discrim (version 0.1.2)

discrim_flexible: General Interface for Flexible Discriminant Models

Description

discrim_flexible() is a way to generate a specification of a flexible discriminant model using features created using multivariate adaptive regression splines (MARS).

Usage

discrim_flexible(
  mode = "classification",
  num_terms = NULL,
  prod_degree = NULL,
  prune_method = NULL
)
# S3 method for discrim_flexible
update(
  object,
  num_terms = NULL,
  prod_degree = NULL,
  prune_method = NULL,
  fresh = FALSE,
  ...
)

Arguments

mode

A single character string for the type of model. The only possible value for this model is "classification".

num_terms

The number of features that will be retained in the final model, including the intercept.

prod_degree

The highest possible interaction degree.

prune_method

The pruning method.

object

A flexible discriminant model specification.

fresh

A logical for whether the arguments should be modified in-place of or replaced wholesale.

...

Not used for update().

Engine Details

Engines may have pre-set default arguments when executing the model fit call. For this type of model, the template of the fit calls are:

discrim_flexible() %>% 
  set_engine("earth") %>% 
  translate()

## Flexible Discriminant Model Specification (classification)
## 
## Computational engine: earth 
## 
## Model fit template:
## mda::fda(formula = missing_arg(), data = missing_arg(), method = earth::earth)

The standardized parameter names in parsnip can be mapped to their original names in each engine that has main parameters. Each engine typically has a different default value (shown in parentheses) for each parameter.

parsnip	earth
num_terms	nprune (all created by forward pass)
prod_degree	degree (1)
prune_method	pmethod (backward)

Details

Flexible discriminant analysis (FDA) uses the work of Hastie et al (1994) to create a discriminant model using different feature expansions. For this function, MARS (Friedman, 1991) hinge functions are used to nonlinearly model the class boundaries (see example below). The mda and earth packages are needed to fit this model.

The main arguments for the model are:

• num_terms: The number of features that will be retained in the final model.

• prod_degree: The highest possible degree of interaction between features. A value of 1 indicates and additive model while a value of 2 allows, but does not guarantee, two-way interactions between features.

• prune_method: The type of pruning. Possible values are listed in ?earth.

These arguments are converted to their specific names at the time that the model is fit. Other options and argument can be set using set_engine(). If left to their defaults here (NULL), the values are taken from the underlying model functions. If parameters need to be modified, update() can be used in lieu of recreating the object from scratch.

The model can be created using the fit() function using the following engines:

• R: "earth" (the default)

References

Friedman (1991), Multivariate Adaptive Regression Splines (with discussion), Annals of Statistics 19:1, 1<U+2013>141. Hastie, Tibshirani and Buja (1994), Flexible Discriminant Analysis by Optimal Scoring, Journal of the American Statistical Association, 1255-1270.

Examples

# NOT RUN {
parabolic_grid <-
  expand.grid(X1 = seq(-5, 5, length = 100),
              X2 = seq(-5, 5, length = 100))

fda_mod <-
  discrim_flexible(num_terms = 3) %>%
  # increase `num_terms` to find smoother boundaries
  set_engine("earth") %>%
  fit(class ~ ., data = parabolic)

parabolic_grid$fda <-
  predict(fda_mod, parabolic_grid, type = "prob")$.pred_Class1

library(ggplot2)
ggplot(parabolic, aes(x = X1, y = X2)) +
  geom_point(aes(col = class), alpha = .5) +
  geom_contour(data = parabolic_grid, aes(z = fda), col = "black", breaks = .5) +
  theme_bw() +
  theme(legend.position = "top") +
  coord_equal()


model <- discrim_flexible(num_terms = 10)
model
update(model, num_terms = 6)
# }