P-BBA: A Master/Slave Parallel Binary-based Algorithm for Mining Frequent Itemsets in Big Data

Frequent itemsets mining is an effective but computational expensive technique especially when dealing with big datasets. Hence, the need for a customizable algorithm to work with big datasets in a reasonable time becomes a necessity. The Binary-based Technique Algorithm (BBT) used a binary representation of the database transactions as well as binary operations in order to simplify the process of identifying the frequent patterns as well as reduce the memory consumption. However, BBT algorithm still suffer the problem of low performance in terms of execution times when dealing with big data. This is due to the fact that the BBT algorithm was designed to run as a single thread of execution. Therefore, there is a need to improve the performance of the Binary-based Technique Algorithm (BBT). In this research, we proposed a Parallel Binary-Based Algorithm (P-BBA) towards solving the above mentioned problem. The objective of the proposed P-BBA is to process big datasets by developing collaborative threads that would work together concurrently and collaboratively and generates the list of frequent itemsets within an acceptable time frame. The algorithm is designed using a Master/Slave thread model to fits in Apache Spark distributed platform. The performance will be evaluated based on the total execution time.