城市公共交通具有降低能源消耗、改善城市环境等优点，在交通资源紧张、交通问题突出的背景下，公交优先发展是减缓城市交通紧张现状、促进城市交通科学合理发展的关键路径已成为共识。然而一方面，传统的公共交通发展过程存在着被动适应的现象，使得公交难以满足日益增长的多样化出行和城市交通可持续发展的要求。另一方面，由于乘客在出行行为特征和空间位置上存在异质性，维持服务质量和乘客满意度对规划管理者和运营商来说是一个巨大的挑战。因此，本文将从细分公交市场的角度出发，深入分析居民公交出行特征，并面向居民出行需求对公交服务网络进行优化研究，主要的研究内容包括以下几个方面：

       首先，研究以厦门市公交运营数据为依托，对公交IC卡数据、公交车辆GPS数据、车辆运营数据和线路站点的地理信息数据等公交基础运行数据进行预处理，在此基础之上分别从常规公交线路上下行与车辆GPS匹配、上车站点推导、下车站点推导、换乘分析共四个方面展开计算和分析，为后续研究提供良好数据基础和支撑。

       其次，研究归纳了一种扩展性较强、可解释性良好的公交市场细分方法，将分析角度分为两个方面：（1）在空间维度内，以乘客家的位置为依据，运用Geohash算法对固定范围内的乘客进行搜索，将研究区域划分为若干个有限空间乘客群，每个群体内的乘客具有空间近邻性。（2）在非空间维度内，通过确定乘客出行行为特征变量集，采用Mini-Batch Kmeans聚类算法，对每个有限空间乘客群作出划分与归类，分析得到的不同子市场间则存在明显异质性，为挖掘具有稳定出行需求的乘客奠定基础。

然后，考虑需要将公交市场细分结果运用到公交服务网络的优化上，研究通过分析细分区域内人群的出行特征，提取出具有稳定出行规律的乘客，在确定其高频出行起讫点后，选取得到三个等级的稳定客流通道。由于这些通道由每个细分区域内的乘客通过高频位移而形成，因此可以为网络优化提供支撑。

       最后，研究根据识别出的稳定客流通道，运用遗传算法构建了由31条线路组成的高频快线网络，对现有公交网络进行补充性优化，较好地弥补了现有公交网络缺乏集散便利的、具有针对性和精确性服务线路的不足。与此同时，还提出了衔接高频快线网络的接续公交服务方案和舒适安全的接驳站点一体化服务空间布局，以更好地满足居民出行需求，完善公交的服务性和便利性。

关键词：空间-行为分析，公交市场细分，稳定客流通道，公交服务网络

ABSTRACT

Urban public transportation has the advantages of reducing energy consumption and improving the urban environment, etc. Under the background of transportation resource shortage and prominent transportation problems, it has become a consensus that the priority development of public transport is the critical path to alleviate the current situation of urban traffic tension and promote the scientific and rational development of urban transportation. However, on the one hand, there is a phenomenon of passive adaptation in the traditional development process of public transportation, making it difficult for public transport to meet the increasing diversified travel and sustainable development of urban transportation. On the other hand, due to the heterogeneity of passengers in travel behavior characteristics and spatial location, maintaining service quality and passenger satisfaction is a huge challenge for planning managers and operators. Therefore, this paper will start from the perspective of pulic transport market segmentation, deeply analyze the characteristics of residents' travel based on pulic transport, and conduct a research on optimizing the pulic transport service network for residents' travel demands. The main research contents include the following aspects:

First, based on the pulic transport operation data of Xiamen, pre-process the basic operation data such as bus IC card data, bus vehicle GPS data, vehicle operation data, and geographic information data of line stations. On this basis, the calculation and analysis are carried out from the following four aspects: the matching of the up and down lines of conventional bus with the vehicle GPS data, the derivation of boarding station, the derivation of getting-off station, and the transfer analysis, which provide a good data foundation and support for subsequent research.

Secondly, the research summarizes a pulic transport market segmentation method with strong expandability and good interpretability. The analysis point of view is divided into two aspects: (1) In the spatial dimension, based on the location of passenger's home, the Geohash algorithm is used to search passengers within a fixed range, so that the research area will be divided into several passenger groups in limited space, and the passengers in each group have spatial proximity. (2) In the non-spatial dimension, by determining the feature variable set of passenger travel behavior, the Mini-Batch Kmeans clustering algorithm is adopted to divide and categorize each passenger group in limited space, and there is obvious heterogeneity among the different sub-markets obtained through analysis, which lays a foundation for mining passengers with stable travel demand.

Then, considering the need to apply the pulic transport market segmentation results to the optimization of the pulic transport service network, the study extracts passengers with stable travel laws by analyzing the travel characteristics of the people in the subdivided area, and after determining their starting and ending points of high-frequency travel, selects three levels of stable passenger flow channels. Since these channels are formed by passengers in each subdivision area through high-frequency displacement, they can provide support for network optimization.

Finally, based on the identified stable passenger flow channels, a high-frequency express line network composed of 31 lines is constructed by genetic algorithm to supplement and optimize the existing public transport network, which better compensates for the lack of convenient, targeted and precise service lines in the existing public transportation network. At the same time, we also propose a continuous public transport service plan that connects to the high-frequency express line network and an integrated service space layout of comfortable and safe connection stations, in order to better meet the travel demands of residents, improve the serviceability and convenience of public transport.

Key Words：space-behavior analysis, public transport market segmentation, stable passenger flow channels, public transport service network