大数据驱动的城市交通流机理研究和语义挖掘

项目名称： 大数据驱动的城市交通流机理研究和语义挖掘

项目编号： No.61472087

项目类型： 面上项目

立项/批准年度： 2015

项目学科： 其他

项目作者： 杨夙

作者单位： 复旦大学

项目金额： 80万元

中文摘要： 当今碳排放、能源损耗、交通拥堵等大城市病困扰着我们，这与城市中大规模人群移动关系密切。本项目基于大数据时空关联分析进行城市交通流机理和语义挖掘的研究。包括：(1) 集群移动多源数据时空图谱关联分析；(2) 基于群体轨迹相似度的主要交通路线及脆弱性发现；（3）区域功能与交通流时空分布关联建模。特色：（1）由于人类对于自身活动规律、以及人类集群移动的社会动机和需求还知之甚少，大数据驱动的城市动力学机理研究、语义挖掘是一个新的领域。（2）城市中用户位置数据呈现多模态，如网络签到、GPS等，建立多模态数据的自动分析和理解的方法，从海量异构数据中自动挖掘出状态、模式、时空演化和关联规律，是城市大数据智能处理的基础科学问题之一。本项目结合了多种模态的数据进行研究，有可能更好地解决单一模态数据难以解决的问题。

中文关键词： 特征提取；特征选择；模式识别；数据挖掘；人工智能

英文摘要： Nowadays, people living in big cities are suffering from air pollution, energy consumption, and traffic jams. Those problems are caused by the large-scale human mobility in a city. This investigation aims to study the causal and semantic aspects regarding how and why traffic flows perform as such and reveal the spatial-temporal correlations from the prespective of big data. The main tasks are: (1) Correlation analysis on the spatial-temporal evoluatoin patterns of multimodal data. (2) Discovery of main routes and traffic bottlenecks from human trace similarity. (3) Region function discovery and spatial-temporal causal modeling of traffic in a mutual learning manner. The novelties lie in the following 2 points: (1) We have little knowledge about the city mobility patterns as well as the underlying social motivation and casue to result in such patterns. Studying it from the causal and semantic perspective is a new area so far. (2) The location data of people in a city are in a multimodal form such as GPS traces and check-in data on mobile internet. Development of theoretical framework to mine concepts, patterns, and correlations from such heterogeneous data should be the main focus of big dat science. In this study, we apply multimodal data in order to tackle the problem that may not be well solved based on the data from only one source.

英文关键词： Feature Extraction;Feature Selection;Pattern Recognition;Data Mining;Artificial Intelligence