Runway Visual Range Prediction Based on Ensemble Learning

With special geographical and meteorological conditions, it is difficult for planes to fly at High altitude airport. Thus the rate of delay and accident remains very high. Flight accidents mostly happened when the planes was taking off or land. These are directly related with the runway visual range, which is crucial to airport operation. This paper propose a method to obtain the meteorological feature by using the deep auto-encoding network. Guided by this method, we use the auto-encoding network to study the meteorological feature and the Deep auto-coding network to explore the potential information in 2 dimensional data. We have made a real-time prediction of the runway visual range, by using the algorithm of machine learning and deep learning and the monitoring of meteorological characteristic through different dimensions. The precision rate is 91% and the TS score is up to 81.14%, 6% higher than the industry level of 75%. The method based on Ensemble Learnings multiple model fusion was studied to improve the overall performance. The final overall TS score of the fusion model have reached 82%. The study of the runway visual range prediction method in this paper help space dispatcher make a comprehensively and efficiently predict of runway visual range. It is expected to improve airport operation and reduce economic losses caused by flight accident efficiently.