ExplaineR

Overview

It enables detailed interpretation of complex classification and regression models through Shapley analysis including data-driven characterization of subgroups of individuals. Furthermore, it facilitates multi-measure model evaluation, model fairness, and decision curve analysis. Additionally, it offers enhanced visualizations with interactive elements. Find out how to use the package by checking out the package tutorials.

Installation

You can install the released version of the package from CRAN:

install.packages("explainer")

And the development version from GitHub (recommended for latest updates):

# install.packages("devtools")
devtools::install_github("PERSIMUNE/explainer")

Usage

Preparing your own data

DAIMS—Datasheets for AI and Medical Datasets provides a checklist, automated data checks, and templates for dataset documentation and data dictionaries. It can help identify preparation issues before model training and interpretation with ExplaineR. DAIMS is optional; passing its checks does not establish that a dataset, study design, or trained model is suitable for a particular analysis.

Example

This example classifies breast cytology samples as benign or malignant using the original predictors in mlbench::BreastCancer. It does not predict future breast-cancer risk. The ID column is excluded and rows with missing values are omitted. No synthetic demographic features are added. See articles for more examples.

# Load necessary packages
library("explainer")

# Set seed for reproducibility
seed <- 246
set.seed(seed)

# load necessary packages
library(mlbench)
library(mlr3learners)
library(ranger)

# Load BreastCancer dataset
utils::data("BreastCancer", package = "mlbench")

# Specify target column and positive class
target_col <- "Class"
positive_class <- "malignant"

# Extract relevant columns from dataset
mydata <- BreastCancer[, -1]
mydata <- na.omit(mydata)

# Create a classification task
maintask <- mlr3::TaskClassif$new(
  id = "my_classification_task",
  backend = mydata,
  target = target_col,
  positive = positive_class
)

# Split the dataset for training
splits <- mlr3::partition(maintask)

# Create a ranger learner for classification
mylrn <- mlr3::lrn("classif.ranger", predict_type = "prob")

# Train the learner on the training set
mylrn$train(maintask, splits$train)

# Generate SHAP values and plot
SHAP_output <- eSHAP_plot(
  task = maintask,
  trained_model = mylrn,
  splits = splits,
  sample.size = 30,
  seed = seed,
  subset = 0.8
)

shap_Mean_wide <- SHAP_output[[2]]

shap_Mean_long <- SHAP_output[[3]]

# Generate SHAP clusters and plot
SHAP_plot_clusters <- SHAPclust(
  task = maintask,
  trained_model = mylrn,
  splits = splits,
  shap_Mean_wide = shap_Mean_wide,
  shap_Mean_long = shap_Mean_long,
  num_of_clusters = 3,
  seed = seed,
  subset = 0.8
)

# Please note that the colors on the legend on SHAP cluster plot is not the same as in the SHAP summary plot, but the markers are the same.
# The markers indicate prediction correctness not the colors (so the colors could be manually changed to black/white on the legend when reporting)

Documentation

• Official Documentation

Research citing ExplaineR

Cited by 14 papers on Google Scholar · As of 3 November 2026.

ExplaineR is cited across medical research, food science, engineering, sustainability, and machine learning. All 14 citing works are listed below, grouped by research area.

Research area	Citing publication
Cancer research	Lung cancer outcomes after immunotherapy
Cancer research	R-based protocols for predicting synthetic lethal interactions
Cell biology	AI for cell biology discovery
Food science and microbiology	Sausage spoilage intensity prediction with explainable ML and data augmentation
Food science and microbiology	VOC profiles for sausage spoilage prediction
Food science and microbiology	Lactococcus abundance prediction in rump steak (preprint)
Biomass and bioprocessing	Predicting enzymatic hydrolysis yields of lignocellulosic biomass
Environmental sustainability	Entrepreneurship and Sustainable Development Goal achievement
Energy policy	Forecasting fossil fuel trends in the EU-27
Industrial manufacturing	AI-driven smart manufacturing in the aluminum industry
Construction materials	Permeability prediction for recycled concrete aggregates
Earthquake engineering	Multitask genetic programming and seismic building vulnerability
Wellbeing and social research	SHAP-value clustering of life satisfaction determinants
Interpretable machine learning	nn2poly: converting neural networks into interpretable polynomials

Used ExplaineR in your research? Share your work to be featured here.

Citation

If you use this package in your research, please consider citing it:

Zargari Marandi, Ramtin. "ExplaineR: an R package to explain machine learning models." Bioinformatics Advances 4, no. 1 (2024): vbae049.doi:10.1093/bioadv/vbae049

Contributing

If you want to contribute to the development of this package, please read CONTRIBUTING.md for guidelines.

Issues

If you encounter any issues or have suggestions, please open an issue.

Major updates

23-09-2024: a new function for multi-class classification is added, check out the tutorial for multi-class classification on how to use the function (eSHAP_plot_multiclass.R).

License

This package is released under the MIT License.

References

Zargari Marandi R, Leung P, Sigera C, Murray DD, Weeratunga P, Fernando D, Rodrigo C, Rajapakse S, MacPherson CR, (2023). Development of a machine learning model for early prediction of plasma leakage in suspected dengue patients. PLoS Negl Trop Dis 17(3): e0010758. doi:10.1371/journal.pntd.0010758

Lang M, Binder M, Richter J, Schratz P, Pfisterer F, Coors S, Au Q, Casalicchio G, Kotthoff L, Bischl B, (2019). mlr3: A modern object-oriented machine learning framework in R. Journal of Open Source Software. doi:10.21105/joss.01903

Molnar C, Bischl B, Casalicchio G, (2018). iml: An R package for Interpretable Machine Learning.JOSS, 3(26), 786. doi:10.21105/joss.00786

Lundberg SM and Lee SI, (2017). A unified approach to interpreting model predictions. Advances in neural information processing systems.arXiv:1705.07874

Ludvig Renbo Olsen and Hugh Benjamin Zachariae, (2022). cvms: Cross-Validation for Model Selection. R package version 1.3.4. https://CRAN.R-project.org/package=cvms

R Core Team (2021). R: A language and environment for statistical computing. R Foundation for Statistical Computing, Vienna, Austria. URL https://www.R-project.org/.

Acknowledgements

Many thanks to my colleagues at CHIP/PERSIMUNE to support this work by their feedback. In addition, thanks to the CRAN team as well as the authors and maintainers of all the utilized packages for their dedication to advancing the R ecosystem.

Funding

This work was supported by the Danish National Research Foundation (DNRF126).