| Subject: | Missing itemset |
| Date: | Mon, 21 Oct 2013 01:21:57 +0300 |
| To: | bug-Tree-FP [...] rt.cpan.org |
| From: | Peter Tsolakis <mlsgisanalyst [...] gmail.com> |
Take this list of transactions, stored in file.txt:
1 2 3
4 5 6
4 6
9 6 8 7
2 4 3 6 7
And run the following script. It will throw an "Illegal division by zero at
Tree/FP.pm line 388." when calling the association_rules method. By
debugging a little bit, it seems that the _fp_growth sub in some
circumstances might miss a set (in this case, with support=0.3, it misses
the {'3'} set).
#!/usr/bin/perl
use warnings;
use strict;
use Tree::FP;
use POSIX qw(ceil);
my %words;
{
open (my $IN,'<','file.txt') or die $!;
while (<$IN>) {
chomp;
$words{$_}++ for split/ /;
}
close $IN;
}
my @sorted=sort {$words{$b}<=>$words{$a}} keys %words;
my $fptree=Tree::FP->new(@sorted);
$fptree->set_support(0.3);
$fptree->set_confidence(0.6);
{
open (my $IN,'<','file.txt') or die $!;
while (<$IN>) {
chomp;
$fptree->insert_tree(split/ /) or die "Error while insert row $.:
",$fptree->err;
}
close $IN;
}
my @rules=$fptree->association_rules;