Abstract-Aircraft performances are a very important input for trajectory calculations. All known models distinguish performances mainly on aircraft type. We know there are many other factors influencing aircraft performances like operator, aerodrome, take-off weight, etc. We propose to build a machine learning model that would predict aircraft performances based on all known attributes about the flight. The model learns from real flights in the past and makes predictions based on accumulated knowledge. For instance: A full flight to a distant holiday destination has different performances than a short regular flight with the same aircraft. Our model would recognize the difference in that case and predict aircraft performances best suited for each flight. aircraft performances; machine learning; data mining