基于支持向量机回归的短时交通流预测模型  

Short-Term Traffic Flow Forecasting Model Based on Support Vector Machine Regression

作　　者：傅贵[1] 韩国强[1] 逯峰 许子鑫 

机构地区：[1]华南理工大学计算机科学与工程学院,广东广州510640 [2]广州市交通管理科学技术研究所,广东广州510640

出　　处：《华南理工大学学报（自然科学版）》2013年第9期71-76,共6页Journal of South China University of Technology(Natural Science Edition)

基　　金：NSFC-广东省政府联合基金资助项目(U1035004);国家自然科学基金青年科学基金资助项目(61003270);广州市科技计划重点支撑项目(11A11080267);广东省计算科学重点实验室开放基金资助项目(201206005)

摘　　要：将交通流预测的理论和方法引入交通控制系统,可提高交通控制系统对交通流变化的自适应能力.为此,文中通过引入核函数把短时交通流预测问题转化为高维空间中的线性回归问题,提出了基于支持向量机回归的短时交通流预测模型,并利用广州市交通流检测系统的数据进行实验.结果表明,文中模型的预测结果与实际数据相吻合,预测误差小于基于卡尔曼滤波的预测方法,从而验证了该模型的可行性和有效性.As the short-term traffic flow forecasting theories and approaches help to improve the ability of traffic control systems to automatically adapt to traffic flow changes,this paper proposes a short-term traffic flow forecasting model based on the support vector machine regression by using a kernel function to transform the issues into a linear regression problem in Hilbert Space.Then,the corresponding experiments are conducted based on the data from the traffic flow detection systems in Guangzhou.It is found that the forecasted results accord well with the actual data,and that the forecasting error of the proposed model is less than those of the prediction methods based on Kalman filtering.Thus,the feasibility and effectiveness of the proposed model are verified.

关 键 词：交通控制 短时交通流 预测模型 机器学习 支持向量机回归 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]