Wine Quality Prediction Using Supervised and Unsupervised Machine Learning Techniques

Wine quality and its assurance are crucial for the wine-making business. The traditional method of measuring the quality of wine is expensive and time-consuming. Innovative measures, such as using machine learning tools, can replace the methods used to determine the quality of wine. Key features or attributes must be identified to predict the quality of the wine. It is important to note that not all wines possess these relevant features due to factors such as origin, chemical composition, etc. For this reason, this work explored identifying the key features required for wine quality. The project also aimed to investigate two machine learning techniques for classifying wine quality when provided with the same data (red and white wine). This project also aims to study the impact of supervised and unsupervised machine learning on the same dataset. We used a random forest and a support vector machine (SVM) to classify wine quality under supervised learning. The unsupervised machine learning algorithm used is principal component analysis (PCA). The models that predicted wine quality the best according to metrics were supervised machine learning models. According to the results of the investigation, supervised machine learning is the best classifier, being a random forest with an accuracy score of 71% for white wine and 74% for red wine, as opposed to unsupervised machine learning, which scored less.