Classification of beverages according to their qualities using different algorithms

In the scope of this thesis, beverages are classified according to their qualities using different algorithms. The data set used is taken from the UCI (University of California, Irvine) Machine Learning Database. The data set is composed of two different data sets that include white and red wine samples. Firstly, the two data sets were merged and the wine samples are classified as red and white. Then, classification process were employed separately on red (6 different qualities) and white wine samples (7 different qualities) that were of different qualities. After that, quality and colour classification were applied on 13 different qualities of red and white wine samples. Support Vector Machine, k-Nearest Neighbours and Random Forest are the classification algorithms that are used in this study. Principle Component Analysis were applied on the data set for reduction purposes, as well as filter-based (Information Gain, Gain Ratio) and spiral-based methods for feature selection; and with using these, the classification process were repeated. The imbalance in the data set were eliminated by using SMOTE (Synthetic Minority Over-sampling Technique) random over-sampling and under-sampling algorithms and the same classification processes were re-applied. The performance measures that are used in the study are the precision, recall, F-measure and receiver operating characteristic.