When you’re flying anywhere you can pretty much turn the whole day into a black hole. “The airport/in-flight Wi-Fi wasn’t working.” “We sat at the gate for an hour.” “We were in a holding pattern.” It’s great. But sometimes, sometimes you actually want to get where you’re going.
Hoping to improve arrival predictions, GE, Alaska Airlines and Kaggle created a challenge called Flight Quest, offering data analysts a slice of $US600,000 if they proposed a more accurate mathematical model for flight patterns. The contest gave participants access to flight data from the National Airspace System. The two months of data contained origin, destination and flight number data, plus weather, wind and position information.
More than 3000 submissions came in from 58 countries. The winner, a Singapore-based team called Gxav &* built a model that improves runway and gate arrival time predications by about 40 per cent. If the model is implemented travellers would still only save about five minutes at the gate, but airlines could save millions of dollars a year, reducing crew costs by $US1.2 million and fuel costs by $US5 million from saving one minute on each departure.
Everyone wants to reduce the hassle of flying and the contest could be the beginning of a positive trend, but combing through all those data may have taken its toll.
Pawel Jankiewicz, a member of the team that came in second said that the data “starts to talk to me. I must know what it means.” [GE, GE Reports via PopSci]