Feature Selection by Using Discrete Imperialist Competitive Algorithm to Spam Detection

Abstract

Spam is a basic problem in electronic communications such as email systems in large scales and large number of weblogs and social networks. Due to the problems created by spams, much research has been carried out in this regard by using classification techniques. Redundant and high dimensional information are considered as a serious problem for these classification algorithms due to their high computation costs and using a memory. Reducing feature space results in representing an understandable model and using various methods. In this paper, the method of feature selection by using imperialist competitive algorithm has been presented. Decision tree and SVM classifications have been taken into account in classification phase. In order to prove the efficiency of this method, the results of evaluating data set of Spam Base have been compared with the algorithms proposed in this regard such as genetic algorithm. The results show that this method improves the efficiency of spam detection.