基于因子分析和最小最大概率机的交通事件检测算法

为了进一步提高交通事件检测的精度与效率,在多角度构建事件检测初始交通变量的基础上,设计了1种基于因子分析和最小最大概率机的交通事件检测算法。通过分析交通事件上下游交通流参数的变化规律,构建了11种初始交通事件检测变量,利用因子分析方法对初始交通变量进行特征提取,实现初始交通变量的有效降维,并分别采用核函数最小最大概率机算法和线性最小最大概率机算法进行交通事件检测。最后,采用美国I‐880数据库的实测数据进行实验验证和对比分析,实验结果表明,FA‐M PM算法较M PM算法事件检测率提高3.5%,误报率降低0.17%,平均检测事件减少了27.5s,且最小最大概率机算法的交通事件检测效果明显优于支持向量机算法和BP神经网络算法。 

An Algorithm of Automated Traffic Incident Detection Based on Factor Analysis and Minimax Probability Machine

In order to further improve the accuracy and efficiency of traffic incident detection ,this paper develops a traffic incident detection algorithm based on factor analysis (FA) and minimax probability machine (MPM) .The initial traffic variables were assigned from the multiple perspectives .This paper developed 11 initial traffic incident detection variables by analyzing the changes of upstream and downstream traffic flows .Factor analysis method is used for feature extraction ,and the dimension of initial traffic variables reduces effectively .The Kernel minimax probability machine and Linear minimax probability machine algorithm are used for traffic incident detection .The proposed method was tested with real traffic flow data from I‐880 database of USA .The experimental results demonstrate that the identification rate of FA‐M PM algorithm is 3 .5% higher than M PM algorithm ,the false identification rate decreases 0 .17% ,and the aver‐age detection time decreases 27 .5 s .The study concludes that MPM algorithm will provide better traffic incident detection than support vector machine (SVM) algorithm and BP neural network (BPNN) algorithm .