Authors: Dhanaraj Karthika Renuka; P. Visalakshi
Addresses: Department of Electronics and Communication Engineering, PSG College of Technology, Coimbatore, Tamil Nadu, India ' Department of Electronics and Communication Engineering, PSG College of Technology, Coimbatore, Tamil Nadu, India
Abstract: In this paper, an efficient spam classification technique is proposed using weighted-based multiple classifier and F-GSO algorithm. At first, input data is given to the feature selection to select the suitable feature for spam classification. Here, firefly and GSO algorithm is effectively hybridised to select the suitable features. Once the best feature is identified through hybrid algorithm, the spam classification is done using the weighted-based multiple classifiers. Here, three categories of classifiers like, rule classifier, lazy classifier and learning classifiers is combined using weight rule. These three classifiers have their own advantages and disadvantages so the hybridisation of classifiers leads to provide overall improvements by rectifying their disadvantages by other algorithms and retaining their advantages. Accordingly, decision tree (rule), lazy classifier (naïve Bayes) and neural network classifier (learning) are combined using voting-based weighted rule. Our experiment result shows the proposed systems have outperformed by having better accuracy value of 98.83%.
Keywords: FGSO; decision tree; lazy classifier and neural network classifier; email spam.
International Journal of Business Intelligence and Data Mining, 2017 Vol.12 No.3, pp.274 - 298
Received: 15 Sep 2016
Accepted: 20 Dec 2016
Published online: 10 Jul 2017 *