基于多分类GA-SVM的高速公路AID模型

智能检测系统已为高速公路交通事件检测提供了有效的途径.为了更加细致地了解高速公路交通运行状态,为突发事件的应急处理提供更加高效、可靠的决策支持, 将支持向量机两分类问题延伸到多分类上来.根据交通事件的发生过程,将其分为自由流状态,交通拥堵加剧状态,交通拥堵消散状态.采集VISSIM对交通事件各阶段进行仿真的原始数据集,运用主成分分析方法对交通输入特性进行降维处理,构建支持向量机多分类事件检测模型,最后用遗传算法选择支持向量机模型参数,获得了满意的检测效果.

Highway automatic incident detection based on multi-class classification and GA-SVM

The automatic incident detection system (AID) has provided an effective way to achieve the freeway incident detection. To have a good command of the traffic running state on highways and provide reliable efficient decision support in disposing an emergency, authors extend two-class classification and SVM to multi-class classification and SVM. According to the process of traffic incident, the traffic state is divided into three classes: the traffic state of free flow, the traffic state of increased congestion, the traffic state of dissipated congestion. This method researches on the VISSIM to obtain the original data set, and selects input traffic features through the principle component analyses (PCA) method, then constructs the multi-class classification and support vector machine (SVM) model, and applies the genetic algorithm (GA) to optimize the model parameters. Finally, the multi-class classification and SVM model with GA has obtained the satisfactory effect.