train_models_performance: Train models

In specmine: Metabolomics and Spectral Data Analysis and Mining

Description

Train various models.

Usage

train_models_performance(dataset, models, column.class, 
validation, num.folds = 10, num.repeats = 10, tunelength = 10, 
tunegrid = NULL, metric = NULL, summary.function = "default", 
class.in.metadata = TRUE, compute.varimp = TRUE)

Arguments

dataset	list representing the dataset from a metabolomics experiment.
models	models to be used in training.
column.class	metadata column class.
validation	validation method.
num.folds	number of folds in cross validation.
num.repeats	number of repeats.
tunelength	number of levels for each tuning parameters.
tunegrid	dataframe with possible tuning values.
metric	metric used to evaluate the model's performance. Can be "Accuracy" or "ROC".
summary.function	summary function. For "ROC" the multiClassSummary function must be used.
class.in.metadata	boolean value to indicate if the class is in metadata.
compute.varimp	boolean value to indicate if the var importance is calculated.

Value

Returns a list with the results from training

performance	The results from the best tunes of the models
vips	The variable importance from the models
full.results	The full results from the tuning parameters of each model
best.tunes	The best tune of each model
confusion.matrices	The confusion matrices of the models (only in classification)
final.models	The final models

Examples

  ## Example of training models
  library(specmine.datasets)
  data(cachexia)
  result = train_models_performance(cachexia, "pls", 
  	   "Muscle.loss", "cv")

specmine documentation built on Sept. 21, 2021, 5:06 p.m.