Temporal Correlation Networks Of American Flight Delays

Every day,flight delays happen in airports around the world,which decreases the efficiency of air transportation and causes great trouble to passengers.Methods in statistical physics have already been used in the research on complex systems successfully,yet have not been applied in the study on flight delays.Havlin proposed the Time Delay Stability(TDS)method for the first time,and applied it into the research of correlations among human physiological signals,which reveals the deterministic relationship between the network structure and physiological state.The purpose of the present investigation is,to reveal the possible correlations among domestic passenger flights of the United States based on the approach of TDS.We mine data of three groups of flights with the longest series(having 12,84,102 flights respectively),then construct temporal correlation networks from the time series of dimensionless velocity based on a suitable threshold ρ.We find the frequently appearing correlation networks in the correlation network of 12 flights,namely the attractors in the evolution process.Theses attractors are highly sparse networks,and some evolutions among these attractors happen in one direction.We do not find any attractors in 84 and 102 flights.We also find the aviation meaning of the network link numbers.There is a deterministic relationship between the link number of the network and the delay ratios for 12 flights.For 84 and 102 flights,the mean degree of the correlation network reaches a peak before the delay ratio arrives at a maximum value when the delay ratio is greater than 65%,then it will decrease in short time.Therefore,the peak of link numbers can be regarded as an indicator of large scale flight delays.Meanwhile,the Hurst indices for the series of link numbers and for the series of delay ratios are quite close,which implies that these series have the similar memory effect.Finally,we determine the proper range for the threshold ρ based on data mining.For 12 flights,the proper threshold ρ should lie in the range 0.5-0.6.For 84 and 102 flights,the proper threshold ρ should be between 0.4 and 0.6.