Train your Filters with Bayesian Email Filtering
LuxSci’s Basic Spam Filtering service has just been augmented to include Bayesian analysis. with Bayesian analysis, each user can train his/her own Spam filters with examples of what that user considers “Spam” and “not Spam”. With enough examples, Bayesian analysis allows for the classification of new messages by their likelihood to be Spam or not and this drastically increases the accuracy of your Spam filtering.
All users of LuxSci’s Basic Spam Filtering system get Bayesian analysis at no additional charge — all you have to do is (1) enabled it and then (2) train it.
How To Get Started with Bayesian Spam Filtering
To get started:
- First, Go to your “Spam Filter Configuration Page” (you have to be logged in)
- Scroll to the “Bayesian Filtering” section.
- Check the box next to “Enable use of personal Bayesian analysis for improve spam filter accuracy.”
- Press “Save Changes”
- Next, Train your filter
- Train it with 200 or more examples of Spam and 200 or more examples of “not Spam”.
- Then, Bayesian analysis will start working with your Spam Filter to improve Spam classification.
How To Train your Bayesian Filter
There are a few ways to train your filter:
- In WebMail, you can select one or more messages in any folder and then choose the “Mark > Mark as Spam” or “Mark > Mark as Not Spam” options in the command (magic wand) menu. This will present your filter with the selected messages to learn as examples of Spam or Nor Spam.(Note, unlike with Bayesian Filtering disabled, using these commands will not delete or move the selected messages … they will only train your filter)
- You can also train your filters in “Bulk”. Simply go to the “Folder Properties” for any email folder and choose the “Train Spam Filtering” tool. Then, select if the folder contains only Spam or only Non-Spam and “Go”. This tool trains the folder with up to 1000 messages from the selected folder.
But, but, but….?
- The system remembers which messages you have trained with for 90 days. Re-training with the same messages in that time period does nothing bad to the filters.
- It is good to train with similar amounts of Spam a and Not Spam messages
- It is good to train with a variety of different Spam and Not Spam Messages
- It is good to continue training over time so that the system adapts to your changing inbound email profile
- You can “reset” and “restart” your training if/when needed.
- Training is individualized … each user must train his/her own filters.
- You can see how many messages have contributed to your Bayesian filter on your “Spam Filter Configuration Page“.
If you have questions about Bayesian Filtering, please contact support.