Modeling Red Wine Quality Based on Physicochemical Tests: A Data Mining Approach

Abstract

Classification of the quality of red wine is done in the hope of making it easier to assess the quality of red wine. Data used for this research is the wine quality data set with 4898 number of instances, obtained from UCI machine learning repository. Classification of the quality of red wine this study was carried out by comparing the three algorithms of data mining, that is random forest, naive bayes and generalized linear model. From the results of this study comparing the three algorithms, the generalized linear model showed the highest accuracy among the other algorithms. It was tested with a generalized linear model with 68.75% accuracy results, this algorithm is ideal for classifying the quality of red wine. In addition, a secondary random forest gives 67.81% accuracy results, while Naive Bayes gives 61.25% accuracy results. Studies conducted to classify the quality of red wine based on its composition use a generalized linear model for the optimal algorithm.