run

聚类算法全集以及内附数据集

#!/bin/bash

alg=${1:-"fuzzy"}
cnt=${2:-10}

rm -f iris.res
if [ $alg == "fuzzy" ]; then
  echo "fuzzy clustering"     | tee -a iris.res
else
  echo "competitive learning" | tee -a iris.res
fi

dom -sa iris.tab - 2> /dev/null | sed "/iris_type/d" > iris.dnc
sed 's/Iris-setosa/0/;s/Iris-versicolor/1/;s/Iris-virginica/2/;' \
iris.tab > iris.trn

init="-icenter -o0.5"
fcp=""
clp="-acomplrn -k30 -t0.3:0.1"
std="-I0.4 -H-4 -Vc3 iris.dnc iris.trn iris.cls"

for (( seed=1 ; seed <= cnt ; seed++ )); do

  if [ $alg == "fuzzy" ]; then
    cli -e30 -S$seed $init $std $fcp 2> /dev/null
  else 
    cli -e30 -S$seed $init $std $clp 2> /dev/null
  fi

  echo "" | tee -a iris.res
  clx -cclx iris.cls iris.trn iris.out 2> /dev/null
  xmat -c iris.out 2> /dev/null | tee -a iris.res

done

cnt=`wc iris.trn | gawk '{ printf("%d", $1-1); }'`

gawk -v k=$cnt '
BEGIN            { n = 0; }
($2 == "errors") { e[n++] = $NF; }
END {
  for (i = 0; i < n; i++) {
    sum += e[i]; sqr += e[i] * e[i]; }
  printf("\nstatistics:\n");
  printf("cnt: %d, ", n);
  t = (n > 0) ? sum /n : 0;
  printf("avg: %5.1f (%4.1f), ", t, t/k*100);
  t = (n > 1) ? sqrt((sqr -sum*sum/n) /(n-1)) : 1;
  printf("std: %5.1f (%4.1f)\n", t, t/k*100);
}' iris.res

rm -f iris.cls iris.dnc iris.out iris.res