Mining Negative Association Rules Study Based On Negative Frequent Itemsets

In recent years, with bright and popularization of Internet and data storage technology development, many areas of database can reserve mass data, through utilizing data mining tools for analyzing and further understanding history data in the data, finding the useful knowledge behind data in computer field become the most active a research field. Obviously, among them association rule mining as one of these important fields, have very important value and very widely application prospect.Association rules are the correlation between data itemsets and other data itemsets, it contains positive association rules, negative correlation rules. At present, association rules research are given widespread attention, and the rules which includ positive and negative projects at the same time in negative association rules are not given enough attention. However, in many applications, the negative factors things is also very important information sources, therefore, it is necessary to study the relationship between negative attributes.Based on the traditional association rules and the progress of negative association rules, we puts forward the mixture patterns definition which contain positive and negative itemsets in association model right or left or left and rught. At present, the existing mining negative association rules are few, and these kinds of negative association rules algorithm essentially based on Apriori algorithm thought, besides,these methods need generally multiple scanning the candidate of frequent itemsets. This paper proposes a new method used to from frequent items, centralized mining negative frequent itemsets algorithm, namely e-NFIS algorithm. To get positive frequent itemsets, this algorithm uses FP_growth algorithm and FP-tree compressing storage data structure, then based on the principle of permutations formula for calculating out the negative frequent itemsets. As the basic principles, methods, it can be avoided to multiple scan databases and form a shorter candidate itemsets than others methods. The cost in time and space to current mostly methods on data mining algorithms have certain advantages. Experiments show that the algorithm are very efficient.In addition, at present, the problems in the existing research papers with positive and negative itemsets in one side of association model or two of association model are discussed by author in this paper. On the basis of the current methods, we make further analysis. Faceing to contrary association rules, we puts forward the filtering methods to effectively choose association rule. Paper discussed the contrary rules. At last,, this paper puts forward improvement method is correct and effective by examples.