基于个体出行特征图谱的多模式公共交通通勤出行需求辨识方法研究

Public transport of the demand response type is the important way to meet the demand of public transit multi-level characteristics，however，the key problem of determining the different modes of passenger demand level still have not been solved in a positive way. The traditional traffic demand survey based on the small sample failed to descript the overall characteristic accurately, as well as the refinement of public transport demand in multi-modes. Huge amounts of public transit data have great advantages in large samples close to the overall level, and provide the setting of complete travel characteristics. Taking the commute transits as objects, the research will focus on solving two problems: (a) How to obtain public transit characters accurately based on the individual travel feature atlas under incomplete information; (b) How to identify the different types of public transit demand precisely based on the commute travel features. Based on the public transportation multi-source data, the research plans to extract the public transportation individual transit chain based on the individual travel feature atlas, by the fusion of the characteristics in time and space dimensions of the travelers. According to the characteristics indicators of commuting trips, the methods to identify the commute travelers will be discussed with semantic technology and machine learning methods. Guided by the system identification theory, the research will study the methods and comprehensive decision-making model for identifying the public transport demand in multi-level, based on the pattern of commuter travel characteristics. In brief, the research will contribute to identify the scale and structure of different public transit commuter demand in multi-level well and truly, which is helpful support in affording the scientific models in network planning, design and operation optimization of public transportation, and improve the public transport planning method system.

需求响应型公共交通是满足公交出行多层次需求的重要途径，如何确定不同模式的乘客需求水平是其中亟待解决的关键问题。以传统居民出行调查为基础的交通需求预测无法描述个体出行特征的差异，难以实现不同类型公共交通出行需求的精细化辨识。虽然公交海量数据带来了大规模样本，但经典统计学方法在不完全信息条件下的个体出行特征提取精度较低。研究拟重点解决基于特征图谱构建的不完全信息条件下个体出行特征提取方法和公共交通通勤出行判别方法两大科学问题。项目以公共交通海量多源数据为基础，通过公交出行数据的时空统合，构建基于个体出行特征图谱分析的公共交通出行链及其特征提取方法；研究基于语义分析和机器学习的公共交通通勤出行判别方法，建立不同类型的通勤需求图谱特征库；并以系统辨识理论为指导，构建多模式公共交通需求辨识方法和决策模型，实现对不同层次公交通勤需求规模的准确鉴别，从而为公共交通网络规划、设计和运营优化提供科学支撑。

随着大型城市规模扩展、人口剧增和职住不平衡等因素的综合作用，居民通勤出行已成为城市交通最为突出的问题之一。多模式公共交通是满足市民出行多层次需求的重要途径，如何在大数据环境下准确刻画公共交通通勤出行特征，如何根据公共交通通勤出行特征精确辨识不同类型的公共交通出行需求是其中亟待解决的关键问题。以传统居民出行调查为基础的交通出行分析无法准确、全面地描述个体出行特征的差异，难以实现不同类型公共交通出行需求的精细化辨识。虽然公共交通海量数据带来了大规模样本，但经典统计学方法在不完全信息条件下的个体出行特征提取精度较低，缺乏对于出行机理和个体出行特征的表达。研究拟重点解决基于特征图谱构建的不完全信息条件下个体出行特征提取方法和公共交通通勤出行判别方法两大科学问题。项目以海量多源的公共交通数据为基础，利用时空关联分析和数据匹配分析法，将不同类型数据源按照对象的时空位置属性进行关联统合，结合多层规划理论构建个体公共交通出行特征与知识图谱。引入公共交通出行链的概念，将个体出行数据按时间发生顺序组成完整的出行过程，利用时空关联匹配方法构建基于个体出行图谱的公共交通出行链及其特征提取方法。基于通勤者在出行时空和模式等方面的特征指标，利用多层BP神经网络模型构建公共交通通勤行为判别模型，模型的辨识精度为94.2%，Kappa系数为0.875，具有较高的群体类型识别精度。基于特征分析从出行时空分布角度选取5个指标刻画通勤者公共交通出行需求稳定性，利用FP-growth关联规则算法挖掘出不同项集长度下各特征属性间的7条关联规则，基于规则指标数值将公共交通通勤需求稳定性划分为高、中、低3类。研究为多元化公共交通网络规划、设计和运营优化提供科学支撑，为制定需求响应型公共交通服务提供了可靠的理论基础和科学依据。