Hi theobroma22

My Initial matrix is 920x60. I used the 70% of this as training set and the rest 30% as test set. I did this 100 different times using sample() function and for k number 1:20. While the error rate i counted as below:

for(y in 1:100){
      train_set = sample(...)
      test_set = .....

      accuracy = numeric()

      # Repeat the prediction for 1:20 k values
      for(k in 1:20){
          # execute the kNN algorithm
          predict = knn( train = train_set , test = test_set , cl = train_labels , k=k)

          # Get the accuracy i.e the mean of the correctly predicted classes for each k value
          accuracy = c(accuracy , mean(predict == test_set$classes))
      }

      # Store each accuracy variable into accuuracy matrix.
      accuracy_matrix[s,] = accuracy
 }

    # Calculate the mean accuracy (mean of mean) for each k value
    mean_accuracy = as.numeric(lapply(accuracy_matrix,mean))

    # Plot the mean accuracy values for each k.
    plot(1-mean_accuracy ,type="l",ylab="Mean error rate",
         xlab="K",main="Error Rate for Treatments With Varying K (1:20) and 100 samples")