Implementation and Analysis of Enhanced Apriori Using MapReduce
Mercy Nyasha Mlambo, Naison Gasela, Michael Bukohwo Esiefarienrhe · 2018
The exponential growth of data due to advancement in network and computer technologies has driven the need for efficient and real time frequent itemset mining algorithms. Furthermore, as systems are producing large data volumes, there is need for best technologies that can mine and effectively analyze data to obtain crucial information. Association rule mining involves the extraction of associations or connections among data from a given data set. In this research we present market basket analysis for retail stores where mined data associations help marketers to sell frequently purchased combinations of products to their prospective and current customers. Decision makers can also make use of the extracted rules to predict future occurrences and act accordingly. In this paper we present an enhancement of the Apriori algorithm based on a scalable environment called Hadoop MapReduce. Our main goal is to reduce the large resource requirements and minimize communication overheads that are incurred in frequent itemset data extraction using localized split frequent itemset generation and early elimination of infrequent data. We also discuss the experimental results obtained from implementing the enhanced parallel Apriori algorithm based on Hadoop MapReduce. Finally we present possible future directions to improve the implementation of Apriori algorithm.