城市道路交通运行态势大数据分析关键技术研究

Focusing on the theory and technology requirements of the management and strategic decision of the urban road traffic, the quality of the multi-source data can be controlled by cleaning the data sources, compensating the data lack and enhancing the sparse data. To the system state observation and travel analysis, the applicability of the data is analyzed and the multi-source heterogeneous traffic big data is fused in classifications. By the feature extraction method of individual and local network, the spatial and temporal dimensions can be reduced to realize the complex network feature extraction. Based on the intelligent algorithms such as deep learning, the level forecast of data driven traffic system and the evolution are analyzed. Through researching on the operational management of road traffic system in big data, the risk of system operation based on the characteristics and evolution of urban road traffic operation is assessed. The collaborative mining method of multi-source spatial and temporal data based on collaborative filtering and probability graph analysis is researched for the local data sparsity of the location data space. At the same time, the visual expression method of urban road traffic information for management and decision is realized. Based on this, the analysis framework of urban road traffic operation management and decision based on big data is explored, and the empirical test analysis is carried out, which provides the basic theory and key technical support for the reasonable operation and sustainable development of modern urban road traffic.

立足我国城市道路交通运行管理及战略决策理论与技术需求，通过对数据资源的清洗、缺失补偿、稀疏数据增强等实现多源数据的质量控制；面向系统状态观测与出行分析进行数据的 适用性分析和抽取进行多源异构交通大数据分类融合；通过个体和局部网络特征提取方法进行 基于时空尺度降维，实现对复杂网络特征提取；基于深度学习等智能算法进行大数据驱动的交 通系统层级预测分析和运行演化分析；研究大数据条件下道路交通系统运行管理知识发现，基 于城市道路交通运行特征和演化规律进行系统运行风险评估，并针对位置数据空间局部数据稀 疏性研究基于协同过滤和概率图分析的多源时空数据协同挖掘方法，同时实现面向管理与决策 的城市道路交通信息可视化表达方法。在此基础上探索基于大数据的城市道路交通运行管理与决策分析框架，并开展典型城市的实证测试分析，为我国现代城市道路交通合理运行和可持续发展提供基础理论和关键技术支撑。

课题立足我国城市道路交通运行管理及战略决策理论与技术需求。首先研究多源数据质量控制与融合感知，提出了一种基于参数估计的出行轨迹等数据重构技术；其次面向城市交通出行的复杂网络特征提取，建立一套基于手机基站数据的出行需求定量分析方法；最后基于深度学习等智能算法进行大数据驱动的交通系统层级预测分析和运行演化分析，提出了一种城市公共交通运营需求辨识和出行效率评价方法。课题的研究成果将为我国现代城市道路交通合理运行和可持续发展提供基础理论和关键技术支撑。