摘要
在粗糙集合理论中 ,近似空间与概念格之间存在着有趣的对应关系 ,利用概念格研究知识的约简和发现 ,更直观和有效 .但已有的概念格模型是基于近似空间的等价类划分的 .等价类划分过于苛刻 ,扩展的基于容差关系的近似空间具有更广泛的意义 ,但目前未见有相应的格模型被提出 .该文提出了容差近似空间的一种格模型 ,称为广义概念格 ,给出了定义 ,描述了建立方法和由它产生规则的原则 ,讨论了空间复杂性问题 ,并且与其它相近方法做了比较 .
There is an interesting relationship between approximate space and concept lattice in the rough set theory. It is more intuitional and more effective to research on reducing and discovering knowledge through concept lattice. But the existing concept lattice models are based on partition of equivalent classes in approximate space. Partition of equivalent classes is too rigorous. Expanded approximate space based on tolerance relation is significant in wider application, but no lattice model for the space was proposed. The paper suggests a lattice model corresponding to tolerance approximate space, known as generalized concept lattice, gives out its definition, describes the method to build it and the principles to produce rules from it, discusses the problem of its space complexity, and compares it with other similar or correlative methods. For a node of generalized concept lattice, its intension is a set of tolerance relation identifiers and its extension is a set of non order pairs among objects. According to the inclusion relation of extension between nodes, the nodes construct generalized concept lattice. An example is used to show the building of the lattice and the producing of rules from the lattice. For approximate space based on partition of equivalent classes, either generalized concept lattice or general concept lattice can be built, but the description fashions of the rules from the two kinds of lattices are different. The rules from the former describe the tolerance relation between attributes, and the rules from the latter describe the relation between attribute values.
出处
《计算机学报》
EI
CSCD
北大核心
2000年第1期66-70,共5页
Chinese Journal of Computers
基金
国家自然科学基金!( 69985 0 0 4)
高等学校博士学科点专项科研基金!( 970 3 5 90 1)
关键词
概念格
容差近似空间
知识发现
粗糙集合理论
rough sets, concept lattice, tolerance approximate space, knowledge discovery
