1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
|
/*
* Copyright (c) 2023 Stalwart Labs Ltd.
*
* This file is part of the Stalwart Mail Server.
*
* This program is free software: you can redistribute it and/or modify
* it under the terms of the GNU Affero General Public License as
* published by the Free Software Foundation, either version 3 of
* the License, or (at your option) any later version.
*
* This program is distributed in the hope that it will be useful,
* but WITHOUT ANY WARRANTY; without even the implied warranty of
* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
* GNU Affero General Public License for more details.
* in the LICENSE file at the top-level directory of this distribution.
* You should have received a copy of the GNU Affero General Public License
* along with this program. If not, see <http://www.gnu.org/licenses/>.
*
* You can be released from the requirements of the AGPLv3 license by
* purchasing a commercial license. Please contact licensing@stalw.art
* for more details.
*/
use crate::tokenizers::osb::OsbToken;
use super::{BayesClassifier, Weights};
// Position 0 represents Unigram weights
const FEATURE_WEIGHT: [f64; 8] = [1.0, 3125.0, 256.0, 27.0, 1.0, 0.0, 0.0, 0.0];
// Credits: ported from RSpamd
impl BayesClassifier {
pub fn classify<T>(&self, tokens: T, ham_learns: u32, spam_learns: u32) -> Option<f64>
where
T: Iterator<Item = OsbToken<Weights>>,
{
if self.min_learns > 0 && (spam_learns < self.min_learns || ham_learns < self.min_learns) {
return None;
}
let mut processed_tokens = 0;
let mut total_spam_prob = 0.0;
let mut total_ham_prob = 0.0;
for token in tokens {
let weights = token.inner;
let total_count = weights.spam + weights.ham;
if total_count >= self.min_token_hits {
let total_count = total_count as f64;
let spam_freq = weights.spam as f64 / f64::max(1.0, spam_learns as f64);
let ham_freq = weights.ham as f64 / f64::max(1.0, ham_learns as f64);
let spam_prob = spam_freq / (spam_freq + ham_freq);
let ham_prob = ham_freq / (spam_freq + ham_freq);
let fw = FEATURE_WEIGHT[token.idx];
let w = (fw * total_count) / (1.0 + fw * total_count);
let bayes_spam_prob = prob_combine(spam_prob, total_count, w, 0.5);
if !((bayes_spam_prob > 0.5 && bayes_spam_prob < 0.5 + self.min_prob_strength)
|| (bayes_spam_prob < 0.5 && bayes_spam_prob > 0.5 - self.min_prob_strength))
{
let bayes_ham_prob = prob_combine(ham_prob, total_count, w, 0.5);
total_spam_prob += bayes_spam_prob.ln();
total_ham_prob += bayes_ham_prob.ln();
processed_tokens += 1;
}
}
}
if processed_tokens == 0 || self.min_tokens > 0 && processed_tokens < self.min_tokens {
return None;
}
let (h, s) = if total_spam_prob > -300.0 && total_ham_prob > -300.0 {
/* Fisher value is low enough to apply inv_chi_square */
(
1.0 - inv_chi_square(total_spam_prob, processed_tokens),
1.0 - inv_chi_square(total_ham_prob, processed_tokens),
)
} else {
/* Use naive method */
if total_spam_prob < total_ham_prob {
let h = (1.0 - (total_spam_prob - total_ham_prob).exp())
/ (1.0 + (total_spam_prob - total_ham_prob).exp());
(h, 1.0 - h)
} else {
let s = (1.0 - (total_ham_prob - total_spam_prob).exp())
/ (1.0 + (total_ham_prob - total_spam_prob).exp());
(1.0 - s, s)
}
};
let final_prob = if h.is_finite() && s.is_finite() {
(s + 1.0 - h) / 2.0
} else {
/*
* We have some overflow, hence we need to check which class
* is NaN
*/
if h.is_finite() {
1.0
} else if s.is_finite() {
0.0
} else {
0.5
}
};
if processed_tokens > 0 && (final_prob - 0.5).abs() > 0.05 {
Some(final_prob)
} else {
None
}
}
}
/**
* Returns probability of chisquare > value with specified number of freedom
* degrees
*/
#[inline(always)]
fn inv_chi_square(value: f64, freedom_deg: u32) -> f64 {
let mut prob = value.exp();
if prob.is_finite() {
/*
* m is our confidence in class
* prob is e ^ x (small value since x is normally less than zero
* So we integrate over degrees of freedom and produce the total result
* from 1.0 (no confidence) to 0.0 (full confidence)
*/
let mut sum = prob;
let m = -value;
for i in 1..freedom_deg {
prob *= m / i as f64;
sum += prob;
}
f64::min(1.0, sum)
} else {
/*
* e^x where x is large *NEGATIVE* number is OK, so we have a very strong
* confidence that inv-chi-square is close to zero
*/
if value < 0.0 {
0.0
} else {
1.0
}
}
}
/*#[inline(always)]
fn normalize_probability(x: f64, bias: f64) -> f64 {
((x - bias) * 2.0).powi(8)
}*/
#[inline(always)]
fn prob_combine(prob: f64, cnt: f64, weight: f64, assumed: f64) -> f64 {
((weight) * (assumed) + (cnt) * (prob)) / ((weight) + (cnt))
}
|
