using air_time as the edges weight. Compare it with
the distance graph in Question 1 to know whether there is a direct correlation between

[Link] 8/9
11/26/23, 11:09 PM Lab Activity - Network [Link] - Colaboratory

air time and distance. Is the graph similar or there are some differences? You can use
the same code as in Question 1. You just need to change the column name
data_airtime = [Link](['origin', 'dest', 'air_time'])
data_airtime

origin dest air_time

0 EWR MEM 144

1 LGA FLL 147

2 EWR SEA 315

3 JFK DEN 221

4 JFK SEA 358

... ... ... ...

94 LGA TPA 147

95 EWR LAX 308

96 JFK BOS 40

97 EWR SJU 200

98 LGA IAH 189

99 rows × 3 columns

data_airtime = data_airtime.dropna()
data_airtime

origin dest air_time

0 EWR MEM 144

1 LGA FLL 147

2 EWR SEA 315

3 JFK DEN 221

4 JFK SEA 358

... ... ... ...

94 LGA TPA 147

95 EWR LAX 308

96 JFK BOS 40

97 EWR SJU 200

98 LGA IAH 189

99 rows × 3 columns

graph2 = [Link]()

graph2.add_nodes_from(data_airtime['origin'])
for index, row in data_airtime.iterrows():
graph2.add_edge(row['origin'], row['dest'], weight=row['air_time'])

weight2 = [[Link][e]['weight'] for e in [Link]]

[Link](graph2, nx.spring_layout(graph2), node_color='#FA8072', edge_color=weight2, width=2, edge_cmap=[Link], with_labels=True)

[Link] 9/9

Common questions

Powered by AI

Network graphs can be used to visualize flight distances between airports by creating nodes for each airport and edges representing the flight paths, with weights assigned to these edges based on the distances. This approach helps identify hub airports with many long-distance connections, highlighting potential areas for network optimization. For instance, JFK has a significant number of long-distance flights, making it a key hub in the network .

Normalizing delay data ensures that only meaningful delays (positive departure delays) contribute to the network visualization, which helps in accurately representing and analyzing delay patterns. In the lab activity, normalization was done by filtering out negative delay values that indicate early departures and selecting only rows where dep_delay was greater than zero .

The graph comparing air time to flight distance can reveal discrepancies where longer flight times do not necessarily correspond to longer distances, possibly due to indirect flight paths or air traffic congestion. By visualizing both data sets using network graphs with edges weighted by air time and distance respectively, one can assess whether airports have consistently proportional air times for measured distances. Differences in these graphs might indicate inefficiencies or indirect routing .

Dropping records with missing delay data can lead to incomplete analysis by ignoring potential patterns or biases in flight performance. Important connections or trends might be overlooked if such records are common among certain routes. Addressing this could involve imputing missing data using statistical methods or correlating with available data to estimate delays. Moreover, adopting models that tolerate missing values might preserve the integrity of network analysis .

The procedure of grouping and averaging flight distance records involves aggregating all records with the same origin and destination and calculating the mean distance for these groups. The purpose is to condense multiple data points into a single representative value, reducing redundancy and focusing on the most typical distances traveled between pairs of airports. This enhances visualization clarity by simplifying the network graph without losing significant data details .

Visualizing only positive departure delays focuses analysis on operational inefficiencies or issues, spotlighting where improvements are necessary to better align actual and scheduled times. However, it might overlook positive performance areas, such as consistent early departures, which could provide lessons for best practices. Thus, while it highlights problematic areas, a comprehensive understanding of operations requires examining both delays and advancements, to holistically assess airport performance and optimize procedures .

NetworkX and Matplotlib are powerful tools for network graph creation due to their integration capabilities and ease of use. NetworkX provides a framework for representing and manipulating complex networks, allowing for the easy addition of nodes and edges from data frames. Matplotlib enables detailed control over the visualization, such as customizing the appearance of nodes and edges, and applying color scales to represent various data dimensions. This combination allows users to create informative visual representations of flight data efficiently .

The edge color scale and width enhance network graph interpretation by providing a visual emphasis on the intensity or quantity of the data attribute represented by the edges, such as flight distance or delay duration. Softer colors and thinner lines might represent less significant or less frequent connections, while bolder colors and thicker lines highlight major routes or significant delays, enabling quick visual differentiation and prioritization of data examination .

JFK is the airport most affected by flight delays, as determined by analyzing the number of delayed departure flights mapped in a network graph. This was visualized by creating nodes for airports and edges weighted by dep_delay values, highlighting JFK as having the highest aggregated delays .

The graph was plotted with weights and labels for edges by iterating over each edge, assigning the corresponding flight distance or delay as the edge weight, and using nx.draw functions to visualize the network. Edge labels were added using nx.draw_networkx_edge_labels to display these weights on the graph. This customization helps easily identify and quantify the significance of each route or delay, providing clearer insights into which routes are crucial or problematic based on their distance or delay metrics .

You might also like