Efficient email classification strategy based on semantic methods

Abstract

Emails have emerged as one of the foremost packages in each day life. The continuous increase in the wide variety of email users has led to a huge boom of unsolicited emails, which might be also known as junk mail emails. Managing and classifying this large variety of emails is an important challenge. In this paper, a green email filtering approach based totally on semantic techniques is addressed. The proposed technique employs the Word Net ontology and applies exceptional semantic-based totally strategies and similarity measures for lowering the huge number of extracted textual features, and as a result, the gap and time complexities are reduced. Most of the approaches delivered to remedy this trouble treated the high dimensionality of emails by the use of syntactic feature selection. Moreover, to get the minimal most appropriate features’ set, function dimensionality reduction has been integrated using characteristic selection strategies which include the Principal Component Analysis (PCA) and the Correlation Feature Selection (CFS). Experimental results on the usual benchmark Enron Dataset showed that the proposed semantic filtering approach combined with the function choice achieves excessive computational performance at high area and time discount rates. A comparative study for numerous classification algorithms indicated that the Logistic Regression achieves the very best accuracy in comparison to Naïve Bayes, Support Vector Machine, J48, Random Forest, and radial basis function networks. By integrating the CFS characteristic choice technique, the average recorded accuracy for the all used algorithms is above 90%, with more than 90 reductions. Besides, the carried-out experiments showed that the proposed paintings have a highly sizeable overall performance with better accuracy and much less time in comparison to other related works.