An Approach to Improve Apriori Algorithm for Extraction of Frequent Itemsets

Abstract

The amount of data generated today regarding volume, generation velocity, and variety is quite immense. This, in turn, has created a great challenge for scientists and researchers. To devise a solution, researchers have suggested a variety of schemes to help alleviate this problem. One of the suggested schemas is Association Rule Mining, and it is primarily focused on finding the associations in transactionlike data. To assist in finding such associations, Frequent Itemsets should be discovered first. Therefore, this research is a new approach to finding Frequent Itemsets and it is based on the Apriori algorithm and Apache Spark distributed platform. Further, we introduce an extended version of Apriori which tends to find Maximal Frequent Itemsets first to help speed up the mining process. The results and comparison to algorithms like YAFIM and HFIM and the original Apriori show the suggested algorithm outperforms them in dense datasets by an average of 38 percent.