Evolving Aviation using Machine Learning: the Wright Brothers' Style
Engineers from Sarajevo are shaping the future of aviation; surely you can't be serious?
"I am serious; and don't call me Shirley."
Chances are you've probably booked a flight online sometime in your life, so you know the drill: you type in the desired departure and arrival destinations and times and wait for the service to calculate the most affordable option. Flight prices for the same distances vary so greatly for different times it's hard not to wonder about the logic behind the price calculation algorithm and, of course, not to get annoyed by the fact that your ticket would be twice as cheap if you could travel four days earlier. When irritated, some people write a scathing online review but others found a startup that will tackle the problem.
Flying in the face of Traditional Flight Booking
AirProxima is an online platform that takes a different approach. It offers one-way trips on private charter aircraft and uses artificial intelligence to give the user the best options out of the millions of variations. While it may not sound intelligent (artificially or otherwise) to book one-way instead of round-trip tickets the truth is actually the opposite. Hundreds of variables dictate the flight price but it is possible to find round-trip options using different operators by splitting the trip into two separate one-way routes. This can significantly reduce the price of your ticket. AirProxima uses machine learning technology to predict flight variables, optimize aircraft routing across the US and create savings for both the passengers and aircraft operators.
Now, let's get back to the beginning of the story.
Founded in 2015, this Silicon Valley startup needed to bridge the gap between the idea and realization and that's where Symphony came in. Our team from the Sarajevo hub has been working with AirProxima since September 2016 on developing the platform to building predictive tools based on machine learning technologies. Thanks to this project, our engineers are collaborating with some of the leading experts in the ML and AI fields. One of the inspiring people Symphony is working with is Ed Crump, the CEO of AirProxima, who is a former Senior Engineering Executive at Amazon and one of the minds behind Alexa (Echo) and FireTV.
Anything but pie in the sky
The Symphony team was asked to architect and develop a platform that will provide both airline operators and passengers with a fully managed flight reservation system. This included a web and mobile flight-booking site for consumers and flight operator tools. The latter support features essential for aviation, including reports on weather conditions, calculating and optimizing flight times, dynamic price structures and maximizing aircraft utilization. The platform has been engineered and developed using Node.js backend server technology as well as Angular.js and React.js for front-end.
After developing the platform, our team started work on a mobile application that will maximize the end-user experience. We built an iOS and Android applications using React Native and cross-platform development. The mobile application integrates features from the existing web application and has user-friendly UI that enables AirProxima customers to easily browse and search available flights and make booking requests.
(Machine) Learning to fly
What makes this project special is the additional step the company took to improve its search algorithm and the two magic words are ‘machine learning’. The private aviation industry is unique in the way that routes can be altered in order to meet the needs of last minute passengers and schedule changes. Yet this makes it harder to plan and predict flight costs. With conventional business models, the passenger is charged more because of repositioning costs that occur when an aircraft flies home empty after dropping off its passengers. The Virtual Fleet concept from AirProxima means that the airplane can remain at the last flight destination until there is a new flight demand. This makes it possible for AirProxima to offer significantly lower prices. They do this by analyzing constantly changing schedules and showing the most affordable options.
The Symphony team has built scheduling algorithms that reduce the downtime of aircrafts and consequently reduce costs for airline operators. In addition, we built a series of machine learning models that predict demand for a specific period taking into account the specific class of aircraft, point of departure and point of arrival. The results from the machine learning models are combined with scheduling algorithms to improve the process of flight scheduling by using historical data and real-time data feeds to optimize charter jet fleets.
Passed with flying colors
What do the numbers say! This service is currently showing substantial operational savings that are 5 to 10 times higher than with other optimization tools and is on a path to delivering more than 25% cost savings for fleets of over 100 aircraft. The most recent thing we've been working on is a demand prediction tool that has demonstrated the ability to predict future charter flights with over 70% accuracy. This is a potential game changer for the whole aviation industry, which is why being a part of it is so fly (and this is the last bad airplane-related pun you have to read today). Now the sky is the limit (we lied earlier).