基于SARIMA和RBF神经网络的机场货运量预测  

Airport cargo forecasting based on SARIMA and RBF neural network

作　　者：邢志伟[1] 李学哲[1,2] 罗谦[2] 冯文星[1,2] 白楠[1,2] 潘野[2] 罗沛 XING Zhiwei LI Xuezhe LUO Qian FENG Wenxing BAI Nan PAN Ye LUO Pei(College ofA eronautical Automation, CA UC, Tianjin 300300, China Second Institute of CAAC, Chengdu 610041,China)

机构地区：[1]中国民航大学航空自动化学院,天津300300 [2]中国民用航空局第二研究所,成都610041

出　　处：《中国民航大学学报》2016年第5期51-55,共5页Journal of Civil Aviation University of China

基　　金：国家科技支撑计划(2012BAG04B02);国家自然科学基金项目(U1233118,U1333122,U1233124);中央高校基本科研业务费专项(3122014P003)

摘　　要：针对机场货运量预测不能满足机场实际运行精度等缺点,提出一种季节性ARIMA和RBF神经网络集成模型预测机场货运量,该模型首先利用季节性ARIMA模型预测机场货运量线性部分,然后用RBF神经网络模型预测机场货运量非线性部分,最后把非线性部分预测结果作为线性部分预测结果的补偿,得到最终预测结果。实验结果表明,新模型可以有效结合季节性ARIMA和RBF神经网络各自的优点;相对单一季节性ARIMA模型和单一RBF神经网络模型预测精度分别提高了6.30%和3.32%,预测精度满足机场实际运行要求。The model of integrated seasonal ARIMA and RBF neural network （SARIMA-RBF）is proposed to solve the problem that airport cargo forecasting accuracy can not meet the actual operation of the airport. In the SARIMA- RBF, the first use of seasonal ARIMA is to forecast the linear part of airport cargo, and then to forecast the non- linear part of airport cargo with RBF neural network, finally the nonlinear forecasting result is taken as the compensation of linear forecasting result to get the final forecasting result. Experimental results show that the new model can be combined with respective advantages of seasonal ARIMA and RBF neural network. The new model compared with single seasonal ARIMA model and single RBF neural network model forecasting accuracy are improved by 6.30% and 3.32%; and its forecasting accuracy can meet the actual operation of the airport.

关 键 词：机场货运量 季节性ARIMA RBF神经网络 集成 预测 

分 类 号：V35[航空宇航科学与技术—人机与环境工程] TP391[自动化与计算机技术—计算机应用技术]