Algorithms Of Incremental Updating And Pruning Based On Weighted Negative Association Rules

Association rules mining is a hot direction of datamining and it has attracted much attention from researehers. Based on introducing weight value and negative association rules, we focuse on incremental updating and pruning of the weighted negative association rules discussion in this paper.Incremental updating of the weighted negative association rules means that rediscover the new relationships or rules from the changed database.But not run the original rule mining algorithm in the changed database. It is a process that not only use the prior knowledge effectively, but also can discover the new knowledge in time.This is because the need of a large number of repeated database scaning and the price of discovering new model becomes very costly in large databases.Maintenance of association rules also put forward a very high request in knowledge discovery system.In fact, the most complex and difficult problem in the rule maintenance is to find the new relationships. In this paper, a new kind of incremental updating algorithm of weighted negative association rules is presented for such efficient maintenance problem of association rules. It is improved that the algorithms are feasiable and effective though analyzing the theory and simulating experiments.In addition, when weighted value introduced in association rules mining, the downward closure property of the support measure in the unweighted case no longer exist and the previous algorithms cannot be applied. And there are excessive and disordered association rules generated by algorithms of weighted negative association rules that have been proposed, many of which are redundant, so that they are difficult for users to understand and make use of. In this paper, we proposed three times pruning algorithms of weighted negative association rules.The first pruning is in the item mining.The second pruning is in the association rules generating.And the third pruning remove the redundant rules effectively in the generated association rules. Experimental results show the number of weighted negative association rules have been reduced greatly.Through analysis and comparison of experimental results,the number of weighted negative association rules reduced a lot.