An R Package to return and compare variety of different model types, complete with hyper-parameter tuning options
Install by using install_github("orionw/ModelComparison") after installing and loading the devtools library
Example usage:
library(ModelComparison)
# prepare the dataset. This function creates a two class Iris dataset.
iris_data <- PrepareIris()
# create the models. This includes SVM's, K-NN, A 4 layer Neural Network, and Linear or Logistic Regression.
comp <- GetModelComparisons(iris_data[, -5], iris_data[, 5], model.list = "all")
# get prediction values for the models
preds = predict(comp, newdata = iris_data[, -5], type="prob")
# Default. Plot AUC, Accuracy, Recall, and Precision
plot(comp, iris_data[, 5], predictions=preds, plot.type=c("All"))
# Choose specific metrics
plot(comp, iris_data[, 5], predictions=preds, plot.type=c("Specificity", "Precision", "AUC", "Recall", "Detection Rate"))
# plot overlapping ROC lines from all models
plot(comp, iris_data[, 5], predictions=preds, plot.type="roc")