Comparison of Three Mining Algorithms for Association Rules

data mining, association rule, AprioriTid, AprioriItemset, AprioriTidList

AprioriTid is a well-known algorithm for mining association rules. AprioriItemset is an algorithm developed at the authors' department. Comparison experiments of the two algorithms are described both for synthetic and real data.

AprioriTid is a well-known algorithm for mining association rules. AprioriItemset is an algorithm developed at the authors' department. Both algorithms differ especially in representation of information about large (frequent) itemsets present in transactions. Comparison experiments were performed both on synthetic and real data. The former indicated hopeful speed of AprioriItemset for worst-case experiments, but experiments on real data showed that real transactions are far from worst-case conditions. As a result, a modification of AprioriItemset called AprioriTidList was developed, which tries to employ positive features of both algorithms. Results of experiments indicate that this new algorithm could be a good candidate for mining association rules. A brief description of the algorithms with a special emphasis on essential differences between them and experiment results on synthetic and real data are presented in the paper.