This study investigates the sensory profiling of rum by integrating advanced analytical tools and machine learning techniques. Different rum samples were analyzed from diverse regions using a combination of analytical tools, including Headspace Solid-Phase Microextraction coupled with Gas Chromatography-Mass Spectrometry (HS-SPME-GC-MS), electronic noses (E-noses), and electronic tongues (E-tongues). These tools allowed us to identify patterns in chemical compositions and sensory characteristics. Machine learning, particularly unsupervised methods like clustering and Principal Component Analysis (PCA), was crucial in revealing clear groupings of rums based on their molecular behavior. , raw material classification (e.g., molasses vs. sugarcane juice), and regional origin. Significant flavor differences were identified, and critical aromatic descriptors were linked to potential chemical molecules. A Random Forest predictive model achieved an accuracy of over 80%, outperforming other algorithms like SVM and neural networks. These findings demonstrate the potential of combining chemical analysis with machine learning to improve rum classification, enhance sensory profiling, and support producers in creating consistent, high-quality products that align with consumer preferences.
PDF 
FLIP 
