Exercise

# Fit a random forest with custom tuning

Now that you've explored the default tuning grids provided by the `train()`

function, let's customize your models a bit more.

You can provide any number of values for `mtry`

, from 2 up to the number of columns in the dataset. In practice, there are diminishing returns for much larger values of `mtry`

, so you will use a custom tuning grid that explores 2 simple models (`mtry = 2`

and `mtry = 3`

) as well as one more complicated model (`mtry = 7`

).

Instructions 1/2

**undefined XP**

- Define a custom tuning grid.
- Set the number of variables to possibly split at each node,
`.mtry`

, to a vector of`2`

,`3`

, and`7`

. - Set the rule to split on,
`.splitrule`

, to`"variance"`

. - Set the minimum node size,
`.min.node.size`

, to`5`

.

- Set the number of variables to possibly split at each node,