Ensemble boosting classifier for multinomial classification

Share:

Description

Makes multiple predictions on the test set using samples chosen from the data set based on past performance. Uses randomForest as the weak classifier and expects a multinomial character outcome variable. Makes a final prediction on the test set based on votes by the classifiers weighted by performance. Implementation of Zhu et al.'s Stagewise Additive Modeling using a Multi-class Exponential loss function (2006).

Usage

1
samme(formula, data, test, m = 5, trace = T, ntree = 500, mtry = NULL)

Arguments

formula

A formula expression for the classifier model. Expects the form outcome ~ predictors with a single outcome variable. The outcome variable is expected to be of type character.

data

A data frame consisting of the training data set. Must include all variables described in the formula including the outcome variable.

test

A data frame consisting of the test set to be predicted. Must include all predictor variables. May include outcome variable.

m

Number of classifiers to vote on final prediction. Defaults to 5.

trace

Setting for randomForest do.trace. Defaults to TRUE.

ntree

Number of trees for each randomForest to grow. Defaults to 500.

mtry

Number of variables randomForest randomly samples as candidates at each split. Defaults to the square root of the number of variables in data.

Value

Character vector consisting of a final prediction for each test set sample.

Author(s)

Shannon Rush shannonmrush@gmail.com

References

Zhu, J., Rosset, S., Zou, H. and Hastie, T. (2005). Multiclass AdaBoost. Technical report, Stanford Univ. http://www.stanford.edu/~hastie/Papers/samme.pdf

See Also

randomForest

Want to suggest features or report bugs for rdrr.io? Use the GitHub issue tracker.