Local Concentration-Based Feature Extraction Approach
This chapter presents the local concentration (LC)-based feature extraction approach for anti-spam, which is considered to be able to effectively extract position-correlated information from messages by transforming each area of a message to a corresponding LC feature [246,248]. After a brief introduction of the background, the structure of the LC model is given. Then described in detail is the implementation of the LC approach for anti-spam, including term selection, generation of detector sets, and construction of local concentration-based feature vectors. Two strategies for defining local areas and analysis of the LC model are also presented. Finally, the experimental validation is given.