An Improved Apriori Algorithm Based on Features

Abstract

In the traditional Apriori algorithm, all the database transaction items are equally important. However, in fact, in order to discover more reasonable association rules, different items should be given different importance. In this paper, an improved algorithm based on Apriori algorithm is proposed, in which every transaction item has its own feature(s) to carry more information. With adding feature(s) to these items, when mining the association rules, just those transaction data with same feature(s) will be scanned and computed. Studies and analysis in book recommendation system show that it takes less time cost and gets more reasonable association rules by using the improved algorithm.