This is the algorithm I've seen posted elsewhere (when I did a search a long time ago in the perl newsgroups), and that I've used myself:

    my($tw, $winner) = (0, undef);
    my @numbers = ( ... );
    my @freq    = ( ... );
    for my $idx (0..$#numbers) {
        $tw    += $freq[ $idx ];
        $winner = $idx if rand($tw) < $freq[ $idx ];
    }
    return $winner;
[download]

There's also an example in the Perl Cookbook, apparently, if you have that.