Re: [Dspam-user] Meaning of Confidence value
Brought to you by:
paulcockings,
sbajic
From: Tom H. <to...@wh...> - 2010-10-01 10:57:45
|
On 01/10/10 12:42, Julien Vehent wrote: > Hi there, > > I was browsing the code to understand the value of X-DSPAM-CONFIDENCE and I'm not sure I understand its meaning completely. > > --- libdspam.c --- > /* Calculate Confidence */ > > if (CTX->algorithms & DSP_MARKOV) { > if (CTX->result == DSR_ISSPAM) > { > CTX->confidence = CTX->probability; > } > else > { > CTX->confidence = 1.0 - CTX->probability; > } > } else { > if (CTX->result == DSR_ISSPAM) > { > CTX->confidence = rob_result; > } > else > { > CTX->confidence = 1.0 - rob_result; > } > } > ------------------ > > As I understand the code, if a message is a spam, confidence = probability. If the message is a ham, confidence = (1 - probability). > How does that gives an information on the level of maturity DSPAM has reached ? > > Thanks, > Julien > Probability is the 'chance' that a message is spam. A very low probability means that a message is innocent. The lower the probability is, the more confident DSPAM is that the message is innocent. For example: - probability=0.99 means 99% confident that a message is spam. - probability=0.01 means 1% confident that the message is spam, or 99% confident that it is innocent. The calculation is there to make the confidence number have the same value for both ham and spam: DSPAM is equally confident, but on a different decision. Regards, Tom |