基于集装箱运用数据的铁路集装箱装卸站聚类融合算法研究  

Clustering Algorithm for Railway Container Handling Stations Based on Container Operation Data

作　　者：王旭 祝凌曦[2] 李诗林 WANG Xu;ZHU Lingxi;LI Shilin(Container&Tarpaulin Management Department,China Railway Container Transport Corp.,Ltd,Beijing 100055,China;School of Traffic and Transportation,Beijing Jiaotong University,Beijing 100044,China)

机构地区：[1]中铁集装箱运输有限责任公司箱布管理部,北京100055 [2]北京交通大学交通运输学院,北京100044

出　　处：《铁道货运》2024年第6期33-44,共12页Railway Freight Transport

基　　金：中国国家铁路集团有限公司科技研究开发计划课题(K2019S010)。

摘　　要：为了在集装箱造箱质量评价中排除集装箱不同运用工况造成的影响,通过分析近10年涵盖了运输品类、办理站特性及装卸频次等多维度因素的集装箱运用和维修数据,针对铁路集装箱办理站在运用方面的差异性,构建了基于集装箱运用数据的装卸站聚类指标体系,并通过数据预处理技术,形成可用于聚类分析的数据集。采用多种聚类算法对处理后的数据集进行综合分析,以识别出具有相似集装箱运用特征的办理站群体。为克服不同算法可能导致的结果偏差,研究引入一种基于“投票原理”的算法融合策略,以消减算法间的依赖性。聚类结果的对比分析表明,融合后的聚类效果优于任何单一算法的效果,可为铁路集装箱维修综合分析提供基础数据。To eliminate the impact of different container applications on the quality evaluation of container manufacturing,this paper analyzed the container operation and maintenance data from the past 10 years that covered multiple factors such as transportation categories,station characteristics,and handling frequency.Based on the differences in the operation of railway container stations,a container cluster index system was constructed using container operation data.A dataset available for cluster analysis was formed through data preprocessing technology.Various clustering algorithms,such as Kmeans,EM,and Canopy,were utilized to comprehensively analyze the processed dataset to identify station groups with similar container operation characteristics.To overcome the potential bias of different algorithms,an algorithm fusion strategy based on the voting principle was introduced to reduce the dependence among algorithms.The results show that the fused cluster effect is superior to the effect of any single algorithm,which can provide basic data for the comprehensive analysis of railway container maintenance.

关 键 词：铁路 综合运输 集装箱维修 数据挖掘 装卸站分析 聚类算法 融合算法 

分 类 号：U294.3[交通运输工程—交通运输规划与管理] U169.3[交通运输工程—道路与铁道工程] TP181[自动化与计算机技术—控制理论与控制工程]