Machine Learning & Airplanes
The aviation industry is looking to the future.
Prototypes of self-piloted passenger air vehicles by Boeing and Airbus have already made their first test flights. But for now, we can put aside Hollywood-esque scenes in which commercial airplanes fly without human control. It will take decades of certification and testing before air travel becomes fully pilotless.
The good news is that today, airlines use data science and machine learning to automate and speed up operations. Here are several practical examples of machine learning techniques for airplanes.
Revenue management
Nowadays, airline tickets vary significantly for the same flight on the same airlines, even for adjacent seats within the same cabin. Consumers pursue the lowest cost when purchasing a ticket while airlines seek to increase ticket prices. These two contradictory requirements are reconciled in practice.
Airlines tend to optimize their operating revenue using different types of techniques, such as price discrimination. Airlines have models that can predict the minimum cost that customers will buy an airline ticket foron a specific future date.
These models rely on some standard features such as past ticket prices, ticket purchase date, and flight departure date. The future of these models integrates social media data to improve the ability to predict ticket prices and demand.
In-flight food supply
Imagine you are boarding an early morning flight. Your first thought after going through airport checks and finally taking your seat may be to order a cup of coffee and some snacks.
However, people seldom order airplane meals, so one thing the supply management researchers must do is estimate the amount and types of food and beverage they are to prepare onboard so as not to be wasteful and minimize unnecessary cargo weight.
According to data in 2018 from IATA, 6.7 million tons of cabin waste is generated by airlines every year. Each passenger generates more than 3 pounds of cabin waste. To make things worse, plastic is a significant component of this cabin waste and causes irreversible environmental problems around the world.
Machine learning algorithms have been applied to help solve this problem. The demands for snacks may depend on the weather, types of passengers onboard, and their willingness to pay. Airlines can utilize predictive models which consider these factors to ultimately help reduce waste.
Fuel consumption optimization
To reduce the environmental effects of airplanes and cut flight costs, airlines apply AI systems to collect and analyze flight data regarding flight distance, altitudes, actual passenger count, aircraft weight, weather, and so on. For example, neural network models can be used to predict an airplane’s fuel usage
After preprocessing the data and training the model, systems can apply it and therefore estimate the amount of fuel that is necessary for one flight. This helps minimize fuel waste and decrease unnecessary aircraft weight and fuel consumption.
Real-time passenger feedback analysis
Flights are boring. AI systems, as well as machine learning models, can quickly react by making it possible for airlines to determine whether there is a chance to positively influence the passengers' journey and turn their unpleasant experience into a satisfactory one.
Nowadays, models have been developed that include data processing, classification, visualization, and sentiment analysis. With a platform for passengers to leave feedback, the data can be linked to both internal and external operational metrics. The systems can then apply natural language process techniques to process and better understand customer experience data.
Utilizing the power of the AI system, a lot of time-consuming work can be accomplished on existing information so that people can free up their hands to do more complex tasks. This technique can be applied to improve the flight quality of passengers.
Machine learning makes it possible to improve customer experience, optimize their employees' workflow, and ensure aviation safety by predicting fuel consumption and prescribed aircraft maintenance. It also allows airlines to use data intelligently to make informed and fast decisions about prices and market positioning.
With all these automated and self-serviced solutions, we fully utilize our existing data and can foresee a bright future for the airline industry.
Written by Yuxiang Huang & Ethan Samuels
Edited by Kevin Ma, Jack Argiro, Thomas Braun, Michael Ding, Alexander Fleiss & Vishal Dhileepan
Reference
Chen, Y., Cao, J., Feng, S., & Tan, Y. (2015, October). An ensemble learning based approach for building airfare forecast service. In 2015 IEEE International Conference on Big Data (Big Data) (pp. 964-969). IEEE.