Article Preview
Top1.1 Introduction
Association Rule Mining (ARM) is one of the most useful and well known technique of data mining Han J. (2011). It is used to extracts frequent patterns, associations, correlations, or causal structures among sets of items from given datasets. Datasets can be considered as transactional databases, relational databases and other information repositories. Formally, association rule mining problem is explained as follows: Let,
T be m number of transactions or set of records like {t1, t2, t3,…………, tm} from given datasets.
I be a set of n different items or attributes {i1, i2, i3, ……..…, in},
An association rule is an implication of the form 
where 
, 
, and 
. The itemset X is called antecedent part (left side) while the itemset Y is called consequent part (right side) and the rule means X implies Y.
Association Rule Mining is related to finding a set of rules considering a large percentage of data and it tends to generate a useful number of rules. However, since the number of transactions are increasingly more, the user no longer looks for all the possible rules but user looks to determine only a subset of important rules. To measure usefulness of association rules, mainly two basic parameters are used, one is namely support of a rule and second is confidence of rule. The support of an itemset 
is the number of records containing I¢. The support of a rule 
is the support of 
and the confidence of a rule is 
. An association rule 
with a confidence of 70% means that 70% of the transactions that contain X also contain Y together. So, the association rule mining is to find important rules which are having support ≥ MinSup and confidence ≥ MinConf Agrawal (1996). Here, MinSup and MinConf are two threshold predefined by users.
However, Association Rule Mining is difficult task when it applies on large scale datasets. The classical methods already developed to solve this problem, like Apriori, Agarwal (1993), FPgrowth, Han J (2000) algorithms, but it requires more CPU time when handling such data. To deal with this problem in reasonable time many higher level algorithms and procedures are applied to ARM. Some methods are based on evolutionary algorithms like Genetic Algorithm and some are based on swarm intelligence.