| line |
stmt |
bran |
cond |
sub |
pod |
time |
code |
|
1
|
|
|
|
|
|
|
package Algorithm::NaiveBayes::Model::Discrete; |
|
2
|
|
|
|
|
|
|
|
|
3
|
1
|
|
|
1
|
|
6
|
use strict; |
|
|
1
|
|
|
|
|
2
|
|
|
|
1
|
|
|
|
|
53
|
|
|
4
|
1
|
|
|
1
|
|
5
|
use base qw(Algorithm::NaiveBayes); |
|
|
1
|
|
|
|
|
2
|
|
|
|
1
|
|
|
|
|
107
|
|
|
5
|
1
|
|
|
1
|
|
482
|
use Algorithm::NaiveBayes::Util qw(rescale); |
|
|
1
|
|
|
|
|
3
|
|
|
|
1
|
|
|
|
|
602
|
|
|
6
|
|
|
|
|
|
|
|
|
7
|
|
|
|
|
|
|
sub do_add_instance { |
|
8
|
200
|
|
|
200
|
0
|
220
|
my ($self, $attributes, $labels, $data) = @_; |
|
9
|
|
|
|
|
|
|
|
|
10
|
200
|
|
|
|
|
247
|
foreach my $label ( @$labels ) { |
|
11
|
200
|
|
100
|
|
|
420
|
my $mylabel = $data->{labels}{$label} ||= {}; |
|
12
|
200
|
|
|
|
|
197
|
$mylabel->{count}++; |
|
13
|
200
|
|
|
|
|
472
|
while (my ($attr, $value) = each %$attributes) { |
|
14
|
400
|
|
|
|
|
1559
|
$mylabel->{attrs}{$attr}{$value}++; |
|
15
|
|
|
|
|
|
|
} |
|
16
|
|
|
|
|
|
|
} |
|
17
|
|
|
|
|
|
|
} |
|
18
|
|
|
|
|
|
|
|
|
19
|
|
|
|
|
|
|
sub do_train { |
|
20
|
1
|
|
|
1
|
0
|
2
|
my ($self, $training_data) = @_; |
|
21
|
1
|
|
|
|
|
2
|
my $m = {}; |
|
22
|
|
|
|
|
|
|
|
|
23
|
1
|
|
|
|
|
16
|
my $instances = $self->instances; |
|
24
|
1
|
|
|
|
|
3
|
my $labels = $training_data->{labels}; |
|
25
|
1
|
|
|
|
|
2
|
my $probs = $m->{probs} = {}; |
|
26
|
|
|
|
|
|
|
|
|
27
|
|
|
|
|
|
|
# Calculate the log-probabilities for each category |
|
28
|
1
|
|
|
|
|
5
|
foreach my $label ($self->labels) { |
|
29
|
2
|
|
|
|
|
20
|
$m->{prior_probs}{$label} = log($labels->{$label}{count} / $instances); |
|
30
|
|
|
|
|
|
|
|
|
31
|
2
|
|
|
|
|
6
|
my $denominator = log($labels->{$label}{count}); |
|
32
|
2
|
|
|
|
|
3
|
while (my ($attribute, $values) = each %{ $labels->{$label}{attrs} }) { |
|
|
6
|
|
|
|
|
30
|
|
|
33
|
4
|
|
|
|
|
17
|
while (my ($value, $count) = each %$values) { |
|
34
|
8
|
|
|
|
|
44
|
$probs->{$attribute}{$label}{$value} = log($count) - $denominator; |
|
35
|
|
|
|
|
|
|
} |
|
36
|
|
|
|
|
|
|
} |
|
37
|
|
|
|
|
|
|
} |
|
38
|
|
|
|
|
|
|
|
|
39
|
1
|
|
|
|
|
6
|
return $m; |
|
40
|
|
|
|
|
|
|
} |
|
41
|
|
|
|
|
|
|
|
|
42
|
|
|
|
|
|
|
sub do_predict { |
|
43
|
1
|
|
|
1
|
0
|
2
|
my ($self, $m, $newattrs) = @_; |
|
44
|
|
|
|
|
|
|
|
|
45
|
|
|
|
|
|
|
# Note that we're using the log(prob) here. That's why we add instead of multiply. |
|
46
|
|
|
|
|
|
|
|
|
47
|
1
|
|
|
|
|
1
|
my %scores = %{$m->{prior_probs}}; |
|
|
1
|
|
|
|
|
5
|
|
|
48
|
1
|
|
|
|
|
5
|
while (my ($feature, $value) = each %$newattrs) { |
|
49
|
2
|
50
|
|
|
|
6
|
next unless exists $m->{probs}{$feature}; # Ignore totally unseen features |
|
50
|
2
|
|
|
|
|
3
|
while (my ($label, $values) = each %{$m->{probs}{$feature}}) { |
|
|
6
|
|
|
|
|
22
|
|
|
51
|
4
|
|
50
|
|
|
17
|
$scores{$label} += ($values->{$value} || 0); |
|
52
|
|
|
|
|
|
|
} |
|
53
|
|
|
|
|
|
|
} |
|
54
|
|
|
|
|
|
|
|
|
55
|
1
|
|
|
|
|
5
|
rescale \%scores; |
|
56
|
1
|
|
|
|
|
4
|
return \%scores; |
|
57
|
|
|
|
|
|
|
} |
|
58
|
|
|
|
|
|
|
|
|
59
|
|
|
|
|
|
|
1; |