Zad9.1

classify.sh

cat in40 | awk '{$1=""; print $0}' > in0
name=`echo -n $1 | sed s/' '//g`
echo $name
mkdir models${name}
for file in labels/*; do
  echo $file
  i=`echo -n $file | sed s/labels.//`
  paste $file in0 -d " " > "in${i}"
  ../svmlight/svm_perf_learn "$1" "in${i}" "models${name}/model${i}"
  rm "in${i}";
done

mkdir predictions${name}
for file in models${name}/*; do
  echo $file
  i=`echo -n $file | sed s/models${name}.model//`
  ../svmlight/svm_perf_classify intest $file "predictions${name}/${i}"
done

formatowanie_danych.R

sink('in40')
sapply(1:nrow(train), function(i) { 
  row <- train[i, ]; 
  cat(ifelse(labels[i, 40], '+1 ', '-1 ')); 
  cat(mapply(function(name, v) paste(name, v, sep=':'), which(row != 0) -> w, round(row[w], 3))); 
  cat('\n')
}) -> foo
sink()

sink('intest')
sapply(1:nrow(test), function(i) { 
  row <- test[i, ]; 
  cat('0 ')
  cat(mapply(function(name, v) paste(name, v, sep=':'), which(row != 0) -> w, round(row[w], 3))); 
  cat('\n')
}) -> foo
sink()

wczytanie_danych.R

labels <- read.csv('./data/raw/trainingLabels.txt', header=T)
as.vector(as.matrix(t(labels))) -> l
d <- cbind(rep(1:nrow(labels), each=ncol(labels)), l)
d[!is.na(d[,2]),] -> d
tapply(1:nrow(d), list(d[,1], d[,2]), function(ii) T) -> dd
dd[is.na(dd)] <- F
labels <- dd
rownames(labels) <- c()
# data <- data / max(data)
save(labels, file="data/labels.RD")

which.max(colSums(labels)) # 40