Description Usage Arguments Value Author(s) Examples
This function provides a 'population' estimate of the average OOB error computed for different mtry values, starting from a sample of N models. These values will be used to compute the mtry associated to the minimum averaged OOB error, that is the optimal parameter we are looking for.
1 | optimizeMTRY(oob_matrix)
|
oob_matrix |
a n x p of n OOB error values (one for each iteration) and p columns (one for each mtry value tested) Each value of a column is the oob error of a model growth with a particular mtry. Typically for each mtry, we will have N different models (N > 30), a sample large enough to provide an estimate of the average OOB error for the corresponding population of models. |
a list of two elements:
mean_matrix a 1 x p matrix which contains the mean of each OOB errors sample (resulting from the training of N different Random Forest models growth with N different mtry values)
ci_matrix a 2 x p matrix in which each column represents the 95% confidence interval of the mean of the population of the OOB errors for each mtry value
sd_matrix a 1 x p matrix which contains the standard deviatiaon of each OOB error sample resulting from the training of N different models built for each value of mtry
Piergiorgio Palla
1 2 3 | ## data(cachexiaData)
## res <- tuneMTRY(cachexiaData, iterations = 50, maxntree = 600, mtry_length = 10, graph = F)
## l <- optimizeMTRY(res$oob)
|
Add the following code to your website.
For more information on customizing the embed code, read Embedding Snippets.