空中交通流量预测的人工神经网络和回归组合方法

为了寻找合适的空中交通流量预测方法,在综合回归预测方法和人工神经网络预测方法优点的基础上,提出采用组合预测方法的思想,并基于多元线性回归模型确定组合方法的权重系数。以北京管制区大王庄导航台流量预测为实例,分析结果表明:组合预测方法对实际流量有好的拟合度,能提高人工神经网络的泛化能力,并减小预测的误差,即总体上不仅优于传统的回归预测方法,也优于单独的人工神经网络预测方法。组合方法为空中交通流量的预测提供了一种可靠而有效的新途径。

Air traffic flow forecasts based on artificial neural networks combined with regression methods

Because of uncertainties in air traffic flow forecasts, regression analyses do not adequately fit the data or accurately predict its trends. Artificial neutral networks also do not adequately forecast future trends despite their high accuracy on training samples. This paper presents a new forecasting model for the time series representing air traffic flow patterns which provides data processing methods combining these two methods. The model was used to forecast the flow patterns in the Dawangzhuang air traffic control station of the Beijing region. The results demonstrate that the new method accurately fits the actual flow, improves the prediction ability of the neutral network and reduces forecast errors. Therefore, the new method is better than traditional regression forecast methods and artificial neutral network methods and provides a reliable and effective way to forecast air traffic flow.