Quantification of alcohol content and identification of fraud in traditional cachaças using NIR spectroscopy

Cachaça, a traditional Brazilian alcoholic beverage, plays a significant economic role in Brazil. Traditional quality control methods are expensive, requiring specialized labor and extended response times. This study proposes innovative and eco-friendly approaches using near-infrared (NIR) spectroscopy for alcohol content quantification and fraud detection, eliminating the need for chemical reagents. A dataset of 462 cachaças from the Brejo microregion of Paraíba, Brazil, was analyzed using variable selection algorithms: PLS, MCUVE-PLS, CARS-PLS, and iSPA-PLS. The MCUVE-PLS-DA/OFF model achieved 97.3 % accuracy in fraud detection, while the iSPA-PLS-DA/OFF model reached an RMSEP of 0.16 v/v, RPD of 59.04 and r of 0.99 for alcohol content determination. These models provide a sustainable and cost-effective alternative to traditional methods, enhancing interpretability and reducing computational costs while ensuring regulatory compliance and consistency in cachaça production.