Probability-based incremental association rule discovery using the normal approximation

Abstract

An incremental association rules mining is one of an association rule mining research topics which finds the relation between set of item in dynamic databases. As data grows up rapidly, the co-occurrence itemset which discovered in the previous mining may be changed and the association rule will be change consequently. Incremental association rule mining research attempts to maintain that rules. Probability-based algorithm, one of an incremental algorithm, applied the principle of Bernoulli trial to predict expected frequent itemsets for reducing collected border itemsets and a number of times to rescan the original database. However, the numerical problem will occur when the algorithm deals with a large database. To manipulate with this problem, the improved probability-based incremental association rule discovery using normal approximation to estimate the probability of occurrence of expected frequent itemset is introduced in this paper. In addition, the confidence interval is applied to ensure that the collecting of expected frequent itemsets is properly kept.