面向时空路网疏散的群体行为态势挖掘与演化研究

项目名称： 面向时空路网疏散的群体行为态势挖掘与演化研究

项目编号： No.61305062

项目类型： 青年科学基金项目

立项/批准年度： 2014

项目学科： 自动化技术、计算机技术

项目作者： 郭丹

作者单位： 合肥工业大学

项目金额： 23万元

中文摘要： 突发事件伴随大量群体行为的时空变化信息，对城市路网疏散体系提出了严峻考验。针对现有疏散模型缺乏对来源数据的审核及实时交互，且过于强调数学求解理论最优的现状，缺乏对模型精准性、可靠性和实用性的关注。我们力求打破真实时空数据与路网疏散模型间的信息壁垒，在时间维、空间维、属性维上实现异构数据关联集成；研究基于时空关联规则提取群体行为模式的理论基础，分析群体事件发展规律与相关数据源参数的定性或定量关系；构建分层贝叶斯网络结构描述群体行为关联特征空间，采用主动扰动学习方法，发现疏散态势中脆弱性隐因素，进而探寻群体行为态势的演化规律和退化机理；最后，建立路径规划计算模型，运用态势知识反馈调节路网模型参数，制定动态路径规划方案实现群体行为最优，兼顾精确度和时空性能方面的要求，使大规模区域内动态疏散路径的精准优化规划成为可能。

中文关键词： 时空约束；关联规则挖掘；隐因素；态势演化；路网疏散

英文摘要： Accompanied by massive spatio-temporal data, evacuation has become a severe challenge of city emergency response systems. Current evacuation route planning lacks interaction with real-world data and most research overemphasize optimal solutions by mathematical modeling. The evacuation route planning problem has to achieve a balance among precision, reliability and applicability efficiency from the real-world data. Motivated by this, we aim to solve this problem by integrating real spatio-temporal data and road network on time, space and attribute dimensions. We adopt spatio-temporal association rules to mine aggregation association patterns online, and provide qualitative or quantitative analysis on relationship between the aggregate situation and spatio-temporal data. We build a hierarchical Bayesian network structure to describe aggregation behavior space. Based on an active intervention learning method, we explore vulnerable latent variables in evacuation situation. We also explore the aggregation situation evolution and degradation mechanism by probability structure model and related algorithms. Finally, we transfer aggregation situation knowledge to parameters in the route planning model. We will design heuristic algorithms to establish a dynamic route planning scheme that meets the requirements of precisio

英文关键词： Spatio-Temporal Constraints；Association rule mining；Latent Variable；Situation Evolution；Evacuation Route Planning