基于DAGSVM的居民出行方式分类研究  

Research on resident travel mode classification based on DAGSVM

作　　者：康忆宁 张登银[1] KANG Yining;ZHANG Dengyin(College of Internet of Things,Nanjing University of Posts and Telecommunications,Nanjing 210003,China)

机构地区：[1]南京邮电大学物联网学院,南京210003

出　　处：《智能计算机与应用》2025年第3期24-32,共9页Intelligent Computer and Applications

基　　金：国家自然科学基金(61872423);江苏省研究生科研创新计划(KYCX23_1054)。

摘　　要：采用移动蜂窝信令数据识别居民出行方式对于规划交通方案、制定交通策略具有十分重要的意义,然而目前大多数研究方法未考虑不同输入组合特征对模型性能的影响,从而导致识别精度不佳。本文提出一种基于DAGSVM(Directed Acyclic Graph Support Vector Machine)的居民出行方式识别方法。首先,采用基于网格的预处理算法对原始数据进行筛选,提取出信令数据的出行特征,并结合K-means聚类特征和隶属度函数特征获得信令用户出行的多维特征数据集;其次,基于上述特征建立用户轨迹的多维度特征集,利用DAGSVM模型评估不同场景下不同特征组合对出行模式识别的影响。仿真结果表明,在输入最优特征组合时,未区分时段的准确率为88.4%,本文方法在高峰时段准确率约为89.85%,在非高峰时段准确率约为90.45%,识别精度得以提升。The use of mobile cellular signaling data to identify resident travel modes is of great significance for planning traffic plans and formulating traffic strategies.However,most current research methods do not consider the impact of different input combination features on model performance,which limits recognition accuracy.This paper proposes a resident travel mode recognition method based on Direct Acyclic Graph Support Vector Machine(DAGSVM).Firstly,a grid based preprocessing algorithm is used to filter the original data,extract the travel features of the signaling data,and combine K-means clustering features and membership function features to obtain a multidimensional feature dataset of the signaling user's travel;Secondly,a multi-dimensional feature set of user trajectories is established based on the above features,and the DAGSVM model is used to evaluate the impact of different feature combinations on travel pattern recognition in different scenarios.The simulation results show that when the optimal feature combination is input,the accuracy of the undifferentiated period is 88.4%,and the accuracy of the proposed method is about 89.85%in peak period and 90.45%in off-peak period,which improves the recognition accuracy.

关 键 词：城市交通 蜂窝信令数据 DAGSVM 出行方式识别 多分类器 

分 类 号：U495[交通运输工程—交通运输规划与管理]