optimize.model: Model Optimization and Metrics

Description

Optimizes each model based upon the parameters provided either by the internal denovo.grid function or by the user.

Usage

optimize.model(trainVars, trainGroup, method, k.folds = 10, repeats = 3, res = 3, grid = NULL, metric = "Accuracy", allowParallel = FALSE, verbose = "none", theDots = NULL)

Arguments

trainVars

Data used to fit the model

trainGroup

Group identifiers for the training data

method

A vector of strings listing models to be optimized

k.folds

Number of folds generated during cross-validation. Default "k.folds = 10"

repeats

Number of times cross-validation repeated. Default "repeats = 3"

res

Resolution of model optimization grid. Default "res = 3"

grid

Optional list of grids containing parameters to optimize for each algorithm. Default "grid = NULL" lets function create grid determined by "res"

metric

Criteria for model optimization. Available options are "Accuracy" (Predication Accuracy), "Kappa" (Kappa Statistic), and "AUC-ROC" (Area Under the Curve - Receiver Operator Curve)

allowParallel

Logical argument dictating if parallel processing is allowed via foreach package

verbose

Character argument specifying how much output progress to print. Options are 'none', 'minimal' or 'full'.

theDots

List of additional arguments provided in the initial classification and features selection function

Value

method: Vector of strings listing models that were optimized
performance: Performance generated internally to optimize model
bestTune: List of paramaters chosen for each model
dots: List of extra arguments initially provided
metric: Criteria that was used for model optimization
finalModels: The fitted models with the 'optimum' parameters
performance.metrics: The performance metrics calculated internally for each resulting prediction
tune.metrics: The results from each tune
perfNames: The names of the performance metrics
comp.catch: If the optimal PLSDA model contains only 1 component, the model must be refit with 2 components. This catches the 1 component parameter so feature selection and further performance analysis can be conducted on the 1 component.