FEATURE SELECTION USING PARTICLE SWARM OPTIMIZATION WITH APPLICATION IN SPAM FILTERING

Using a finite set of features to determine an e-mail as spam or non-spam is a very popular but important topic. However, in most cases, the feature selection is empirically verified. In order to automatically determine the proper number of features to classify spam e-mails, this paper investigates the possibility of using a particle swarm optimization algorithm to find more relevant subset of the set of features. The selected subset contains the least number of features that most contribute to classification accuracy. The experimental results show that the proposed approach can be used to select the most proper discriminative features and to increase the performance of spam e-mails classification.