基于改进FCM均值聚类算法的城市道路状态判别方法

针对传统FCM算法与阈值法两种交通流状态判别方法在适用性上的不足,通过分析交通流数据的分布特征,以各状态数据的离散性变化差异作为参考进行状态的划分,在FCM算法的基础上加入历史先验数据与后验概率进行初始聚类中心的优化,并将传统欧氏距离替换为对多维度数据之间变化关系与空间分布更加敏感的马氏距离进行算法改良,使交通状态判别结果更加接近实际交通运行状况。并使用实验数据进行算法验证,验证了方法的稳定性与结果的有效性,判别结果与数据表现也更加接近。

Study on Urban Road Congestion Identification Based on An improved FCM Clustering Algorithm

Aiming at the deficiency of the traditional fuzzy c algorithm and the threshold method, the distribution characteristics of the traffic flow data are analyzed, and the difference of the variability of the state data is taken as the reference. The FCM Based on the algorithm, we add the historical a priori data and the posterior probability to optimize the initial clustering center, and replace the traditional Euclidean distance with the Mahalanobu distance which is more sensitive to the change relation and discrepancy between the multidimensional data. In the final result determination stage no longer use the integer to make the results closer to the actual traffic conditions. And the experimental data are used to verify the algorithm. The stability and validity of the method are verified, and the result of the discrimination is closer to the data than traditional FCM algorithm.