How Machine Learning Is Changing Commercial Flight – Simple Flying

Artificial Intelligence is rolling out across the aviation industry to a greater and greater extent. It could even hold the key to a speedier post-pandemic recovery. Let’s take a look at how its branch of machine learning is already impacting everyday aspects of travel, including how tickets are priced, point-to-point routes, fuel consumption optimization, and biometric boarding.
Machine learning and other AI are having a bigger and bigger impact on everything from ticket prices and route planning to food waste and fuel consumption. Photo: Getty Images
“AI is coming and it will have no mercy for any obstacles on its way. Companies can choose to resist and maintain status quo to extend their survival period, or embrace AI and be part of the ongoing revolution,”  – IATA, AI in Aviation White Paper, 2018.
What is machine learning?
Machine earning, or ML, is a type of cumulative Artificial Intelligence, or AI. It is a method of data analysis that allows computer systems to learn through experience. Algorithms and statistical models analyze patterns, which they then use to improve themselves. This (hopefully) leads to better and better decisions with minimal human intervention.
The larger the volumes of data, the more accurate the algorithm’s evolution through ML will be. And what generates an amazing amount of data (apart from Netflix-views and internet searches) every single day? Airlines and their passengers. And the latter are beginning to put it to very good use.
Ticket pricing is determined by massive amounts of data best handled by evolving algorithms. Photo: Getty Images
Dynamic ticket pricing
The airline industry is considered one of the most advanced in using complex pricing strategies. Most travelers seek to obtain their airline tickets for the lowest price possible. Airlines, on the other hand, want to maximize their revenue. Machine learning algorithms can help both parties in their quest for the best deal.
To determine optimum ticket prices, airlines need to forecast demand on a specific time of year and a particular day. Furthermore, they need to account for factors such as holidays, events, and festivals.They also need to keep on top of what the competition is doing, at what time people are more inclined to purchase what kind of tickets, and predicted fuel prices.

To stay in the proximity of the continuously moving sweet spot, today’s data systems are making billions of predictions per day to this effect. When ML is implemented to its current potential, a single model can make about two million evaluations – per second.
Of course, these algorithms can also work in the customer’s favor by figuring out at what time of day and which day of the month ticket prices will be lower. As airline application of ML increases, we are bound to see even more optimal airfare finder services pop up.


Holidays, such as the Chinese lunar new year, affect demand and ticket prices change according to algorithms. Photo: Getty Images
Route planning
When determining route and frequency demand for specific city-pairs, especially with the rise in point-to-point travel, carriers must consider hundreds of factors. Demographics, industry connections, time of the week and day, season, holidays, events, fuel price, etc., all decide whether or not a route will be profitable and when.
To determine optimal routes and schedules, ML can handle much more data than traditional analytical tools. It can analyze search engine data, booking agent data, social media posts and comments, along with recruitment and professional sites, to determine both leisure and business travel demand.
The ability to analyze large amounts of data and continually draw new conclusions from them will be invaluable to airlines and agile revenue management as the industry emerges into a post-pandemic landscape much different than the one of 2019.
New routes, such as these ones announced by JetBlue last year, are planned on the back of data analysis. Photo: JetBlue
Onboard sales and food supply
What a person eats and drinks on board an aircraft varies greatly, not only from individual to individual and types of travel but also depending on destination and time of day. 20% of all food produced by in-flight catering is wasted every single year.
To minimize both food waste and financial losses, carriers need to analyze previous data of onboard sales and adapt their offerings. The more customizable the in-flight experience becomes, the more sophisticated algorithms airlines need to perfect the supply vs. demand snack situation.
In June last year, easyJet, which said in its 2020 annual report that it aims to become the world’s most data-driven airline, hired British AI-firm Black Swan Data to help it analyze customer food consumption.
“For someone like easyJet, it’s likely that 40pc of their fresh food will be wasted,” Steve King, Black Swan Data’s CEO, told the Telegraph at the time. “The aviation world is insane because there’s no data, it’s like a website where you delete the data every day. It wouldn’t work.”
Machine learning will help airlines lessen their food waste. Photo: Getty Images
Fuel consumption
Along with labor, fuel is an airline’s biggest operating cost, accounting for close to a quarter of expenses. Not only that, but aviation is responsible for about 2.4% of global fossil fuel CO2 emissions. To become more efficient, better calculations for exactly how much fuel is needed for a specific flight are required. Enter machine learning.
According to AI Trends, before Southwest Airlines started a pilot project for predictive models in 2016, the airline was producing 1,200 fuel demand forecasts every month – working with spreadsheets. It would take analysts three days each month to compose the forecasts, which in many cases turned out to be less accurate than ideal.
The computer system generated 9,600 forecasts for each of the close to 100 airports Southwest serves for each month, in 60% less time. While the exact cost savings were never publicized, Doug Gray, Director of Southwest’s analytical data services, confirmed that they were “substantial.”
Algorithms can help airlines calculate fuel consumption with better accuracy. Photo: Getty Images
Biometric boarding
Machine learning techniques are also applied to biometrics. For facial recognition to work it needs to be trained using ML algorithms. This is done by processing thousands upon thousands of images searching for patterns of features.
The technology has been around for some time. However, the access to massive amounts of facial data (all those selifes and tagged photos on the internet) and cheaper computing power has seen the multi-layered deep learning Artificial Neural Networks involved in the process make giant leaps in the past few years. Delta Air Lines recently launched the first US domestic digital identity test at Detroit Metropolitan Wayne County Airport.
The convergence of AI and biometrics could not have come at a better time as airports and airlines all over the world are implementing contactless procedures to keep travel as safe as possible for both employees and and passengers.
Biometrics and facial recognition technology have accelerated during the crisis. Photo: Delta Air Lines
Artificial intelligence is not only here to stay, but it can help the aviation industry recover much faster from what continues to be the most severe crisis in its history. As IATA says, you either resisting and hang on just a little bit longer, or you embrace it and become part of the revolution.
Where do you see AI and machine learning having the biggest impact on aviation in the near future? Let us know in the comments.