Wine Price and Prediction Using Machine Learning

Abstract

The wine industry is a growing field across all countries around the globe. Wine selling and production are seen as a benefit for the economy of a country and its culture. Although most of the wine is sold and bought in the usual local supermarket for immediate consumption, many enthusiasts like to invest and appreciate more exquisite wines. More exquisite wines may be sold in different values depending on the region where they are being sold. If the wine is of difficult access in a region, if it is of a unique kind, and so on. These impactful factors could lead to a single bottle of wine to take absurdly huge prices. In this dissertation, the quality and price of wine attributes will be explored systematically by using a data-driven approach to generate prediction models. It is expected that these models will potentially help owners of wine, consumers or producers to provide reliable indicators on the price and quality of a targeted wine. In terms of the related work, there are research articles going as far back as 2008, which already apply a systematic approach to these factors, some of which include data science approaches. One of the most notable works, done by Drs. Michelle Yeo, Tristan Fletcher and John Shawe-Taylor and published by Cambridge University Press in the "Journal of Wine Economics, Volume 10", "Machine Learning in Fine Wine Price Prediction", applies machine learning models such as Gaussian process regression and multi-task learning to predict wine prices, and bases the model on wine historical price data. On the other hand in an article of "Wineinformatics: Regression on the Grade and Price of Wines through Their Sensory Attributes" authored by James Palmer and Bernard Chen, a dataset of consumers' reviews, with more than 105,085 wines reviews, is used to predict the quality and price of wine through Support Vector Regression models. Drs Paulo Cortez, Juliana Teixeira, António Cerdeira, Fernando Almeida, Telmo Matos, and José Reis applied the same model, on an article on "Using Data Mining for Wine Quality Assessment". Where "Vinho Verde" is explored through a dataset of physicochemical laboratory tests. In this dissertation, a new method is applied to predict the wine's price while simultaneously predicting its quality through a dataset focused on the global wine industry data.