基于多层复杂网络时空关联跟踪的不完备交通大数据抗差处理研究

项目名称： 基于多层复杂网络时空关联跟踪的不完备交通大数据抗差处理研究

项目编号： No.61472113

项目类型： 面上项目

立项/批准年度： 2015

项目学科： 计算机科学学科

项目作者： 夏莹杰

作者单位： 浙江大学

项目金额： 78万元

中文摘要： 数据驱动型智能交通系统是当今智能交通系统（ITS）领域的研究热点。其中，数据质量是影响数据驱动型ITS研究的关键问题之一。在当前城市智能交通基础设施条件下，交通数据不完备性问题普遍存在，并随着交通大数据及其相关应用的普及而愈加显著。本项目针对交通大数据不完备性问题，提出利用多层复杂网络建模交通大数据，充分表达多维度交通数据及其时空关联性；设计对多层复杂网络中交通数据时空关联跟踪的算法及性能评估方法，实现利用完备数据推断不完备数据的状态概率分布；确定用于目标跟踪的完备数据时空关联尺度，结合交通大数据并行计算方法，最终实现面向全时段全路网采集的交通大数据的高效抗差处理。本项目拟解决的关键技术包括交通大数据建模、不完备数据推断算法设计和计算效率优化，它们从模型、算法、理论等不同层面支撑不完备交通数据抗差处理研究，并可为城市道路交通传感器部署提供有效的决策支持，因此具有重要的理论意义和应用价值。

中文关键词： 交通大数据；数据质量；海量数据管理；复杂网络；目标跟踪

英文摘要： Data-driven intelligent transportation systems (ITS) blazes a trend in current research of ITS. Data quality is one of the key issues which affect the research of data-driven ITS. Under the current condition of urban intelligent transportation infrastructures, the imperfection problem of traffic data becomes common, and more and more significant as the popularization of big traffic data and relevant applications. For the imperfection of big traffic data, this project presents a method of (1) utilizing multi-layer complex networks to build a data model to adequately show the multi-dimensional traffic information and spatio-temporal correlations; (2) designing an algorithm model and performance evaluation method to track the spatio-temporal correlations of traffic information in the multi-layer complex networks; (3) determining the spatio-temporal correlation scale of data which is used for object tracking; and using parallel computing to efficiently robust processing big traffic data which are collected in the whole road network at all times. The key techniques that this project aims to solve include the modeling of big traffic data, the designing of imperfect data inference algorithm and the optimization of computational efficiency. These techniques support to robustly process imperfect big traffic data from different aspects, such as model, algorithm, and theory. Meanwhile, they can also provide effective decision-making support for traffic sensor deployment in the urban road network. Therefore, this project has strong theoretical significance and application worthiness.

英文关键词： Big traffic data;Data quality;Massive data management;Complex network;Object tracking