Title: Sequential Pattern Mining (SPM) for user-inputted data sets: an empirical framework using bitwise operations

Authors: M. Manohar; R. Dinesh; M.S. Sowmya

Addresses: Department of Information Science & Engineering, School of Engineering and Technology, Jain University, Bangalore, Karnataka, India ' Department of Information Science & Engineering, School of Engineering and Technology, Jain University, Bangalore, Karnataka, India ' Department of Information Science & Engineering, School of Engineering and Technology, Jain University, Bangalore, Karnataka, India

Abstract: Sequential pattern mining is used to discover temporal relationships between item sets within a large data set. The downside of these approaches is the computation time and memory requirement, which increase exponentially with the data set size. We propose a new algorithm for sequential pattern mining using Apriori-based frequent itemset. In this work, a whole transaction is represented using binary number. The main advantage of the proposed method is in eliminating the necessity to scan the whole data set, for every new set of transactions, which is the limitation in existing sequential pattern mining algorithms. The result of the proposed method is analysed, which shows that the proposed algorithm provides support for large data set analysis, taking care of both execution time and memory usage. Also, we have proposed a pilot approach on how the proposed sequential pattern mining algorithm would work in a parallel environment.

Keywords: Apriori algorithm; association rules; binary representation; bitwise operators; data mining; frequent itemsets; parallel mining; sequential pattern mining; SPM; vertical projection.

DOI: 10.1504/IJKEDM.2015.074083

International Journal of Knowledge Engineering and Data Mining, 2015 Vol.3 No.3/4, pp.337 - 361

Received: 04 Jun 2015
Accepted: 19 Nov 2015

Published online: 06 Jan 2016 *

Full-text access for editors Full-text access for subscribers Purchase this article Comment on this article