clrn

it is the Data Mining Algorithm source code.

#!/bin/bash

sel=${1:-all}
cnt=${2:-20}
off=${3:-0}
add=$4

#-----------------------------------------------------------------------

function run ()
{                               # --- run one clustering trial
  echo tuples: $2 >> $data.tmp  # do the clustering
  cli -a$1 -k$2 $args 2> /dev/null
  rm -f $data.tmp               # traverse the training epochs
  for (( i = 0; i <= nepo; i += step )); do
    clc -x $data.$i $best $data.tab 2> /dev/null >> $data.tmp
  done                          # compare clustering results
  gawk -v step=$step 'BEGIN { n = 0 }
    ($0 ~ "^mean squared") { printf("%d %s\n", n, $NF); n += step; }
  ' $data.tmp >> $data.tmp      # collect the comparison results
  rm -f $data.[0-9]*            # remove the result files
}  # run ()

#-----------------------------------------------------------------------

function modes ()
{                               # --- try different training modes
  data=${1:-iris}               # get the data name
  ncls=${2:-2}                  # get the number of clusters
  nepo=${3:-20}                 # get the number of epochs
  if   (( nepo >= 600 )); then inc=50;
  elif (( nepo >= 480 )); then inc=40;
  elif (( nepo >= 240 )); then inc=20;
  elif (( nepo >= 120 )); then inc=10;
  elif (( nepo >=  60 )); then inc=5;
  elif (( nepo >=  48 )); then inc=4;
  elif (( nepo >=  24 )); then inc=2;
  else                         inc=1; fi
  echo shape: $4 >> $data.tmp   # record the cluster shape
  for (( k = 0; k < cnt; k++ )); do
    seed=$(( (7*k)+1+off ))     # compute the seed value
    args="-S$seed -$4c$ncls -e1000 $add $data.dom $data.tab $data.ref"
    cli $args 2> /dev/null      # create the reference result
    args="-S$seed -$4c$ncls $5 -Pe$nepo $add $data.dom $data.tab $data."
    list=`gawk 'END { printf("%d", n = NR-1); \
      for (i = 1; ++i <= 6; ) printf(" %d", n/i);
      printf("\n"); }' $data.tab`
    run altopt 0                # start with standard clustering
    for n in $list; do          # compute list of tuple numbers
      run complrn $n; done      # do one rune with each number of tuples
  done
  rm -f $data.ref               # delete the reference result
}  # modes ()

#-----------------------------------------------------------------------

function average ()
{                               # --- average data over trials
  gawk '
  BEGIN { max = -1; n = 0; }
  {
    if      ($1 == "shape:")    # if cluster shape specification
      shape = $2;               # note the cluster shape
    else if ($1 == "tuples:") { # if number of tuples
      m = shape $2;             # construct the mode name
      for (i = n; --i >= 0; )   # add mode to the list if necessary
        if (tuples[i] == m) break;
      if (i < 0) tuples[i = n++] = m; }
    else {                      # if result line
      if ($1 > max) max = $1;   # update the maximum epoch
      sums[m,$1] += ($2 > 0) ? log($2)/log(10) : -16;
      cnts[m,$1]++;             # store the trial value and count it
      totals[$1]++;             # count the trial for the epoch
    }
  }
  END {                         # output collected data
    for (i = 0; i <= max; i++) {
      if (totals[i] <= 0) continue;
      printf("> %10d diff\n", i);
      for (k = 0; k < n; k++) {
        m = tuples[k];
        if (cnts[m,i] <= 0) continue;
        printf("%-12s %g\n", m, sums[m,i]/cnts[m,i]);
      }
    }
  }' $1.tmp >> ${2:-$1}.res
}  # average ()

#-----------------------------------------------------------------------

if [[ $sel == "all" || $sel == "iris" ]]; then
  rm -f iris.tmp
  modes iris 3 30
  modes iris 3 30 v
  modes iris 3 80 V
  rm -f iriscl.res
  average iris iriscl
  rm -f iris.tmp
fi

if [[ $sel == "all" || $sel == "wsel3" ]]; then
  rm -f wsel.tmp
  modes wsel 3 20
  modes wsel 3 20 v
  modes wsel 3 30 V
  rm -f wselcl3.res
  average wsel wselcl3
  rm -f wsel.tmp
fi

if [[ $sel == "all" || $sel == "wsel6" ]]; then
  rm -f wsel.tmp
  modes wsel 6  40
  modes wsel 6  80 v
  modes wsel 6 120 V
  rm -f wselcl6.res
  average wsel wselcl6
  rm -f wsel.tmp
fi

if [[ $sel == "all" || $sel == "wine3" ]]; then
  rm -f wine.tmp
  modes wine 3 20
  modes wine 3 20 v
  modes wine 3 30 V
  rm -f winecl3.res
  average wine winecl3
  rm -f wine.tmp
fi

if [[ $sel == "all" || $sel == "wine6" ]]; then
  rm -f wine.tmp
  modes wine 6  40
  modes wine 6  80 v
  modes wine 6 120 V
  rm -f winecl6.res
  average wine winecl6
  rm -f wine.tmp
fi

if [[ $sel == "all" || $sel == "breast" ]]; then
  rm -f breast.tmp
  modes breast 2  10
  modes breast 2  20 v
  modes breast 2 100 V
  rm -f brstcl.res
  average breast brstcl
  rm -f breast.tmp
fi

if [[ $sel == "all" || $sel == "abalone3" ]]; then
  rm -f abalone.tmp
  modes abalone 3  30
  modes abalone 3  30 v
  modes abalone 3 160 V
  rm -f abacl3.res
  average abalone abacl3
  rm -f abalone.tmp
fi

if [[ $sel == "all" || $sel == "abalone6" ]]; then
  rm -f abalone.tmp
  modes abalone 6  80
  modes abalone 6  80 v
  modes abalone 6 200 V
  rm -f abacl6.res
  average abalone abacl6
  rm -f abalone.tmp
fi