My generic "it's a number" regex, the first case you might not want:

my $is_number = qr/^
                   [+-]?                             # optional sign
                   (                                 # then:
                   \d{1,3}(\,\d\d\d)+(\.(\d+)?)?     # n,nnn.nn
                   |\d+\.\d+                         # n.n
                   |\d+\.                            # n.
                   |\.\d+                            # .n
                   |\d+                              # n
                  )
                   ([eE][+-]?\d+)?                   # optional expone
+nt
                   $/xo;
[download]

Running your data through it:

#!/usr/bin/perl

use strict;
use warnings;

my $regex = qr/^[+-]?(\d{1,3}(\,\d\d\d)+(\.(\d+)?)?|\d+\.\d+|\d+\.|\.\
+d+|\d+)([eE][+-]?\d+)?$/o;

my @data = (-123.004,-.008,0,-0,.0987,1.0,12345,'d','test');

foreach (@data) {
  if ($_ =~ /$regex/) {
    print "$_: true\n";
  } else {
    print "$_: false\n";
  }
}

exit;
[download]

Yields:

-123.004: true
-0.008: true
0: true
0: true
0.0987: true
1: true
12345: true
d: false
test: false
[download]

I've never tried to optimize it, but it seems to work well just the same.