基于出租车轨迹数据的城市居民出行特征挖掘  

Mining of Urban Residents Travel Behavior Characteristics Based on Taxi Trajectory Data

作　　者：罗丽朝 朱兴林 孙亮 姚亮 刘泓君 LUO Lichao;ZHU Xinglin;SUN Liang;YAO Liang;LIU Hongjun(College of Transportation&Logistics Engineering,Xinjiang Agricultural University,Urumqi,Xinjiang 830052,China;Urumqi Research Center of Urban Comprehensive Transportation Projects(Urumqi Construction Center of Rail Transit Projects),Urumqi 830000,China)

机构地区：[1]新疆农业大学交通与物流工程学院,乌鲁木齐830052 [2]乌鲁木齐市城市综合交通项目研究中心(乌鲁木齐市轨道交通项目建设中心),乌鲁木齐830000

出　　处：《交通工程》2023年第2期114-121,共8页Journal of Transportation Engineering

摘　　要：针对城市交通中出租车寻客与调度困难等问题,提出一种基于KAN N-DBSCAN聚类算法与BP神经网络的热点区域需求预测模型,通过挖掘居民出行聚集模式进而对需求量预测分析.首先在对乌鲁木齐市出租车GPS轨迹数据预处理的基础上,通过数理统计法分析居民出行行为规律与空间出行特征;其次建立基于自适应DBSCAN聚类算法的热点区域识别模型,优化了聚类中参数标定过程,实现了分时段载客热点区域的探测;最后构建BP神经网络进一步对热点区域的居民出行需求进行预测.研究表明:居民出行活动热点区域受不同时间段的影响,交通枢纽以及商业购物区是出行的高热度区域;通过BP神经网络与随机森林模型对比发现,BP神经网络更适用于进行热点区域的交通需求预测.Aiming at the problems of finding and dispatching taxis in urban traffic,this paper proposes a demand forecasting model for hotspot areas based on KAN N-DBSCAN clustering algorithm and BP neural network.Firstly,based on the preprocessing of the GPS trajectory data of taxis in Urumqi,this paper analyzes the residents travel behavior and spatial travel characteristics by using mathematical statistics.Secondly,a hotspot area identification model based on the adaptive DBSCAN clustering algorithm is established to optimize the clustering process.Finally,the BP neural network is used to further predict the travel demand of residents in hotspot areas.The research shows that the hotspot areas of residents'travel activities are affected by different time periods,and transportation hubs and commercial shopping areas are popular areas for travel in a day.Through the comparison of BP neural network and random forest model,it is found that BP neural network is more suitable for forecasting travel demand in hotspot areas.

关 键 词：出租车GPS 热点区域 DBSCAN聚类 需求预测 BP神经网络 

分 类 号：U121[交通运输工程]