Study On Mining Association Rules With Negative Items And Quantitative Attributes

Association rule mining is an important research direction of data mining, which can explore valuable relations between different fields and attributes from data sets and has important theoretical value and broad application prospect.The theoretical foundation of association rules mining is systematically discussed in this paper. And then we categorize and analyze the existing association rules mining algorithms and mainly discuss quantitative association rules and Boolean association rule with negative items. Based on the above-mentioned study we fulfill two innovative tasks:1. A new approach of quantitative association rule mining based on frequent 2-itemsets is proposed. It can be applied not only for typical basket analysis, but also for those with quantitative attributes cases that the present basket analysis algorithm fails to achieve. For instance, it can be used for rules extracting of bundled decision problems in Merchandise Bundled Sales problems with quantitative.2. Based on the study of the quantitative association rules and Boolean association rules with negative items, the quantitative association rules with negative items is brought forward. We detailed analysis the problems causing by quantitative properties and negative items, and propose the corresponding solutions.First, quantitative attributes and its negative items are discussed. Secondly, because the vast infrequent itemsets can increase lots of boring rules, two_level_supports is brought in to impose restrictions on the generating of frequent itemsets and infrequent itemsets; In the generative process of the candidate itemsets, candidate_gen function is put forward, which is the new candidate itemsets generating function by modifying the original apriori_gen function, in order to prevent redundant itemsets caused by negative itemsets. The algorithm not only effectively injects negative items into quantitative association rules, but also mine interesting association rules from the infrequent itemsets.