Empirical Dataset and Computational Implementation of the SODIF Hybrid AHP–Entropy–TOPSIS Model for Supplier Selection

This repository contains the results of the empirical phase of the dissertation titled “Hybrid Multicriteria Model Integrated with Business Intelligence to Optimize Supplier Selection.” It provides all materials required to ensure transparency and reproducibility of the computational procedures and results.

The repository includes the dataset zenodo_dataset.csv, containing the evaluation of five supplier alternatives across four criteria (Cost, Quality, Delivery, and Flexibility). It also includes the R compilation report of the proposed prototype (Hybrid_Multicriteria_Methodology), which implements a semi-automated hybrid approach for supplier selection. These files enable full reproduction of the hypothetical case presented in the dissertation.

The proposed SODIF model follows an organizational structure inspired by the Analytic Hierarchy Process (AHP), particularly in the hierarchical arrangement of criteria and alternatives. However, unlike traditional AHP-based approaches, the model does not rely on subjective pairwise comparisons. Instead, it applies entropy-based weighting to objectively derive the importance of criteria directly from the data. Subsequently, the TOPSIS method is used to aggregate the weighted performance values and compute the relative closeness of each supplier to the ideal solution, generating the final ranking.

Additionally, a script reporting execution times and a performance analysis of the model (first simulation) is provided, supporting the computational assessment of the proposed approach.

A second simulation is also included for cross-validation and comparative purposes with the study of Chen (2020). This simulation is based on the original dataset reported in that study; however, minor adaptations were introduced in the structuring and coding of criteria to ensure compatibility with the proposed computational framework and to standardize the implementation workflow. In particular, the original nomenclature of criteria was simplified and normalized (e.g., “Rate of qualified products” to “Quality”, and “Delivery on time ratio” to “Punctuality”) to improve code readability and ensure consistency across simulations. These adjustments are purely nominal and do not alter the conceptual definition or measurement of the criteria.

In this second simulation, the model is executed using Chen’s dataset under a criterion-by-criterion exclusion procedure (compilation file provided). This process enables the analysis of ranking behavior under controlled perturbations and supports the construction of the dataset required for the subsequent sensitivity analysis. The sensitivity analysis itself was conducted in SPSS, and the corresponding file is also included.

Legal notice: No license is granted. All rights reserved. Copying, modification, or redistribution of any part of this material is prohibited without prior written permission from the author. The materials are made available solely for academic record and reproducibility purposes.