基于长短期记忆网络与支持向量回归的空中交通流量预测研究

由于航空运输的高速发展,空中交通流量日益增大,空中交通流量预测难度不断增加。为提高空中交通流量预测的精确度,通过运用支持向量回归(SVR)和长短期记忆网络(LSTM)2种方式对广播式自动相关监视(ADS-B)数据进行处理。结果表明,LSTM预测方法对交通流误差有较好的控制,具有较好的预测精度。SVR模型和LSTM模型都能很好地捕捉非线性特征与交通流之间的内在关系。发现航路点流量时间序列曲线的趋势性越明显,预测曲线的拟合度越高;时间序列曲线的波动幅度越大,预测曲线拟合度越差。LSTM预测方法的均方根误差为2.56,SVR预测方法的均方根误差为3.59,航路点流量时间序列趋势的明显性直接影响模型的预测精准度。

Due to the rapid development of air transportation, the air traffic flow is increasing day by day, and the air traffic flow forecast is becoming more and more difficult, in order to improve the accuracy of air traffic flow prediction.Support vector regression(SVR) and long short-term memory(LSTM) network are used to process Automatic Dependent Surveillance-Broadcast(ADS-B) data. The results show that the LSTM prediction method has better control of traffic flow error and better prediction accuracy. Both SVR model and LSTM model can well capture the internal relationship between nonlinear features and traffic flow. It is found that the more obvious the trend of the route point flow time series curve is,the higher the fitting degree of the prediction curve is, and the greater the fluctuation range of the time series curve is, the worse the fitting degree of the prediction curve is. The root mean square error of LSTM prediction method is 2.56. The results show that the explicit trend of route point flow time series directly affects the prediction accuracy of the model.