Quality prediction of wine is a significant feature of the wine business and directly affects revenue, competition, and customer happiness. However, traditional approaches—like sensory assessment and chemical analysis—have generally been considered to have limits in terms of precision and effectiveness. This research compared which Random Woods, XGBoost logarithmic regression, Choice Tree Classifier, and k-nearest-neighbor algorithms were most effective in building and evaluating artificial intelligence classifiers for predicting the taste of red and white wines based on biological and physical attributes. We selected features and building strategies to improve the model on a dataset with 1,599 wine specimens with 13 molecular and physical properties. The trained models were then evaluated with reliability, precision, recollection, F1-score, and ROC-AUC metrics. In this regard, our findings indicated that class imbalance had been handled since the Random Forest and XGBoost gave the highest accuracy rates at 0.950158 and 0.950158, respectively. Unsurprisingly, the most important features that affect wine quality are pH, pH level, and sugar concentration. This case study Exhibit demonstrates how machine learning can appropriately and effectively forecast the quality of wine, providing producers with insightful information and thereby helping to build systems that will finally pay dividends to the wine business and the customers by making precise and successful prediction systems.
Read full abstract