Font Size: a A A

Study On Associative Classification Based On Closed Frequent Itemsets

Posted on:2010-10-24Degree:MasterType:Thesis
Country:ChinaCandidate:D X QinFull Text:PDF
GTID:2178360278460292Subject:Computer software and theory
Abstract/Summary:PDF Full Text Request
With the rapidly development of information industrialization, the application of database deepening quickly,Data Mining has recently become the hotspot. Of each branches of data mining, association and classification are two very active fields with broad application. And there are some comparability between association and classification to find strongly relative item sets. So a new classification method—classification based on association was proposed. However, the approach also suffers from one major deficiency: a training data set often generates a huge set of rules, many of which are redundant. It is challenging to build and use the classifier.And a new algorithm ACCF: classification based on frequent closed itemsets is advanced in this paper.Firstly, this paper explores main theory and algorithms of association rule and classification. Secondly, popular classification based on association algorithm CBA and CMAR are introduced. Thirdly, we discuss the concept of closed pattern and extend an effective closed frequent pattern mining method, CHARM. Then this put forwards and proposed the new algorithm ACCF: Associative Classification Based on Closed Frequent Itemsets.ACCF based on the concept of closed frequent pattern.Since the set of closed frequent itemsets can be orders of magnitude smaller than the complete set of frequent itemsets, and it can derive the whole set of frequent itemsets, also the association rules generated by frequent closed itemsets can derive whole association rules. The new algorithm also improves the old way to prune the redundant rules and the ancient method to classify an unseen case. Our extensive experiments on 18 databases form UCI machine learning database repository show that ACCF can generate a smaller set of classificative association rules, with higher quality and no redundancy and has better average classification accuracy in comparison with the traditional and representational algorithm,i.e.CBA.
Keywords/Search Tags:Association Rule Mining, Classification, Associative Classification, Frequent Itemset, Frequent Closed Itemset
PDF Full Text Request
Related items