MCDABENCH: An R Package for Benchmarking Multi-Criteria Decision Making Methods and Scenarios

This paper introduces mcdabench, an R package for Multi-Criteria Decision Making (MCDM), which contains several useful tools, including data normalization, objective weighting, implementation of various well-known methods and their comparisons, sensitivity and stability analyses, benchmarking, aggregation, and visualization. The methods amount to 38 widely used MCDM methods from the literature. Sensitivity and stability analyses utilize measures such as the rank stability index and rank volatility. These features are enhanced by two novel contributions: Generalized gradual-weighting and random weight perturbations, both of which are important for evaluating the robustness of MCDM methods. The benchmark function applies the MCDM methods and returns a combined ranking matrix, which the comparison function then uses to calculate pairwise comparison metrics and perform statistical tests. The statistical tests include Spearman's rank correlation, Sałabun-Urbaniak similarities, and novel contributions such as Shannon’s entropy-based permutation tests and Jensen-Shannon Divergence-based bootstrap tests. The aggregation function provides a variety of methods, such as Borda Count, Kemeny-Young, Copeland, and the Markov method, to generate a final ranking. In addition, a useful "top-k" technique is used to focus on the top-k alternatives. The mcdabench package also provides built-in visualization functions to present analysis results in clean, detailed demonstrations.