Spam Filtering Based on Supervised Latent Semantic Features Extraction

Spam text is an universal phenomenon on the “open web”, including large-scale email systems and the growing number of Blogs. Handling this information overload is becoming an increasingly challenging problem, A promising approach is the using of content-based filtering. In this paper, our focus is placed on finding effective dimension reduction method for email Spam filtering, we apply a supervised latent semantic features to extract the features according to their filtering powers. Using Support Vector Machines and k-Nearest Neighbors classification model, Experiments on TREC 2005 shows that supervised latent semantic features extraction can improve the junk mail detection performance .