Font Size: a A A

Study On Algorithm Of Generalizing Association Rules With Negative Items

Posted on:2007-12-16Degree:MasterType:Thesis
Country:ChinaCandidate:Y PengFull Text:PDF
GTID:2178360185474875Subject:Computer system architecture
Abstract/Summary:PDF Full Text Request
In recent years, with the popularization of computers and Internet, as well as the development of database technology, enormous data has been accumulated in various areas of application. Data mining has been one of the most active areas of current research, which can reveal hidden useful information through analysising and understanding these data. Asociation rules mining is one of the important data mining models, which has important theoretical and broad application prospects.On the relevance of data, association rules have positive or negative association rules. At present, positive association rules have been widely concerned, but association rules with negative attributes or negative items are not given sufficient attentiones. However, in many areas of application, the negative factors are also very important source of information, it is necessary to study the correlation between negative attributes of things.The paper brings forward the generalizing negative association rules, based on changing the definition of traditional positive association rules and emerging negative association rules. Not only the existing algorithms of mining negative association rules and association rules with negative items are very few, but also they are essentially based upon the iterative algorithms of Apriori idea, which needs multiple times scanning data sets and generating large amounts of frequent candidate sets. Paper presentes a new algorithm of mining frequent item sets with negative items. The algorithm is based on the frequent pattern tree, which uses for reference a compressed storage data structure i.e. frequent pattern tree of FP_growth algorithm. It mines frequent item sets with negative items through extending frequent patterns on the tree. The algorithm is similar to the basic idea of FP_growth , therefore it has no need to scan the database repeatedly, and not the generation of large aounts of candidate sets advantages. In addition, comparing to direct using FP_growth algorithms, this algorithm has no need to expand negative item to original database, and construc or destruct additional data structures, which only make some changes on the original frequent pattern tree, so it has certain advantages in time and space costs. Experiments show that the algorithm has better efficiency than the existing similar mining algorithms and direct FP_growth algorithms.In addition, the paper puts forward a confidence property in reference Apriroi property, and introduces interest as the third parameter of associaton rules. Based on the confidence property and interest threshold, this paper uses for reference Apriori algorithm and makes a method for extracting out generalizing associaton rules with negative items from the frequent itemsets with negative items. In addition, the paper discusses the contradictions of generalizing associaton rules with negative items. Experimental results show that the improved algorithm that this paper presents is valid.
Keywords/Search Tags:Negative Item, Frequent Itemset, Frequent Panttern Tree, Interest
PDF Full Text Request
Related items