Research On Airport Delay Analysis And Prediction Using Data Mining Techniques

With the increase of air traffic,flight delay has been a hot issue in the air transport industry at home and abroad.Due to the complexity of the delay,no effective ways have been put forward to avoid the large-scale delay.Therefore,flight delay not only needs reducing the possibility of occurrence from the source,but also needs to be effectively evaluated in the airport before the large-scale flight delay happens.However,it is difficult to build an accurate flight delay prediction model because of the complexity of air traffic transportation system,the diversity of prediction models and the large amount of data in the aviation system.On this background,this paper tries to analyze the delay time series of the airport from the perspective of data science.The characteristics of the airport delay time series have been studied and the data mining methods to predict the departure delay level of the airport have been used.At first,a new way to study the airport delay is put forward.Departure delays of the specific airport and temporal factors associated with it within a prescribed period are taken as objects.Flight Plan and Meteorological Terminal Aviation Routine Weather Report(MEATR)of Guangzhou Baiyun International Airport in 2016 are collected as sample data in this paper.All kinds of the flight plan time nodes are extracted from the sample data.Relevant indexes and calculation formula of the airport departure delay are defined based on the single flight.The cumulative values of all flights within the statistical period are calculated as experimental data.Secondly,the analysis of similar features of the airport delay time sequence is studied using Dynamic Time Warping(DTW).Based on the sequence similarity measure,other temporal factors associated with the airport delay in similar circumstances are explored the Pearson correlation coefficient.In addition,K-means algorithm is used to study the features of the airport departure delay.Then,based on the above experimental data and features,Classification and Regression Tree(CART)and BP neural Network are used to predict the sample data in three different scenarios.Finally,the effective classification method and prediction algorithm of airport departure delay are obtained by comparing the three classification methods as well as the prediction accuracy of the two algorithms.In this manner,the validity of the data processing method in this paper is also verified.