Yes, in general re-reading an input file multiple times is a bad idea. I/O is very "expensive" both in terms of CPU time, but also in terms of clock time.

A simple way to read the input file once, could be:

#!/usr/bin/perl
use strict;
use warnings;
use Data::Dumper;

my @matrix; # Array of Array, ["africa1",18,1,48,23]

<DATA>;  # throw away first line
while (<DATA>)
{
   next if /^\s*$/;                   # skip blank lines   
   push (@matrix, [split(/\s+/,$_)]); # create AoA (2-D matrix)
}

print Dumper \@matrix;

__DATA__
Name    f1    f2    f3    f4
africa1    18    1    48    23
usa2    48    23    60    23
africa2    17    3    49    25
africa3    20    6    52    30
usa1    55    20    56    25
china1    35    37    55    87
china2    40    33    50    73
africa4 18    2    47    23
[download]

It appears to me that you are comparing columns. I think that you want an output that is based "best match(es)" for each each column. I added "africa4" above.

What is supposed to happen if: my @array = (18,35,59,70); Should this first array value of 18 result in two matches in the first data column?, africa1 and africa4??

It is completely plausible that 16 and 20 are both equally close to 18. What should the output be in those cases?