行为轨迹数据高性能时空聚类及社会分析

项目名称： 行为轨迹数据高性能时空聚类及社会分析

项目编号： No.41471326

项目类型： 面上项目

立项/批准年度： 2015

项目学科： 天文学、地球科学

项目作者： 秦昆

作者单位： 武汉大学

项目金额： 80万元

中文摘要： 与日俱增的浮动车数据(点)、GPS轨迹数据(线)，及街道网络数据等，将构成蕴含人们行为模式和活动规律的行为轨迹大数据，隐藏着具有强时空相关性的时空聚类模式，并要求进行高性能求解。行为轨迹数据是人类活动的记录，与社会学密切相关。本项目通过对行为轨迹数进行时空相关性分析，研究其时空聚类特性。将时空相关性统计与谱聚类、密度聚类等方法相结合，提出考虑时空相关性的时空聚类方法，挖掘时空聚类模式(客流热点区域、出租车接客模式、拥堵模式、出行模式等)。应用并行计算和弹性计算等策略，进行时空聚类方法的高性能求解。进一步结合社会调查数据，分析行为轨迹数据与社会因素(经济收入、受教育程度、工作状况等)的关系。本项目研究将为城市交通管理与社会管理提供基础，将在行为轨迹数据高性能时空聚类及社会分析方面产生创新性成果，并为促进地理信息科学的空间人文社会学、高性能地理计算、时空数据挖掘等方向的学科发展做出基础贡献。

中文关键词： 时空轨迹分析；空间聚类；时空聚类；知识发现；空间统计学

英文摘要： The increasing floating car data (point), GPS trajectory data (line), and the associated street network data have formed big geographic spatio-temporal data, which contain behavior patterns and rules, and hide some spatio-temporal clustering patterns with strong spatio-temporal correlations and require a high performance solution. Behavior trajectory data are the records of human activities, which are closely related to sociology. Based on the spatio-temporal correlation analysis about the behavior trajectory data, this project will study the spatio-temporal clustering characteristics of behavior trajectory data and will propose new spatio-temporal clustering algorithms by combining the spatio-temporal autocorrelation statistics and the clustering methods of spectral clustering and density-based clustering, and discover meaning patterns(behavior patterns, traffic hotspots, taxi pick up patterns, congestion mode, etc.) by using the proposed spatio-temporal clustering methods. This project will research the high performance computing problems of the spatio-temporal clustering by applying parallel computing and scalable computing, and further analyze the relationships between behavior trajectory and social factors (the income, educational level and the happiness index, etc.). This study will provide the basis for urban transportation management and social management, and will generate innovative research results in the high performance spatio-temporal clustering of behavior trajectory data and the correlated social analysis, which could make some fundamental contributions to the development of disciplines of geographic information science in the aspects of spatially integrated humanities and social science, high performance geographic computing, ant spatio-temporal data mining, etc.

英文关键词： spaio-temporal trajectory analysis;spatial clustering;spatio-temporal clustering;knowledge discovery;spatial statistics