featR
Filter, wrapper, and embedded feature-selection methods behind a consistent set of functions that share one calling convention and one return type: correlation and chi-squared filters, information gain, LASSO and elastic net, Bayesian model comparison, Boruta, recursive feature elimination, random forest importance, multivariate adaptive regression splines, support vector machine recursive feature elimination, stepwise selection, and principal component / singular value decomposition helpers. The implemented methods follow Tibshirani (1996) <doi:10.1111/j.2517-6161.1996.tb02080.x>, Zou and Hastie (2005) <doi:10.1111/j.1467-9868.2005.00503.x>, Friedman (1991) <doi:10.1214/aos/1176347963>, Breiman (2001) <doi:10.1023/A:1010933404324>, Guyon, Weston, Barnhill and Vapnik (2002) <doi:10.1023/A:1012487302797>, Kursa and Rudnicki (2010) <doi:10.18637/jss.v036.i11>, and Vehtari, Gelman and Gabry (2017) <doi:10.1007/s11222-016-9696-4>. Heavy modeling engines are optional and only required by the functions that use them.
- Feature Selection with the Boruta Package
- Gene selection for cancer classification using support vector machines
- Unnamed Publication
- Multivariate Adaptive Regression Splines
- Practical Bayesian model evaluation using leave-one-out cross-validation and WAIC
- Regression Shrinkage and Selection Via the Lasso
- Regularization and Variable Selection Via the Elastic Net
This page was built for software: featR
