Aviation Delay Estimation using Deep Learning

Abstract

Flight delays cause wastage of time and money to airports, airlines and passengers. Estimation of delays and factors affecting delays help in significant reduction of losses in aviation industry on daily basis. Taking advantage of historical Airline data, weather data at various locations and deep learning algorithms, we can achieve better real time results. In this paper, the model has been trained using the Air traffic data: Flight On-Time performance data obtained from U.S Bureau of Transportation statistics and Weather data: Daily Summaries data obtained from NOAA - National Oceanic and atmospheric administration. The dataset created is a combination of Flight schedules and weather information over a period of 12 months. A deep learning algorithm known as Gated Recurrent Unit network has been proposed due to the recurrent and time-series nature of dataset.