root/feedmelinks/bin/bayesian-learner.pl

#!/usr/local/bin/perl
use XML::Simple;
use Data::Dumper;
use Algorithm::NaiveBayes;
use POSIX;

my $learner = Algorithm::NaiveBayes->new;

@datasets = ( "USERS.xml", "SPAMMERS.xml" );
foreach $dataset ( @datasets ) {

        $xml = new XML::Simple( KeyAttr=>[] );
        $data = $xml->XMLin( $dataset, KeepRoot => 1);

        $trueset_classification =  $data->{user_group}->{type};
        print "Training the learner on *KNOWN* $trueset_classification: ";

        my $c = 0;
        foreach $u (@{$data->{user_group}->{user}}) {
                $c++;
                print "." if( $c % 5 == 0 ); # status dot
                $learner->add_instance( attributes => $u->{heuristics}, label => $trueset_classification );
        }
        print "\n\n";
}

print "Training complete. Now let's test the learner on some users who may or may not be spammers:\n\n";
$learner->train; # get smart!

$xml = new XML::Simple( KeyAttr=>[] );
$unknowns = $xml->XMLin( "UNKNOWN.xml", KeepRoot => 1);
foreach $u (@{$unknowns->{user_group}->{user}}) {
        my $result = $learner->predict(attributes => $u->{heuristics});
        print "  " . $u->{name} . " - probability of being in the group of: \n";
        while( my ($k, $v) = each %$result ) {
                print "    $k: " . POSIX::floor(100 * $v ) . "%\n";
        }
        print "\n";
}