A classification approach with different feature sets to predict the quality of different types of wine using machine learning techniques

In the past few years, with the availability of lot of wine brands it is difficult to identify the good quality wines. Good quality wine depends on the so many important factors such as chemical, scientific as well as technical factors. However in the previous study the researchers always focus on the subjective study to define the quality of wine. The result based on the subjective study takes much time as well as it is not effective compared to the objective study with the analytical methods. In the last few year's machine learning techniques caught lot of attention in every field. Most of the machines learning techniques are able to produce highly accurate result that compels most of the data scientist to implement it in case of predictive analytics. In the past few works related to wine data has been studied using different classifiers, however so far nobody has compared the performance metrics of the different classifiers with different feature sets to predict the quality of different type of wine by considering several factors. In this paper a new approach has been proposed by considering different feature selection algorithm such as Principal Component Analysis (PCA) as well as Recursive Feature Elimination approach (RFE) approach for feature selection and nonlinear decision tree based classifiers for analyzing the performance metrics. We found accuracies ranging from 94.51% to 97.79% with different feature sets using Random Forest classifier. This analysis will help the wine experts to know the important factors to consider while selecting the good quality wine.