RESUMEN
In this work, we investigate alternative quantification methods based on optical spectroscopy and variable selection approach, using as a case study the aging process of cachaça. The cachaça was aged in an Amburana barrel and the samples were analyzed with a traditional analytical method for total phenolic quantification (Folin-Ciocalteu), with 2-D fluorescence and infrared spectroscopy. We applied a methodology based on Ant Colony Optimization to optimize variable selection and model fitting to predict total phenolic content. Our results demonstrated that fluorescence spectroscopy was more sensitive than IR in the quantification of total phenolic compounds for both global and local models, presenting good results (R2 superior to 0.979), significantly reducing the number of original variable (1995) for only 4 pairs of Ex/Em. Variable selection combined with spectroscopy reveals potential because this technique eliminates the need for sample preparation and allows the construction of customized sensors for application in the food industry.