Predicting Whisky Odors from Molecules
Author Information
Author(s): Satnam Singh, Doris Schicker, Helen Haug, Tilman Sauerwald, Andreas Grasskamp
Primary Institution: Fraunhofer Institute for Process Engineering and Packaging IVV, Freising, Germany
Hypothesis
Can machine learning algorithms accurately predict the sensory qualities of whiskies based on their molecular composition?
Conclusion
The study successfully demonstrates that machine learning can predict the sensory scores of whiskies with high accuracy, outperforming human panelists.
Supporting Evidence
- Machine learning algorithms achieved up to 100% accuracy in predicting whisky types based on molecular features.
- OWSum and CNN methods outperformed the average performance of human panelists in predicting odor descriptors.
- The study utilized a comprehensive dataset of 16 whisky samples for analysis.
Takeaway
This study shows that computers can help us guess what whiskies smell like by looking at the tiny parts they are made of.
Methodology
The study combined automated analytical tools, sensory data from 11 panelists, and machine learning algorithms to classify whiskies based on their molecular composition.
Potential Biases
The reliance on a small number of panelists may introduce bias in sensory evaluations.
Limitations
The dataset is limited and may not represent all types of American and Scotch whiskies.
Participant Demographics
11 experienced panelists were involved in the sensory evaluation.
Digital Object Identifier (DOI)
Want to read the original?
Access the complete publication on the publisher's website