calculate_residual_classification_trajectory: Build residual classification trajectory.

In WinVector/CVRTSEncoder: Categorical Variable Residual Trajectory Specteral Encoder

Description

Build a cross-validated residual trajectory for a model. The core idea is: other models factor the quantity to be explained into an explainable versus residual portion (with respect to the given model). Each of these components are possibly useful for modeling.

Usage

calculate_residual_classification_trajectory(data, vars, ..., dep_var,
  dep_target = TRUE, cross_plan = vtreat::kWayStratifiedY(nrow(data),
  3, data, data[[dep_var]] == dep_target),
  fitter_c = xgboost_fit_predict_c, fitter_r = xgboost_fit_predict_r,
  cl = NULL)

Arguments

data	The data.frame of data to fit.
vars	character vector, explanatory variable names.
...	not used, force arguments to be bound by name
dep_var	character, the name of dependent variable.
dep_target	scalar, the value considered to be the target category of dep_var.
cross_plan	a vtreat-style cross validation plan for data rows (list of disjoint tran/app lists where app partitions the data rows).
fitter_c	fit/predict classification signature function
fitter_r	fit/predict regression signature function
cl	parallel cluster for processing

Value

a matrix with the same number of rows as data representing the cross-validated modeling residual trajectories.

Examples

data <- iris
vars <- c("Sepal.Length", "Petal.Length",
          "Sepal.Width", "Petal.Width")
dep_var <- "Species"
dep_target <- "versicolor"
augments <- calculate_residual_classification_trajectory(
  data = data,
  vars = vars,
  dep_var = dep_var,
  dep_target = dep_target
)

WinVector/CVRTSEncoder documentation built on June 7, 2019, 9:53 a.m.