Hybrid Intelligent Method for Association Rules Mining Using Multiple Strategies

Related Works

Many algorithms for generating association rules have been proposed in literature. Some well known exacts algorithms are AIS Agrawal and Imielinski (1993), Apriori, Agrawal, R. and Ramakrishan (2004), Eclat Zaki (2000) and FP-Growth Han, J., Pei, J.(2004). AIS is very space consuming and requires too many passes over the whole database. Apriori is the best known algorithm for association rules mining. It is based on breadth first search strategy to count the supports of itemsets and uses a candidate generation function to exploit the downward closure property of support. FP-growth uses a FP-tree structure to compress the database and a divide-and-conquer approach, to decompose the mining tasks and the database as well. In Zaki,M J. (1999) the authors present an interesting survey about different exact and polynomial algorithms. However, because of the fast web development and growth of databases, they have become very quickly inefficient. Indeed, even if these polynomial algorithms can still calculate the association rule in a very short time, they remain limited face to our goal which is to extract all the rules from large databases in real-time.