项目名称: 不确定性数据流自适应聚类分析及演化分析方法研究
项目编号: No.61202274
项目类型: 青年科学基金项目
立项/批准年度: 2013
项目学科: 计算机科学学科
项目作者: 杨悦
作者单位: 哈尔滨工程大学
项目金额: 25万元
中文摘要: 为迎合当前军事、经济、电信及科学计算等关键领域对其广泛涌现的不确定性数据流数据分析处理的强烈需求,本课题研究适合的不确定性数据流聚类分析及聚类演化分析方法,具有重要的理论价值和现实意义。课题开创性地研究基于Shannon熵超椭球体指标的不确定性度量方法,实现数据不确定性程度的量化描述,并在此基础上构建新的不确定性数据流模型,继而研究该模型下基于不确定加权密度树的不确定性数据流网格密度自适应聚类分析方法;同时研究基于聚类分布密度指标的不确定性数据流聚类有效性评价方法,并利用密度指数直方图实现最佳聚类数目的确定;课题创新性地研究基于灰度马尔可夫预测模型的聚类演化分析技术,实现不确定性数据流聚类演化过程的实时追踪与分析。本课题旨在通过以上研究,实现对不确定性数据流这一新兴重要数据形式进行高效准确的聚类分析处理,并通过聚类演化分析为不确定性数据流事件检测方法研究提供新的思路和探索。
中文关键词: 聚类分析;不确定性数据流;不确定性度量;衰减窗口;直方图
英文摘要: The research of clustering analysis and clustering evolution analysis method on uncertain data stream in this project is conducted in order to meet the data analysis strong demand on uncertain data stream which widespreadly emerge as the new data form in areas of economic, military, telecommunications and computing science, etc. This research project has important theoretical value and practical significance. The new uncertainty measure method based on Shannon entropy ellipsoid index multi-dimensional expanding will be researched to describe data uncertainty degree quantitatively. The uncertain data stream will construct new model under the uncertainty degree. Then we will research the method of uncertain data stream self-adaptive clustering method based on grid-density under the uncertain weighted density tree structure. At the same time, the clustering validity evaluation method will be researched based on clustering distribution density index which is organically combine of cluster compactness, inter-cluster separation degree and uncertainty. And determine the optimal number of clustering by density index histogram. This project will research clustering evolution analysis method innovatively based on gray Markov prediction model to tracking and analysis the clustering evolution process of uncertain data strea
英文关键词: Clustering Analysis;Uncertain data Stream;Uncertainty Measurement;Attenuation Window;Histogram