Research On Key Technologies In Construction And Application Of Concept Lattice

Concept lattice is a newly prosperous knowledge representation model and is the core data structure of Formal Concept Analysis. Formal Concept Analysis emphasizes human cognition, offers a new way about knowledge representation which totally differs from traditional statistical method, and is becoming an important research field of artificial intelligence, and has been widely used in machine learning, data mining, information retrieve and other areas. This paper focuses on key technologies in the construction and application of concept lattice. The content of this paper includes:(1) Concept lattice incremental construction. Prove concept generation and insert place only relate to newly generated concepts instead of the entire concepts in the lattice. Put forward concept lattice incremental construction algorithm, and experiments show that new concepts are few and don't increase with the growth of the lattice.(2) Concept lattice union. Vertical union and horizontal union can be carried on sub-context, vice versa; vertical split and horizontal split can be carried on a context. Accordingly, sub-lattice of the sub-context can be united to create a new lattice which is affiliate to the context created by the sub-contexts. Both vertical union and horizontal union don't change the father-son relationships that already exist in the sub-lattice. Based on this fact and the definition of lattice union, put forward a union algorithm of concept lattices, which uses the sub-lattices'structure, only makes a few adjustment on the original sub-lattices. Compared with insert one sub-lattice's concepts into another sub-lattice one by one, experiments and analysis show efficiency is improved.(3) Concept lattice maintenance. Divide the changes of formal context into two types, one is object-attribute relation's add and delete, another is object or attribute's add and delete. And then study concept lattice maintenance that is caused by these two types of changes of formal context respectively. In the maintenance caused by object-attribute relations'add, put forward the term"father-son concept pair"to identify maintenance place, and to deal with relation adjustment. In the maintenance caused by object-attribute relations'delete, put that after identify maintenance place, generate father-son concept pair to take place the concepts which need to altered, and delete redundant concepts in father-son concept pair. In the maintenance caused by objects or attributes'delete, put forward the term"critical concept", and use it to adjust relationship between concepts. Finally, study maintenance issue which is caused by concept absence of concept lattice, and propose a method to carry out maintenance which adopts the completeness feature of concept lattice.(4) Concept lattice and association rule extraction. Firstly, generally discuss partial order on concept lattice, and intent reduction and association rule extraction. Discover and prove that formal context's association rule and its supplement formal context'association rule is inverse.Secondly, study association rules'incremental extraction on concept lattice, which fall into three types: caused by objects'add, caused by attribute'add and caused by object-attribute relations'add. Discover and prove association rules'update law. Then, study association rules'maintenance on concept lattice, which fall into three types: caused by objects'delete, caused by attribute'delete and caused by object-attribute relations'delete. Finally, study decision rule extraction on concept lattice, first establish concept lattice which include all condition attributes, then add decision attributes to found relationships between condition attributes and decision attributes, finally extract simplified decision rule based on intent reduction.In this paper, study key technologies in the construction and application of concept lattice, and get several valuable results. These results as well as their application and promotion in concept lattice expansion model will have a very good prospects and value.