EFFICIENT TRANSACTION REDUCTION IN ACTIONABLE PATTERN MINING FOR HIGH VOLUMINOUS DATASETS BASED ON BITMAP AND CLASS LABELS

Frequent pattern mining in databases plays an indispensable role in many data mining tasks namely, classification, clustering, and association rules analysis. When a large number of item sets are processed by the database, it needs to be scanned multiple times. Consecutively, multiple scanning of the database increases the number of rules generation, which then consume more system resources. Existing CCARM (Combined and Composite Association Rule Mining) algorithm used minimum support in order to generate combined actionable association rules, which in turn suffer from the large number of generating rules. Explosion of a large number of rules is the major problem in frequent pattern mining that adds difficult to find the interesting frequent patterns. This paper presents an efficient transaction reduction technique named TR-BC to mine the frequent pattern based on bitmap and class labels. The proposed approach reduces the rule generation by counting the item support and class support instead of only item support. Moreover, the database storage is compressed by using bitmap that significantly reduces the number of database scan. The rules are reduced by horizontal and vertical transaction and then finally combined rules are generated by eliminating the redundancy. Experimental results validate the performance of the proposed approach and expose that proposed method is more effective and efficient than previously proposed algorithm. Index terms - bitmap, item support, database scan, class support, class labels, database transaction, and combined rules.