Neural Network approach to spam detection in emails

Asked Nov 04 '12 at 14:09

Active Nov 04 '12 at 14:17

Viewed 475 times

I'm building a spam detection system using neural networks. I'm not able to understand how to proceed with what I have currently.

I have- Unread Mails being flagged as read and converted to mail vectors using tf-idf weighing. So basically, My Email Message looks like

Email : (Word1,Score1),(Word2,Score2)...

After doing(parsing , stemming,stopword removal and tf-idf conversion). I have read about feedback network trained via backpropogation and it seems to be the approach followed most commonly. Basically, How do i reduce the dimensionality further of the vectors I have and how to feed it as an input. Also, how does hidden layer behave and how does the number of hidden layer neurons affect the performance of neural network.Also , How is a feature vector different from what I have ? How do I form a feature vector?

Thanks.Looking forward to some clarity.

edited Nov 04 '12 at 14:17

asked Nov 04 '12 at 14:09

Hooli

1

http://stackoverflow.com/questions/770238/neural-networks-for-email-spam-detection is where you'll find more clarity. – cggaurav Nov 04 '12 at 14:19
Thanks cggaurav. I went through it. Helps :) – Hooli Nov 04 '12 at 14:30

Neural Network approach to spam detection in emails

0 Answers0