基于岭回归BP神经网络的管制工作负荷预测方法.pdfVIP

第15卷第1期 交通运输系统工程与信息 Vol. 15 No. 1 2015 年 2 月 Journal of Transportation Systems Engineering and Information Technology February 2015 文章编号：1009-6744（2015）01-0123-07 中图分类号：V35 文献标志码：A 基于岭回归—BP神经网络的管制 工作负荷预测方法 * 温瑞英，王红勇 (中国民航大学 空中交通管理学院，天津300300) 摘 要： 基于空中交通复杂程度刻画管制工作负荷是当前空中交通管理领域的研究热 点.本文采集了厦门空管站的雷达数据，计算得出10 个空中交通复杂性评价指标数值，通 过共线性诊断发现复杂性指标间存在较强的多重共线性.在利用岭迹图对复杂性评价指 标进行筛选的基础上，建立岭回归—BP 神经网络组合模型对管制员工作负荷进行预测， 并通过实测陆空通话数据进行验证.结果表明，本文提出的岭回归—BP 神经网络组合模 型收敛速度快、训练时间少；组合模型的均方误差、均方根误差、平均绝对误差、平均绝对 相对误差等4 项性能指标都相对较小，预测精度较高. 关键词： 航空运输；管制员工作负荷；岭回归；神经网络；交通复杂性 A Forecasting Method of Controller ’s Workload Based on Ridge Regression—BP Neural Network WEN Rui-ying ，WANG Hong-yong （Air Traffic Management College, Civil Aviation University of China, Tianjin 300300, China) Abstract:Abstract: It is becoming a new hot topic in the field of air traffic management that evaluating the controller ’s workload by the traffic complexity factors. Based on the radar data of Xiamen air traffic control station, 10 typical complexity evaluation factors were calculated. The strong multi-co-linearity among various complexity factors is discovered through co-linearity diagnosis. Using the ridge trace plot of ridge regression, the complexity evaluation factors are selected, and the combined model of ridge regression and neural network are established to predict the controller ’s workload. The forecasting results are verified by the pilot/controller voice communication data. It shows that the combination model of ridge re