基于近邻传播聚类的铁路客运节点类别划分

Classification of railway passenger transport nodes based on affinity propagation cluster

为简化旅客列车开行方案优化编制问题,采用聚类法对铁路客运节点进行类别划分。选取2014年日均旅客发送量排名前100位客运节点的相关属性数据为实例,以铁路客运节点的类别划分依据作为属性变量,首先运用分层聚类中的凝聚法对属性变量进行聚类,然后根据简化的客运节点变量指标,运用近邻传播算法对客运节点样本进行聚类,并引用CH、KL、IGP三种聚类有效性指标对聚类结果加以分析。研究结果表明,将100个客运节点分为五个类别时,具有最好的聚类效果,可为旅客列车开行方案的设计奠定基础。

To simplify the problem of optimizing passenger line plan, this paper adopted clustering method to classify railwaypassenger transport nodes. The method selected 100 top passenger transport nodes in order of daily passenger dispatch volumesin 2014 as an example,and took classifying foundation of passenger transport nodes as property variables. Firstly,it clusteredproperty variables by hierarchical clustering, and clustered passenger transport nodes samples by affinity propagation algorithmsaccording to the simplified nodes indexes. Finally,it analyzed three clustering effectiveness indexes contained CH,KL and IGPindexes to the clustering consequence. The result shows that it is of the best effect while the 100 passenger transport nodes mentionedabove are divided into 5 sorts,and it lays a foundation for the design of passenger train plan.