baseliner

The goal of baseliner is to have very basic models that can serve as a nice baseline for all your ML endeavours. It is nice to be able to have a baseline that you can beat.

Installation

And the development version from GitHub with:

# install.packages("devtools")
devtools::install_github("koaning/baseliner")

Features

We missed some of the dummy regressors and classifiers available in scikit learn. This repo contains these simple models that are fit for benchmarking purposes and some extra ones as well.

Example

This is a basic example which shows you how to solve a common problem:

library(baseliner)
# get yer dataframe ready
dataf <- data.frame(y = 1:10, x = 1:10)
# make a dummy model for benchmarking
mod <- dummy_regressor(y ~ ., data=dataf, strategy="mean")
# get 'yer test data and predict 
test_data <- data.frame(x = 1:5)
pred <- predict(mod, test_data) 
# well done! you now have an easy benchmark to beat