Research of Mining Positive and Negative Weighted Association Rules Based on Chi-Squared Analysis

Recently, mining negative association rules has received some attention and proved to be useful. Several algorithms have been proposed. However, there are some questions with those algorithms, for example, misleading rules will occur when the positive and negative rules are mined simultaneously. The chi-squared test can avoid the problem in the paper because of the mature theory basis. It is based on the statistics. In addition, if the minimum support is low so that many redundant rules are generated; but it is likely to lose a lot of useful information once it is set too high. Therefore, the every item is set a weight because there is different importance between items. Thus, it also can avoid above two cases. The negative rules mining is associated with weight, an algorithm PNWC is proposed. The experiment results show that the strong association rules are mined and the misleading rules are pruned. It suggests that the algorithm is correct and efficient.