Jagged Itemset Counting for Mining Frequent Itemsets

Abstract

Frequent Itemsets (FIs) are very useful and meaningful information for business intelligence and business analytics. The evidence of occurrences of Frequent Itemsets or Frequent patterns present in the business transactions plays a vital role in decision making. But it is a tedious and lengthier process in finding them. It is a challenging situation when mining this piece of valuable information from extensive databases such as ‘BigData’ in a faster manner. Though many researchers proposed many algorithms for Frequent Itemset Mining it suffers to defeat against either the size of frequent itemset or size of the database or the time taken to finish the task. To surpass this difficulty, this paper proposes the Jagged Itemset Counting (JIC) algorithm which uses Geometric Progression sequence numbers for labeling items to mine the frequent items easily. Geometric Progression Label Number (GPLN)is used to label the item and Cumulative Geometric Progression Label Number (CGPLN) is used to label the itemsets. And this algorithm requires two passes over the transaction database in which the first pass includes preprocessing and finding singleton (1-k) frequent itemsets. In the next pass, the rest of the n-k frequent itemsetsare found using the JIC algorithm and then finally compared the efficiency of this algorithm with the famous algorithms like Apriori and Eclat. The proposed algorithm outperformed the Apriori and Eclat algorithms and shows better performance in execution time even at the least minimum support.