Research Of Weighted Negative Association Rules Mining Technology

Association rules mining is a form of datamining and it has attracted much attention from researehers. The related knowledge of data mining is introduced, including conception and classcical algorithm, especially that of the association rules.However, traditional Algorithm Apriori treats each item in the database as uniformity. But in the real world, the importance of each item is often different. Therefor, the mined association rules though the past algorithms may be not interesting for the users. The weighed association rules datamining is proposed to solve the above problem.On the other hand, the users are interested in low frequent and strong correlate rules, which are nagative association rules. The positive and negative rules are mined simultaneously based on mining for weighted association rules, so many contrary and uninteresting rules are mined. On the condition, the third measure is added to the framework of"support-confidence"to elimate the contrary rules. The algorithm of mining the positive and negative weighted association rules based on correlation is proposed; the algorithm of mining the positive and negative association rules based on interest is proposed; the algorithm of mining the positive and negative association rules based on Chi-square is proposed.But there is another case that the frequency of every item is different from each other, the above algorithms may be not effecitive. If occurrence frequence of the items is low, then the weighted support will be very low. In that case, the few rules can be mined. The model of multipal support is proposed in the paper. In the model, every item has different support constraint, thus, the rules can be mined more effectively.Association rules are mined from the frequent itemsets in the past. But the negative rules are consisted in the infrequent itemsets. The algorithm generation of infrequent itemsets and negative rules is proposed and the correlation measure is used to delete the contrary rules.The above theory of the data mining is applied to discovering association rules and the corresponding algorithms are designed. It is improved that the algorithms are feasiable and effective though analyzing the theory and simulating experiments.