Dynamic Construction Of Concept Lattice And Its Application In Discovering Association Rules

Concept lattice is a concept hierarchical structure based on binary relations. It essentially describes the relations between objects and attributes, embodis the unification of intension and extension of concepts, and is an effective tool of data analysis and rule extraction. Association rules are used for discovering the relationship between different items in database. It is one of the most important research content in data mining.In this paper firstly it presents an algorithm for constructing concept lattice dynamically based on an in-depth analysis of algorithms of constructing concept lattice and mining association rules. Then, it detailedly expounds the application of concept lattice in mining association rules.In the algorithms of dynamically building concept lattice, when database changes and needs to insert a node, in order to find fathers of new nodes, the traditional algorithms needs compare all the nodes with them, so it is not convenient for the dynamic modification of concept lattice. This paper has made improvements: if a node has a sub-node that is a super father of the new generated node, then it is impossible that this node is the direct father of it. So the search process will be transferred to other nodes, no need to scan all its sub-nodes, thus it greatly reduces the number of nodes searched, and efficiently adapts to the dynamic changes in database. Through specific examples it analyses the performance of the algorithm, and verifies the correctness and stabilization of the algorithm by experiment.In mining association rules based on concept lattice, firstly it compares the algorithm of mining association rules based on quantitative concept lattice with Apriori algorithm, and shows the performance advantages on time and space of the first one. As little attention on importance of various attributes, we neglect users'attention of certain attribute in current process of mining association rules. This paper presents an algorithm for mining association rules based on weighted quantitative concept lattice, and verifies the correctness and validity through specific examples and careful analysis.