C-Command Software Forum

Accuracy Falling in past 90 days

I’ve been a happy user since 9/5/13, with very few False Negatives (spam in my inbox) until the past 90 days or so. Over the past few months, obvious spam is getting through,. I am dutiful about using “Train as Spam” with all such errors, but one or two bad messages a month “getting through” has become one or two a day lately. I hate the idea of starting over with a new corpus. Is that the answer?

Running latest OSX (10.9.4) and Mail (7.3) and SpamSieve (2.9.15) versions, stats follow.

**Filtered Mail
19,517 Good Messages
10,028 Spam Messages (34%)
31 Spam Messages Per Day

SpamSieve Accuracy
28 False Positives
501 False Negatives (95%)
98.2% Correct

Corpus
1,781 Good Messages
2,423 Spam Messages (58%)
385,115 Total Words

Rules
1,585 Blocklist Rules
9,761 Whitelist Rules

Showing Statistics Since
9/5/13, 3:00 PM**

Please send in your log file via e-mail so that I can take a look at what’s happening.

Done. Thanks.
I appreciate the support, and the rapid response here. Best Mac investment I have made, period.

I can’t imagine dealing manually with the 10,000 spam messages (1/3 of my incoming mail!) SpamSieve has gobbled up over the past 10 months.

Thanks! However, I don’t think we actually received your log file, unless you sent it under a different name.

Oops, logfile came from avatar at stevebasile dot com

I thought that might be it. Just to close the loop on this forum thread, the problem ended up being that SpamSieve’s training was messed up because some mistakes had gone uncorrected. To fully fix this, you would need to reset both the rules and the corpus by quitting SpamSieve and deleting the folder:

/Users/<username>/Library/Application Support/SpamSieve/

Then re-train SpamSieve with a smaller number of recent messages.