Addding target column to test dataset



I am working on the project Digit Recognizer, i divided my train dataset into train and test in the ratio 75 and 25 percent.

Now my test dataset does not contain the column label, so i wish to know how to used it to do my prediction. How do i add the column label?


Can you share the code snippet for the operations you have done?

data <- read.csv ("train.csv")
test <- read.csv ("test.csv")


min(data[2:785]) #0 for balck and 255 for white
#lets take a look at t2 samples first n fouth
sample_4 <- matrix(as.numeric(data[4,-1]), nrow = 28, byrow = TRUE)
rotate <- function(x) t(apply(x, 2, rev))
image(rotate(sample_4), col = grey.colors(255))

sample_7 <- matrix(as.numeric(data[7,-1]), nrow = 28, byrow = TRUE)
rotate <- function(x) t(apply(x, 2, rev))
image(rotate(sample_7), col = grey.colors(255))

# Transform target variable "label" from integer to factor, in order to perform classification

data$label <- as.factor(data$label)


proportion <- prop.table(table(data$label)) * 100
cbind(count=table(data$label), proportion=proportion)

central_block <- c("pixel376", "pixel377", "pixel404", "pixel405")
par(mfrow=c(2, 2))
for(i in 1:9) {
  hist(c(as.matrix(data[data$label==i, central_block])),
         main=sprintf("Histogram for digit %d", i),
         xlab="Pixel value")

if (!require("caret"))
library (caret)
#split the data into txo partitions 75% for training and 25% for testing 
train_perc = 0.75
train_index <- createDataPartition(data$label, p=train_perc, list=FALSE)

data_train <- data[train_index,]
data_test <- data[-train_index,]

# Multinomial logistic regression
model_lr <- multinom(label ~ ., data=data_train, MaxNWts=10000, decay=5e-3, maxit=100)
#make predictions                       
prediction_lr <- predict(model_lr, test, type = "class")


Isn’t column label what you are trying to predict? That’s why it is not present in the test data set. I’ve not looked at the data but from the title “Digit Recognizer”, I’m guessing this is a multi class classification problem. Joining the train and test data sets is required for variable transformations or creating new features simultaneously. But once you have done the feature engineering separate train and test and use any techniques to predict “Column Labels”. I think techniques like SVM or GBM should work well.


@soumadiptya is right. because the column “label” is the target, and thats what you have to predict.

To clarify, I will briefly explain the Digit Recognition problem.

Here, we need to identify the digit in given images. We have total 70,000 images, out of which 49,000 are part of train images with the label of digit and rest 21,000 images are unlabeled (known as test images). Now, We need to identify the digit for test images.

For example,

Suppose the image is as given below, your label for that image is 8