Machine Learning Modeling of Wine Sensory Profiles and Color of Vertical Vintages of Pinot Noir Based on Chemical Fingerprinting, Weather and Management Data

Abstract

Important wine quality traits such as sensory profile and color are the product of complex interactions between the soil, grapevine, the environment, management, and winemaking practices. Artificial intelligence (AI) and specifically machine learning (ML) could offer powerful tools to assess these complex interactions and their patterns through seasons to predict quality traits to winegrowers close to harvest and before winemaking. This study considered nine vintages (2008–2016) using near-infrared spectroscopy (NIR) of wines and corresponding weather and management information as inputs for artificial neural network (ANN) modeling of sensory profiles (Models 1 and 2 respectively). Furthermore, weather and management data were used as inputs to predict the color of wines (Model 3). Results showed high accuracy in the prediction of sensory profiles of vertical wine vintages using NIR (Model 1; R = 0.92; slope = 0.85), while better models were obtained using weather/management data for the prediction of sensory profiles (Model 2; R = 0.98; slope = 0.93) and wine color (Model 3; R = 0.99; slope = 0.98). For all models, there was no indication of overfitting as per ANN specific tests. These models may be used as powerful tools to winegrowers and winemakers close to harvest and before the winemaking process to maintain a determined wine style with high quality and acceptability by consumers.

Review of the Effects of Grapevine Smoke Exposure and Technologies to Assess Smoke Contamination and Taint in Grapes and Wine

“…Recent models based on AI and specifically using machine learning have been based on the estimation steps for chemometrics [5,6,90], smoke-related compounds [5,6], aroma profiles [15,103], and sensory information of final wines [6,15,90,[108][109][110], which give more information of changes at different stages and deal with the better understanding of intermediary processes while doing amelioration techniques at various stages (Tables 4 and 5). In this way, the AI models developed to offer further intermediary information rather than a direct AI model, which gives specific targets that can be considered a black box.…”

Machine Learning in Chemical Product Engineering: The State of the Art and a Guide for Newcomers

“…Chemometrics methods, such as ANOVA or PCA, are most often applied before ML methods to select and compress the original data [156][157][158][159][160][161][162][163][164]. Feature extraction methods such as Fourier analysis or Si-PLS are also used for extracting relevant information or optimal spectral interval from high dimensional spectra measurements [156,158].…”

“…A typical example concerns wine, where important quality traits, such as the sensory profile and color are a product of complex interactions between the soil, grapevine, environment, management and winemaking practices. ANN has been shown to be an efficient tool in assessing these complex interactions and predicting wine sensory properties from NIR spectra and from weather and water management information [163]. This example is illustrative of the way AI can present an opportunity for winemakers to adjust vinification techniques in order to obtain a more consistent wine style, predict market and consumer acceptance for pricing adjustments and provide better description of wines on labels for accurate information to consumers.…”

“…Finally, it is worth noting new emerging technologies that, combined with ML, enable performing analyses in a more standardized and rapid manner, such as robotic pourers with computer vision, electronic tongue or nose sensors or low-cost NIR spectroscopic devices and color sensors, attachable to smartphones with applications in food and beverages [163,169,170].…”

Integrating a Low-Cost Electronic Nose and Machine Learning Modelling to Assess Coffee Aroma Profile and Intensity

“…Furthermore, this is most convenient when a single multitarget model is required, which is a feature that is not possible in other ML algorithms [ 47 , 48 ]. Some studies that have presented multitarget ANN models include the identification of proteomics from beer foamability analysis [ 49 ], prediction of beer aromas and physicochemical data using foam-related parameters obtained from a robotic pourer [ 50 ], prediction of beer acceptability of sensory attributes using e-nose data [ 22 ], and prediction of sensory profiles of wine using near-infrared spectroscopy [ 51 ].…”