Computational Comparison of Reference-Based Similarity Measures for Partitioning Soft Spaces

Abstract

In this paper, we present a comparative computational analysis of five reference- based similarity measures for soft sets, namely the Jaccard, Sørensen–Dice, Cosine, Overlap and Matching coefficient similarities. An experimental soft space is generated using a nutritional awareness model, and the partitioning behavior of the similarity measures is investigated using equal-width and equal-frequency partitioning strategies. The consistency of the rankings produced by the similarity measures is further examined using Spearman's rank correlation coefficient. The computational results show that the Jaccard similarity provides the finest partition under the equal-width strategy, while all five measures produce highly consistent rankings. The proposed framework offers an effective approach for analyzing and comparing the partitioning behavior of reference-based similarity measures on soft spaces.