Abstract

Supervised learning is widely used in biology for prediction, and ensemble learning, including stacking, is a promising technique for increasing and stabilizing the prediction accuracy. In this study, we developed an R package for stacking. This package depends on the R package caret and can handle models supported by caret. Stacking involves cross-validation of training data with multiple base learners, and the predicted values are used as explanatory variables for the meta-learner. In the prediction, the testing data were fed into the base models, and the returned values were averaged for each base learner. The averaged values were then fed into the meta-model, and the final predictions were returned. Using this package, the training and prediction procedures for stacking can be conducted using one-row scripts. The R package stacking is available at the Comprehensive R Archive Network (CRAN) (https://cran.r-project.org/) and GitHub (https://github.com/Onogi/stacking). R scripts to reproduce the presented results are also reposited at GitHub.

Full Text
Published version (Free)

Talk to us

Join us for a 30 min session where you can share your feedback and ask us any queries you have

Schedule a call