An Entropy-Based Fuzzy Time Series Method for Forecasting Airport Passenger Throughput

Abstract

Airport passenger throughput forecasting (APTF) plays an important role in the intelligent air transportation systems (IATS), which has great influence on the operation, controlling and planning of civil aviation. Due to the complicacy and uncertainty oflATS, however, passenger throughput is difficult to be forecasted precisely if traditional statistical methods are applied. In order to improve the precision of APTF, a refined entropy-based fuzzy time series method is put foreword in this paper. First, according to the features of the given datum, the minimize entropy principle approach (MEPA) are adopted here to determine the length of each interval in the universe of discourse. Then, a time-invariant fuzzy relation matrix is built based on the constructed first-order fuzzy time series model, and sequentially the minimum invariant time value of which the data approaches steady state is obtain using the entropy of fuzzy set. Finally, the forecasting results are calculated based on the max-min composition operation and the principle of maximum degree of membership. To illustrate the whole forecasting process, we use the monthly data of FAA from Jan. 1998 to Jun.2000 and compare results obtained with those of other approaches. It is found that the root mean square error of forecast can be improved from 39.36 for the Wu's method and 34.85 for Chen's method and 32.74 for Tsaur-Yang method to 21.98 for the proposed method, which shows that our method is doing much better.