位运算和核属性快速识别下的粗糙集属性约简算法研究  

Rough Set Attribute Reduction Algorithm Using Bit Arithmetic and Core Attributes Quick Identification

作　　者：赵洁[1] 梁俊杰[2] 董振宁[1] 唐德玉[3] 

机构地区：[1]广东工业大学,广州510520 [2]华南理工大学,广州510006 [3]广东药学院信息工程学院,广州510006

出　　处：《小型微型计算机系统》2015年第2期316-321,共6页Journal of Chinese Computer Systems

基　　金：国家自然科学基金项目(70971027;71401045)资助;教育部人文社会科学研究青年基金项目(10YJCZH234)资助;广东省自然科学博士启动基金项目(S2011040004285;S2012040007439;S2012040007286)资助

摘　　要：首先基于多个位向量算法和改进的数组键哈希表,设计快速正区域算法;之后深入研究全局正区域与核属性之间关系,提出并证明核属性的识别特征优化求核过程,设计求核算法,从而快速识别核属性;并在最大增益属性算法选择属性的过程中通过舍弃已识别属性不断缩减搜索空间从而提高效率;最后基于上述子算法得到高效属性约简算法.通过UCI中16个决策集验证算法是正确、高效、完备的,且其效率不受实体数量、属性多少影响,尤其适合处理大型决策表.Firstly,based on multiple algorithms of bit vector and improved Hash array keys, basic algorithms such as positive region. Thenthe relationship between global positive regionwith core attributes is studied to propose and prove effective pruning theory, so that thecore attributes seeking algorithm is designed to identify core attributes quickly. Using attributes gain to measure attributes＇ impor- tance and the search space is kept reducing so the efficient attributes reduction algorithm is designed. Through 16 decision sets of UCI, the algorithms of this paper are proved to be correct, efficient and complete. No matter the number of entities and attributes are more or less, algorithms of this paper show good performance and particularly suitable for processing large decision table.

关 键 词：约简 核属性 粗糙集 正区域 位运算 

分 类 号：TP311[自动化与计算机技术—计算机软件与理论]