项目名称: 基于属性偏序模式发现原理的多维混合数据模式分类研究
项目编号: No.61273019
项目类型: 面上项目
立项/批准年度: 2013
项目学科: 自动化技术、计算机技术
项目作者: 洪文学
作者单位: 燕山大学
项目金额: 62万元
中文摘要: 模式分类是模式识别、数据挖掘、机器学习和信息融合等研究领域中重要而基础性问题。目前,定性与定量混合数据的模式分类是一个还没有得到充分研究的前沿性课题。本项目基于集合论和形式概念分析两种数学形式化描述方法,在构建分块形式背景属性偏序结构图的基础上,开展多维混合数据的模式分类数学原理的研究。研究主要内容有:构建表达混合数据模式发现过程所需要的多立方体模型(数据立方体、概念立方体和概念网络立方体)和立方体切片模型;混合数据形式背景优化方法;模式分类属性约简数学方法;分块形式背景属性偏序结构图生成的数学原理;基于属性偏序结构图和有监督聚类原理的模式发现方法;基于子类划分的模式分类数学定理;混合数据模式自动分类系统设计等。本项目原创性学术思想在已经开展的预实验中得到了验证。本项目工程与数学紧密结合、理论与工程应用研究结合特色鲜明,项目的开展与完成对解决混合数据模式分类研究中的科学问题具有重要意义。
中文关键词: 模式分类;多维混合数据;属性偏序;模式发现;形式概念分析
英文摘要: Pattern classification is the important and fundamental issue for pattern recognition, data mining, machine learning and information fusion research. At present, pattern classification of the qualitative and quantitative mixed-mode data is not yet well-researched cutting-edge topics. Based on the mathematical formal description in set theory and formal concept analysis and building the structure graphs of attribute partial order for formal contexts, study of mathematical principles for pattern classification of the multi-dimensional mixed data is to carry out in this project. Main study include: building the required cube models for pattern discovery of the mixed- data (data cube, the concept cube and the concept network cube) and the cube slicing models, the optimization methods for the mixed data formal context, pattern classification attribute reduction mathematical methods, mathematical principles of generation for a partial order structure graphs, pattern discovery methods in the attribute partial order structure graphs and the based on the principle of supervised clustering, division based on sub-class pattern classification mathematical theorems, pattern automatic classification system design for mixed-data application. The original academic thought has been verified in preliminary experiments. The proj
英文关键词: Pattern Classification;Multidimentional Heterogeneous Data;Attribute Partial order;Pattern Finding;Formal concept analysis