FittedGridSearch is an object containing fitted predictive models across
a tuning grid of hyper-parameters returned by GridSearch$fit() as well as
relevant model information such as the best performing model, best
hyper-parameters, etc.
Public fields
best_idxAn integer specifying the index of
$modelsthat contains the best-performing model.best_metricThe performance metric of the best model on the validation data.
best_modelThe best performing predictive model.
best_paramsA named list of the hyper-parameters that result in the optimal predictive model.
tune_paramsData.frame of the full hyper-parameter grid.
modelsList of predictive models at every value of
$tune_params.metricsNumeric list; Cross-validation performance metrics on each fold.
predictionsA list containing the predicted hold-out values on every fold.
Methods
FittedGridSearch$new()
Create a new FittedGridSearch object.
Usage
FittedGridSearch$new(tune_params, models, metrics, predictions, optimize_score)Arguments
tune_paramsData.frame of the full hyper-parameter grid.
modelsList of predictive models at every value of
$tune_params.metricsList of performance metrics on the validation data for every model in
$models.predictionsA list containing the predicted values on the validation data for every model in
$models.optimize_scoreEither "max" or "min" indicating whether or not the specified performance metric was maximized or minimized to find the optimal predictive model.