Abstract: This work provides a hybrid technique for processing fuzzy binary soft set data that combines supervised reference matching with unsupervised clustering. Without the requirement for manual parameter calibration, patients are classified based on symptom patterns applying cosine similarity and an automatically generated dendrogram threshold. The cluster centroids are then aligned with the COVID-19 and flu reference disease profiles using averaged cosine similarity, allowing for a direct and understandable classification. The method achieves high accuracy, scalability, and interpretability while bridging the gap between fully supervised models and pure clustering.

Keywords: Fuzzy binary soft sets, cosine similarity, Clustering, MCDM problems.

VOLUME 9 ISSUE 12 2025: 165 – 172