REPID: Regional Effect Plots with implicit Interaction Detection

This repository gives access to an implementation of the methods presented in the paper submission “REPID: Regional Effect Plots with implicit Interaction Detection”, as well as all code that was used for the simulations and the real-world example.

This repository is structured as follows:

    ├── R/                       # All implemented methods and general helper functions                          
    |   ├── simulations/         # Scripts for simulation examples in paper
    |   |   ├── analysis/        # Scripts used to create figures and tables in the paper for simulation examples
    |   |   ├── batchtools/      # Scripts used to create data for more complex simulation examples (Sec. 4.2, Appendix B)
    |   ├── real_world_examples/ # Scripts used for modelling and to create figures for real-world examples in Section 5 and Appendix B
    ├── data/                    # Location where all generated data are stored
    │   ├── batchtools/          # Location where generated data of batchtools experiments are stored
    │   ├── sim_vine_vs_repid/   # Location where generated data of 1. simulation example (VINE vs. REPID) are stored
    │   ├── sim_weak/            # Location where generated data of 2. simulation example (Weaknesses of other methods) are stored
    │   ├── titanic/             # Location where pre-processed data of titanic example are stored
    |   ├── california_housing/  # Location where pre-processed data of california housing example are stored
    |   notebooks/               # Python and R notebooks to generate data for VINE and titanic example
    ├── LICENSE
    └── README.md               

Reproduce Experiments

Steps to reproduce the experiments of Section 4.2 and Appendix B.2.

1. Install all required packages.

# from CRAN
install.packages(c("ranger", "dplyr", "batchtools", "mlr", "ggplot2", "gridExtra", "tidyr", "reshape2",
"ggpubr", "BBmisc", "data.table", "stringi", "stringr", "checkmate", "kernlab", "xtable", "devtools",
"tidyverse", "Rmalschains", "iml","kmlShape","dtw","egg","rlist","mgcv","mvtnorm", "vip", "data.table",
"e1071", "RColorBrewer", "R6", "sfsmisc", "mlr3", "xgboost"))


# from github
devtools::install_github("giuseppec/featureImportance")

2. Create an experimental registry, add experiments and problem and run simulations via script R/simulations/generate_data.R. Data produced by the scripts is stored in the subfolder data/batchtools/interaction_detection as a separate registry.

3. Prepare data for analysis by running the script R/simulations/reduce_experiments.R.

4. To reproduce figures and tables of Section 4.2 and Appendices B.2 and B.3, run the script R/simulations/analysis/analysis_sim_complex.R. Figures produced within the script are stored in figures.

To reproduce the experiments of Section 3 and 4.1, run the scripts R/simulations/analysis/analysis_sim_vine_vs_repid.R and R/simulations/analysis/analysis_sim_weak.R respectively. To reproduce the results of the real-world examples, run the scripts in R/real_world_example/