项目名称: 基于时空上下文数据的关联关系挖掘与推理技术研究
项目编号: No.61472149
项目类型: 面上项目
立项/批准年度: 2015
项目学科: 自动化技术、计算机技术
项目作者: 余辰
作者单位: 华中科技大学
项目金额: 83万元
中文摘要: 时空数据关联挖掘是对时空数据中非显性知识、时空关系等模式的自动提取。其在交通、生物、公共安全、气候、人口普查等领域等有广泛的用途。基于地理位置的时空上下文数据,是生活中最普通也是最重要的基础数据,同时也是一种复杂的异构环境应用。本项目基于多种时空轨迹数据,采用语义化的上下文描述模型,研究时空位置信息的关联关系推理方法与挖掘技术。拟从四个方面展开:1)多源时空数据的多维度表示理论与模型研究,建立时空数据及时空变化的统一语义描述模型;2)基于概率图模型的关联关系挖掘,从统计学角度阐明时空数据与时空变化的模式特征与关联关系;3)基于行为语义的上下文推理技术研究,根据挖掘出的行为模式特征,进行未知事件的推理预测;4)面向群体用户的空间关联模型研究,挖掘群体活动的行为模式以及对事件的演进路径进行分析。
中文关键词: 时空数据挖掘;上下文推理;多维度建模;概率图模型;事件演进
英文摘要: Spatial-temporal based big data mining aims to extract hidden knowledge and spatial-temporal relationships from human's daily context information. This technology is widely used in transportation, biological, public security, climate, population census and other fields. The location-based spatial-temporal data is the most common and important context in our lives, but also a complex, heterogeneous environments. This project will use semantic context description model to study the spatial and temporal reasoning method of location information associated with mining. There are four aspects proposed: 1 ) Multi-source and multi-dimensional representation theory and spatial-temporal data modeling, based on which we are trying to establish a unified semantic description of spatial- temporal data model. 2 ) The graph modeling, we try to clarify complex big data with temporal and spatial variation of pattern features and association from a point of time and space. 3 ) Behavior based semantic reasoning techniques, according to find out the patterns of behavior characteristics , predicting unknown events. 4 ) Group relationship mining study, learning group behavior activities patterns and user generated content techniques.
英文关键词: Spatial-temporal data mining;context reasoning;multi-dimensional modeling;probabilistic graphical models;gradual progress