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
|
=head1 NAME
heri-stat - calculates precision, recall, F1 and some other things
=head1 SYNOPSIS
B<heri-stat> [-R] [-mrca] [-u label] [-t threshold]
I<outcomes_file> I<predictions_file>
B<heri-stat> -1 [-R] [-mrca] [-u label] [-t threshold]
[I<files>...]
B<heri-stat> -g mode [-R] [-xy] I<outcomes_file> I<predictions_file>
B<heri-stat> -1 -g mode [-R] [I<files>...]
B<heri-stat> [-h]
=head1 DESCRIPTION
The first and second types of B<heri-stat> invocation takes
classification dataset and predictions on input, and calculate
precision, recall and F1. Unless option B<-1> was applied, B<heri-stat>
reads correct classes from I<outcomes_file> (one class per line) and
predicted classes from I<predictions_file> (one class per line). It
is allowed for I<predictions_file> to contain two tokens per line. The
first one is a predicted class, the second one is a score,
e.g. probability.
The third and forth type of B<heri-stat> invocation takes regression
outcomes and predictions on input (one value per line) and calculate
mean absolute error (MAE),
mean squared error (MSE) and/or mean absolute error (MAE).
If B<-1> was applied, two or three tokens per line are expected on input:
correct value (or class for classification), predicted value (or class),
and optional score.
=head1 OPTIONS
=over 6
=item B<-h, --help>
Display help information.
=item B<-R, --raw>
Raw tab-separated output.
=item B<-m, --micro-avg>
Disable micro averaged P/R/F1 output.
=item B<-r, --macro-avg>
Disable macro averaged P/R/F1 output.
=item B<-c, --per-class>
Disable output of per-class statistics.
=item B<-a, --accuracy>
Disable output of accuracy.
=item B<-1, --single>
2 or 3 tokens per line are expected on input.
=item B<-u, --unclassified> I<label>
Set the label for "unclassified" object. If specified, micro-averaged
P/R/F1 is calculated instead of accuracy.
Also, statistics for this class is not calculated.
=item B<-t, --treshold> I<value>
Consider classes with score less than the specified I<value> as unclassified.
=item B<-g, --regression> I<mode>
Regression outcomes and predictions are expected on input. If I<mode> contains
symbol I<a>, MAE is output, I<s> -- MSE, and I<r> -- RMSE is output.
=back
=head1 HOME
L<http://github.com/cheusov/herisvm>
=head1 SEE ALSO
L<heri-eval(1)>
L<heri-split(1)>
|
