bolasso
The goal of bolasso is to apply robust feature selection via lasso and bootsrtapping.
Installation
You can install bolasso from github with:
# install.packages("devtools")
devtools::install_github("david26694/bolasso")Example
This is a basic example which shows you how to apply bolasso in a regression setting:
## basic example code
library(bolasso)
# We want to predict vs with the other mtcars variables
predictors <- mtcars[,c("mpg", "disp", "hp", "drat", "wt")]
outcome <- mtcars[, "vs"]
# Apply bolasso with 10 bootstrap samples
model <- bolasso(predictors, outcome, n_bootstraps = 10)
# Show variable selection results
summary(model)
#> # A tibble: 5 x 4
#> feature selection_ratio coefficient selected_bolasso
#> <chr> <dbl> <dbl> <lgl>
#> 1 hp 1 -0.00273 TRUE
#> 2 disp 0.9 -0.00139 TRUE
#> 3 drat 0.5 NA FALSE
#> 4 mpg 0.5 NA FALSE
#> 5 wt 0.5 NA FALSE
# It can also run on recipes, and with formulas
library(recipes)
#> Loading required package: dplyr
#> Warning: package 'dplyr' was built under R version 3.5.2
#>
#> Attaching package: 'dplyr'
#> The following objects are masked from 'package:stats':
#>
#> filter, lag
#> The following objects are masked from 'package:base':
#>
#> intersect, setdiff, setequal, union
#>
#> Attaching package: 'recipes'
#> The following object is masked from 'package:stats':
#>
#> step
rec <- recipe(mpg ~ ., mtcars)
rec <- step_log(rec, disp)
model_rec <- bolasso(rec, mtcars, n_bootstraps = 20)
summary(model_rec)
#> # A tibble: 10 x 4
#> feature selection_ratio coefficient selected_bolasso
#> <chr> <dbl> <dbl> <lgl>
#> 1 wt 1 -1.73 TRUE
#> 2 disp 0.9 -4.39 TRUE
#> 3 carb 0.85 -0.785 TRUE
#> 4 hp 0.65 NA FALSE
#> 5 drat 0.6 1.40 TRUE
#> 6 am 0.35 NA FALSE
#> 7 gear 0.35 NA FALSE
#> 8 qsec 0.35 NA FALSE
#> 9 vs 0.3 NA FALSE
#> 10 cyl 0.2 NA FALSE
# It has different types of prediction, like numeric, class and prob
predict(model_rec, mtcars, type = 'numeric')
#> # A tibble: 32 x 1
#> .pred
#> <dbl>
#> 1 21.5
#> 2 21.1
#> 3 26.1
#> 4 19.6
#> 5 17.1
#> 6 19.3
#> 7 15.4
#> 8 22.2
#> 9 22.8
#> 10 19.9
#> # … with 22 more rows