文章基本信息

标题：GNVDF: A GPU-accelerated Novel Algorithm for Finding Frequent Patterns Using Vertical Data Format Approach and Jagged Array
本地全文：下载
作者：P.Sumathi ; S.Murugan
期刊名称：International Journal of Modern Education and Computer Science
印刷版ISSN：2075-0161
电子版ISSN：2075-017X
出版年度：2021
卷号：13
期号：4
页码：24-41
DOI：10.5815/ijmecs.2021.04.03
语种：English
出版社：MECS Publisher
摘要：In the modern digital world, online shopping becomes essential in human lives. Online shopping stores like Amazon show up the "Frequently Bought Together" for their customers in their portal to increase sales. Discovering frequent patterns is a fundamental task in Data Mining that find the frequently bought items together. Many transactional data were collected every day, and finding frequent itemsets from the massive datasets using the classical algorithms requires more processing time and I/O cost. A GPU accelerated Novel algorithm for finding the frequent patterns using Vertical Data Format (GNVDF) has been introduced in this research article. It uses a novel pattern formation. In this, the candidate i-itemsets is divided into two buckets viz., Bucket-1 and Bucket-Bucket-1 contain all the possible items to form candidate-(i+1) itemsets. Bucket-2 has the items that cannot include in the candidate-(i+1) itemsets. It compactly employs a jagged array to minimize the memory requirement and remove common transactions among the frequent 1-itemsets. It also utilizes a vertical representation of data for efficiently extracting the frequent itemsets by scanning the database only once. Further, it is GPU-accelerated for speeding up the execution of the algorithm. The proposed algorithm was implemented with and without GPU usage and compared. The comparison result revealed that GNVDF with GPU acceleration is faster by 90 to 135 times than the method without GPU.
关键词：Frequent Patterns;GNVDF;Graphical Processing Unit;Novel Pattern Formation;Vertical Data Format;and Jagged Array