Stratified Features via Clustering — tl_stratified_models • tidylearn

Create cluster-specific supervised models for heterogeneous data

Usage

tl_stratified_models(
  data,
  formula,
  cluster_method = "kmeans",
  k = 3,
  supervised_method = "tree",
  ...
)

Arguments

data: A data frame
formula: Model formula
cluster_method: Clustering method
k: Number of clusters
supervised_method: Supervised learning method (default: "tree", which handles both regression and classification). "linear" silently fits lm() to a factor response rather than refusing it, so it is not a safe default here.
...: Additional arguments

Value

A list with class "tidylearn_stratified" containing:

cluster_model: The fitted clustering model.
supervised_models: Named list of tidylearn models, one per cluster.
formula: The model formula.
data: The original training data.

Examples

# \donttest{
models <- tl_stratified_models(mtcars, mpg ~ ., cluster_method = "kmeans",
                                k = 3, supervised_method = "linear")
#> Note: Response 'mpg' has 6 unique numeric values. Treating as regression. Convert to factor for classification.
#> Note: Response 'mpg' has 8 unique numeric values. Treating as regression. Convert to factor for classification.
# }