The main objective of cooperative Transferable-Utility games (TU-games) is to allocate a good among the agents involved. The package implements major solution concepts including the Shapley value, Banzhaf value, and egalitarian rules, alongside their extensions for structured games: the Owen value and Banzhaf-Owen value for games with a priori unions, and the Myerson value for communication games on networks. To address the inherent exponential computational complexity of exact evaluation, the package offers both exact algorithms and sampling-based approximation methods, enabling the analysis of large-scale games. Additionally, it supports core set-based solutions, allowing computation of the vertices and the centroid of the core.
You can install the released version of the package from CRAN with:
install.packages("TUvalues")And the development version from GitHub with:
devtools::install_github("mariaguilleng/TUvalues")The repository also contains the code of the experiments in the paper:
Camacho, J., Gonçalves-Dosantos, J. C., & Guillen, M. D. TUvalues: Exact and Approximate Allocation Rules for Transferable-Utility Games in R. SoftwareX (under review).
These folders are not part of the CRAN package:
examples/: the usage examples of Section 3.3, with
exact (exact_example.R) and approximate
(approximate_example.R) calculations.benchmarks/: scalability and approximation-accuracy
experiments (Section 3.2). run_benchmarks.R runs all the
games and builds the figures and tables.package_comparison/: comparison with TUGLab, CoopGame
and GameTheory (Section 3.1).Run the scripts from the repository root, for example:
Rscript benchmarks/run_benchmarks.RBesides TUvalues, the experiments use ggplot2, readxl, writexl, callr, geometry, microbenchmark, TUGLab, CoopGame and GameTheory.