事件检测支持向量机模型与神经网络模型比较

针对交通领域中的事件检测(无事件模式和有事件模式)模式识别问题,描述了支持向量机(SVM)的基本方法,建立了基于线性(linearfunction)、多项式(polynomialfunction)和径向基(radialbasisfunction)3种核函数的事件检测SVM模型,并与PNN、MLF模型进行了理论比较。采用I-880线圈数据集和事件数据集建立并验证SVM、PNN和MLF模型,结果发现:无论对于向北、向南或混合方向的事件检测,SVM模型的检测率(DR)、误报率(FAR)和平均检测时间(MTTD)指标均比MLF模型好;PNN模型的DR比SVM(P)模型的高,但FAR和MTTD指标不比SVM(P)模型好;在3个SVM模型中,SVM(P)检测效果最好,SVM(L)最差。SVM算法与神经网络算法相比具有避免局部最小,实现全局最优化,更好的泛化效果的优点,是高速公路事件检测的一种很有潜力的算法。

Comparison of SVM and Neural Network Model for Incident Detection

This paper investigates the fundamentals of support vector machine(SVM) in traffic incident detection,presents three SVM model with different linear,polynomial and radial basis function and compare with PNN,MLF models.Comparison of SVM,PNN and MLF simulation results by I-880 real field database show that DR,FAR,MTTD is achieved by SVM are better than MLF;DR is achieved by PNN is better than SVM(P),but FAR,MTTD are inferior than SVM(P)whether in three direction freeway incident detection.SVM(P) is the best and SVM(L) is the worse among three SVM models.SVM is an effective algorithm in incident detection and superior to neural network in global optimization and good generalization ability.