remluvr has asked for the wisdom of the Perl Monks concerning the following question:

Hi everyone.
I have a new question. I've written some code and I need to modify it in order to make it work with a new input I have. My code was a simple one, used to compute Average Precision on some data.
Now my data has changed because I have a list like the following. I need for it to work on separate elements, so to say, I need it to compute Average precision for different element.
My input is the following: 

1    acacia-n    hyper    tree-n    0.838364743354488
2    acacia-n    hyper    plant-n    0.740581661563839
3    acacia-n    mero    wood-n    0.687569086370938
4    acacia-n    mero    flower-n    0.650909477491374
5    acacia-n    coord    oak-n    0.610092594991099
6    acacia-n    coord    pine-n    0.537690234715029
7    acacia-n    mero    branch-n    0.510899195919917
8    acacia-n    mero    root-n    0.491198721010063
9    acacia-n    coord    willow-n    0.481680704877001
1    ant-n    hyper    animal-n    0.634580215370739
2    ant-n    hyper    insect-n    0.53621081509255
3    ant-n    mero    head-n    0.535980012877533
4    ant-n    mero    body-n    0.505827598873949
5    ant-n    coord    bee-n    0.481918895790599
1    apricot-n    hyper    fruit-n    0.76748797529242
2    apricot-n    coord    apple-n    0.685155565667883
3    apricot-n    mero    juice-n    0.560337418489082
4    apricot-n    coord    banana-n    0.559525300446683

[download]

My output (given that I compute AP for hyper) should be like the following:

acacia-n hyper 1
ant-n hyper 1
apricot-n hyper 1

[download]

I've picked lucky cases in which AP is 1!
My code at the moment is the following: 



my $conteggio_trovato=0;
my $conteggio_hyper=0;
my $precision_glob=0;
my $precision=0;
my $average_precision=0;
my ($rank, $u, $relaz, $v, $score);
while (<INPUT>) {
    ($rank, $u, $relaz, $v, $score) = split;
    if ($relaz eq "hyper"){
        print "REl:".$relaz;
        $conteggio_hyper++;
        my $precision = $conteggio_trovato/$rank;
        print "precision:".$precision."\n";
        $precision_glob=$precision_glob+$precision;
    }
    $conteggio_trovato++;
  #  $precision_glob=$precision_glob+$precision;
}
#$precision_glob=$precision_glob+$precision;
$average_precision=$precision_glob/$conteggio_hyper;
#print "precision totale: ".$precision_glob."\n";
#print "Conteggio: ".$conteggio_hyper."\n";
print $u."\t"."hyper"."\t".$average_precision."\n";

[download]

How can I modify it?
Thanks!
Giulia </code>

Replies are listed 'Best First'.
Re: Iterate over multiple data
by kcott (Archbishop) on Mar 08, 2012 at 19:32 UTC

    I can see you've had problems with the code. It would have been useful to provide an algorithm for calculating what you refer to as Average Precision. I've made a guess at what you've attempting to achieve in the following script. If it's not right, please provide more concrete information. 

    #!/usr/bin/env perl

use strict;
use warnings;

my %result = ();
my $wanted_relaz = q{hyper};

my $last_u = q{};
my $u_count = 0;
my $relaz_count = 0;
my $precision = 0;

my ($rank, $u, $relaz);

while (<DATA>) {
    ($rank, $u, $relaz, undef) = split;

    if ($u ne $last_u) {
        if ($last_u) {
            $result{$last_u} = $precision / $relaz_count;
        }

        $last_u = $u;
        $u_count = 0;
        $relaz_count = 0;
        $precision = 0;
    }

    ++$u_count;

    if ($relaz eq $wanted_relaz) {
        ++$relaz_count;
        $precision += $u_count / $rank;
    }
} 

$result{$u} = $precision / $relaz_count;

for my $key (sort keys %result) {
    print qq{$key\t$wanted_relaz\t$result{$key}\n};
}

__DATA__
1    acacia-n    hyper    tree-n    0.838364743354488
2    acacia-n    hyper    plant-n    0.740581661563839
3    acacia-n    mero    wood-n    0.687569086370938
4    acacia-n    mero    flower-n    0.650909477491374
5    acacia-n    coord    oak-n    0.610092594991099
6    acacia-n    coord    pine-n    0.537690234715029
7    acacia-n    mero    branch-n    0.510899195919917
8    acacia-n    mero    root-n    0.491198721010063
9    acacia-n    coord    willow-n    0.481680704877001
1    ant-n    hyper    animal-n    0.634580215370739
2    ant-n    hyper    insect-n    0.53621081509255
3    ant-n    mero    head-n    0.535980012877533
4    ant-n    mero    body-n    0.505827598873949
5    ant-n    coord    bee-n    0.481918895790599
1    apricot-n    hyper    fruit-n    0.76748797529242
2    apricot-n    coord    apple-n    0.685155565667883
3    apricot-n    mero    juice-n    0.560337418489082
4    apricot-n    coord    banana-n    0.559525300446683

    [download]

    Here's the output: 

    $ pm_avg_precision.pl
acacia-n    hyper    1
ant-n    hyper    1
apricot-n    hyper    1

    [download]

    -- Ken

[reply]
[d/l]
[select]
Re: Iterate over multiple data
by JavaFan (Canon) on Mar 08, 2012 at 17:47 UTC
    Can you explain what you want the output to be, and how the calculations go? And what are the different elements in this case?
[reply]

      My output (given that I compute AP for hyper) should be like the following:

      acacia-n hyper 1
ant-n hyper 1
apricot-n hyper 1

      [download]

      Basically, for each different element of the first row of data I submitted as input, I should repeat the calculations my code does. The only problem I have is that I don't know how to make the code iterate only while acacia, and do the calculations on that data, then do them only for ant, and then do them only for apricot.
      I hope I made it a little clearer.
      Thanks again,
      Giulia

[reply]
[d/l]
        Seems you want to use a hash, keyed on $u where you can keep track of the different values you need.
[reply]
[d/l]

Back to Seekers of Perl Wisdom