plot.ModelComparison: Plot Comparisons

In orionw/BestModel: A package to run and compare standard machine learning models with easy visualization

Description

This function evalutates many different machine learning models and returns plots comparing them.

Usage

## S3 method for class 'ModelComparison'
plot(object, labels, training.data = "none",
  predictions = "empty", plot.type = "ROC", format.data = TRUE, ...)

Arguments

object	The ModelComparison object
labels	The labels of the training set
training.data	The dataset to be trained on. If predictions are provided this is not needed.
predictions	The list of predictions from the models, optional if training.data is not provided.
plot.type	A vector of metrics (as characters) that are the values seen in the plot (examples include ROC, AUC, Accuracy, etc.) Note: ROC cannot be plotted with other metrics.
format.data	Whether the data should be transformed into one-hot encoding if it needs to be. The default is TRUE. If you would like to predict on unchanged data that is not in the right format, set this to false at your own risk.
...	Other arguments for plotting

Examples

# prepare the dataset
titanic <- PrepareNumericTitanic()

# create the models
comp <- GetModelComparisons(titanic[, -1], titanic[, 1], model.list = "all")

# Default.  Plot AUC, Accuracy, Recall, and Precision
plot(comp, titanic[, 1], titanic[, -1], plot.type=c("All"))

# Choose specific metrics
plot(comp, titanic[, 1], titanic[, -1], plot.type=c("Specificity", "Precision", "AUC",
"Recall", "Detection Rate"))

# plot overlapping ROC lines
plot(comp, titanic[, 1], titanic[, -1], plot.type="roc")

orionw/BestModel documentation built on Aug. 17, 2019, 7:29 p.m.