A Criterion-Driven Consistency Indicator for Evaluating Multicriteria Sorting and Clustering Results

This study investigates the role of data structure in multicriteria sorting by integrating supervised and unsupervised approaches. Specifically, a hybrid framework combining TOPSIS-Sort-B and cluster analysis is proposed to define class boundaries and evaluate sorting quality. Unlike traditional studies that focus primarily on methodological performance, this work emphasizes the impact of criteria conflict and trade-offs on class formation and stability. A unified performance-based labeling scheme is introduced, and a Criterion-Driven Consistency Indicator (CDCI) is used to quantify intra-class similarity. This indicator assesses the extent to which alternatives within the same class exhibit similar performance across criteria, offering a complementary perspective to conventional distance-based metrics. The proposed framework is validated through multiple case studies with distinct structural characteristics, including a highly structured dataset, a trade-off-intensive electric vehicle dataset, and an intermediate supplier selection problem. The results show that sorting outcomes are largely driven by the intrinsic structure of the data rather than by the choice of method. Datasets with low criteria conflict yield high class consistency and clear separation, whereas strong trade-offs lead to reduced cohesion and overlapping class boundaries, especially for intermediate alternatives. Overall, the study demonstrates that incorporating criteria-level information is essential for the robust evaluation of multicriteria sorting. The proposed approach enhances interpretability, reduces subjectivity in class definition, and provides new insights into the relationship between data structure and sorting consistency.