CAARMSAD: Combinatorial Approach of Association Rule Mining for Sparsely Associated Data Bases

Classification based on association rule mining has now become a very powerful technique of data mining for extraction of accurate knowledge from very large databases. First it finds all association rules in a training database then uses these rules to classify an unknown object of test database. Support_threshold and confidence_threshold are used to specify the user's interests. This paper proposes an algorithm CAARMSAD to mine the association rules in sparsely associated databases by applying minimum efforts. It uses the concepts of combinatorial mathematics to generate only those candidate itemsets whose sizes are up to the probable size of maximal frequent itemsets. Since CAARMSAD reduces the large number of candidate itemsets generations, it is very efficient for association rule mining in sparsely associated databases