摘要
针对短时交通流在线预测时存在的计算复杂性问题,提出了一种最小二乘支持向量机在线式短时交通流预测方法。该方法简化了在线学习过程中Lagrange乘子的求解过程,利用训练数据集滑动时间窗口的移动来控制新样本的加入和旧样本的移除,通过线性运算完成Lagrange乘子的更新,进而完成预测模型的在线更新。测试结果表明,相对已有方法,所提方法在保证预测精度的条件下,能够将在线模型更新时间平均降低约62.64%,是一种有效的在线式短时交通流预测方法。
Aiming at the computational complexity of the online prediction of short time traffic flow,this paper proposed a least square support vector machine online short term traffic flow prediction method based on sliding time window.The method simplified the solving process of Lagrange multiplier in the process of online learning,and controlled the new samples to join and remove old samples by the movement of the sliding time window of the training data set.Furthermore,the method could use a few linear operations to update the Lagrange multiplier,and then completed the online updating of the prediction model.The test results indicate that,compared with the existing algorithms,the proposed algorithm can reduce the average time of online model updating about 62.64%under the conditions of guaranteed the accuracy of the prediction,and it is an effective online short traffic flow prediction method.
作者
康军
段宗涛
唐蕾
温兴超
Kang Jun;Duan Zongtao;Tang Lei;Wen Xingchao(School of Information Engneering,Chang’an University,Xi’an 710064,China;Shaanxi Road Traffic Intelligent Detection&Equipment Engineering Research Center,Xi’an 710064,China)
出处
《计算机应用研究》
CSCD
北大核心
2018年第10期2965-2968,共4页
Application Research of Computers
基金
国家自然科学基金资助项目(61303041)
国家交通运输部基础研究项目(2014319812150)
陕西省工业科技攻关项目(2014K05-28
2015GY002
2016GY-078)
关键词
短时交通流预测
统计学习
最小二乘支持向量机
在线式学习算法
滑动时间窗口
short term traffic flow prediction
statistical learning
least square-support vector machine(LS-SVM)
online learning algorithm
sliding time window
