Tanks for all your answers!
We know that there is a lot of factors thats can influence a flight. Maybe we’re not pilots, but we think that the human factor is one of the most important. That’s why it would be very difficult to predict time of arrival of a plane by looking the natural factors.
So let me explain you one of the base principal of our algorithms.
Let’s take two airports, Atlanta and Nashville. Between these two airports, there is a local storm. Then, a first flight that departs at 8:00, is going to have to climb up, or to gain speed to pass before the storm etc… Then, our ‘sisia’ project is going to know it when this flight will execute his changes, and at each data that we’re going to have on this flight, the system will place a mark on the importance of the change (that’s the part where we need statistics). With these marks, we are able to crate a map that shows places where planes had to change something in their flights. Then a second flight departs at 8:06 from the same airport of the first flight, and it is following a similar path of the first flight but is not necessarily landing at Nashville. So, with our data, we are able to say that the second flight is going to cross a ‘red’ sector for 30 kilometers, where a lot of other planes had to change a bit their path to pass the storm. Then if we are logical, the flight that departs at 8:06 have a lot of chances to change a bit his way like the others, and proportionally to the during of time the plane is going to be in the ‘red’ zone and the intensity of this zone, we are going to try to evaluate a late time for this plane.
In brief, our system is not going to use meteo data and all that stuff to predicts late for a plane, our system is going to use the brief history of a sector, to do the same thing.
Tank you for all your answers!
(I hope our web site is going to be online soon!)
Benoit, ‘sisia project’
P.S. : We already know that it’s different from a plane to the others, that’s why we’re probably going to use only boeing planes.