tidymodels

tidymodels

Tidy machine learning framework.

Workflow

library(tidymodels)

# 1. Split data
split <- initial_split(df, prop = 0.8, strata = target)
train <- training(split)
test <- testing(split)

# 2. Create recipe
recipe <- recipe(target ~ ., data = train) %>%
  step_normalize(all_numeric_predictors()) %>%
  step_dummy(all_nominal_predictors())

# 3. Specify model
model <- rand_forest(trees = 100) %>%
  set_engine("ranger") %>%
  set_mode("classification")

# 4. Create workflow
wf <- workflow() %>%
  add_recipe(recipe) %>%
  add_model(model)

# 5. Fit
fit <- wf %>% fit(data = train)

# 6. Predict
predictions <- predict(fit, test)

Recipes

recipe(target ~ ., data = train) %>%
  # Imputation
  step_impute_mean(all_numeric_predictors()) %>%
  step_impute_mode(all_nominal_predictors()) %>%
  step_impute_knn(all_predictors()) %>%
  
  # Transformation
  step_normalize(all_numeric_predictors()) %>%
  step_scale(all_numeric_predictors()) %>%
  step_center(all_numeric_predictors()) %>%
  step_log(value, base = 10) %>%
  step_sqrt(value) %>%
  step_BoxCox(all_numeric_predictors()) %>%
  step_YeoJohnson(all_numeric_predictors()) %>%
  
  # Encoding
  step_dummy(all_nominal_predictors()) %>%
  step_other(category, threshold = 0.05) %>%
  step_novel(all_nominal_predictors()) %>%
  step_unknown(all_nominal_predictors()) %>%
  
  # Feature engineering
  step_interact(~ x1:x2) %>%
  step_poly(x, degree = 2) %>%
  step_ns(x, deg_free = 3) %>%
  step_date(date, features = c("dow", "month", "year")) %>%
  
  # Selection
  step_zv(all_predictors()) %>%
  step_nzv(all_predictors()) %>%
  step_corr(all_numeric_predictors(), threshold = 0.9) %>%
  step_pca(all_numeric_predictors(), num_comp = 5) %>%
  step_select(x1, x2, x3)

Models (parsnip)

# Linear models
linear_reg() %>% set_engine("lm")
logistic_reg() %>% set_engine("glm")
logistic_reg(penalty = 0.1, mixture = 0.5) %>% set_engine("glmnet")

# Trees
decision_tree() %>% set_engine("rpart")
rand_forest(trees = 100, mtry = 5, min_n = 10) %>% set_engine("ranger")
boost_tree(trees = 100, learn_rate = 0.1) %>% set_engine("xgboost")

# SVM
svm_rbf(cost = 1, rbf_sigma = 0.1) %>% set_engine("kernlab")
svm_linear() %>% set_engine("kernlab")

# Neural network
mlp(hidden_units = 10, penalty = 0.01) %>% set_engine("nnet")

# Set mode
set_mode("classification")
set_mode("regression")