Flight Delay Prediction Study Based On Spline Curves

Adding to the fast development today is the appearance of miscellaneous data produced in different fields. Dealing with these data requires us to highly make use of these big data to categorize, analyze and predict. By utilizing large amount of real time data, this thesis proposes two novel prediction models based on spline functions, ARIMA model and Multiple Regression.The first model is called General Long Term Departure Prediction Model, which categorizes delay factors into season, departure period and random factors. A weighted smoothing spline is employed to season and departure period simulation respectively and an ARIMA model is used for random factors. This model shows not only good prediction but also the possibility of long term prediction.The second model is focusing on real time arrival prediction. It covers most significant factors such as departure weather, carrier, National Aviation System, security and previous late aircraft. The model is based on spline functions and Multiple Regression and is created by applying large amount of corresponding real time data. It captures up to date weather data and flight schedule data to complete a real time prediction. Besides the theory part, it also presents a practical and beautiful web application for real time flight arrival prediction based on our second model. The results demonstrate goodness of fit.