ml4e: Machine Learning for Everyone

Utilities make machine learning projects easier. Note this project is highly development state.

Installation

ml4e is not available through CRAN. To install from github, use:

devtools::install_github("five-dots/ml4e")

Quick Introduction

library(ml4e)

# Formulas defines label variable (reponse variable) and features
formulas <- new_formulas(Species ~ .)

# Seeds for hyper parameter optimizaion and cross validation
seeds <- new_seeds()$add_random()

# Resample method for cross validation. Currenlty supports functions from rsample package
resample_calls <- new_resample_calls(
  cv4 = quote(rsample::vfold_cv(data, v = 4)),
  cv4_strata = quote(rsample::vfold_cv(data, v = 4, strata = "Species"))
)

# Datasets must have a resample_calls (defined above) and can have test_ids to calculate scores.
datasets <- new_datasets(
  iris,
  .test_ids = sample(nrow(iris), 30),
  .resample_calls = resample_calls
)

# Models, in this case include 5 models
models <- new_models(
  new_model("glmnet_cv"),
  new_model("kknn"),
  new_model("kernlab"),
  new_model("ranger"),
  new_model("xgboost_es")
)

# Gather all options into one project
project <- new_project(formulas, datasets, seeds, models,
                       show_progress = FALSE)

# Run cross validations of all options combinations
project$run_cv()

# Confirm CV and test set scores of all combinations
project$get_scores() %>% dplyr::mutate_if(is.numeric, round, 4)

resample	seed	model	cv_acc	cv_logloss	test_acc	test_logloss
cv4	sed₀₁	glmnet_cv	0.975	0.1009	0.9667	0.1029
cv4	sed₀₁	kernlab	0.9667	0.1384	0.9333	0.1613
cv4	sed₀₁	kknn	0.95	0.1116	0.9333	0.1451
cv4	sed₀₁	ranger	0.9583	0.1424	0.9333	0.1485
cv4	sed₀₁	xgboost_es	0.9583	0.1563	0.9333	0.1599
cv4	sed₀₂	glmnet_cv	0.9417	0.1174	0.9333	0.1101
cv4	sed₀₂	kernlab	0.95	0.1346	0.9333	0.1479
cv4	sed₀₂	kknn	0.9417	0.1145	0.9333	0.1438
cv4	sed₀₂	ranger	0.9417	0.1777	0.9333	0.1407
cv4	sed₀₂	xgboost_es	0.9417	0.2066	0.9333	0.1525
cv4_strata	sed₀₁	glmnet_cv	0.9667	0.0829	0.9667	0.0987
cv4_strata	sed₀₁	kernlab	0.9417	0.1603	0.9333	0.1697
cv4_strata	sed₀₁	kknn	0.9417	0.1084	0.9333	0.156
cv4_strata	sed₀₁	ranger	0.9583	0.1557	0.9333	0.1452
cv4_strata	sed₀₁	xgboost_es	0.95	0.1613	0.9333	0.1457
cv4_strata	sed₀₂	glmnet_cv	0.9583	0.0805	0.9667	0.105
cv4_strata	sed₀₂	kernlab	0.9583	0.1341	0.9333	0.1387
cv4_strata	sed₀₂	kknn	0.9583	0.679	0.9333	0.1869
cv4_strata	sed₀₂	ranger	0.95	0.1614	0.9333	0.1426
cv4_strata	sed₀₂	xgboost_es	0.9583	0.1823	0.9333	0.1261