MADRID: Clusterização hierárquica integrada à modelagem multicritério para tomada de decisão com apoio à identificação de critérios relevantes no cenário de aplicação

Abstract

ABSTRACT The study proposes the development of the Multicriteria Analysis Directed by Ranking and Integrated Data-clustering (MADRID) method, an innovative approach that integrates Clustering, an unsupervised Machine Learning technique, with Multicriteria Decision Making (MCDM), focusing on optimizing the acquisition process of new assets for the Brazilian Navy to strengthen its operational capabilities to meet the guidelines established in the "Plano Estratégico da Marinha (PEM-2040)", the National Defense Policy and Strategy, and the National Defense White Paper, ensuring national sovereignty, the protection of jurisdictional waters, and the fulfilment of strategic missions. The proposed method applies clustering using the hierarchical agglomerative single linkage method, employing the hybrid DIBR-MABAC method to rank clusters based on the values of their centroids, ranking the alternatives within each cluster, and performing the final ranking of the method, which consists of the alternatives contained in the final decision matrix. This matrix includes all the alternatives from the highest-ranked cluster and the best-evaluated alternatives from the other clusters. The case study assessed a group of six unmanned aerial vehicles (UAVs) based on seven criteria. To conduct the sensitivity analysis, this study chose to compare rankings generated by varying the criteria weights. For this purpose, two scenarios were created where the weights were modified. The MADRID method ranked the UAVs and correlated highly with the results obtained from both scenarios, demonstrating its robustness and applicability. A computational tool was also developed to implement the MADRID method, providing an intuitive and user-friendly interface for practical application. During the presentation of the computational platform, the software was applied to a decision matrix with hypothetical aircraft data, achieving a correlation of 0.96 with the COCOSO method and 0.97 with the TOPSIS method, highlighting MADRID as a practical and precise tool in decision support. Finally, the MADRID method is a relevant contribution to evaluating defense products and decision-making processes, with potential applications across various fields of knowledge.