Research On Distributed Association Rules Mining Based On Concept Lattices

Distributed association rules mining, emerging for sovling the association rules mining in distributed environment, is one of the most active topic in the KDD field, which is significant in theory and application. Concept lattice represents data and knowledge by the relations betweent the extension and intension of concept, and by the generalization/specialization relationship between concepts, which is suitable for the study of association rules mining. The maturity of framework of lattice, leading up to the rapidly increase of size of lattice along with the increase of database's size, impedes concept lattice's the application in KDD. Based on studying the concept lattice and its extension, the distributed mining of association rules based on pruned concept lattices was discussed in this dissertation.The main points of this dissertation are listed as following:(1) Based on studying the representation of data and knowledge based on concept, the pruned concept lattice model (PCL) is discussed in detail. The corresponding formation algorithm of PCL is given too.(2) Comparing the representation of frequent itemsets based on pruned concept lattice and Apriori-like model. The theory analysis shows the former is more concise and more exact in representation of association in original dataset.(3) Study the association rules mining from multiple data sources based on pruned concept lattices, and the corresponding algorithm UMPL is provided. The experimental evidence shows the validity of this algorithm.(4) The prototype system that is used to mining association rules based on concept lattice is implemented.