nVApriori : A novel approach to avoid irrelevant rules in association rule mining using n-cross validation technique

Association rule mining finds interesting associations or correlationsin a large pool of transactions. Apriori based algorithms are two stepalgorithms for mining association rules from large datasets. They findthe frequent item sets from transactions as the first step and thenconstruct the association rules. Though these algorithms generatemultiple rules, most of the rules become irrelevant to the transactions.The exercise becomes costly in terms of memory usage and decisionmaking is also not precise. This research addresses this drawback bydeveloping ways to reduce irrelevant rules. This paper proposes the ncrossvalidation technique to filter such irrelevant rules. The proposedalgorithm is called nVApriori (n-cross Validation based Apriori)algorithm. The proposed nVApriori algorithm uses a partition basedapproach to support the association rule validations. The proposednVApriori algorithm has been tested with two synthetic datasets and tworeal datasets. The performance analysis is compared with Apriori, mostfrequent rule mining algorithm and non redundant rule miningalgorithm to study the efficiency. This proposed work aims at reducing alarge number of irrelevant rules and produces a new set of rules havinghigh levels of confidence.