from molecules to clusters: unsupervised learning insights into perfume composition

This study presents a novel application of unsupervised machine learning techniques to analyze the molecular and evaporative characteristics of perfumery compounds. a dataset comprising molecular descriptors, structural notations, and physical properties of scent compounds has been prepared using three extensive sql databases, and some well-known methodological approaches including principal component analysis (pca) and factor analysis (fa) for dimensionality reduction and hierarchical clustering (hc) are implemented to identify intrinsic olfactory families without relying on pre-existing classes.