Introduction to SAPP

Overview

SAPP (Sector-Adjusted Points Plot) is an R package that visualizes feature dominance in a two-dimensional space. It combines PCA for dimensionality reduction with a novel sector-adjustment mechanism to show which features drive predictions for which observations.

Quick Start

Load the package and prepare your data:

library(sappviz)
data(iris)
X <- iris[, 1:4]

Fit a linear model and compute importances:

model <- lm(Petal.Width ~ Sepal.Length + Sepal.Width + Petal.Length, data = iris)
imp <- abs(coef(model)[-1])
names(imp) <- c("Sepal.Length", "Sepal.Width", "Petal.Length")

Compute per-observation influence and plot:

inf <- influence_feature(X, model)
plot_sapp(X, imp, inf, alpha = "auto")
#> Auto-Alpha selected: alpha = 0.54
#> Warning: Using `size` aesthetic for lines was deprecated in ggplot2 3.4.0.
#> ℹ Please use `linewidth` instead.
#> ℹ The deprecated feature was likely used in the sappviz package.
#>   Please report the issue at <https://github.com/FaresAminu/sappviz/issues>.
#> This warning is displayed once per session.
#> Call `lifecycle::last_lifecycle_warnings()` to see where this warning was
#> generated.
#> Warning: The following aesthetics were dropped during statistical transformation: size.
#> ℹ This can happen when ggplot fails to infer the correct grouping structure in
#>   the data.
#> ℹ Did you forget to specify a `group` aesthetic or to convert a numerical
#>   variable into a factor?

Interpretation

Colors: The feature that most influenced each observation.
Point size: How strongly that feature dominated.
Black markers: Sector centers, based on global importances.
Contours: Where each feature tends to dominate.

Advanced Usage: Random Forest

SAPP supports tree-based models via SHAP values (requires fastshap):

library(randomForest)
rf <- randomForest(Species ~ ., data = iris, importance = TRUE)
imp_rf <- importance(rf)[, "MeanDecreaseGini"]
names(imp_rf) <- colnames(X)
inf_rf <- influence_feature(X, rf)
plot_sapp(X, imp_rf, inf_rf, alpha = "auto")

References

Lundberg, S. M., and Lee, S.-I. (2017). A Unified Approach to Interpreting Model Predictions. NIPS.