rf_classify_multilabel.R

cols <- names(train)[1:12]
system.time(clf <- randomForest(factor(quality) ~ ., data=train[,cols], ntree=20, nodesize=5, mtry=9))
  #  user  system elapsed
  # 0.366   0.006   0.372

table(test$quality, predict(clf, test[cols]))
  #     3   4   5   6   7   8   9
  # 3   0   0   4   5   0   0   0
  # 4   0   0  35  22   1   0   0
  # 5   0   0 375 137   6   1   0
  # 6   0   1 129 531  52   1   0
  # 7   0   0  13 113 150   2   0
  # 8   0   0   0  19  13  18   0
  # 9   0   0   0   3   0   0   0
sum(test$quality==predict(clf, test[cols])) / nrow(test)
  # [1] 0.6609