SHAP values in association with feature values
Examples
# \donttest{
library("explainer")
seed <- 246
set.seed(seed)
# Load necessary packages
if (!requireNamespace("mlbench", quietly = TRUE)) stop("mlbench not installed.")
if (!requireNamespace("mlr3learners", quietly = TRUE)) stop("mlr3learners not installed.")
if (!requireNamespace("ranger", quietly = TRUE)) stop("ranger not installed.")
# Load BreastCancer dataset
utils::data("BreastCancer", package = "mlbench")
target_col <- "Class"
positive_class <- "malignant"
mydata <- BreastCancer[, -1]
mydata <- na.omit(mydata)
sex <- sample(
c("Male", "Female"),
size = nrow(mydata),
replace = TRUE
)
mydata$age <- as.numeric(sample(
seq(18, 60),
size = nrow(mydata),
replace = TRUE
))
mydata$sex <- factor(
sex,
levels = c("Male", "Female"),
labels = c(1, 0)
)
maintask <- mlr3::TaskClassif$new(
id = "my_classification_task",
backend = mydata,
target = target_col,
positive = positive_class
)
splits <- mlr3::partition(maintask)
mylrn <- mlr3::lrn(
"classif.ranger",
predict_type = "prob"
)
mylrn$train(maintask, splits$train)
SHAP_output <- eSHAP_plot(
task = maintask,
trained_model = mylrn,
splits = splits,
sample.size = 2, # also 30 or more
seed = seed,
subset = 0.02 # up to 1
)
#> Warning: Ignoring unknown aesthetics: text
shap_Mean_long <- SHAP_output[[3]]
myplot <- ShapFeaturePlot(shap_Mean_long)
#> Warning: Removed 4 rows containing non-finite outside the scale range (`stat_smooth()`).
# }