glmnet: Interfaces for glmnet package for data science pipelines.

In rbertolusso/intubate: Interface to Popular R Functions for Data Science Pipelines

Description

Interfaces to glmnet functions that can be used in a pipeline implemented by magrittr.

Usage

ntbt_glmnet(data, ...)
ntbt_cv.glmnet(data, ...)

Arguments

data	data frame, tibble, list, ...
...	Other arguments passed to the corresponding interfaced function.

Details

Interfaces call their corresponding interfaced function.

Value

Object returned by interfaced function.

Author(s)

Roberto Bertolusso

Examples

## Not run: 
library(intubate)
library(magrittr)
library(glmnet)


## NOTE: glmnet package does not implement formula interface. As I need it
##       for teaching purposes, this is the first pure non-formula library
##       included, as a proof of concept that not only the functions with
##       formula + data variant can be successfully interfaced to use in
##       a pipeline.


library(ISLR)
data("Hitters")
Hitters <- na.omit(Hitters)

dta <- list(x = model.matrix(Salary ~ ., Hitters)[, -1],  ## Remove intercept
            y = model.frame(Salary ~ ., Hitters)[, 1])
grid <- 10^seq(10, -2, length = 100)

## ntbt_glmnet: fit a GLM with lasso or elasticnet regularization

## Original function to interface
attach(dta)
## Ridge Regression
ridge <- glmnet(x, y, alpha = 0, lambda = grid)
plot(ridge)

## The Lasso
lasso <- glmnet(x, y, alpha = 1, lambda = grid)
plot(lasso)
detach()

## The interface puts data as first parameter
## Ridge Regression
ridge <- ntbt_glmnet(dta, x, y, alpha = 0, lambda = grid)
plot(ridge)

## The Lasso
lasso <- ntbt_glmnet(dta, x, y, alpha = 1, lambda = grid)
plot(lasso)

## so it can be used easily in a pipeline.
## Ridge Regression
dta %>%
  ntbt_glmnet(x, y, alpha = 0, lambda = grid) %>%
  plot()

## The Lasso
dta %>%
  ntbt_glmnet(x, y, alpha = 1, lambda = grid) %>%
  plot()


## ntbt_cv.glmnet: Cross-validation for glmnet

## Original function to interface
attach(dta)
## Ridge Regression
set.seed(1)
cv.ridge <- cv.glmnet(x, y, alpha = 0)
plot(cv.ridge)

## The Lasso
cv.lasso <- cv.glmnet(x, y, alpha = 1)
plot(cv.lasso)
detach()

## The interface puts data as first parameter
## Ridge Regression
set.seed(1)
cv.ridge <- ntbt_cv.glmnet(dta, x, y, alpha = 0)
plot(cv.ridge)

## The Lasso
cv.lasso <- ntbt_cv.glmnet(dta, x, y, alpha = 1)
plot(cv.lasso)

## so it can be used easily in a pipeline.
## Ridge Regression
set.seed(1)
dta %>%
  ntbt_cv.glmnet(x, y, alpha = 0) %>%
  plot()

## The Lasso
dta %>%
  ntbt_cv.glmnet(x, y, alpha = 1) %>%
  plot()

## End(Not run)

rbertolusso/intubate documentation built on May 27, 2019, 3 a.m.