| Title: | Weighted Mean SHAP for Feature Selection in ML Grid and Ensemble |
|---|---|
| Description: | This R package introduces Weighted Mean SHapley Additive exPlanations (WMSHAP), an innovative method for calculating SHAP values for a grid of fine-tuned base-learner machine learning models as well as stacked ensembles, a method not previously available due to the common reliance on single best-performing models. By integrating the weighted mean SHAP values from individual base-learners comprising the ensemble or individual base-learners in a tuning grid search, the package weights SHAP contributions according to each model's performance, assessed by multiple either R squared (for both regression and classification models). alternatively, this software also offers weighting SHAP values based on the area under the precision-recall curve (AUCPR), the area under the curve (AUC), and F2 measures for binary classifiers. It further extends this framework to implement weighted confidence intervals for weighted mean SHAP values, offering a more comprehensive and robust feature importance evaluation over a grid of machine learning models, instead of solely computing SHAP values for the best model. This methodology is particularly beneficial for addressing the severe class imbalance (class rarity) problem by providing a transparent, generalized measure of feature importance that mitigates the risk of reporting SHAP values for an overfitted or biased model and maintains robustness under severe class imbalance, where there is no universal criteria of identifying the absolute best model. Furthermore, the package implements hypothesis testing to ascertain the statistical significance of SHAP values for individual features, as well as comparative significance testing of SHAP contributions between features. Additionally, it tackles a critical gap in feature selection literature by presenting criteria for the automatic feature selection of the most important features across a grid of models or stacked ensembles, eliminating the need for arbitrary determination of the number of top features to be extracted. This utility is invaluable for researchers analyzing feature significance, particularly within severely imbalanced outcomes where conventional methods fall short. Moreover, it is also expected to report democratic feature importance across a grid of models, resulting in a more comprehensive and generalizable feature selection. The package further implements a novel method for visualizing SHAP values both at subject level and feature level as well as a plot for feature selection based on the weighted mean SHAP ratios. |
| Authors: | E. F. Haghish [aut, cre, cph] |
| Maintainer: | E. F. Haghish <[email protected]> |
| License: | MIT + file LICENSE |
| Version: | 0.4 |
| Built: | 2024-10-24 04:29:32 UTC |
| Source: | CRAN |
extracts the model IDs from H2O AutoML object or H2O grid
h2o.get_ids(automl)h2o.get_ids(automl)
automl |
a h2o |
a character vector of trained models' names (IDs)
E. F. Haghish
## Not run: library(h2o) h2o.init(ignore_config = TRUE, nthreads = 2, bind_to_localhost = FALSE, insecure = TRUE) prostate_path <- system.file("extdata", "prostate.csv", package = "h2o") prostate <- h2o.importFile(path = prostate_path, header = TRUE) y <- "CAPSULE" prostate[,y] <- as.factor(prostate[,y]) #convert to factor for classification aml <- h2o.automl(y = y, training_frame = prostate, max_runtime_secs = 30) # get the model IDs ids <- h2o.ids(aml) ## End(Not run)## Not run: library(h2o) h2o.init(ignore_config = TRUE, nthreads = 2, bind_to_localhost = FALSE, insecure = TRUE) prostate_path <- system.file("extdata", "prostate.csv", package = "h2o") prostate <- h2o.importFile(path = prostate_path, header = TRUE) y <- "CAPSULE" prostate[,y] <- as.factor(prostate[,y]) #convert to factor for classification aml <- h2o.automl(y = y, training_frame = prostate, max_runtime_secs = 30) # get the model IDs ids <- h2o.ids(aml) ## End(Not run)
This function normalizes a vector based on specified minimum and maximum values. If the minimum and maximum values are not specified, the function will use the minimum and maximum values of the vector.
normalize(x, min = NULL, max = NULL)normalize(x, min = NULL, max = NULL)
x |
numeric vector |
min |
minimum value |
max |
maximum value |
normalized numeric vector
E. F. Haghish
Weighted average of SHAP values and weighted SHAP confidence intervals provide a measure of feature importance for a grid of fine-tuned models or base-learners of a stacked ensemble model. Instead of reporting relative SHAP contributions for a single model, this function takes the variability in feature importance of multiple models into account and computes weighted mean and confidence intervals for each feature, taking the performance metric of each model as the weight. The function also provides a plot of the weighted SHAP values and confidence intervals. Currently only models trained by h2o machine learning software or autoEnsemble package are supported.
shapley( models, newdata, plot = TRUE, performance_metric = "r2", standardize_performance_metric = FALSE, performance_type = "xval", minimum_performance = 0, method = c("lowerCI"), cutoff = 0, top_n_features = NULL, n_models = 10 )shapley( models, newdata, plot = TRUE, performance_metric = "r2", standardize_performance_metric = FALSE, performance_type = "xval", minimum_performance = 0, method = c("lowerCI"), cutoff = 0, top_n_features = NULL, n_models = 10 )
models |
H2O search grid, AutoML grid, or a character vector of H2O model IDs.
the |
newdata |
h2o frame (data.frame). the data.frame must be already uploaded on h2o server (cloud). when specified, this dataset will be used for evaluating the models. if not specified, model performance on the training dataset will be reported. |
plot |
logical. if TRUE, the weighted mean and confidence intervals of the SHAP values are plotted. The default is TRUE. |
performance_metric |
character, specifying the performance metric to be used for weighting the SHAP values (mean and 95 "r2" (R Squared). For binary classification, other options include "aucpr" (area under the precision-recall curve), "auc" (area under the ROC curve), and "f2" (F2 score). |
standardize_performance_metric |
logical. if TRUE, performance_metric, which is used as weights vector is standardized such that the sum of the weights vector would be equal to the length of the vector. the default value is FALSE. |
performance_type |
character, specifying where the performance metric should be retrieved from. "train" means the performance of the training process should be reported, "valid" indicates that the performance of the validation process should be reported, and "xval" means the cross-validation performance to be retrieved. |
minimum_performance |
the minimum performance metric for a recognizable model. any model with performance equal or lower than this argument will have weight of zero in computing the weighted mean and CI SHAP values. the default value is zero. |
method |
character, specifying the method used for identifying the most important features according to their weighted SHAP values. The default selection method is "lowerCI", which includes features whose lower weighted confidence interval exceeds the predefined 'cutoff' value (default is 0). Alternatively, the "mean" option can be specified, indicating any feature with normalized weighted mean SHAP contribution above the specified 'cutoff' should be selected. Another alternative options is "shapratio", a method that filters for features where the proportion of their relative weighted SHAP value exceeds the 'cutoff'. This approach calculates the relative contribution of each feature's weighted SHAP value against the aggregate of all features, with those surpassing the 'cutoff' being selected as top feature. |
cutoff |
numeric, specifying the cutoff for the method used for selecting the top features. the default is zero, which means that all features with the "method" criteria above zero will be selected. |
top_n_features |
integer. if specified, the top n features with the highest weighted SHAP values will be selected, overrullung the 'cutoff' and 'method' arguments. specifying top_n_feature is also a way to reduce computation time, if many features are present in the data set. The default is NULL, which means the shap values will be computed for all features. |
n_models |
minimum number of models that should meet the 'minimum_performance' criterion in order to compute WMSHAP and CI. If the intention is to compute global summary SHAP values (at feature level) for a single model, set n_models to 1. The default is 10. |
a list including the GGPLOT2 object, the data frame of SHAP values, and performance metric of all models, as well as the model IDs.
E. F. Haghish
## Not run: # load the required libraries for building the base-learners and the ensemble models library(h2o) #shapley supports h2o models library(shapley) # initiate the h2o server h2o.init(ignore_config = TRUE, nthreads = 2, bind_to_localhost = FALSE, insecure = TRUE) # upload data to h2o cloud prostate_path <- system.file("extdata", "prostate.csv", package = "h2o") prostate <- h2o.importFile(path = prostate_path, header = TRUE) set.seed(10) ### H2O provides 2 types of grid search for tuning the models, which are ### AutoML and Grid. Below, I demonstrate how weighted mean shapley values ### can be computed for both types. ####################################################### ### PREPARE AutoML Grid (takes a couple of minutes) ####################################################### # run AutoML to tune various models (GBM) for 60 seconds y <- "CAPSULE" prostate[,y] <- as.factor(prostate[,y]) #convert to factor for classification aml <- h2o.automl(y = y, training_frame = prostate, max_runtime_secs = 120, include_algos=c("GBM"), # this setting ensures the models are comparable for building a meta learner seed = 2023, nfolds = 10, keep_cross_validation_predictions = TRUE) ### call 'shapley' function to compute the weighted mean and weighted confidence intervals ### of SHAP values across all trained models. ### Note that the 'newdata' should be the testing dataset! result <- shapley(models = aml, newdata = prostate, performance_metric = "aucpr", plot = TRUE) ####################################################### ### PREPARE H2O Grid (takes a couple of minutes) ####################################################### # make sure equal number of "nfolds" is specified for different grids grid <- h2o.grid(algorithm = "gbm", y = y, training_frame = prostate, hyper_params = list(ntrees = seq(1,50,1)), grid_id = "ensemble_grid", # this setting ensures the models are comparable for building a meta learner seed = 2023, fold_assignment = "Modulo", nfolds = 10, keep_cross_validation_predictions = TRUE) result2 <- shapley(models = grid, newdata = prostate, performance_metric = "aucpr", plot = TRUE) ####################################################### ### PREPARE autoEnsemble STACKED ENSEMBLE MODEL ####################################################### ### get the models' IDs from the AutoML and grid searches. ### this is all that is needed before building the ensemble, ### i.e., to specify the model IDs that should be evaluated. library(autoEnsemble) ids <- c(h2o.get_ids(aml), h2o.get_ids(grid)) autoSearch <- ensemble(models = ids, training_frame = prostate, strategy = "search") result3 <- shapley(models = autoSearch, newdata = prostate, performance_metric = "aucpr", plot = TRUE) ## End(Not run)## Not run: # load the required libraries for building the base-learners and the ensemble models library(h2o) #shapley supports h2o models library(shapley) # initiate the h2o server h2o.init(ignore_config = TRUE, nthreads = 2, bind_to_localhost = FALSE, insecure = TRUE) # upload data to h2o cloud prostate_path <- system.file("extdata", "prostate.csv", package = "h2o") prostate <- h2o.importFile(path = prostate_path, header = TRUE) set.seed(10) ### H2O provides 2 types of grid search for tuning the models, which are ### AutoML and Grid. Below, I demonstrate how weighted mean shapley values ### can be computed for both types. ####################################################### ### PREPARE AutoML Grid (takes a couple of minutes) ####################################################### # run AutoML to tune various models (GBM) for 60 seconds y <- "CAPSULE" prostate[,y] <- as.factor(prostate[,y]) #convert to factor for classification aml <- h2o.automl(y = y, training_frame = prostate, max_runtime_secs = 120, include_algos=c("GBM"), # this setting ensures the models are comparable for building a meta learner seed = 2023, nfolds = 10, keep_cross_validation_predictions = TRUE) ### call 'shapley' function to compute the weighted mean and weighted confidence intervals ### of SHAP values across all trained models. ### Note that the 'newdata' should be the testing dataset! result <- shapley(models = aml, newdata = prostate, performance_metric = "aucpr", plot = TRUE) ####################################################### ### PREPARE H2O Grid (takes a couple of minutes) ####################################################### # make sure equal number of "nfolds" is specified for different grids grid <- h2o.grid(algorithm = "gbm", y = y, training_frame = prostate, hyper_params = list(ntrees = seq(1,50,1)), grid_id = "ensemble_grid", # this setting ensures the models are comparable for building a meta learner seed = 2023, fold_assignment = "Modulo", nfolds = 10, keep_cross_validation_predictions = TRUE) result2 <- shapley(models = grid, newdata = prostate, performance_metric = "aucpr", plot = TRUE) ####################################################### ### PREPARE autoEnsemble STACKED ENSEMBLE MODEL ####################################################### ### get the models' IDs from the AutoML and grid searches. ### this is all that is needed before building the ensemble, ### i.e., to specify the model IDs that should be evaluated. library(autoEnsemble) ids <- c(h2o.get_ids(aml), h2o.get_ids(grid)) autoSearch <- ensemble(models = ids, training_frame = prostate, strategy = "search") result3 <- shapley(models = autoSearch, newdata = prostate, performance_metric = "aucpr", plot = TRUE) ## End(Not run)
This function applies different criteria to visualize SHAP contributions
shapley.domain( shapley, domains, plot = "bar", method = "AUTO", legendstyle = "continuous", scale_colour_gradient = NULL, print = FALSE )shapley.domain( shapley, domains, plot = "bar", method = "AUTO", legendstyle = "continuous", scale_colour_gradient = NULL, print = FALSE )
shapley |
object of class 'shapley', as returned by the 'shapley' function |
domains |
character list, specifying the domains for grouping the features' contributions. Domains are clusters of features' names, that can be used to compute WMSHAP at higher level, along with their 95 better understand how a cluster of features influence the outcome. Note that either of 'features' or 'domains' arguments can be specified at the time. |
plot |
character, specifying the type of the plot, which can be either 'bar', 'waffle', or 'shap'. The default is 'bar'. |
method |
character, specifying the method used for identifying the most important features according to their weighted SHAP values. The default selection method is "AUTO", which selects a method based on number of models that have been evaluated because lowerCI method is not applicable to SHAP values of a single model. If 'lowerCI' is specified, features whose lower weighted confidence interval exceeds the predefined 'cutoff' value would be reported. Alternatively, the "mean" option can be specified, indicating any feature with normalized weighted mean SHAP contribution above the specified 'cutoff' should be selected. Another alternative options is "shapratio", a method that filters for features where the proportion of their relative weighted SHAP value exceeds the 'cutoff'. This approach calculates the relative contribution of each feature's weighted SHAP value against the aggregate of all features, with those surpassing the 'cutoff' being selected as top feature. |
legendstyle |
character, specifying the style of the plot legend, which can be either 'continuous' (default) or 'discrete'. the continuous legend is only applicable to 'shap' plots and other plots only use 'discrete' legend. |
scale_colour_gradient |
character vector for specifying the color gradients for the plot. |
print |
logical. if TRUE, the WMSHAP summary table for the given row is printed |
ggplot object
E. F. Haghish
## Not run: # load the required libraries for building the base-learners and the ensemble models library(h2o) #shapley supports h2o models library(shapley) # initiate the h2o server h2o.init(ignore_config = TRUE, nthreads = 2, bind_to_localhost = FALSE, insecure = TRUE) # upload data to h2o cloud prostate_path <- system.file("extdata", "prostate.csv", package = "h2o") prostate <- h2o.importFile(path = prostate_path, header = TRUE) ### H2O provides 2 types of grid search for tuning the models, which are ### AutoML and Grid. Below, I demonstrate how weighted mean shapley values ### can be computed for both types. set.seed(10) ####################################################### ### PREPARE AutoML Grid (takes a couple of minutes) ####################################################### # run AutoML to tune various models (GBM) for 60 seconds y <- "CAPSULE" prostate[,y] <- as.factor(prostate[,y]) #convert to factor for classification aml <- h2o.automl(y = y, training_frame = prostate, max_runtime_secs = 120, include_algos=c("GBM"), # this setting ensures the models are comparable for building a meta learner seed = 2023, nfolds = 10, keep_cross_validation_predictions = TRUE) ### call 'shapley' function to compute the weighted mean and weighted confidence intervals ### of SHAP values across all trained models. ### Note that the 'newdata' should be the testing dataset! result <- shapley(models = aml, newdata = prostate, plot = TRUE) ####################################################### ### PLOT THE WEIGHTED MEAN SHAP VALUES ####################################################### shapley.plot(result, plot = "bar") shapley.plot(result, plot = "waffle") ## End(Not run)## Not run: # load the required libraries for building the base-learners and the ensemble models library(h2o) #shapley supports h2o models library(shapley) # initiate the h2o server h2o.init(ignore_config = TRUE, nthreads = 2, bind_to_localhost = FALSE, insecure = TRUE) # upload data to h2o cloud prostate_path <- system.file("extdata", "prostate.csv", package = "h2o") prostate <- h2o.importFile(path = prostate_path, header = TRUE) ### H2O provides 2 types of grid search for tuning the models, which are ### AutoML and Grid. Below, I demonstrate how weighted mean shapley values ### can be computed for both types. set.seed(10) ####################################################### ### PREPARE AutoML Grid (takes a couple of minutes) ####################################################### # run AutoML to tune various models (GBM) for 60 seconds y <- "CAPSULE" prostate[,y] <- as.factor(prostate[,y]) #convert to factor for classification aml <- h2o.automl(y = y, training_frame = prostate, max_runtime_secs = 120, include_algos=c("GBM"), # this setting ensures the models are comparable for building a meta learner seed = 2023, nfolds = 10, keep_cross_validation_predictions = TRUE) ### call 'shapley' function to compute the weighted mean and weighted confidence intervals ### of SHAP values across all trained models. ### Note that the 'newdata' should be the testing dataset! result <- shapley(models = aml, newdata = prostate, plot = TRUE) ####################################################### ### PLOT THE WEIGHTED MEAN SHAP VALUES ####################################################### shapley.plot(result, plot = "bar") shapley.plot(result, plot = "waffle") ## End(Not run)
This function specifies the top features and prepares the data for plotting SHAP contributions for each row, or summary of absolute SHAP contributions for each feature.
shapley.feature.selection( shapley, method = "lowerCI", cutoff = 0, top_n_features = NULL, features = NULL )shapley.feature.selection( shapley, method = "lowerCI", cutoff = 0, top_n_features = NULL, features = NULL )
shapley |
shapley object |
method |
character, specifying the method used for identifying the most important features according to their weighted SHAP values. The default selection method is "lowerCI", which includes features whose lower weighted confidence interval exceeds the predefined 'cutoff' value (default is relative SHAP of 1 Alternatively, the "mean" option can be specified, indicating any feature with normalized weighted mean SHAP contribution above the specified 'cutoff' should be selected. Another alternative options is "shapratio", a method that filters for features where the proportion of their relative weighted SHAP value exceeds the 'cutoff'. This approach calculates the relative contribution of each feature's weighted SHAP value against the aggregate of all features, with those surpassing the 'cutoff' being selected as top feature. |
cutoff |
numeric, specifying the cutoff for the method used for selecting the top features. the default is zero, which means that all features with the "method" criteria above zero will be selected. |
top_n_features |
integer. if specified, the top n features with the highest weighted SHAP values will be selected, overrullung the 'cutoff' and 'method' arguments. |
features |
character vector, specifying the feature to be plotted. |
normalized numeric vector
E. F. Haghish
This function applies different criteria to visualize SHAP contributions
shapley.plot( shapley, plot = "bar", method = "lowerCI", cutoff = 0, top_n_features = NULL, features = NULL, legendstyle = "continuous", scale_colour_gradient = NULL )shapley.plot( shapley, plot = "bar", method = "lowerCI", cutoff = 0, top_n_features = NULL, features = NULL, legendstyle = "continuous", scale_colour_gradient = NULL )
shapley |
object of class 'shapley', as returned by the 'shapley' function |
plot |
character, specifying the type of the plot, which can be either 'bar', 'waffle', or 'shap'. The default is 'bar'. |
method |
character, specifying the method used for identifying the most important features according to their weighted SHAP values. The default selection method is "lowerCI", which includes features whose lower weighted confidence interval exceeds the predefined 'cutoff' value (default is relative SHAP of 1 Alternatively, the "mean" option can be specified, indicating any feature with normalized weighted mean SHAP contribution above the specified 'cutoff' should be selected. Another alternative options is "shapratio", a method that filters for features where the proportion of their relative weighted SHAP value exceeds the 'cutoff'. This approach calculates the relative contribution of each feature's weighted SHAP value against the aggregate of all features, with those surpassing the 'cutoff' being selected as top feature. |
cutoff |
numeric, specifying the cutoff for the method used for selecting the top features. |
top_n_features |
integer. if specified, the top n features with the highest weighted SHAP values will be selected, overrullung the 'cutoff' and 'method' arguments. |
features |
character vector, specifying the feature to be plotted. |
legendstyle |
character, specifying the style of the plot legend, which can be either 'continuous' (default) or 'discrete'. the continuous legend is only applicable to 'shap' plots and other plots only use 'discrete' legend. |
scale_colour_gradient |
character vector for specifying the color gradients for the plot. |
ggplot object
E. F. Haghish
## Not run: # load the required libraries for building the base-learners and the ensemble models library(h2o) #shapley supports h2o models library(shapley) # initiate the h2o server h2o.init(ignore_config = TRUE, nthreads = 2, bind_to_localhost = FALSE, insecure = TRUE) # upload data to h2o cloud prostate_path <- system.file("extdata", "prostate.csv", package = "h2o") prostate <- h2o.importFile(path = prostate_path, header = TRUE) ### H2O provides 2 types of grid search for tuning the models, which are ### AutoML and Grid. Below, I demonstrate how weighted mean shapley values ### can be computed for both types. set.seed(10) ####################################################### ### PREPARE AutoML Grid (takes a couple of minutes) ####################################################### # run AutoML to tune various models (GBM) for 60 seconds y <- "CAPSULE" prostate[,y] <- as.factor(prostate[,y]) #convert to factor for classification aml <- h2o.automl(y = y, training_frame = prostate, max_runtime_secs = 120, include_algos=c("GBM"), # this setting ensures the models are comparable for building a meta learner seed = 2023, nfolds = 10, keep_cross_validation_predictions = TRUE) ### call 'shapley' function to compute the weighted mean and weighted confidence intervals ### of SHAP values across all trained models. ### Note that the 'newdata' should be the testing dataset! result <- shapley(models = aml, newdata = prostate, plot = TRUE) ####################################################### ### PLOT THE WEIGHTED MEAN SHAP VALUES ####################################################### shapley.plot(result, plot = "bar") shapley.plot(result, plot = "waffle") ## End(Not run)## Not run: # load the required libraries for building the base-learners and the ensemble models library(h2o) #shapley supports h2o models library(shapley) # initiate the h2o server h2o.init(ignore_config = TRUE, nthreads = 2, bind_to_localhost = FALSE, insecure = TRUE) # upload data to h2o cloud prostate_path <- system.file("extdata", "prostate.csv", package = "h2o") prostate <- h2o.importFile(path = prostate_path, header = TRUE) ### H2O provides 2 types of grid search for tuning the models, which are ### AutoML and Grid. Below, I demonstrate how weighted mean shapley values ### can be computed for both types. set.seed(10) ####################################################### ### PREPARE AutoML Grid (takes a couple of minutes) ####################################################### # run AutoML to tune various models (GBM) for 60 seconds y <- "CAPSULE" prostate[,y] <- as.factor(prostate[,y]) #convert to factor for classification aml <- h2o.automl(y = y, training_frame = prostate, max_runtime_secs = 120, include_algos=c("GBM"), # this setting ensures the models are comparable for building a meta learner seed = 2023, nfolds = 10, keep_cross_validation_predictions = TRUE) ### call 'shapley' function to compute the weighted mean and weighted confidence intervals ### of SHAP values across all trained models. ### Note that the 'newdata' should be the testing dataset! result <- shapley(models = aml, newdata = prostate, plot = TRUE) ####################################################### ### PLOT THE WEIGHTED MEAN SHAP VALUES ####################################################### shapley.plot(result, plot = "bar") shapley.plot(result, plot = "waffle") ## End(Not run)
Weighted mean of SHAP values and weighted SHAP confidence intervals provide a measure of feature importance for a grid of fine-tuned models or base-learners of a stacked ensemble model at subject level, showing that how each feature influences the prediction made for a row in the dataset and to what extend different models agree on that effect. If the 95 vertical line at 0.00, then it can be concluded that the feature does not significantly influences the subject, when variability across models is taken into consideration.
shapley.row.plot( shapley, row_index, features = NULL, plot = TRUE, print = FALSE )shapley.row.plot( shapley, row_index, features = NULL, plot = TRUE, print = FALSE )
shapley |
object of class 'shapley', as returned by the 'shapley' function |
row_index |
subject or row number in a wide-format dataset to be visualized |
features |
character vector, specifying the feature to be plotted. |
plot |
logical. if TRUE, the plot is visualized. |
print |
logical. if TRUE, the WMSHAP summary table for the given row is printed |
a list including the GGPLOT2 object, the data frame of SHAP values, and performance metric of all models, as well as the model IDs.
E. F. Haghish
## Not run: # load the required libraries for building the base-learners and the ensemble models library(h2o) #shapley supports h2o models library(shapley) # initiate the h2o server h2o.init(ignore_config = TRUE, nthreads = 2, bind_to_localhost = FALSE, insecure = TRUE) # upload data to h2o cloud prostate_path <- system.file("extdata", "prostate.csv", package = "h2o") prostate <- h2o.importFile(path = prostate_path, header = TRUE) set.seed(10) ### H2O provides 2 types of grid search for tuning the models, which are ### AutoML and Grid. Below, I demonstrate how weighted mean shapley values ### can be computed for both types. ####################################################### ### PREPARE AutoML Grid (takes a couple of minutes) ####################################################### # run AutoML to tune various models (GBM) for 60 seconds y <- "CAPSULE" prostate[,y] <- as.factor(prostate[,y]) #convert to factor for classification aml <- h2o.automl(y = y, training_frame = prostate, max_runtime_secs = 120, include_algos=c("GBM"), seed = 2023, nfolds = 10, keep_cross_validation_predictions = TRUE) ### call 'shapley' function to compute the weighted mean and weighted confidence intervals ### of SHAP values across all trained models. ### Note that the 'newdata' should be the testing dataset! result <- shapley(models = aml, newdata = prostate, performance_metric = "aucpr", plot = TRUE) ####################################################### ### PREPARE H2O Grid (takes a couple of minutes) ####################################################### # make sure equal number of "nfolds" is specified for different grids grid <- h2o.grid(algorithm = "gbm", y = y, training_frame = prostate, hyper_params = list(ntrees = seq(1,50,1)), grid_id = "ensemble_grid", # this setting ensures the models are comparable for building a meta learner seed = 2023, fold_assignment = "Modulo", nfolds = 10, keep_cross_validation_predictions = TRUE) result2 <- shapley(models = grid, newdata = prostate, performance_metric = "aucpr", plot = TRUE) ####################################################### ### PREPARE autoEnsemble STACKED ENSEMBLE MODEL ####################################################### ### get the models' IDs from the AutoML and grid searches. ### this is all that is needed before building the ensemble, ### i.e., to specify the model IDs that should be evaluated. library(autoEnsemble) ids <- c(h2o.get_ids(aml), h2o.get_ids(grid)) autoSearch <- ensemble(models = ids, training_frame = prostate, strategy = "search") result3 <- shapley(models = autoSearch, newdata = prostate, performance_metric = "aucpr", plot = TRUE) #plot all important features shapley.row.plot(shapley, row_index = 11) #plot only the given features shapPlot <- shapley.row.plot(shapley, row_index = 11, features = c("PSA", "AGE")) # inspect the computed data for the row 11 ptint(shapPlot$rowSummarySHAP) ## End(Not run)## Not run: # load the required libraries for building the base-learners and the ensemble models library(h2o) #shapley supports h2o models library(shapley) # initiate the h2o server h2o.init(ignore_config = TRUE, nthreads = 2, bind_to_localhost = FALSE, insecure = TRUE) # upload data to h2o cloud prostate_path <- system.file("extdata", "prostate.csv", package = "h2o") prostate <- h2o.importFile(path = prostate_path, header = TRUE) set.seed(10) ### H2O provides 2 types of grid search for tuning the models, which are ### AutoML and Grid. Below, I demonstrate how weighted mean shapley values ### can be computed for both types. ####################################################### ### PREPARE AutoML Grid (takes a couple of minutes) ####################################################### # run AutoML to tune various models (GBM) for 60 seconds y <- "CAPSULE" prostate[,y] <- as.factor(prostate[,y]) #convert to factor for classification aml <- h2o.automl(y = y, training_frame = prostate, max_runtime_secs = 120, include_algos=c("GBM"), seed = 2023, nfolds = 10, keep_cross_validation_predictions = TRUE) ### call 'shapley' function to compute the weighted mean and weighted confidence intervals ### of SHAP values across all trained models. ### Note that the 'newdata' should be the testing dataset! result <- shapley(models = aml, newdata = prostate, performance_metric = "aucpr", plot = TRUE) ####################################################### ### PREPARE H2O Grid (takes a couple of minutes) ####################################################### # make sure equal number of "nfolds" is specified for different grids grid <- h2o.grid(algorithm = "gbm", y = y, training_frame = prostate, hyper_params = list(ntrees = seq(1,50,1)), grid_id = "ensemble_grid", # this setting ensures the models are comparable for building a meta learner seed = 2023, fold_assignment = "Modulo", nfolds = 10, keep_cross_validation_predictions = TRUE) result2 <- shapley(models = grid, newdata = prostate, performance_metric = "aucpr", plot = TRUE) ####################################################### ### PREPARE autoEnsemble STACKED ENSEMBLE MODEL ####################################################### ### get the models' IDs from the AutoML and grid searches. ### this is all that is needed before building the ensemble, ### i.e., to specify the model IDs that should be evaluated. library(autoEnsemble) ids <- c(h2o.get_ids(aml), h2o.get_ids(grid)) autoSearch <- ensemble(models = ids, training_frame = prostate, strategy = "search") result3 <- shapley(models = autoSearch, newdata = prostate, performance_metric = "aucpr", plot = TRUE) #plot all important features shapley.row.plot(shapley, row_index = 11) #plot only the given features shapPlot <- shapley.row.plot(shapley, row_index = 11, features = c("PSA", "AGE")) # inspect the computed data for the row 11 ptint(shapPlot$rowSummarySHAP) ## End(Not run)
This function normalizes a vector based on specified minimum and maximum values. If the minimum and maximum values are not specified, the function will use the minimum and maximum values of the vector.
shapley.test(shapley, features, n = 5000)shapley.test(shapley, features, n = 5000)
shapley |
object of class 'shapley', as returned by the 'shapley' function |
features |
character, name of two features to be compared with permutation test |
n |
integer, number of permutations |
normalized numeric vector
E. F. Haghish
## Not run: # load the required libraries for building the base-learners and the ensemble models library(h2o) #shapley supports h2o models library(autoEnsemble) #autoEnsemble models, particularly useful under severe class imbalance library(shapley) # initiate the h2o server h2o.init(ignore_config = TRUE, nthreads = 2, bind_to_localhost = FALSE, insecure = TRUE) # upload data to h2o cloud prostate_path <- system.file("extdata", "prostate.csv", package = "h2o") prostate <- h2o.importFile(path = prostate_path, header = TRUE) ### H2O provides 2 types of grid search for tuning the models, which are ### AutoML and Grid. Below, I demonstrate how weighted mean shapley values ### can be computed for both types. set.seed(10) ####################################################### ### PREPARE AutoML Grid (takes a couple of minutes) ####################################################### # run AutoML to tune various models (GBM) for 60 seconds y <- "CAPSULE" prostate[,y] <- as.factor(prostate[,y]) #convert to factor for classification aml <- h2o.automl(y = y, training_frame = prostate, max_runtime_secs = 120, include_algos=c("GBM"), # this setting ensures the models are comparable for building a meta learner seed = 2023, nfolds = 10, keep_cross_validation_predictions = TRUE) ### call 'shapley' function to compute the weighted mean and weighted confidence intervals ### of SHAP values across all trained models. ### Note that the 'newdata' should be the testing dataset! result <- shapley(models = aml, newdata = prostate, plot = TRUE) ####################################################### ### Significance testing of contributions of two features ####################################################### shapley.test(result, features = c("GLEASON", "PSA"), n=5000) ## End(Not run)## Not run: # load the required libraries for building the base-learners and the ensemble models library(h2o) #shapley supports h2o models library(autoEnsemble) #autoEnsemble models, particularly useful under severe class imbalance library(shapley) # initiate the h2o server h2o.init(ignore_config = TRUE, nthreads = 2, bind_to_localhost = FALSE, insecure = TRUE) # upload data to h2o cloud prostate_path <- system.file("extdata", "prostate.csv", package = "h2o") prostate <- h2o.importFile(path = prostate_path, header = TRUE) ### H2O provides 2 types of grid search for tuning the models, which are ### AutoML and Grid. Below, I demonstrate how weighted mean shapley values ### can be computed for both types. set.seed(10) ####################################################### ### PREPARE AutoML Grid (takes a couple of minutes) ####################################################### # run AutoML to tune various models (GBM) for 60 seconds y <- "CAPSULE" prostate[,y] <- as.factor(prostate[,y]) #convert to factor for classification aml <- h2o.automl(y = y, training_frame = prostate, max_runtime_secs = 120, include_algos=c("GBM"), # this setting ensures the models are comparable for building a meta learner seed = 2023, nfolds = 10, keep_cross_validation_predictions = TRUE) ### call 'shapley' function to compute the weighted mean and weighted confidence intervals ### of SHAP values across all trained models. ### Note that the 'newdata' should be the testing dataset! result <- shapley(models = aml, newdata = prostate, plot = TRUE) ####################################################### ### Significance testing of contributions of two features ####################################################### shapley.test(result, features = c("GLEASON", "PSA"), n=5000) ## End(Not run)
This function applies different criteria simultaniously to identify the most important features in a model. The criteria include: 1) minimum limit of lower weighted confidence intervals of SHAP values relative to the feature with highest SHAP value. 2) minimum limit of percentage of weighted mean SHAP values relative to over all SHAP values of all features. These are specified with two different cutoff values.
shapley.top(shapley, lowerci = 0.01, shapratio = 0.005)shapley.top(shapley, lowerci = 0.01, shapratio = 0.005)
shapley |
object of class 'shapley', as returned by the 'shapley' function |
lowerci |
numeric, specifying the lower limit of weighted confidence intervals of SHAP values relative to the feature with highest SHAP value. the default is 0.01 |
shapratio |
numeric, specifying the lower limit of percentage of weighted mean SHAP values relative to over all SHAP values of all features. the default is 0.005 |
data.frame of selected features
E. F. Haghish
## Not run: # load the required libraries for building the base-learners and the ensemble models library(h2o) #shapley supports h2o models library(shapley) # initiate the h2o server h2o.init(ignore_config = TRUE, nthreads = 2, bind_to_localhost = FALSE, insecure = TRUE) # upload data to h2o cloud prostate_path <- system.file("extdata", "prostate.csv", package = "h2o") prostate <- h2o.importFile(path = prostate_path, header = TRUE) ### H2O provides 2 types of grid search for tuning the models, which are ### AutoML and Grid. Below, I demonstrate how weighted mean shapley values ### can be computed for both types. set.seed(10) ####################################################### ### PREPARE AutoML Grid (takes a couple of minutes) ####################################################### # run AutoML to tune various models (GBM) for 60 seconds y <- "CAPSULE" prostate[,y] <- as.factor(prostate[,y]) #convert to factor for classification aml <- h2o.automl(y = y, training_frame = prostate, max_runtime_secs = 120, include_algos=c("GBM"), # this setting ensures the models are comparable for building a meta learner seed = 2023, nfolds = 10, keep_cross_validation_predictions = TRUE) ### call 'shapley' function to compute the weighted mean and weighted confidence intervals ### of SHAP values across all trained models. ### Note that the 'newdata' should be the testing dataset! result <- shapley(models = aml, newdata = prostate, plot = TRUE) ####################################################### ### Significance testing of contributions of two features ####################################################### shapley.top(result, lowerci = 0.01, shapratio = 0.005) ## End(Not run)## Not run: # load the required libraries for building the base-learners and the ensemble models library(h2o) #shapley supports h2o models library(shapley) # initiate the h2o server h2o.init(ignore_config = TRUE, nthreads = 2, bind_to_localhost = FALSE, insecure = TRUE) # upload data to h2o cloud prostate_path <- system.file("extdata", "prostate.csv", package = "h2o") prostate <- h2o.importFile(path = prostate_path, header = TRUE) ### H2O provides 2 types of grid search for tuning the models, which are ### AutoML and Grid. Below, I demonstrate how weighted mean shapley values ### can be computed for both types. set.seed(10) ####################################################### ### PREPARE AutoML Grid (takes a couple of minutes) ####################################################### # run AutoML to tune various models (GBM) for 60 seconds y <- "CAPSULE" prostate[,y] <- as.factor(prostate[,y]) #convert to factor for classification aml <- h2o.automl(y = y, training_frame = prostate, max_runtime_secs = 120, include_algos=c("GBM"), # this setting ensures the models are comparable for building a meta learner seed = 2023, nfolds = 10, keep_cross_validation_predictions = TRUE) ### call 'shapley' function to compute the weighted mean and weighted confidence intervals ### of SHAP values across all trained models. ### Note that the 'newdata' should be the testing dataset! result <- shapley(models = aml, newdata = prostate, plot = TRUE) ####################################################### ### Significance testing of contributions of two features ####################################################### shapley.top(result, lowerci = 0.01, shapratio = 0.005) ## End(Not run)
This function performs a weighted permutation test to determine if there is a significant difference between the means of two weighted numeric vectors. It tests the null hypothesis that the difference in means is zero against the alternative that it is not zero.
test(var1, var2, weights, n = 2000)test(var1, var2, weights, n = 2000)
var1 |
A numeric vector. |
var2 |
A numeric vector of the same length as |
weights |
A numeric vector of weights, assumed to be the same for both |
n |
The number of permutations to perform (default is 2000). |
A list containing the observed difference in means and the p-value of the test.