Parallel Multithreaded Apriori Algorithm for Vertical Association Rule Mining

Abstract: Association rule mining is one of the important concepts in data mining domain for analyzing customer's data. The association rule mining is a process of finding correlation among the items involved in different transactions. Traditionally association rule mining is implemented horizontally. For this we have plenty of different algorithms in research like Apriori based, FP tree based so on. Recently we have a new method in association rule mining which generates vertical association rules. In horizontal association rule mining we read transaction items record by record basis and computes support of each frequent item or candidate item. We repeat this process to generate frequent item sets. The vertical association rule mining evaluates support frequency of each item column wise for this it uses bitmap matrix this saves support sets of frequent item sets in memory which is used to calculate candidate item sets. In our system it is proposed to combine both horizontal mining and vertical mining in generating association rules. The horizontal and vertical mining are implemented in parallel using multithreading concept. For this we propose a modified parallel multithreaded Apriori algorithm. The algorithm saves time and decreases memory space as the process is running because of bitmap representation of dataset and bitmap compression algorithms.

Keywords: Data mining, Association Rule Mining, Frequent item sets, Candidate item sets, Horizontal mining, Vertical mining, Apriori, Bitmap Apriori, and Parallel multithreaded Apriori