摘要
主要探讨支持向量机理论在路段行程时间预测中的应用。具体的方法是,首先将研究路段根据路段交通状态和车辆检测器设置情况进行分段,然后以前几个时段的各个小路段的交通流量、平均速度和车道占有率和整个路段的行程时间为输入,以下一时段的整个路段的行程时间为输出,选取高斯径向基函数作为核函数,建立了基于支持向量机的路段行程时间预测模型,从而探讨支持向量机在路段行程时间预测中的应用效果。最后,利用交通仿真软件的模拟数据进行验证,并与BP神经网络计算结果比较,计算结果的对比表明本文提出的方法预测效果更好。
Application of theory of support vector machine for forecasting link travel time is discussed. First, link under research is divided into several segments according to locations of traffic flow sensors. Then tratfic flows, velocities, lane occupancy rates of these link segments and link travel time in severa/preceding periods of time are set as input, link travel time in the next period of time is set as output, gauss radial basis function is chosen as kernel function, thus a forecasting model based on support vector machine is presented so as to discuss use of support vector machine for forecasting link travel time. Illustrative numerical tests are conducted using the simulation-based data, where support vector machine and BP neural network are adopted to forecast link travel time, and the errors of these two methods are examined. The result based on support vector machine is superior to that of BP neural network.
出处
《公路交通科技》
CAS
CSCD
北大核心
2007年第9期96-99,共4页
Journal of Highway and Transportation Research and Development
基金
国家自然科学基金资助项目(50578009)
国家重点基础研究发展计划资助项目(2006CB705500)
关键词
智能运输系统
行程时间预测
支持向量机
路段行程时间
人工智能
Intelligent Transport Systems
travel time forecasting
support vector machine
link travel time
artificial intelligence
