Verdict of association rule using systematic approach of time slicing for efficient pattern discovery

Abstract

In Data mining, Association rule learning is a popular and well researched method for discovering interesting relations between variables in large databases. The main task of association rule mining is to mine association rules by using minimum support thresholds, which could be explicitly specified by the users. Minimum support threshold is the one which differentiates frequently observed patterns from infrequent patterns from large number of transactional databases. In algorithms like association rule mining, sequential pattern mining, structured pattern mining, correlation mining, and associative classification, minimum support threshold is set up, by the user, to uncover the frequent patterns. Detecting a complete set of association rules is the desired aspect in data mining. But whenever the user specifies minimum support threshold, there is an ample chance of losing some association rules. This may lead to incompatible decisions. To overcome this problem, systematic algorithm has been proposed in this paper. In this algorithm, the user is not allowed to specify any minimum support threshold values to find the frequent patterns; instead the system itself generates the minimum threshold values, thus plugging the loophole of other algorithms. Using this approach, the user is well aware of entire information aiding him to take correct informed decisions. We also introduce the concept of timing algorithm along with the systematic algorithm, which will statically assign a unique value to each record of the transactional database. This technique is mainly used to save time by scanning through the entire transactional database only once rather than making multiple scans. The benefit of one scan database leads to better performance and minimization of time.