Evaluation of Arabica Coffee Fermentation Using Machine Learning.

This study explores the variances in the organic, chemical, and sensory attributes of fermented coffee beans, specifically examining how post-harvest processes influence cup quality. Coffee fruits from the Catuaí IAC-144 variety were processed using both natural coffee (NC) and pulped coffee (PC) methods. The fruits were then subjected to self-induced anaerobic fermentation (SIAF) using one of the following fermentation methods: solid-state fermentation (SSF) or submerged fermentation (SMF). Within these methods, either spontaneous fermentation (SPF) or starter culture fermentation (SCF) was applied. Each method was conducted over periods of 24, 48, and 72 h. For this purpose, two-hundred-liter bioreactors were used, along with two control treatments. Numerous parameters were monitored throughout the fermentation process. A comprehensive chemical profiling and sensory analysis, adhering to the guidelines of the Specialty Coffee Association, were conducted to evaluate the influence of these fermentation processes on the flavor, aroma, and body characteristics of the coffee beverage across multiple dimensions. Data analysis and predictive modeling were performed using machine learning techniques. This study found that NC exhibited a higher production of acids (citric, malic, succinic, and lactic) compared to PC, resulting in distinct chemical and sensory profiles. The decision tree showed that fructose and malic and succinic acids were identified as the main factors enhancing sensory notes during cupping. SMF promoted higher concentrations of lactic acid, while SSF led to increased ethanol content. Consequently, the SIAF process enhances the sensory quality of coffee, adding value to the product by generating diverse sensory profiles.