Part $($b$)$

Create data frames train and test from the data frame $df.$ The data frame train should consist of observations $($i$.$e$.$ rows$)$ of $df$ whose indices

are precisely those $400$ random indices from vector $tr.$ The data frame test should consist of the complement of $tr($you can use df$[-$tr$,]$ for

taking the complementary indices, i$.$e$.$ precisely those indices which are not in $tr).$

Hint: First two rows of train should be

## check whether train and test data are correct $($hidden tests$)$

We now create vector kvec of possible values for the tuning parameter $($i$.$e$.$ hyperparameter$)k$ in the $kNN$ model. We will take values $1,2,$dots,$350.$ We could

also include values up to the size of the training sample train $($which is $400),$ but it's not difficult to see that the performance of the model for such

values of $k$ would be lower than performance for all other values of $k.$

Apart from kvec, we also define $nk$ to be the length of kvec $($i$.$e$.350).$ Run the following cell to create kvec and $nk.$

## run this code to create kvec and define $nk$

#loop over values of $k,$ fit on train, predict on test

kvec $=1$:$350$

$nk=$ length $(kvec)$