Beefy Boxes and Bandwidth Generously Provided by pair Networks
Your skill will accomplish
what the force of many cannot
 
PerlMonks  

Comment on

( #3333=superdoc: print w/ replies, xml ) Need Help??

Indeed, there were a number of other parameters that could be tweaked, and by doing so, I was able to get better results than that.

However, in the end, it turns out there are some special properties of your problem that allow much simpler and more effective solutions. Namely, the fact that your data points are one-dimensional. (I'm assuming they are.)

It means, for example, that (1 3),(2 4) is never an optimal clustering. Neither is (1 2),(1 2).

From these, we can define the following constraints on clusters:

  1. clusters are simply segments within the ordered data set.
  2. all repetitions of a number must be kept together.

In the following solution, I'm using variance as a measure of the "coherence" or "binding strength" or whatever you want to call it within each cluster. You could use other measures; I wouldn't be surprised if variance doesn't necessarily give the best results. I've seen discussions of clustering that talk about maximizing variance between clusters, in addition to minimizing it within clusters. I have my doubts that that would help in a simple one-dimensional problem like this one.

I've used Statistics::Lite for a variance function; but you could easily substitute any other module of your preference.

use Statistics::Lite; use strict; use warnings; my @d = # I assume the order as given in the OP is not important: sort { $a <=> $b } ( 12, 14, 16, 18, 18, 20, 20, 20, 20, 20, 20, 20, 22, 24, 24, 24, 24, +30, 30, 30, 32, 35, 35, 35, 35, 35, 35, 35, 36, 40, 42, 46, 48, 48, 5 +0, 50, 50, 50, 54, 54, 55, 56, 56, 58, 58, 60, 60, 60, 60, 63, 64, 67 +, 67, 68, 70, 70, 76, 80, 86, 86, 86, 90, 90, 99, 100, 100, 100, 100, + 104, 105, 128, 150, 150, 154, 169, 190, 200, 200, 200, 250, 280, 291 +, 291, 300, 325, 325, 330, 450, 460, 550, 566, 600, 700, 750, 770, 95 +0, 1226, 1250, 2000, 15, 22, 24 ); my %count; $count{$_}++ for @d; my @e = sort { $a <=> $b } keys %count; # an individual is an array of arrays of nums. # each element of the (top-level) array represents a cluster. # the order of all the numbers, if you were to concat all the arrays, # is strictly numeric ascending. # in a 1-d space only, it never makes sense to cluster the # numbers (1,2,3,4) as (1,3), (2,4). # the only mutation possible is shifting a number off the # beggining of one array and pushing it onto the previous array. # (and the other way). sub random_bipartition { my( $min_size, $ar ) = @_; my $sel = @$ar - (2*$min_size); my $p = $min_size + int rand( $sel ); $p > $#{$ar} ? ( [ @{$ar} ], [ ] ) : ( [ @{$ar}[0 .. $p] ], [ @{$ar}[$p+1 .. $#{$ar}] ] ) } sub Ind::new_randomized { my $nc = shift; my $n = $nc <= 2 ? 1 : $nc <= 4 ? 2 : $nc <= 8 ? 3 : 4; # 16 max my @a = ( \@e ); @a = map { random_bipartition( 1<<$n, $_ ) } @a while $n--; # deep + magic :-) my $i = 0; while ( @a > $nc ) { unshift @{$a[$i+1]}, @{$a[$i]}; splice @a, $i, 1; $i++; } \@a } sub Ind::clone { my $ind = shift; [ map { [@$_] } @$ind ] } sub Ind::as_string { my $ind = shift; my $varsum; my $s; for my $cl ( @$ind ) { my $var = 0; if ( @$cl ) { my @d; push @d, ($_) x $count{$_} for @$cl; $var = int( Statistics::Lite::variance(@d)||0 ); $varsum += $var; } $s .= "$var ( @$cl )\n"; } $s .= "Total variance: $varsum\n"; $s } sub Ind::fitness { my $ind = shift; my $sum = 0; my $empty = 0; for my $cl ( @$ind ) { if ( @$cl ) { my @d; push @d, ($_) x $count{$_} for @$cl; $sum += (Statistics::Lite::variance(@d)||0); } else { $empty++; } } # harshly penalize individuals with the wrong number (too few) of +clusters: $sum *= ( 1 + $empty / 10 ); 1_000_000 - $sum # convert to Larger Is Better } sub Ind::mutate { my( $ind, $n ) = @_; $n ||= 1; my $lo = int rand( @$ind - 1); if ( int rand 2 ) { # up my($to,$from) = ( $ind->[$lo+1], $ind->[$lo] ); unshift @$to, pop @$from while $n-- && @$from } else { # down my($to,$from) = ( $ind->[$lo], $ind->[$lo+1] ); push @$to, shift @$from while $n-- && @$from } $ind } # this clones an element of @pop sub clone { [ $_[0]->[0], Ind::clone($_[0]->[1]) ] } sub do_run { my $n_clusters = shift; my @pop = sort { $b->[0] <=> $a->[0] } map { [ Ind::fitness($_), $_ ] } map { Ind::new_randomized($n_clusters) } 1 .. 60; for my $gen ( 1 .. 200 ) { # kill the bottom 30: splice @pop, @pop-30, 30; # make 10 new ones: push @pop, map { [ Ind::fitness($_), $_ ] } map { Ind::new_randomized($n_clusters) } 1 .. 10; # clone the top 20: push @pop, map clone($_), @pop[0 .. 19]; # mutate the top 20: for my $e ( @pop[0 .. 19] ) { Ind::mutate( $e->[1], 1 + int rand 4 ); $e->[0] = Ind::fitness( $e->[1] ); } # sort by fitness again: @pop = sort { $b->[0] <=> $a->[0] } @pop; } $pop[0][1] # best individual } for my $nc ( 2 .. 8 ) { print "\n$nc Clusters:\n"; my $winner = do_run( $nc ); print Ind::as_string( $winner ), "\n"; }
We're building the house of the future together.

In reply to Re^3: Making sense of data: Clustering OR A coding challenge by jdporter
in thread Making sense of data: Clustering OR A coding challenge by belg4mit

Title:
Use:  <p> text here (a paragraph) </p>
and:  <code> code here </code>
to format your post; it's "PerlMonks-approved HTML":



  • Posts are HTML formatted. Put <p> </p> tags around your paragraphs. Put <code> </code> tags around your code and data!
  • Read Where should I post X? if you're not absolutely sure you're posting in the right place.
  • Please read these before you post! —
  • Posts may use any of the Perl Monks Approved HTML tags:
    a, abbr, b, big, blockquote, br, caption, center, col, colgroup, dd, del, div, dl, dt, em, font, h1, h2, h3, h4, h5, h6, hr, i, ins, li, ol, p, pre, readmore, small, span, spoiler, strike, strong, sub, sup, table, tbody, td, tfoot, th, thead, tr, tt, u, ul, wbr
  • Outside of code tags, you may need to use entities for some characters:
            For:     Use:
    & &amp;
    < &lt;
    > &gt;
    [ &#91;
    ] &#93;
  • Link using PerlMonks shortcuts! What shortcuts can I use for linking?
  • See Writeup Formatting Tips and other pages linked from there for more info.
  • Log In?
    Username:
    Password:

    What's my password?
    Create A New User
    Chatterbox?
    and the web crawler heard nothing...

    How do I use this? | Other CB clients
    Other Users?
    Others meditating upon the Monastery: (7)
    As of 2014-07-28 04:57 GMT
    Sections?
    Information?
    Find Nodes?
    Leftovers?
      Voting Booth?

      My favorite superfluous repetitious redundant duplicative phrase is:









      Results (186 votes), past polls