Wine Type Classification Using Random Forest Model

Abstract

Wine Type Classification indicates that its indexes can ascertain the wine category. Therefore, it can be applied in modern industrial wine production and identification to reduce the rates of inferior products or to terminate the sale of homemade hooch or watered-down cheap alcohol. This paper explores Random Forest to classify wine. Since there are null values in the data, we first input the wine quality dataset and drop out the null values. Standard scaling is ignored because it expands the differences of data and the original datas are special for its distribution to deviation. Then, principal components analysis (PCA) is applied to reduce the dimensions of variable attributes. Finally, we perform random forest to the dataset to see the precision and F1 scores. We compare our methods with logistic regression, SVM, and naive Bayes model. The accuracies of these methods are 0.884375, 0.88125, and 0.884375, respectively. Our result shows that the random forest strategy generates promising accuracy of wine classification. Therefore, Random Forest can predict the industrial product quality and even can recognize the wine type with a high precision rate.

SFC: A Sampling from Clusters for Reduction of Dataset Size

“…The RF ensemble training technique is used to create a classifier model that divides the dataset into a number of smaller-scale datasets [30][31]. In the RF, Bootstrap aggregation is used for it is proved to be better in most situations of reducing variance to reduce the probability of over-fitting.…”