基于NARX动态神经网络的民航客运量预测研究

Research on Passenger Volume Forecast of Civil Aviation Based on NARX Dynamic Neural Network

民航客运量的准确预测对民航交通规划建设具有现实意义。论文首先对民航客运量数据的变化趋势和特点进行分析,得到客运量时序数据的发展趋势及波动规律。其次,论文提出基于NARX动态神经网络的民航客运量预测方法,该模型可以从非平稳的时序数据中提起历史数据的发展趋势以及周期波动的规律。论文利用2008年~2017年的历史客运量时序数据对未来两年的客运量进行仿真预测并验证。同时,论文还选用了ARIMA模型,Holt-Winters模型进行对比仿真实验。仿真结果表明,NARX动态神经网络的预测精度最高,R2、MAE、RMSE分别为0.91、81(万人)、102(万人),可以有效提高民航客运量预测的准确度。

The accurate prediction of civil aviation passenger traffic is of practical significance to civil aviation transportation planning and construction.This paper first analyzes the change trend and characteristics of civil aviation passenger traffic data,and obtains the development trend and fluctuation law of passenger traffic time series data.Secondly,this paper proposes a passenger traffic volume prediction method based on NARX dynamic neural network.This model can bring up the development trend of histori⁃cal data and the law of periodic fluctuations from non-stationary time series data.This article uses historical passenger traffic time series data from 2008 to 2017 to simulate and predict passenger traffic in the next two years.At the same time,this article also se⁃lected the ARIMA model,Holt-Winters model for comparative simulation experiments.The simulation results show that the NARX dynamic neural network has the highest prediction accuracy,R2,MAE,and RMSE are 0.91,81(10,000),and 102(10,000),re⁃spectively,which can effectively improve the accuracy of civil aviation passenger traffic forecast.