An Optimization based Algorithm for Association Rule Mining in Large Databases
K.Kala Head & Associate Professor, Department of Computer Science, Nachiappa Swamigal arts and Science College,Koviloor, India. Corresponding Author.
Manuscript received on 03 August 2019. | Revised Manuscript received on 09 August 2019. | Manuscript published on 30 September 2019. | PP: 3746-3754 | Volume-8 Issue-3 September 2019 | Retrieval Number: C4932098319/2019©BEIESP | DOI: 10.35940/ijrte.C4932.098319
Open Access | Ethics and 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: In recent trends, the information and science technology has been developed which makes most of the research in Association Rule Mining (ARM) to be focused on enhancing the computational efficiency.But the searching for most appropriate data in large databases become more difficult.The major difficulties in this is determining the threshold values for confidence and support which causes effect on the quality of the association rule mining. Thus a novel optimization based algorithm for association rule mining is proposed in this work for the purpose of enhancing the computational efficiency and also to determine the appropriate threshold values automatically. Initially the input raw data is taken from the large transactional database. Then this input raw data is preprocessed and converted into binary type data. After this, the Modified Binary Particle Swarm Optimization (MBPSO) algorithm is used which involves five phases namely 1) Encoding – where a string based encoding is followed that encodes every single item into respective string chromosome, 2) Fitness Evaluation – where the ‘maximization’ has been determined using association rule type parameters, 3) Population generation – where an initial population has been generated using particles with best fitness values, 4) Searching – where best particles are searched in order to prevent it from falling apart from search space whenever the position is updated, 5) Stopping circumstance – as it is essential to complete the particle evaluation, by finding respective confidence and support as the minimal threshold values. This would be helpful in deriving valuable information in ARM. The performance of this proposed methodology is validated and compared with the tradition association rule mining algorithms. This proposed methodology offers better results with increased efficiency and proves its superiority.
Key words: Data Mining, Association Rule Mining, Modified Binary Particle Swarm Optimization, Confidence and Support Value, Fitness Function.
Scope of the Article: Data Mining