NEURAL NETWORKS SELF-ORGANIZING MAPS AND LEARNING VECTOR QUANTIZATION IN E-MAILS SPAM IDENTIFICATION
Alisson Marques da Silva, Gray Farias Moita, Paulo E. M. Almeida · International Conference on Information Systems, Technology and Management · 2010
This paper proposes the implementation of an anti-spam system through which the email messages will be filtered. Messages will be processed too allow for transforming of complex information into simpler parts allowing a better performance regarding their classification. Methods for feature extraction were used to select the more relevant ones of each set of messages, and those were used to compose an input vector to the classifying agent. The experiments were carried out for these vectors with 25 and 50 features and also for ones obtained after using a multiple linear regression technique (Stepwise Regression) to reduce the number of the vector characteristics. Artificial neural networks SOM and LVQ were used as classifying agents. Different configuration networks were investigated. Hence, the attained results can be regarded quite promising. In some of the experiments, 99% of correct classification was achieved.