An Efficient Approach to Find Frequent Item Sets in Large Database

Abstract

In recent years, Data Mining [1] [2] is a constitutive technique for finding useful knowledge from large transactional dataset. Association rule mining [3] is one of the resuscitative data mining techniques. It finds the interesting patterns from large datasets to maximize the profit of the future business. Several algorithms are available to find frequent patterns. Apriori and FP-Tree [4] [5] algorithms are most common techniques for discovering frequent item sets. The Apriori algorithm uses a breadth-first search approach to find all significant frequent patterns. This is performed by candidate generation method which takes several number of database scans. The FP-Tree algorithm scans the whole database twice to discover significant frequent patterns without generation of candidate. So the main motive behind this proposed approach to discover frequent patterns from transactional database in minimum execution time. The proposed TR-FC-GCM (Transaction Reduction - Frequency Count - Generate Combination Method) finds all significant frequent patterns by generating all possible combinations of an item with single database scan and also works better for null and full datasets. The comparative results of TR-FC-GCM, Apriori and FP-Tree algorithms with different transactions and thresholds, it clearly shows that TR-FC-GCM algorithm outperforms than Apriori and FP-Tree algorithms.