This blog is based on comparative studies on email spam filtering (Code included).
Original paper: link
1. Email spam filtering architecture in main companies
- Gmail:
Rules→features→likelihood; art spam detection machine learning algorithms such as logistic regressionand neural networks; optical character recognition (OCR); linked factors which are based on Google search results; text filters to eradicate hazards posed by spammers depending on the senders and their history. - Yahoo:
URL filtering, email content and spam complaints from users. - Outlook:
encrypted mail; distinctive methods (not in detail)
2. Performance evaluation measures
- Classification accuracy (Acc)
comparative number of messages rightly classified.
NH = Number of non-spam messages to be classified
NS = Number of spam messages to be classified