The shimex package - SHIny Models EXplorer

The shimex package enables interactive exploration of machine learning predictive models using the shiny application. Thanks to the graphical interface, the application streamlines the process of model exploration.

The Explainers presented in the app:

• Ceteris Paribus from ingredients,
• BreakDown from iBreakDown,
• LIME from DALEXtra, iml, localmodel,
• SHAP from iBreakDown, shapper, iml.

Installation

# Development version from GitHub:
# install.packages("devtools")
devtools::install_github("ModelOriented/shimex")

Demo

library(randomForest)

# Create a model
model_rm <- randomForest(life_length ~., data = DALEX::dragons, ntree = 200)

# Wrap it into an explainer 
explainer <- DALEX::explain(model_rm)

# Choose an observation (it is possible to modify it through the app)
observation <- DALEX::dragons[1, ]

# Create shimex app for observation
shimex::create_shimex(explainer, observation)

Results can be found following the link.