Toward a Distinguishing Approach for Improving the Apriori Algorithm

Abstract

Association rule mining, one of the most important branches of data mining, which focused on detecting frequent patterns of itemsets. Apriori is the first algorithm proposed for association rule mining. This algorithm has the best response and can detect all frequent itemsets from transaction databases. Apriori is of time complexity order two to the power n at worst case, n is the number of items in the database. At each step, the database is scanned to detect frequent itemsets. As a result, this algorithm has a very large response time for large databases. There are two ways to reduce the response time of this algorithm. First, prune the itemsets which candidate for checking. Second, reduce the dimension of the database. We used the second solution and reduce the dimension of the database considering that if a set is frequent, all of its subsets are frequent with more frequencies in the database. In the proposed algorithm, database scanned one time, and then frequent itemsets are detected by the reduced database. Our algorithm improved an apriori response time. To evaluate the algorithm, precision and recall measures have been used. According to the experimental in most cases, the algorithm can provide precision and recall above ninety percent.