Lazy Learning Associative Classification in MapReduce Framework
S. P. Siddique Ibrahim1, M. Sivabalakrishnan2, S.P. Syed Ibrahim3
1S. P. Siddique Ibrahim, Assistant Professor, Kumaraguru College of Technology, Coimbatore (Tamil Nadu), India.
2Dr. M. Sivabalakrishnan, Associate Professor, School of Computing Science and Engineering, VIT University, Chennai (Tamil Nadu), India.
3Dr. S.P. Syed Ibrahim, Professor, School of Computing Science and Engineering, VIT University, Chennai (Tamil Nadu), India.
Manuscript received on 11 December 2018 | Revised Manuscript received on 22 December 2018 | Manuscript Published on 09 January 2019 | PP: 168-172 | Volume-7 Issue-4S November 2018 | Retrieval Number: E2027017519/19©BEIESP
Open Access | Editorial and Publishing Policies | Cite | Mendeley | Indexing and Abstracting
© The Authors. Blue Eyes Intelligence Engineering and Sciences Publication (BEIESP). This is an open access article under the CC-BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/)
Abstract: He core objective of the work is to propose a distributed environment based lazy learning Associative Classification (AC). Associative Classification is a hybrid version of data mining tasks which integrated both Association Rule Mining (ARM) and Classification technique to construct accurate classifier. Unfortunately, the AC used for learning these classifier are less popular in real time for building application due to its higher computation time complexity and memory constraints in large volume of datasets. Moreover, single processor’s CPU resources and memory are limited, which makes the algorithm incompetent to handle such datasets. To overcome such downsides, we proposed a distributed and parallel computing for lazy learning associative classification for accelerating algorithm performance by projecting the testing instances with large training datasets. In this work, we have implemented MapReduce based algorithms which reduce the computation by eliminates the need of constructing generalized classifier. It also well handled rare rules and generated institutive rules. The proposed algorithm may be suitable in area such as network intrusion detection, fraud detection, crowd analysis, rare disease prediction and crime analysis. Our algorithm has been compared with well known existing algorithms in relations of precision and running time. The experiments result has strengthened the proposed algorithm well handle the rare rules in distributed environment and is making better performance even the size of the datasets is huge.
Keywords: Association Rule Mining, Rare Rules, Lazy Learning, Associative Classification.
Scope of the Article: Classification