Me=$0
Here=`pwd`
cd $HOME/work/mine/trunk/doc/06/markland

worker() {
    printf "\n\n----| $1 |-------\n\n"
	silap -R $0.$1.risk -n 500 $1.csv    |tee /tmp/$$
     gawk -F, '
            function trim(x) { sub(/^ */,"",x); sub(/ *$/,"",x); return x }
			Pass==1 { Names[++Names[0]]=$1 }
            Pass==2 { 
                      $1 = trim($1) 
					  sub(/\%/,"",$3);
                      Data[$1]=$3
                    }
            END     {
					 for(i=1; i<=Names[0]; i++)  {
							j= Names[i];
							print (j in Data) ? Data[j] : 0
                    }}
            ' Pass=1 $Here/15.in Pass=2 /tmp/$$  > $0.$1.selected
}
for i in 0  1 2 3 
do
	worker 	cluster${i}.data433wekaClusteredSorted         
	cat cluster${i}.data433wekaClusteredSorted.csv > all.csv
done 

worker all


paste $0.*selected | sort -n | cat -n > /dev/null

for i in *.risk
do
	echo "=== $i ==="
	gawk '{X[$0]++; N++} END {for(i in X) print 100*X[i]/N " " i}' $i | sort -n
done

exit

wc	cluster*data433wekaClusteredSorted.csv   
ls $0.selected.*

gnuplot <<EOF
set key bottom right

set terminal postscript eps 15
set noxtics
set size 0.75,0.5
set title "Clusters"
set output "$Here/../doc/selectedClusters.eps"
set ylabel "%times selected"
set xlabel "tasks, sorted on majority frequency"
set xrange [-5:60]
plot "$0.dat" using 1:2 title "c0" with linesp,\
	 "$0.dat" using 1:3 title "c1" with linesp,\
     "$0.dat" using 1:4 title "c2" with linesp,\
	 "$0.dat" using 1:5 title "c3" with linesp
EOF