Predicting Wine Properties Based on Weather Conditions Using Machine Learning Techniques

Wine quality depends on different factors from cultivation to production. The main factors affecting the quality are weather and climate, growing practices of the vineyard and techniques used by winemakers. This paper explores the effectiveness of several machine learning algorithms to predict the quality based on various features. The wine dataset is prepared from a certification and quality assessment laboratory, containing various physicochemical characteristics such as alcohol content, volatile acids, total extracts, residual sugar, among others. Weather conditions, including precipitation levels, daily average temperature, temperatures exceeding 10°C, and relative air humidity, exhibit varying impacts on vineyards during different growth stages. Our analysis shows that Random Forest with SMOTE method outperforms all other classifiers with 73% accuracy. Similar outcomes are achieved using the RUSBoost ensemble method. Furthermore, we investigate how weather conditions impact the characteristics of white and red wines from diverse regions in North Macedonia, each with its own unique climate and soil conditions. Results indicate that high temperatures without precipitation during the ripening period positively affect wine quality. The analysis yielded a Pearson coefficient of -0.11 for the correlation between air humidity and alcohol content, and 0.19 for the correlation between average temperature and residual sugar levels.