Finding Cyclic Frequent Itemsets

Mining various types of association rules from supermarket datasets is an important data mining problem. One similar problem involves finding frequent itemsets and then deriving rules from frequent itemsets. The supermarket data is temporal. Considering time attributes in the supermarket dataset some association rules can be extracted which may hold for a small time interval and not throughout the data gathering period. Such rules are called as local association rules and corresponding frequent itemsets as locally frequent itemsets. Mahanta et al proposes an algorithm for extracting all locally frequent itemsets where each locally frequent itemset is associated with sequence time intervals in which it is frequent. The sequence of time intervals associated with a locally frequent itemsets may exhibit some interesting properties e.g. the itemsets may be cyclic in nature. In this paper we propose an alternative method of finding such cyclic frequent itemsets. The efficacy of the method is established through experimental results.