Analysis of Time Series of Statistical Air Traffic Data

Commercial ADS-B data providers deliver massive amounts of continuous data for aircraft traffic worldwide. They are therefore an ideal source for statistical analysis of traffic flows and anomaly detection. By clearly identifying the aircraft using the ICAO code, unique trajectories are obtained worldwide that allow the pattern-of-life of each aircraft to be understood. Furthermore, with modern machine learning technologies and parallel processing frameworks such as Spark, heat maps can be generated and areas of interest, e.g. airports, can be found, which in turn are a starting point for further analysis. The investigation of the change over time of these statistics is part of this paper, as well as the detection of anomalies in the statistics with the usage of an autoencoder deep learning network.