基于网络核密度的网约车上下客热点识别

为分析网约车上下客热点的时空分布特性，利用网约车订单数据，构建基于网络核密度估计的上下客热点识别模型，采用回归模型对热点进行聚类和分级。通过研究区域划定、数据清洗和筛选，引入以路网距离为度量的网络核密度估计方法，基于非均质网络方向延展和网络距离衰减效应，对工作日和非工作日的特征时段内网约车上下客点的核密度值进行估计。采用零膨胀负二项回归模型对核密度值进行聚类，识别出研究区域的热点路段分布及其等级。通过与平面核密度估计结果对比分析，本文提出的网络核密度估计方法体现了上下客热点在路段和交叉口的分布特点，表征了实际的交通需求与路网结构的关系。研究结论为优化城市网约车的运营管理、提高城市居民出行效率提供理论依据。

Passengers' Hot Spots Identification of Online Car-hailing Based on Network Kernel Density

In order to analyze temporal and spatial distribution characteristics of online car-hailing hot spots, data of trip orders is used to build a hot- spot identification model based on network kernel density estimation, in which a regression model is used to cluster and classify the hot spots. Through the delineation of the research area, data cleaning, and screening, the network kernel density estimation method measured by road network distance is introduced. Based on the heterogeneous network direction extension and network distance attenuation effect, the kernel density of pick-up and drop-off points at specific periods during working days and weekends are estimated. The zeroexpansion negative binomial regression model is used for kernel density clustering, and the distribution of hot road sections is identified and classified. By comparing with the results of kernel density estimation, the network kernel density estimation method reflects the distribution characteristics of passenger hot-spots on road sections and intersections, and characterizes the relationship between actual traffic demand and road network structure. The conclusions help to optimize the operation and management of online car-hailing and improve the travel efficiency of urban residents.