基于改进谱聚类算法的交通区域划分方法

Traffic regional division method based on improved spectral clustering algorithm

针对传统谱聚类算法在路网划分时承载路网信息较少和聚类中心优化问题,提出一种基于改进谱聚类算法的城市路网划分算法。通过转移概率模拟交通路网动态运行特征,利用马尔可夫链对谱聚类相似图进行重构,增强相似图的健壮性,与遗传算法结合,通过遗传算法优化初始聚类中心,提高谱聚类全局寻优能力。实验结果表明,改进后的算法比基准算法具有较好的聚类效果,能够有效划分城市路网。

To solve the problems of the traditional spectral clustering algorithm in road network division,such as the lack of bea-ring road network information and the optimization of clustering center,an urban road network division algorithm based on the improved spectral clustering algorithm was proposed.The dynamic operation characteristics of the traffic network were simulated by means of transfer probability,and Markov chain was used to reconstruct the spectral clustering similarity graph,which enhanced the robustness of the similar graph.Combined with genetic algorithm,the initial clustering center was optimized using genetic algorithm to improve the global optimization ability of spectral clustering.Experimental results show that the improved spectral clustering algorithm has better clustering effects than the benchmark algorithm,and can effectively divide the urban road network.