Allow 'mtry'

[tslumley]

Would you be interested in a patch to allow users to adjust mtry? It's a major tuning parameter for random forests, so I think it would help. What I have done is to add a mtry_rule argument, which is a function taking an vector of variable names as input and returning a value of mtry, called with union(v,completed). The current rule is roughly

function(x) floor(sqrt(length(x)-1)))

and I've tried others such as

function(x) min(5, length(x)-1)

[mayer79]

Your approach is very good. ranger actually already allows mtry to be a function of the number of covariables, so we only have to stick to this one.

E.g.

ranger(Sepal.Length ~ ., data = iris, mtry = function(x) x%/%2) 

gives mtry = 2.

Interestingly, the default passed to the internal C++ interface is mtry=0 and is internally set to something like floor(sqrt(m)). However, an mtry function is not allowed to return 0.

Suggestion 1: We add the argument mtry = NULL and allow only the "function" interface of ranger otherwise.

Suggestion 2: Same but we allow also numeric values for mtry, e.g. 1. When calling ranger, the value is capped from above with the number of covariables. We could mention in the argument description something like "since the number of covariables might change during the process, consider passing a single argument function of the number of covariables, e.g. function(m) = m %/% 3.

[mayer79]

Now in master branch.