Accidents on the runway triggered the development and implementation of mitigation strategies. Therefore, the airline industry is moving toward proactive risk management, which aims to identify and predict risk percursors and to mitigate risks before accidents occur. For certain predictions Machine Learning techniques can be used. Although many studies have explored and applied novel Machine Learning techniques on different aircraft Radar and operational Taxi data, the identification and prediction of abnormal Runway Occupancy Times and the observation of related percursors are not well developed. In our previous papers, three feasible methods were introduced: Lasso, Multi-Layer Perceptiona and Neural Networks to predict the Taxi-Out Time on the taxiway and the time to Fly and True Airspeed profile on final approach. This paper presents a new Machine Learing method, where we merge these feasible Machine Learning techniques for prediction the abnormal Runway Occupancy times of unique radar data patterns. Additionally we use in this study the Regression Tree method to observe key related precursors extracted from the top 10 features. Compared with existing methods, the new method no longer requires predefined criteria or domain knowledge. Tests were conductioned using runway and final approach aircraft radar data consisting of 78,321 Charles de Gaulle flights and were benchmarked against 500,000 Vienna flights. @en

currently, at many airports, the runway throughput is the limiting factor for the overall capacity. Among the most important constraining parameters is the separation minima expressed in distance. On the top of these minima, the difference of the leader and follower aircraft speed profiles imposes to consider buffer to cope with compression effect. Currently, Air Traffic Control Officers (ATCO’s) take these buffers on the basis of their training and experience. However, this experience will not be sufficient to safety deploy advanced concepts, like pair-wise separations, that increase variability in the separations to be delivered and therefore in the compression buffer to be considered. Systematic analysis of years of radar tracks has allowed to better predict the buffers to apply by characterising the time to fly (T2F) given a separation distance and True Airspeed (TAS) profile as a function of meteorological parameters. This paper presents how Machine Learning (ML) techniques may be used for predicting the T2F and TAS profile on final approach. Different ML techniques will be assessed on their forecast performance, computational time and amount of data needed for delivering a reliable prediction. The techniques will be applied on 2 different major European airports traffic and will be benchmarked against Optimized Runway Delivery (ORD) study using a Model Based Approach (MBA) for deriving the T2F and TAS. As a result the most efficient ML techniques will be applied on two case studies for predicting the T2F and TAS. @en