Improvement in Execution Time of Apriori Algorithm

Abstract

Data mining which is also known as Knowledge Discovery in the databases (KDD) is an important research area in today’s time. One of the important techniques in data mining is frequent pattern discovery. Finding co-occurrence relationships between items is the focus of this technique. The active research topic for KDD is association rule mining and many algorithms have been developed on this. This algorithm is used for finding associations in the item-sets. Its application areas include medicine, World Wide Web, telecommunication and many more. Efficiency has been an issue of concern for many years in mining association rules. Till date the researchers of data mining have worked a lot on improving the quality of association rule mining and have succeeded to a great extent. There are many algorithms for mining association rules. Apriori algorithm is the mostly used algorithm which is used to determine the item-sets, which are frequent, from a large database. It extracts the association rules which in turn are used for knowledge discovery. Apriori is based on the approach of finding useful patterns from various datasets. There are lot many other algorithms that are used from association rule mining and are based on Apriori algorithm. Although it is a traditional approach, it still has many shortcomings. It suffers from the deficiency of unnecessary scans of the database while looking for frequent item-sets as there is frequent generation of candidate item-sets that are not required. Also there are sub item-sets generated which are redundant and algorithm involves repetitive searching in the database. This work has been done to reduce the redundant generation of sets. The large dataset is scanned only once. As a result the overall time of execution is reduced. Also the number of transactions to be scanned are reduced.