README.md

In alessandromagrini/caretAddOn: Additional Functionalities for the 'caret' Package

caretAddOn

Additional functionalities for the R package 'caret'

This package contains additional functionalities for the caret package, including new models, new summaries, graphical tools, general-purpose functions, and datasets. Reference for the caret package:

M. Kuhn (2022). caret: Classification and Regression Training. R package version 6.0-92. CRAN page: https://CRAN.R-project.org/package=caret. Github page: https://github.com/topepo/caret

New models: - loglm: linear regression on log transformed response variable, with prediction made through either the conditional expected value (bias.adj=TRUE) or the conditional median (bias.adj=FALSE) - svm_linear: support vector machine/regression with linear kernel - svm_radial: support vector machine/regression with radial kernel

New summaries: - customSummaryClass: summary for classification tasks, including AUC and optimized specificity/sensitivity in case of class imbalance - customSummaryReg: summary for regression tasks, including RMSE, MAE, MAPE, and R-squared

New graphics: - multiPairPlot: bivariate graphics displaying the relationship between one response variable and several explanatory variables - corPlot: correlogram, conceived to check collinearity among the explanatory variables - trainPlot: graphic displaying any one metric as a function of any one hyperparameter - importancePlot: graphic displaying variable importance metrics computed through the function importanceCalc (see below) - rocPlot: ROC curve (only for binary classification) - densPlot: graphic displaying kernel density estimations for class probabilities (only for binary classification), conceived to assess the discriminative power - predPlot: scatterplot of observed values versus predicted values (only for regression tasks)

New general-purpose functions: - bestTune: best tuning of hyperparameters - fitted (S3 method for class train): cross-validation prediction for each unit - importanceCalc: computation of variable importance metrics. For models of class lm or glm, the proportion of explained deviance is used in place of the absolute t-statistic. - vifCalc: computation of variance inflation factors, conceived to check collinearity among the explanatory variables - addTerms: addition of polynomial and/or logarithmic terms to a formula - stepCV: backward selection of explanatory variables through cross-validation - stepAIC_train: stepwise selection of explanatory variables is performed through information criteria, then cross-validation is run to compute performance metrics (faster than stepCV) - enetCoef: coefficients at best tuning of hyperparameters for an object of class glmnet

alessandromagrini/caretAddOn documentation built on June 10, 2025, 10:25 p.m.