基于个体出行图谱的公共交通通勤行为辨别方法研究

不同公共交通类型乘客的出行特征存在显著差异，实现公共交通通勤乘客准确辨识，有助于获取精细化的公共交通出行特征，更好地满足不同类型乘客的出行需求.基于北京市公共交通刷卡和线站数据，对公共交通多源数据进行关联匹配并提取出行链.利用北京市连续1个月的公共交通刷卡出行数据，采用多层规划理论构建了个体出行知识图谱，提取了出行天数、出行空间均衡度等7类特征指标.通过RP调查获得乘客出行行为类别.以特征指标为输入，乘客分类为输出，构建了面向公共交通乘客分类的BP神经元网络模型.验证表明，模型平均分类精度为94.5%，Kappa系数为0.879.本文研究有助于准确识别不同类别的公共交通乘客，为优化公共交通运营及公共交通精准化服务提供支撑.

Public Transport Commuter Identification Based on Individual Travel Graph

To obtain elaborate travel characteristics and better meet travel demands for different public transport passengers, it is necessary to find ways identifying public transport commuter accurately. Based on public transport smart card transaction and network data, travel chain is obtained by data processing and matching. Taking travel behavior data of April, 2017 in Beijing, China, individual travel graph is constructed by adopting multi- layer planning theory. Seven feature indexes are extracted from individual travel graph and set as input for passenger classification model. Revealed Preference survey is conducted to collect travel behavior category attributes, which is the output of classification model. A back propagation neuron networks based public transport passenger classification model is constructed. Validation results indicate that the average classification accuracy and Kappa coefficient are 94.5% and 0.879, respectively. The study results contribute to identify public transport passengers of different types accurately and further support to optimize public transport operating and improve service level precisely.