Air Transport Network Analysis with Python

This project studies domestic air transport networks in the United States, China, the United Kingdom and Australia using one month of flight data.

I built it to look beyond the obvious question of which airports are busiest. The analysis compares traffic concentration, bridge airports, hub dependence and the size of each country's network core.

Problem Statement

In an air network, airport importance is not only about flight volume. An airport can also matter because it connects distant regions or sits between parts of the network that would otherwise be weakly connected.

This project asks:

1. Which airports and routes carry the most domestic traffic?
2. Which airports act as bridges?
3. Which countries are more dependent on a small number of hubs?
4. What does the network structure suggest about resilience?

Dataset

The project uses two files in the data/ folder.

File	What it contains
data/Airports.csv	Airport code, airport name, country and coordinates
data/Flight Data.xlsx	Source airport, target airport, countries and route weight

Only domestic flights are used, so the source country and target country must match. I also corrected visible coordinate errors in the airport table before creating the maps.

Methods

I built an undirected weighted graph for each country. Airports are nodes, and domestic routes are weighted edges. If the same route appears more than once, the route weights are summed.

The main metrics are:

1. Weighted degree, to rank airports by traffic volume.
2. Betweenness centrality, to find bridge airports.
3. Degree assortativity, to measure how strongly hubs connect to smaller airports.
4. k+ core size, to estimate the central group of airports in the network.

I also added SQL queries for the same type of checks in a relational format: country summaries, top hubs and busiest route pairs.

Results

Country	GDP rank used as context	Assortativity	Core size	Main reading
United States	1	-0.198	41	Large core, strong hub and spoke behaviour, and several bridge airports
China	2	-0.397	23	The most hub focused network in this comparison
United Kingdom	6	-0.119	16	Compact network with less extreme hub dependence
Australia	15	-0.230	9	Small core shaped by long distances and remote area connectivity

The clearest pattern is that volume and structural importance are not the same. Some airports are important because they handle a lot of traffic. Others are important because they connect regions that would otherwise be harder to reach.

China has the strongest negative assortativity, which points to a more hub dependent structure. The United States has the largest core, meaning that important network functions are spread across more airports. Australia has a smaller core, but distance makes several airports strategically important.

Country Notes

United States. The network is large and geographically spread out. It has a clear hub and spoke pattern, but some airports with moderate degree still have high betweenness because they connect isolated regions.

China. The network is dense in the east and sparser in the west. Its strong negative assortativity suggests that major hubs carry a lot of the network structure.

United Kingdom. The network is compact, with shorter domestic distances and a weaker link between degree and betweenness. That points to a more distributed domestic structure.

Australia. Long distance links matter more here. Several airports have value because they connect remote areas, even when they are not the largest by direct traffic volume.

Figures

Cleaned spatial networks	Weighted degree ranking

Assortativity	Core and periphery

How to Run

python -m venv .venv
.venv\Scripts\activate
pip install -r requirements.txt

Open the notebooks in this order:

1. notebooks/01_data_cleaning_and_network_construction.ipynb
2. notebooks/02_network_metrics_analysis.ipynb
3. notebooks/03_operational_interpretation.ipynb

The map function uses a Natural Earth basemap when internet access is available. If the basemap cannot be downloaded, the notebooks still run and plot the routes and airport points without the background map.

SQL

The sql/ folder shows how I would reproduce the main checks if the data were loaded into PostgreSQL style tables.

File	Purpose
01_create_tables.sql	Table structure and indexes
02_domestic_flight_summary.sql	Domestic flight summaries by country
03_top_airport_hubs.sql	Top airport hubs using weighted degree logic
04_route_volume_analysis.sql	Busiest domestic route pairs

The Python notebooks remain the main analysis because the network metrics need NetworkX.

Dependencies

Python, pandas, NumPy, NetworkX, GeoPandas, Matplotlib, openpyxl and SQL. Install the package ranges with:

pip install -r requirements.txt

My Contribution

I did the full workflow: data cleaning, coordinate correction, graph building, metric calculation, SQL checks, figures and interpretation.

The part I focused on most was interpretation. I wanted the project to say more than "these are the biggest airports". The final analysis connects the metrics to hub dependence, bridge airports, route concentration and network resilience.

Repository Structure

air-transport-network-analysis-python/
  data/
  figures/
  notebooks/
  sql/
  src/
  README.md
  requirements.txt

References

1. Guo et al. (2019), Global air transport complex network: multi scale analysis
2. Ersoz and Aldemir (2024), complex network analysis of European air transport
3. Nominal GDP country ranking used as context in the original analysis

License

The project code is released under the MIT License.