Help for package SHAPBoost

Title:

The SHAPBoost Feature Selection Algorithm

Version:

1.0.0

Description:

The implementation of SHAPBoost, a boosting-based feature selection technique that ranks features iteratively based on Shapley values.

License:

MIT + file LICENSE

Encoding:

UTF-8

RoxygenNote:

7.3.2

Imports:

xgboost, SHAPforxgboost, methods, caret, Matrix

Suggests:

flare, survival

URL:

https://github.com/O-T-O-Z/SHAPBoost-R

BugReports:

https://github.com/O-T-O-Z/SHAPBoost-R/issues

NeedsCompilation:

Packaged:

2025-09-22 09:04:24 UTC; o.t.ozyilmaz

Author:

Ömer Tarik Özyilmaz

[aut, cre, cph], Tamas Szili-Török

[aut, cph]

Maintainer:

Ömer Tarik Özyilmaz <o.t.ozyilmaz@umcg.nl>

Repository:

CRAN

Date/Publication:

2025-09-29 16:40:02 UTC

SHAPBoostEstimator Class

Description

This class implements the SHAPBoost algorithm for feature selection. It is designed to be extended by specific implementations such as SHAPBoostRegressor and SHAPBoostSurvival. Any new method should implement the abstract methods defined in this class.

Fields

evaluator: The model that is used to evaluate each additional feature.
metric: A character string representing the evaluation metric.
xgb_params: A list of parameters for the XGBoost model.
number_of_folds: The number of folds for cross-validation.
epsilon: A small value to determine convergence.
max_number_of_features: The maximum number of features to select.
siso_ranking_size: The number of features to consider in the SISO ranking.
siso_order: The order of combinations to consider in SISO.
reset: A logical indicating whether to reset the weights.
num_resets: The number of resets allowed.
fold_random_state: The random state for reproducibility in cross-validation.
verbose: The verbosity level of the output.
stratification: A logical indicating whether to use stratified sampling. Only applicable for c-index metric.
collinearity_check: A logical indicating whether to check for collinearity.
correlation_threshold: The threshold for correlation to consider features as collinear.

Examples

if (requireNamespace("flare", quietly = TRUE)) {
  data("eyedata", package = "flare")
  shapboost <- SHAPBoostRegressor$new(
    max_number_of_features = 1,
    evaluator = "lr",
    metric = "mae",
    siso_ranking_size = 10,
    verbose = 0
  )
  X <- as.data.frame(x)
  y <- as.data.frame(y)
  subset <- shapboost$fit(X, y)
}

SHAPBoostRegressor is a reference class for regression feature selection through gradient boosting.

Description

This class extends the SHAPBoostEstimator class and implements methods for initializing, updating weights, scoring, and fitting estimators.

Fields

evaluator: The model that is used to evaluate each additional feature. Choice between "lr" and "xgb".
metric: The metric used for evaluation, such as "mae", "mse", or "r2".
xgb_params: A list of parameters for the XGBoost model.
number_of_folds: The number of folds for cross-validation.
epsilon: A small value to prevent division by zero.
max_number_of_features: The maximum number of features to consider.
siso_ranking_size: The size of the SISO ranking.
siso_order: The order of the SISO ranking.
reset: A boolean indicating whether to reset the model.
xgb_importance: The importance type for XGBoost.
num_resets: The number of resets for the model.
fold_random_state: The random state for folds.
verbose: The verbosity level for logging.
stratification: A boolean indicating whether to use stratification. Only applicable for c-index metric.
use_shap: A boolean indicating whether to use SHAP values.
collinearity_check: A boolean indicating whether to check for collinearity.
correlation_threshold: The threshold for correlation to consider features as collinear.

Examples

if (requireNamespace("flare", quietly = TRUE)) {
  data("eyedata", package = "flare")
  shapboost <- SHAPBoostRegressor$new(
    max_number_of_features = 1,
    evaluator = "lr",
    metric = "mae",
    siso_ranking_size = 10,
    verbose = 0
  )
  X <- as.data.frame(x)
  y <- as.data.frame(y)
  subset <- shapboost$fit(X, y)
}

SHAPBoostSurvival is a reference class for survival analysis feature selection through gradient boosting.

Description

This class extends the SHAPBoostEstimator class and implements methods for initializing, updating weights, scoring, and fitting estimators.

Fields

evaluator: The model that is used to evaluate each additional feature. Choice between "coxph" and "xgb".
metric: The metric used for evaluation, such as "mae", "mse", or "r2".
xgb_params: A list of parameters for the XGBoost model.
number_of_folds: The number of folds for cross-validation.
epsilon: A small value to prevent division by zero.
max_number_of_features: The maximum number of features to consider.
siso_ranking_size: The size of the SISO ranking.
siso_order: The order of the SISO ranking.
reset: A boolean indicating whether to reset the model.
xgb_importance: The importance type for XGBoost.
num_resets: The number of resets for the model.
fold_random_state: The random state for folds.
verbose: The verbosity level for logging.
stratification: A boolean indicating whether to use stratification. Only applicable for c-index metric.
use_shap: A boolean indicating whether to use SHAP values.
collinearity_check: A boolean indicating whether to check for collinearity.
correlation_threshold: The threshold for correlation to consider features as collinear.

Examples

if (requireNamespace("survival", quietly = TRUE)) {
  shapboost <- SHAPBoostSurvival$new(
    max_number_of_features = 1,
    evaluator = "coxph",
    metric = "c-index",
    verbose = 0,
    xgb_params = list(
      objective = "survival:cox",
      eval_metric = "cox-nloglik"
    )
  )
  
  X <- as.data.frame(survival::gbsg[, -c(1, 10, 11)])
  y <- as.data.frame(survival::gbsg[, c(10, 11)])
  subset <- shapboost$fit(X, y)
}