Flight Delay Prediction Based On Aviation Big Data And Machine Learning

Along with the development of civil aviation industry,air traffic flow has experienced rapid growth.It is predicted that the number of aircraft will be doubled by 2030.To better suit the development of civil aviation industry,next generation air traffic surveillance technology ADS-B has been widely deployed.Meanwhile,the flight delay has been the major problem hindering the development of civil aviation industry.Related flight delay prediction studies have been focused on applying machine leaning methods to predict fight delay on single airport or air route.This paper explores a wide scope of factors which may potentially influence the flight delay,proposes several machine learning based prediction model,and compares the general performance of proposed models.To build a dataset for the proposed scheme,Automatic Dependent Surveillance-Broadcast(ADS-B)ground stations are deployed,ADS-B messages are received,pre-processed,and integrated with other information such as weather condition from Internet.Different from previous works,this paper divides flight delay into higher resolution classification task and regression task,proposes three Long Short-Term Memory(LSTM)based and two ensemble learning based prediction models.The experimental results show that the LSTM is capable of handing the obtained aviation time sequence data,but overfitting problem occurs in limited dataset.The random forest and Gradient Boosting Decision Tree(GBDT)based classification model can efficiently distinguish flight delay or not,and GBDT based regression model can predict almost 65% flights with the error within 30 minutes.