Answers are grounded in this paper and cite their sources.
In modern world, wine has become a part and pencil of life and culture. With the improvement of production techniques, wine making has been turned into as a form of art and a branch of science. Italian wine is very popular because of its variation in taste. The taste of wine depends on different ty pes of cultivars. This paper attempts to classify the cultivars on the basis of different chemical constituents recorded as wine data. To accomplish this task, we used linear discriminant analysis (LDA), multinomial logistic regression (MLR), random forest (RF) and support vector machine (SVM) classification techniques. We have analyzed these in the absence of outliers and in the presence of different rate of outliers. In both of the cases, bootstrapping is used due to small data. We have used the accuracy, sensitivity and specificity as the measuring criteria of classification techniques. In absence of the outlier, LDA gives maximum classification accuracy, sensitivity and specificity. When the percentage of outlier is increases, the performance of RF tends to get better than LDA. Generally, we can suggest LDA when such type of data is obtained in the absence of outliers and RF in the presence of outliers.
No takes yet. Share an insight, caveat, or question.
Ahammed et al. (2018) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: