基于SFLA-FCM聚类的城市交通状态判别研究*

针对城市道路交通状态判别的问题，提出了一种混合蛙跳算法（SFLA）与模糊C-均值算法（FCM）相结合的SFLA-FCM聚类算法。SFLA是一种全新的后启发式群体进化算法，具有高效的计算性能和优良的全局搜索能力。SFLA-FCM使用SFLA的优化过程代替FCM的基于梯度下降的迭代过程，有效地避免了FCM对初值敏感及容易陷入局部极小的缺陷。将该算法用于城市交通流数据的聚类分析结果表明，与单一FCM聚类算法相比，SFLA-FCM聚类算法更准确，效果更佳，能够快速而有效地对城市交通流状况进行判别，为动态交通拥堵预警和交通诱导策略的制定提供依据。

Study of urban traffic state judgment based on SFLA-FCM clustering

In order to solve the problem of urban traffic state judgment,this paper proposed a novel traffic interval programming method(SFLA-FCM)based on SFLA and FCM.SFLA-FCM used SFLA to replace the iteration process of FCM based on the gradient descent and avoided the disadvantages of local optimality and initialization dependence.Then applied it to a set of real traffic flow data. The clustering result shows that this method enjoys good performance in fast and effective distinguishing urban traffic congestion and will contribute to dynamic traffic congestion warning and traffic guidance.