A Novel Time-Series Data Analytics Approach to Improve the Forecast of Irregular Operations for Airlines and Air Traffic Management

A novel time-series data analytics approach can improve the accuracy of real-time forecasts of the time-line and magnitude for irregular operations for airlines and for Air Traffic Flow Management (ATFM). Contemporary methods used by airlines for implementing irregular operations ldquo;playbookrdquo; exhibitnbsp; forecasting of these events that frequently over/under estimate the timing and magnitude of these events (i.e. too late to be effective, or implemented when it was not required.) This results in excess costs and lost efficiency.Further, the ability to forecast the timing and magnitude of irregular operations will be critical for the implementation of end-to-end Trajectory-based Operations particularly in an environment in which airline ldquo;business rulesrdquo; and tactical preferences are shared by the airlines with ATFM to improve the efficiency of operations.With improved forecasting of irregular operations, airlines and ATFM can addressnbsp;demand-capacity imbalances in the presence of uncertainty at critical choke-points in the NAS.nbsp;The project will extend research conducted on times-series predictive analytics for real-time forecasting the presence, the time-line, and the magnitude of irregular operations in the NAS and for individual airline networks. The approach that has yielded success is to use a novel Machine Learning framework that uses recursive forecasting Bayesian updating. The recursive predictions and the Bayesian update provide a significant boost to the accuracy of neural networks and other Machine Learning algorithms. This project will develop and demonstrate the novel approach different airline networks and the National Airspace System (NAS).nbsp;