Bubble-net hunting strategy of whales based optimized feature selection for e-mail classification

Spam E-mailis a kind of electronic spam in which unsolicited messages are sent by E-mail. It is themost severe problem world-wide for decades. One of the best approach to identify spam E-mails is filtering E-mails by classification. In many applications feature selection isthe most widely used and essential task in many classification techniques to reduce the dimensionality of feature space. In this paper a Nature Inspired Meta-Heuristic Algorithm, that exploits the SVM principles for finding optimized structures of the Enron-Spam dataset having high similarity, is proposed. We adopted the WOA to obtainan optimal feature subset for E-mail classification. Four different kernel functions are exploited, that includes Linear, Quadratic, Polynomial and RBF in classification to test the best kernel function for SVM. Different evaluation measurements such as Precision, Accuracy, Recall and F-measureare calculated to find the performance of the proposed technique. The investigated results are analysed and compared with those from other techniques published in spam E-mails filtering. All the analysed and compared results show that proposed technique is very competitive for E-mail classification.